View Full Version : The NEXT LAST R600 Rumours & Speculation Thread
LOL just the g80, wanted to see the benefits of increased bandwidth vs. core clocks :)
Evildeus
16-Apr-2007, 17:04
LOL just the g80, wanted to see the benefits of increased bandwidth vs. core clocks :)
Yeah i know, just teasing :razz:
Mcmlxxx IV
16-Apr-2007, 17:42
http://www.xtremesystems.org/forums/showpost.php?p=2131444&postcount=29
The X2900 XT (512mb/256bit) is slower then 8800 gtx.
The 2900 XT/512mb does 10200 in 3D 06 stock clocks vs 8800 gtx @ 10900
both using QX6700 @ 3.6 testing.
256bit? Typo perhaps? This Dynasty guy is apparently one of those hardcore overclockers who often knows whats going on. :smile:
I'm dying here! Please release the cards! :lol:
Those scores look pretty low for that system. My friend got a mid 12k, I believe (maybe higher, I forget what it was), with a stock GTX and a Core 2 duo running at 3.8. A quad core running at 3.6 should get somewhat higher than that with a GTX, let alone the r600 XT.
Not that i believe that the 256 bit rumours are true but which nr's are you using to get 8800 GTX rivaling performance ? 3D Mark ?
Just the general consensus and Gibbo's hints, although I know that Geo will not be pleased about the latter ;)
I just don't see a very good reason for him to spout lies in the face of an impending launch.
Just the general consensus and Gibbo's hints, although I know that Geo will not be pleased about the latter ;)
I just don't see a very good reason for him to spout lies in the face of an impending launch.
Wouldn't be the first time :lol:
Geeforcer
16-Apr-2007, 19:53
Did he test one himself or is it an "I know a guy" situation? Because if it's the latter you are only as reliable as your current source.
Just the general consensus and Gibbo's hints, although I know that Geo will not be pleased about the latter ;)
I just don't see a very good reason for him to spout lies in the face of an impending launch.
Ok, well, the only nr's i've seen so far is 3D Mark. And it's using 1280*1024, no FSAA by default. Hardly the most bandwidth intesive test..
Did he test one himself or is it an "I know a guy" situation? Because if it's the latter you are only as reliable as your current source.
He claimed he has one and is testing it himself
icecold1983
16-Apr-2007, 21:17
razor can u paste the benches here, or put them in a file that doesnt require microsoft word?
its on pdf now, the link has been modified too.
Just the general consensus and Gibbo's hints, although I know that Geo will not be pleased about the latter ;)
I just don't see a very good reason for him to spout lies in the face of an impending launch.
Heh, I'm starting to desagree with Gibbo :razz:
jimmyjames123
16-Apr-2007, 23:38
Disagree in what way?
icecold1983
16-Apr-2007, 23:41
Heh, I'm starting to desagree with Gibbo :razz:
disagree that it wont match it in the sense that it will exceed it, or be slower?
Just the general consensus and Gibbo's hints, although I know that Geo will not be pleased about the latter ;)
I just don't see a very good reason for him to spout lies in the face of an impending launch.
Well that's what he always did, so why would it be different now?
disagree that it wont match it in the sense that it will exceed it, or be slower?
I dont' think his 7 out of 10 when the g80 gtx is 5 is correct its way to high. That means what 3dmark the r600 xt is getting 14-15k.
Has any of the NDAed people who got a card so far been really enthausiastic? Not as far as I could find. They're at most mildly happy. Which could simply be because they got a card and can make all those snarky remarks.
Where are the "OH, WOW! I WISH I COULD TELL YOU GUYS ABOUT THIS!!!" ?
Has any of the NDAed people who got a card so far been really enthausiastic? Not as far as I could find. They're at most mildly happy. Which could simply be because they got a card and can make all those snarky remarks.
Where are the "OH, WOW! I WISH I COULD TELL YOU GUYS ABOUT THIS!!!" ?
What makes you think anyone who has the bubbly personality to make that kind of post has one yet? I think somewhere or other there have been rumored dates for the tech day when most reviewers would get one --and if memory serves none of those rumored dates have arrived yet.
13th April, a finnish hardware site (muropaketti) did mention about going to the techday in tunisia "in couple weeks" (just to make clear, not meaning "within 2 weeks", but rather " about couple weeks from the date"), so that would place it to next week probably
bigtabs
17-Apr-2007, 02:28
@Razor re: FEAR bench : What graphics settings is that?
1600x1200 x16 AA x16Q AF summery, test setings, and conclusion is on the bottom of the file :wink:
I dont' think his 7 out of 10 when the g80 gtx is 5 is correct its way to high. That means what 3dmark the r600 xt is getting 14-15k.
When did he upgrade it to 7? Last I heard it was 5-6 compared to a 5 for 8800 GTX.
R600XT scored 12K in 3DMark06, slightly better than a GeForce 8800 GTX. It is benched on the latest 8.361-RC4 drivers, quite an improvement over the older sets of drivers. The core clock of R600XT stood at 750MHz while the memory is at 820MHz. It looks all decent for a US$399 card http://www.vr-zone.com/?i=4898
I have few questions?
A. Why memory only @ 820MHz which makes 1640MHz GDDR3 "Original plan was 900MHz"
B. Is AMD going to ship thousands or millions of 80nm HD-2900XT in May 2007 "I think AMD only knows that answer:("
But price $400 for 2900XT-512MB which will compete with $600 GF-8800GTX-768MB will be excellent. :)
(Edit: While back I made poll; should I purchase ATI/AMD R600, but hey for $400 dollars and it will compete with 8800GTX, I would say that is awesome! http://snappoll.com/poll/185282.php )
Thanks,
Best regards from trouble maker, :) "Just kidding I'm not trouble"
Shtal.... :) :)
Geeforcer
17-Apr-2007, 07:41
B. Is AMD going to ship thousands or millions of 80nm HD-2900XT in May 2007 "I think AMD only knows that answer:("
You think wrong, for I know the answer. Drum roll please...
...
...
...
AMD is not going to ship "millions of 80nm HD-2900XT in May 2007". And no, not because they might ship "billions" instead. That's not happening either. "Trillions" are also out of the picture. In fact they, are not going to ship "millions of 80nm HD-2900XT" in the whole 2007. Factlier still, they will not ship "millions (plural) of 80nm HD-2900XT" ever.
Please make out all checks for my consulting and tarot card reading services to Geefrocer (with two "e"s). @ B3DForums.
You think wrong, for I know the answer. Drum roll please...
...
...
...
AMD is not going to ship "millions of 80nm HD-2900XT in May 2007". And no, not because they might ship "billions" instead. That's not happening either. "Trillions" are also out of the picture. In fact they, are not going to ship "millions of 80nm HD-2900XT" in the whole 2007. Factlier still, they will not ship "millions (plural) of 80nm HD-2900XT" ever.
Please make out all checks for my consulting and tarot card reading services to Geefrocer (with two "e"s). @ B3DForums.
Well, I read somewhere that AMD will do a lot amount of shipment!
I'm not sure what is definition a lot :(
It could mean on 65nm a lot, but on 80nm will be low quantity.
Well, I read somewhere that AMD will do a lot amount of shipment!
I'm not sure what is definition a lot :(
It could mean on 65nm a lot, but on 80nm will be low quantity.
Please Shtal, read the posts on this forum, no one ever said AMD will ship millions of 2900's, only fudzilla wrote about it and they said amd was busy filling the orders for OEM's (which could go up to a million gpu's) for RV610/30's
As they are the low end parts, you know on which process they will arrive
Please Shtal, read the posts on this forum, no one ever said AMD will ship millions of 2900's, only fudzilla wrote about it and they said amd was busy filling the orders for OEM's (which could go up to a million gpu's) for RV610/30's
As they are the low end parts, you know on which process they will arrive
Sorry :(
flopper
17-Apr-2007, 07:57
amd will sell a lot to oem and also midsegment cards.
High end cards isnt that big of a market, most rather buy a ps3 for the same price.
The question is, will ati have a card that shines in dx10 but just do ok in dx9?
or just if i am using it for 1600x1200 then it be bandwidth and scaling for AA and AF and whatever adaptive it might hold.
Geeforcer
17-Apr-2007, 08:06
Exactly what neliz said rather well, plus...
Shtal... I know you are new to the forums, at least as an active participant. And that English is not your first language (at least I think so, my apologies if I am being presumptions). And that contribution of each member is valued and welcome. All that said, a little common sense along with a very basic understanding of how things work makes for a more interesting and more productive discussion.
I mean, if you want, I can explain why AMD is not shipping the aforementioned millions of HD 2900XTs this May. I am just slightly surprised I have to.
icecold1983
17-Apr-2007, 08:31
Exactly what neliz said rather well, plus...
Shtal... I know you are new to the forums, at least as an active participant. And that English is not your first language (at least I think so, my apologies if I am being presumptions). And that contribution of each member is valued and welcome. All that said, a little common sense along with a very basic understanding of how things work makes for a more interesting and more productive discussion.
I mean, if you want, I can explain why AMD is not shipping the aforementioned millions of HD 2900XTs this May. I am just slightly surprised I have to.
dude we all know why they arent shipping in may, if they did it would completely kill their image as a company for not launching everything as a family. all oems would cease doing business with them immediately and they would go bankrupt within 2 months.
dude we all know why they arent shipping in may, if they did it would completely kill their image as a company for not launching everything as a family. all oems would cease doing business with them immediately and they would go bankrupt within 2 months.
teh noes! and AMD stocks were just on a rise! :(
Exactly what neliz said rather well, plus...
Shtal... I know you are new to the forums, at least as an active participant. And that English is not your first language (at least I think so, my apologies if I am being presumptions). And that contribution of each member is valued and welcome. All that said, a little common sense along with a very basic understanding of how things work makes for a more interesting and more productive discussion.
I mean, if you want, I can explain why AMD is not shipping the aforementioned millions of HD 2900XTs this May. I am just slightly surprised I have to.
Yes, my English is not first language, I'm from Estonia "family origin _Half-Germany ", what is it has to do with that ?"
Yes, my English is not first language, I'm from Estonia "family origin _Half-Germany ", what is it has to do with that ?"
that most of the time when the talk gets technical, non native English speakers will end up missing the point of a conversation because most time is spent understanding the technical aspects.
Geeforcers point is that it is clear that English is not your first language and it translates into ... well.. your above post actually.
Ok, I made a mistake about millions stuff, sometimes I'm not very bright when I say stuff!
But hey I like ATI Radeon products, my first personal ATI video card was Rage Fury 128 when it came out, that was before I switch from Matrox to ATI.
I owe/ I have my own Business "Computer Repair company", I'm pretty knowledgeable in that field. Ok enough with my background :)
Yes, my English is not first language, I'm from Estonia "family origin _Half-Germany ", what is it has to do with that ?"
I think what GeeForcer actually meant was that it is very unlikely that AMD has managed to prepare millions of fully-fledged R600 GPUs for this launch. Also, it is somewhat unlikely that they will be selling millions of cards priced at $399.
Don't pay any heed to their antics, they're just having fun on AMD's expense ;)
OT: I'm from Estonia as well, an English language and literature major working as a tech writer for an international software company.
EDIT: Shtal, I sent you a PM, let's keep this thread on topic ;)
OT: I'm from Estonia as well, an English language and literature major working as a tech writer for an international software company.
hello friend!!!! :)
vertex_shader
17-Apr-2007, 10:58
http://we.pcinlife.com/thread-749836-1-1.html
http://img341.imageshack.us/img341/639/attachment239ce1nt9.jpg
Someone noticed this picture is fake?
Just save the picture zoom for the GFX field and after the R600 Series there is a clearly visible 8 :smile:
icecold1983
17-Apr-2007, 11:10
zoomed all the way in, i dont see it
epicstruggle
17-Apr-2007, 11:14
zoomed all the way in, i dont see it
Ditto, dont see it either.
vertex_shader
17-Apr-2007, 11:19
zoomed all the way in, i dont see it
Link (http://img297.imageshack.us/img297/4669/picturehd2.jpg) :wink:
I just noticed the 2 NV icon in the dekstop (left in the picture) :smile:
Link (http://img297.imageshack.us/img297/4669/picturehd2.jpg) :wink:
I see Jesus
http://hungryspoon.com/random/r600.png
You have distrubingly good eyes (or maybe it's my laptops 16bit lcd :razz:).
I didn't notice it until I zoomed this edit out.
The real question is, does it line up with the 'original', or is it simply random chance?
CarstenS
17-Apr-2007, 11:30
http://hungryspoon.com/random/r600.png
You have distrubingly good eyes (or maybe it's my laptops 16bit lcd :razz:).
I didn't notice it until I zoomed this edit out.
The real question is, does it line up with the 'original', or is it simply random chance?
In any way, there's no product ending with an "8"...
vertex_shader
17-Apr-2007, 11:38
In any way, there's no product ending with an "8"...
Maybe its not the last number, and almost perfectly photoshoped :smile:
CarstenS
17-Apr-2007, 11:42
Maybe its not the last number, and almost perfectly photoshoped :smile:
But why should anyone do a great ps-job on the last digits and then becoming sloppy on one of the earlier ones? Doesn't make sense to me.
vertex_shader
17-Apr-2007, 11:48
But why should anyone do a great ps-job on the last digits and then becoming sloppy on one of the earlier ones? Doesn't make sense to me.
Maybe who fake the picture not noticed.
someone needs to compare it to a picture of the results of a real 8800. If the 8's line up exactly, we can say it's a fake.
When did he upgrade it to 7? Last I heard it was 5-6 compared to a 5 for 8800 GTX.
Later on in the thread, when he got new drivers or something like that.
TheInq reports the Radeon HD launch date as May 14th:
http://www.theinquirer.net/default.aspx?article=38983
Rebel44
17-Apr-2007, 13:17
TheInq reports the Radeon HD launch date as May 14th:
http://www.theinquirer.net/default.aspx?article=38983
I just got confirmation of this info from someone who signed NDA:???:
bigtabs
17-Apr-2007, 13:34
Looks like with the new drivers (158.16) just released the 8800GTX 3dmark06 scores need revising. People are reporting 700 - 1000 point increase. :shock:
So they were holding back quite a bit to spoil the R600 launch.
Looks like with the new drivers just released the 8800GTX 3dmark06 scores need revising. People are reporting 700 - 1000 point increase. :shock:
So they were holding back quite a bit to spoil the R600 launch.
Those drivers seem to boost 3DM quite a bit while causing problems in real games. So what does that tell you?
bigtabs
17-Apr-2007, 13:38
Do they? I wasn't aware of that, what a shame for 8800GTX owners. Hope the reviewers pick that up when it comes to comparison time.
Bouncing Zabaglione Bros.
17-Apr-2007, 13:39
Those drivers seem to boost 3DM quite a bit while causing problems in real games. So what does that tell you?
Same old, same old?
bigtabs
17-Apr-2007, 13:39
Well, maybe they didn't hold something back then. :smile:
http://www.theinquirer.net/default.aspx?article=38983
Inq thinks launch day is May 14th, with tech day three weeks earlier, which would make it next week.
Then they go on to make a rather astonishing statement:
The problem is that company then plans a three-week silence period in the run up to launch. Nothingh from teh shindig should be posted until May 14th, our sources reveal. In today's world of immediate information, the three weeks in question sound very weird.
We learned that this could be due to late delivery of 65nm RV610LE/610XT/630Pro/630XT chips, but everything should be available in massive quantities. Of course, if there isn't for another mysterious delay that seems to happen regulary to AMD crew these days.
Theo, you ignorant slut. G80 tech day was October 18th. Launch day was November 8th. Take off your shoes and your pants and count to 21, Theo. Three weeks is not the least bit unusual for the flagship of a new generation.
Evildeus
17-Apr-2007, 14:53
Theo, you ignorant slut. G80 tech day was October 18th. Launch day was November 8th. Take off your shoes and your pants and count to 21, Theo. Three weeks is not the least bit unusual for the flagship of a new generation.
Calm down Geo, it's not good fort your heart. You should ask theo (and he should considering how much buzz he does on B3D) to pay for your next vacations (or perhaps all B3D forum members :twisted: ) ;)
Joshua Luna
17-Apr-2007, 17:40
Theo, you ignorant slut. G80 tech day was October 18th. Launch day was November 8th. Take off your shoes and your pants and count to 21, Theo. Three weeks is not the least bit unusual for the flagship of a new generation.
And that is the problem! He only has 20 digits! It is hard to count to 21 when you keep getting stuck on 20 :wink:
compres
17-Apr-2007, 17:42
Theo, you ignorant slut. G80 tech day was October 18th. Launch day was November 8th. Take off your shoes and your pants and count to 21, Theo. Three weeks is not the least bit unusual for the flagship of a new generation.
Well said. People should inform themselves before making that kind of remark, although that was a little harsh(LOL), I assume you guys know each other and it's not a big deal.
Evil Geo, bad! :lol:
Now that Fudo is gone, someone has to fill the slot, right? ;)
Hey compres, didn't know we're neighbours!
And that is the problem! He only has 20 digits! It is hard to count to 21 when you keep getting stuck on 20 :wink:
I think that's why Geo told him to lose the pants too, if you get the drift ;)
And that is the problem! He only has 20 digits! It is hard to count to 21 when you keep getting stuck on 20 :wink:
Hence the bit about pants. . . :wink:
Well said. People should inform themselves before making that kind of remark, although that was a little harsh(LOL), I assume you guys know each other and it's not a big deal.
I probably should have thrown a smilie in there somewhere. But the "you ignorant slut" bit is an old Saturday Night Live gag that hopefully communicated I was 49% kidding there, tone-wise.
I probably should have thrown a smilie in there somewhere. But the "you ignorant slut" bit is an old Saturday Night Live gag that hopefully communicated I was 49% kidding there, tone-wise.
Heh you are a old timer :) haven't seen that SNL gag in along time, that was Dan Akaryod, think I spelled his name right.
clement
17-Apr-2007, 18:30
And that is the problem! He only has 20 digits! It is hard to count to 21 when you keep getting stuck on 20 :wink:
Funny I can count to 32 on 5 fingers. :wink:
clement
17-Apr-2007, 18:32
Funny I can count to 32 on 5 fingers. :wink:
I mean 31
http://img186.imageshack.us/img186/9267/amddemo1vb3.th.jpg (http://img186.imageshack.us/img186/9267/amddemo1vb3.jpg) http://img186.imageshack.us/img186/3029/amddemo3ox6.th.jpg (http://img186.imageshack.us/img186/3029/amddemo3ox6.jpg) http://img402.imageshack.us/img402/7273/amddemo5qg0.th.jpg (http://img402.imageshack.us/img402/7273/amddemo5qg0.jpg)
http://img402.imageshack.us/img402/1135/amddemo7proceduralms6.th.jpg (http://img402.imageshack.us/img402/1135/amddemo7proceduralms6.jpg) http://img413.imageshack.us/img413/5381/amddemo8proceduralzi9.th.jpg (http://img413.imageshack.us/img413/5381/amddemo8proceduralzi9.jpg) http://img402.imageshack.us/img402/6289/amddemo6oc2.th.jpg (http://img402.imageshack.us/img402/6289/amddemo6oc2.jpg)
http://img266.imageshack.us/img266/4356/amddemo2gu7.jpg
http://img264.imageshack.us/img264/7231/amddemo4yg4.jpg
Source: http://ati.amd.com/developer/gdc/2007/Andersson-Tatarchuk-FrostbiteRenderingArchitecture(GDC07_AMD_Session). pdf
It's gonna be an interesting few weeks to come. The HD2900XT 512 (GDDR3) will be on allocation, while the other cards aren't, which seems to indicate that there will be plenty of RV cards at launch. The XTX is still scheduled for later on but no indication on the releasedate.
bdotobdot2
17-Apr-2007, 20:22
I mean 31
Depends if you start at 1 or 0
I mean, if you want, I can explain why AMD is not shipping the aforementioned millions of HD 2900XTs this May. I am just slightly surprised I have to.
The table (see attached) in this Xbit article (http://www.xbitlabs.com/news/video/display/20070413235044.html) summarizing JPR's Q406 discrete GPU market may be enlightening. I was surprised to see the volume in the "Performance" category, though I'm probably thinking of street (more like online discount store) prices whereas the table probably tracks only retail (brick and mortar and associated online storefront) prices. I say probably b/c I have no idea. Maybe the quarter was an aberration (here's B3D's take on it (http://www.beyond3d.com/content/articles/22)).
Those workstation cards are serious ripoffs. They use a slow, common GPU, with a slightly modified driver, and sell for at least 5 times the price.
Geeforcer
17-Apr-2007, 21:51
The table (see attached) in this Xbit article (http://www.xbitlabs.com/news/video/display/20070413235044.html) summarizing JPR's Q406 discrete GPU market may be enlightening. I was surprised to see the volume in the "Performance" category, though I'm probably thinking of street (more like online discount store) prices whereas the table probably tracks only retail (brick and mortar and associated online storefront) prices. I say probably b/c I have no idea. Maybe the quarter was an aberration (here's B3D's take on it (http://www.beyond3d.com/content/articles/22)).
But doesn't that table bare my point out?
trinibwoy
17-Apr-2007, 22:10
Those workstation cards are serious ripoffs. They use a slow, common GPU, with a slightly modified driver, and sell for at least 5 times the price.
Maybe that's because you're not paying just for the hardware.
DemoCoder
17-Apr-2007, 22:32
I think you're going to find that at R600 launch, NVidia will release drivers which significantly improved usage of the missing MUL, and for some applications, they made use shader detection to influence compiler heuristics, or shader replacement with hand-tuned code to maximally exploit missing MUL, all in all, *not* a repeat of the previous 3DM fiasco, because it's not an instance of turning down IQ or using approximations, but of helping the compiler do a better job.
NV in my book has provided more legit "mid-life driver boosts" than dodgy ones (yes, there have been both), so could be. They've surely had enough time to work on it. One could wonder if the otherwise curious speed with which they are ripping through driver rev numbers might have something to do with it. 158.xx? Wtf? In a hurry to get somewhere?
I think you're going to find that at R600 launch, NVidia will release drivers which significantly improved usage of the missing MUL, and for some applications, they made use shader detection to influence compiler heuristics, or shader replacement with hand-tuned code to maximally exploit missing MUL, all in all, *not* a repeat of the previous 3DM fiasco, because it's not an instance of turning down IQ or using approximations, but of helping the compiler do a better job.
That would be: considering they need that advantage?
But doesn't that table bare my point out?
Yup. I meant it as an illustration of your comment, not a contradiction.
BTW, it's "bear."
It's gonna be an interesting few weeks to come. The HD2900XT 512 (GDDR3) will be on allocation, while the other cards aren't, which seems to indicate that there will be plenty of RV cards at launch. The XTX is still scheduled for later on but no indication on the releasedate.
The 2900XTX schedule later release probably due to encounter 8800Ultra!
Geeforcer
17-Apr-2007, 23:06
Yup. I meant it as an illustration of your comment, not a contradiction.
BTW, it's "bear."
Ah, I see. Thanks. And... oops.:oops:
icecold1983
17-Apr-2007, 23:37
if the 8800 ultra is 1000, ati doesnt even need to counter it as the only people aware of the situation wont even care about a 1000 card, and all the review sites will likely slam nvidia for launching an 8800 oc at 1000.
vertex_shader
18-Apr-2007, 12:57
I think you're going to find that at R600 launch, NVidia will release drivers which significantly improved usage of the missing MUL, and for some applications, they made use shader detection to influence compiler heuristics, or shader replacement with hand-tuned code to maximally exploit missing MUL, all in all, *not* a repeat of the previous 3DM fiasco, because it's not an instance of turning down IQ or using approximations, but of helping the compiler do a better job.
Looks like already did with the 158.18 driver Link (http://www.fudzilla.com/index.php?option=com_content&task=view&id=554&Itemid=1)
icecold1983
18-Apr-2007, 13:18
theres no performance increases in any of the games i play
vertex_shader
18-Apr-2007, 13:25
theres no performance increases in any of the games i play
Synthetic benchmarks faster with the new nv magic drivers, play with that :wink:
vertex_shader
18-Apr-2007, 13:49
if the 8800 ultra is 1000, ati doesnt even need to counter it as the only people aware of the situation wont even care about a 1000 card, and all the review sites will likely slam nvidia for launching an 8800 oc at 1000.
Check what happend with the 7800gtx512mb times, many reviewer used the 799$ msrp card (after first low stock no one able to buy the card in store) 7800gtx512mb in tests, and make the conclusions nv has the king of the hill.
(something like this happend before too, when the reviewers used the 6800 extreme ultra against the other fantom card x800xtpe, but last time the x1800xt was not a fantom card when the 7800gtx512mb released).
Many review site manipulative, and biggest part of the users never see the reality, this users buy the the 7800gtx 256mb after they realized the 7800gtx512mb nowhere in stock, and not buy the faster x1800xt.
Not too many independent review sites left, and this is very bad for the users (newbies in hardware world) who are read the reviews and nothing real there just manipulation.
Skinner
18-Apr-2007, 14:09
theres no performance increases in any of the games i play
I get some nice gains with 165.01 in Stalker, Oblivion (64 to 72 fps at a safe point) and Source engine games. COH is faster to I heard.
(With a 8800GTX that is).
icecold1983
18-Apr-2007, 14:55
well i use the 158.xx for xp 32 bit, and no gains at all in ut2k4, coh, oblivion, fear demo, fear xp demo, armed assault demo, bf 2142/2 demo. only game i havent yet tried is dark messiah, which ill try in a little bit. but seeing as how theres not a single increase in performance in any of the other games, i dont expect one in dm either. i couldnt care less about 3dmark.
icecold1983
18-Apr-2007, 14:57
Check what happend with the 7800gtx512mb times, many reviewer used the 799$ msrp card (after first low stock no one able to buy the card in store) 7800gtx512mb in tests, and make the conclusions nv has the king of the hill.
(something like this happend before too, when the reviewers used the 6800 extreme ultra against the other fantom card x800xtpe, but last time the x1800xt was not a fantom card when the 7800gtx512mb released).
Many review site manipulative, and biggest part of the users never see the reality, this users buy the the 7800gtx 256mb after they realized the 7800gtx512mb nowhere in stock, and not buy the faster x1800xt.
Not too many independent review sites left, and this is very bad for the users (newbies in hardware world) who are read the reviews and nothing real there just manipulation.
i thought the gtx 512 launched at 650 or 700? thats a big ways off from 1000. it also had a sizable performance increase, much bigger than a 650 to 700 mhz 8800 will bring.
Silent_Buddha
18-Apr-2007, 15:01
The table (see attached) in this Xbit article (http://www.xbitlabs.com/news/video/display/20070413235044.html) summarizing JPR's Q406 discrete GPU market may be enlightening. I was surprised to see the volume in the "Performance" category, though I'm probably thinking of street (more like online discount store) prices whereas the table probably tracks only retail (brick and mortar and associated online storefront) prices. I say probably b/c I have no idea. Maybe the quarter was an aberration (here's B3D's take on it (http://www.beyond3d.com/content/articles/22)).
Intersting it might be something like this that also helped push AMD into delaying the R600 from Feb/March to April/May (techday/release). Considering roughly 90% of revenue was made on mid-range and lower (250 USD or less) parts, they weren't really losing out on a lot from lost R600 retail sales vs the possible boost it could have on the midrange and lower parts.
There's also the possibility that they see a lowered demand for enthusiast parts in general and came to the conclusion that even enthusiasts are getting tired of paying 600-800 for a video card. Which might be another thing prompting them to release HD 2900 XT at such a low price point. Although, I'm still going with the theory that HD 2900 XTX will be released a month or so after at 100-200 dollars more.
Regards,
SB
vertex_shader
18-Apr-2007, 15:24
i thought the gtx 512 launched at 650 or 700? thats a big ways off from 1000. it also had a sizable performance increase, much bigger than a 650 to 700 mhz 8800 will bring.
I remember wrong, msrp was 649$, in store the price was 799$, than in ebay 1000-1400$.
Maybe the 1000$ or the "its just a oc the to sky 8800gtx" rumor are false, when not, than collectors/rich users/fanboys or users without brain buy it :smile:
Skinner
18-Apr-2007, 15:33
well i use the 158.xx for xp 32 bit, and no gains at all in ut2k4, coh, oblivion, fear demo, fear xp demo, armed assault demo, bf 2142/2 demo. only game i havent yet tried is dark messiah, which ill try in a little bit. but seeing as how theres not a single increase in performance in any of the other games, i dont expect one in dm either. i couldnt care less about 3dmark.
Well I forgot to mention I use Vista(64), so that's propbably it, just improving Vista perf.
Ontopic, what should be possible to improve the IQ, apart from higher AA?
LOCALIZED REFRIGERATOR APPARATUS FOR A THERMAL MANAGEMENT DEVICE (http://v3.espacenet.com/textdoc?DB=EPODOC&IDX=US2007044483&F=0)
http://www.cupidity.f9.co.uk/b3d91.gif
This seems to be a vapour chamber (30) on top of the die (28) with a peltier cooler (32) making the vapour condense at a low temperature, with a fan (40) to cool the main heatsink (38).
So that earlier rumour about a Vapochill style cooler was only half right? A peltier cooler consumes quite a lot of power - perhaps that's where the concept of <=300W power consumption comes from.
Jawed
SirPauly
18-Apr-2007, 16:49
Interesting. Thanks Jawed
Mcmlxxx IV
18-Apr-2007, 17:02
Check what happend with the 7800gtx512mb times, many reviewer used the 799$ msrp card (after first low stock no one able to buy the card in store) 7800gtx512mb in tests, and make the conclusions nv has the king of the hill.
(something like this happend before too, when the reviewers used the 6800 extreme ultra against the other fantom card x800xtpe, but last time the x1800xt was not a fantom card when the 7800gtx512mb released).
Many review site manipulative, and biggest part of the users never see the reality, this users buy the the 7800gtx 256mb after they realized the 7800gtx512mb nowhere in stock, and not buy the faster x1800xt.
Not to nitpick, but the x800xtpe was not a phantom card. They existed and were on sale (I have one in my second computer), though they were hard to find. The 6800 Ultra Extreme never existed. It was never even an official Nvidia SKU. It was a card clocked at 450/1200, I believe, that was sent to reviewers to spoil the x800 launch. They never went on sale.
vertex_shader
18-Apr-2007, 17:27
Not to nitpick, but the x800xtpe was not a phantom card. They existed and were on sale (I have one in my second computer), though they were hard to find. The 6800 Ultra Extreme never existed. It was never even an official Nvidia SKU. It was a card clocked at 450/1200, I believe, that was sent to reviewers to spoil the x800 launch. They never went on sale.
The low number SKU's i call as fantom cards too, this is why i call 7800gtx 512mb and the x800xtpe as fantom cards.
The card was exist in very low number and only for some AIB( EVGA 6800 Ultra Extreme (http://images.anandtech.com/reviews/video/nvidia/geforce6800ue/evga6800ue.jpg) ), but still many review site used it against the x800xtpe.
Fresh new info posted on my favorite dutch forum from the same reliable source as always:
8800GTX 768MB = 18514 / 12957 -> 599$
R600XT 512MB = 18352 / 12817 -> 399$
8800GTS 640MB = 16892 / 11182 -> 399$
8800GTS 320MB = ~16892 / ~11182 -> 299$
Driver version unknown
X1950Pro drops in price to fight GF8600GTS
X1950XT will also drop in price. X1950Pro and X1650XT will still exist until Q3.
HD2900XT uses 225W. With overdrive it will be more than that.
Voucher for HL Episode 2, Team Fortress2 and Portal included
Only HD2900XT will be available at launch. RV630 and RV610 have been delayed.
Clockspeed: 740e/825m
Drivers for DX10 CrossFire aren't ready yet.
3 versions of the XT. 512 GDR3 9.5" for channel. 1GB GDDR4 12.5" for SI's and a 1GB GDDR4 9.5" on special order.
Geeforcer
18-Apr-2007, 21:46
Fresh new info posted on my favorite dutch forum from the same reliable source as always:
8800GTX 768MB = 18514 / 12957 -> 599$
R600XT 512MB = 18352 / 12817 -> 399$
8800GTS 640MB = 16892 / 11182 -> 399$
8800GTS 320MB = ~16892 / ~11182 -> 299$
Driver version unknown
Thanks a lot ofr the info CJ! What ARE the numbers after the name and SKU?
HD2900XT uses 225W. With overdrive it will be more than that.Since most current PSUs only only have 6-pin PCIe, I am curious if overdriver will be an option with them.
X1950Pro drops in price to fight GF8600GTS
X1950XT will also drop in price. X1950Pro and X1650XT will still exist until Q3.
...
Only HD2900XT will be available at launch. RV630 and RV610 have been delayed.
Clockspeed: 740e/825mDo you know how by how much? The pricing/postioning moves seem to suggest that the delay in not trivial. And with AMD proclaiming that the real reason why R600 was on hold is because "family launch" is so important (a suggestion that already had more holes in it then a chunk of swiss on a machine gun range), their credebility will take yet another blow.
Drivers for DX10 CrossFire aren't ready yet.Well they better be come launch day, or those spiffy soundbites will come to soundbite them in the rear.
trinibwoy
18-Apr-2007, 22:33
Voucher for HL Episode 2 included....
:grin:
1GB GDDR4 9.5" on special order.
:sad:
Thanks CJ!
Skrying
18-Apr-2007, 22:38
Three models of one card? I already have a bit of an issue with the 320MB/640MB 8800GTS, but this... I just don't see why AMD thinks this is at all the right way to go. If they feel 512MB is enough for the HD 2900XT then keep it at 512MB, maybe a shift later to GDDR4 for power consumption reduction is fine on that. But to have a special order release of a 1GB GDDR4 model is just weird, not only that but at what cost? Availability? On top of that confusion it will be seen as giving the OEMs the "better" card in some ways, as I am rather sure that many might be willing to have a huge card if that means getting double the memory and it coming in GDDR4 flavor.
So, GDDR4 is in such tight supply as of now? Interesting...
icecold1983
18-Apr-2007, 22:43
so no family launch afterall? wonder what 'original' excuse they will think up now.
Sound_Card
18-Apr-2007, 22:49
RV610 and RV630 being delayed does not sound right at all. Really F'd up on AMD's part if true.:roll:
INKster
18-Apr-2007, 22:50
so no family launch afterall? wonder what 'original' excuse they will think up now.
I was thinking the exact same thing.
AMD seemed very frontal when they justified the delay with the option to do a simultaneous launch for the whole family instead..., and now this. :roll:
vertex_shader
18-Apr-2007, 22:55
HD2900XT uses 225W. With overdrive it will be more than that.
How can be exact 225watt?
Maximum power come from the pci-e and from the 2x6pin psu connector 1x75watt+2x75watt=225watt, but this is the theoretical maximum.
vertex_shader
18-Apr-2007, 22:59
I was thinking the exact same thing.
AMD seemed very frontal when they justified the delay with the option to do a simultaneous launch for the whole family instead..., and now this. :roll:
Mainstream parts etail/retail launch delayed.
INKster
18-Apr-2007, 23:00
How can be exact 225watt?
Maximum power come from the pci-e and from the 2x6pin psu connector 1x75watt+2x75watt=225watt, but this is the theoretical maximum.
Except we have already seen a 1x6 coupled with a 1x8 out there.
Don't know which version will end up in the retail XT version though.
vertex_shader
18-Apr-2007, 23:07
Except we have already seen a 1x6 coupled with a 1x8 out there.
Don't know which version will end up in the retail XT version though.
Rumors say the card runs fine with the 2x6pin connector too but its have 1x8pin (compatible with 6pin) and 1x6pin psu connector, 6pin+8 pin needed only for overclocking, with newer PSU the card can be used default with 1x8pin (150watt) too.
vertex_shader
18-Apr-2007, 23:16
RV610 and RV630 being delayed does not sound right at all. Really F'd up on AMD's part if true.:roll:
Will be funny to see after the delays and official reasons/comments/hints one SKU released in launch the hd2900xt, looks like this will be for some weeks amd dx10 retail lineup, and this card need to save AMD from the total user dissapontment. :smile:
Cuthalu
18-Apr-2007, 23:28
Why should overclocking require a psu with 8-pin connector? I'm using atm x1800xt with a psu that doesn't have pci-e power connector at all - that's why we have adapters.
INKster
18-Apr-2007, 23:38
Why should overclocking require a psu with 8-pin connector? I'm using atm x1800xt with a psu that doesn't have pci-e power connector at all - that's why we have adapters.
Because it may stress the amps on the PSU rails quite a bit more ?
Russell
19-Apr-2007, 00:12
RV610 and RV630 being delayed does not sound right at all. Really F'd up on AMD's part if true.:roll:
I think I posted this awhile ago. Fudo started talking about this a few weeks ago and it seems to be semi-confirmed. Nonetheless, its hardly surprising.
How can be exact 225watt?
Maximum power come from the pci-e and from the 2x6pin psu connector 1x75watt+2x75watt=225watt, but this is the theoretical maximum.
Considering it comes from a Dutch site, mistranslations aren't entirely impossible. It's pretty safe to say that this information confirms what we already know about the 225W maximum draw.
Voucher for HL Episode 2, Team Fortress2 and Portal included
Hells yes!
I think I posted this awhile ago. Fudo started talking about this a few weeks ago and it seems to be semi-confirmed. Nonetheless, its hardly surprising.
Considering it comes from a Dutch site, mistranslations aren't entirely impossible. It's pretty safe to say that this information confirms what we already know about the 225W maximum draw.
Hells yes!
About the vouchers, you will probably have to choose one of the three. And to make the wait for these games more bearable, you'll be able to pick another already released game from Valve.
Some more specifics on the power consumption numbers. The HD 2900XT uses a maximum of 225W at standard clocks. Recommended PSU is 500W. If you plan on using Overdrive, the HD 2900XT will draw more than 225W power. If you plan on using a CrossFire HD 2900XT setup, AMD recommends a 750W PSU. But if you use a Vista setup it's no use anyway, since DX10 CrossFire support for Vista will come at a later date. And 24xAA has been confirmed as well as Fudo's report that the current drivers don't support HDMI audio yet. But that will be implemented before the launch.
So looking at 3DM05/06 numbers it looks like AMD's gonna fight the pricewar by offering GTX performance at a much lower pricepoint.
Skrying
19-Apr-2007, 00:52
Choose one of the three? I thought they were being released together in a pack? That's A LOT more "meh" for the voucher then. Especially since EP2 for example should only cost $20 on its on, the other's the same probably.
You could be right about it. I'm not sure though. It seems more likely to bundle all three of them seeing as Portal was a special project that comes bundled with Episode 2.
Russell
19-Apr-2007, 01:00
Choose one of the three? I thought they were being released together in a pack? That's A LOT more "meh" for the voucher then. Especially since EP2 for example should only cost $20 on its on, the other's the same probably.
Yeah, they're together for sure. Though I must say, I'd only buy the set for TF2 anyways, so that's what I'd pick if one had to choose (which I don't believe we will anyways).
EDIT: All this, naturally, assumes this is fact. I know it's still rumor.
SYSTEM AND METHOD FOR HIGHER LEVEL FILTERING BY COMBINATION OF BILINEAR RESULTS (http://appft1.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&d=PG01&p=1&u=%2Fnetahtml%2FPTO%2Fsrchnum.html&r=1&f=G&l=50&s1=%2220070080972%22.PGNR.&OS=DN/20070080972&RS=DN/20070080972)
A method and system for higher level filtering uses a native bilinear filter, typically found in a texture mapper, and combines a plurality of bilinear filter results from the bilinear filter to produce a higher level filtered texel value. A native bilinear filter is operative to generate bilinear filtered texel values by performing a plurality of bilinearly filtered texture fetches using bilinear filter fetch coordinates. The method and system combines the plurality of bilinear filtered texel values with a plurality of weights to generate the higher level filtered texel value.
I'm struggling to understand when this kind of filtering would be used:
[0003] A typical three-dimensional graphics processing system natively provides a two-by-two bilinear filter. A single address fetch by the filter results in 4 texels being returned. When an application requires a higher level of taps (higher level filtering), such as a 4.times.4 filter, this native bilinear filter is inadequate.
Is this referring to anisotropic filtering, or is it merely referring to filtering based on more samples from a single surface? (I think the latter.)
This bit is peculiar:
[0004] A solution to overcome the limitations of the native bilinear filter is utilizing pure pixel shader code. In this embodiment, a pixel shader that includes one or more ALUs includes specific code, such as executable instructions operated on a graphics processing unit, that performs the filtering functionality. In any situation where scaling occurs, such as the source and destination are not the same size or the output pixel does not align exactly to input texels, different weight sets are needed to compensate for the possible different sub-pixel alignments. Utilizing pixel shader code is generally slow because, among other things, two texture fetches are needed for each tap, one fetch is needed to fetch the weight of the texel value and one fetch is needed to fetch the texel value to be weighted. Furthermore, using pure pixel shader code, typically only "pick nearest" filtering can be used by the texture fetches, meaning that only one texel is returned for each fetch request and another 16 fetches can be required for the corresponding weights for a total of 32 fetches. It will be recognized that if a weight applies equally to four texels, fewer weight fetches may be needed (e.g., perhaps only four fetches to obtain the weights for 16 texels) however, the total number of fetches may still be unnecessarily high.
If this document is about anisotropic filtering, it gives me a vision of prior Radeon architectures using pixel shading to perform the filtering. Sounds mad and unlikely.
So, anyway, I'm confused and wondering what this is really about.
Maybe worth reading in collaboration with:
Methods and apparatus for retrieving and combining samples of graphics information (http://v3.espacenet.com/textdoc?DB=EPODOC&IDX=US2007046682&F=0)
Disclosed are methods and apparatus for accomplishing the fetching or sampling of channels of pixels or texels such as neighboring pixels or texels or non-neighboring pixels or texels in a simultaneous operation in order to achieve optimization of the performance of a texture pipeline. In particular, logic is disclosed including selector logic configured to retrieve data including a plurality of channels from each of a plurality of pixels or texels and operable to select one channel from the plurality of channels of the data from each of the pixels or texels. The logic also includes combination logic configured to combine two or more of the selected channels into a single vector, such as an RGBA vector representing the color.
which seems to be about fetching texels in an unconventional order and packing them into vectors (much like fetch-4 packs four different single-channel texels into a vector).
Jawed
Andrew Lauritzen
19-Apr-2007, 05:11
I dunno, but that first segment struck me first as doing something like bicubic filtering using bilinear lookups - which is pretty easy actually (I believe there was even a Gems 2 chapter on it). Also since it can be done quite efficiently in shader code, that theory would seem to contradict the later paragraph.
Still the first quoted paragraph could definitely be referring to something like higher-order filtering although the patent-speak may be confusing the issue.
Rangers
19-Apr-2007, 05:33
So after all this the R600 is going to be about the same performance as 8800GTX but cheaper. Kind of underwhelming after all this time.
It does set up interesting memory scenarios though, where I think Nvidia is kind of caught in tweener land memory wise. A tweener is like, a football player that's too small to play linebacker but too big to play safety, or something like that. Anyways, I think AMD could have edge there, 512 MB is enough at this performance level, but saves them some money. So yeah, X2900XT sets up as a nice buy, but not going to blow anybody away.
Clockspeed: 740e/825m
What does this mean? Is the clockspeed 740 or 825?
What does this mean? Is the clockspeed 740 or 825?
Yes. Engine speed is 740. Memory speed is 825.
I'm going to assume this has been posted but if not or for those who haven't seen it here it is again.
http://www.ixbt.com/news/hard/index.shtml?08/03/55
Recommended PSU is 500W.
So there's still hope for my 2 year old Hiper 480W? That's sweet. :smile: I've been very pleased with this PSU.
$399 for the 2900XT is just awesome too.
Fresh new info posted on my favorite dutch forum from the same reliable source as always:
8800GTX 768MB = 18514 / 12957 -> 599$
R600XT 512MB = 18352 / 12817 -> 399$
8800GTS 640MB = 16892 / 11182 -> 399$
8800GTS 320MB = ~16892 / ~11182 -> 299$
Are saying that 8800GTX Scores 12957 in 3DMark06?
I thought it only pushed to 11k using intel-Quad CPU.
icecold1983
19-Apr-2007, 07:40
so all this time and amd isnt even launching a card faster than a 7 month old 8800 till god knows when. what a joke. what good is 24x aa on a card that isnt faster than the 8800?
Twinkie
19-Apr-2007, 07:42
This is quite odd. A "flagship card" that performs on par with the competition that has been out for 6 months, while being cheaper. I dont recall this kind of a match up happening very often.
Then one has to wonder what the 8800 ultra is for, when frankly R600 doesnt look like it improves much on the performance sector (compared to what G80 has done 6 months ago) as hoped for by many enthusiasts. (all based on rumour though)
Im kind of interested by nVIDIA's response to the $399 R600 threat. nVIDIA doesnt like price cuts to my knowledge, but this time it might force them to do some price cuts especially on the GTX model unless theyre rolling out with their refresh part very soon.
But another mystery for the R600 launch is that why are they using GDDR3? doesnt seem to make much sense because their current flagship "X1950XTX" uses GDDR4. Sort of like a step back in that sense. GDDR4 yields arent good? or performance benefits just doesnt compensate for the cost of GDDR4? (isnt GDDR4 cheaper than GDDR3?)
http://www.ocworkbench.com/2006/nvidia/GeForce-8800GTX/3dmark06.jpg
http://www.ocworkbench.com/2006/nvidia/GeForce-8800GTX/b2.htm
This is correct score for 8800GTX using Intel-QX6700....
If they used new / just released Quad Intel-QX6800 it would probably be score between 12k-12.2k 3DMark06 score for 8800GTX.
And (R600) HD-2900XT probably score 12.5k in 3DMark06.
SugarCoat
19-Apr-2007, 08:30
This is quite odd. A "flagship card" that performs on par with the competition that has been out for 6 months, while being cheaper. I dont recall this kind of a match up happening very often.
Then one has to wonder what the 8800 ultra is for, when frankly R600 doesnt look like it improves much on the performance sector (compared to what G80 has done 6 months ago) as hoped for by many enthusiasts. (all based on rumour though)
Im kind of interested by nVIDIA's response to the $399 R600 threat. nVIDIA doesnt like price cuts to my knowledge, but this time it might force them to do some price cuts especially on the GTX model unless theyre rolling out with their refresh part very soon.
But another mystery for the R600 launch is that why are they using GDDR3? doesnt seem to make much sense because their current flagship "X1950XTX" uses GDDR4. Sort of like a step back in that sense. GDDR4 yields arent good? or performance benefits just doesnt compensate for the cost of GDDR4? (isnt GDDR4 cheaper than GDDR3?)
Like what happened with the 7800GTX when the X1800XT launched im sure we'll see some pretty significant price reductions, down to $450-500MSRP for vanilla models not including any rebates that any AIBs may do.
TheInq actually reported that the 8800Ultra would be almost non-existant. More of a limited run card much like the X1950XTX 'Uber edition' or 7800 512mb so i dont even think i'd call it an actual Retail card. Personally i expect the performance battles to initially be waged by AIB overclocked models if anything. Like the X1800XT vs the 7800GTX its all going to come down to superior features in my opinion when determining the best card. I dont know about most high end card purchasers but i'd wager most of them dont have $1000+ monitors with wildly extreme resolutions where one card may really differ itself from the other thus i wonder how much of a real world bandwidth advantage the R600 will really have.
Personally i'll be most interested in performance per watt figures and watching the new tech demos.
If that is the case, who cares about G80Ultra to beat/win against R600XT. But ATI/AMD should have R600XTX for backup.
I wounder how long will it take for Nvidia to make real refresh G81 and for AMD to make ATI-R650.
If they used new / just released Quad Intel-QX6800 it would probably be score between 12k-12.2k 3DMark06 score for 8800GTX.
And (R600) HD-2900XT probably score 12.5k in 3DMark06.
I'm just guessing here... but maybe they used an even newer yet to be released CPU...
If that is the case, who cares about G80Ultra to beat/win against R600XT. But ATI/AMD should have R600XTX for backup.
I wounder how long will it take for Nvidia to make real refresh G81 and for AMD to make ATI-R650.
R650 should be well on it's way by now, just like R580 was nicely on track eventhough R520 got delayed. A few months ago the "next generation R600" was scheduled for release in Q3 just like RV670.
Im kind of interested by nVIDIA's response to the $399 R600 threat. nVIDIA doesnt like price cuts to my knowledge, but this time it might force them to do some price cuts especially on the GTX model unless theyre rolling out with their refresh part very soon.
The GTX has 256 MB more memory, so that makes sense that it's more expensive (though not $200, but it will surely be lowered a bit to match the XT price better). No-brainer.
its all going to come down to superior features in my opinion when determining the best card.
So what features would that be that we don't already have (besides maybe higher levels of AA, which is hardly a new feature, just more horsepower needed)? Really, I keep hearing about some "great new features" all the time and have no idea what you people are expecting.
R650 should be well on it's way by now, just like R580 was nicely on track eventhough R520 got delayed. A few months ago the "next generation R600" was scheduled for release in Q3 just like RV670.
Don't know about it being on track but if the midrange is delayed till June then it might not be so much on track.
Despite being late it's still a good card having GTX performance for much less and I think the R650 at 900+ MHz with 1 GB of GDDR4 should put them in good competition with anything nvidia has to throw there way.
http://www.ocworkbench.com/2006/nvidia/GeForce-8800GTX/b2.htm
Pretty old article though, Nov 9. I'm not sure how much faster the newer drivers are in 3dmark-06, but they should be faster after 6 months.
Brand loyalty is a bitch sometimes. Must be exhausting.. :lol:
Promise us you won't jump off a bridge if g80 actually beats R600 in a few benchies. ;)
Bouncing Zabaglione Bros.
19-Apr-2007, 12:53
Don't know about it being on track but if the midrange is delayed till June then it might not be so much on track.
I think if the XT is going to be in a lower price bracket and not a whole lot faster the the G80, then that leaves AMD with a gap at the top end, it terms of both performance and addressing that market segment. I think they'll be eager to get a 65nm XTX with 1 gb of DDR4 and higher clocks as soon as possible.
I wonder if midrange has been delayed because AMD are using some of that fab capacity for R650?
Rebel44
19-Apr-2007, 13:00
I dont care about mainstream I want 2900XTX 1GB GDDR4:cool:
and I think the R650 at 900+ MHz with 1 GB of GDDR4 should put them in good competition with anything nvidia has to throw there way.
I sincerely severely doubt that. I'm quite confident that nV already could start churning out G9x (or whatver it will be called) in a matter of weeks if needed. Same thing I told numerous times since R520 debacle, obviously my predictions were spot-on as pretty much always. And I'm only using my common sense, since I have no insider info whatsoever, just lots of experince in the SW/HW industry.
This delay was another christmass present for nV, they can happily try stuff out and develop new tech while ATI has to struggle with getting stuff done. That's a priceless advantage every CEO or investor drools over in his wet dreams.
I'd be very surprised if the (high-end gfx) ATI part of AMD still exists in 2009/2010 timeframe. Unless some miracle happens and R650 turns out to be a real killer, but I'll believe it when I see it on the shelves and not before that.
Oh, and guys please stop parroting that "OMG, much lower prices for R6xx parts" fairy tale, it's not even funny anymore. Wake up and turn your brains on, really. First, nV doesn't even need to compete with prices, having the lead they have now and they could go even lower than ATI easily without making losses, if so needed.
EDIT: and before someone calls me nV fan, I'm not. But I'm a huge fan of their management, I'd give them a paycheck with blank space for filling in their wished payment for themselves if I was the "big boss".
Rangers
19-Apr-2007, 13:51
I sincerely severely doubt that. I'm quite confident that nV already could start churning out G9x (or whatver it will be called) in a matter of weeks if needed. Same thing I told numerous times since R520 debacle, obviously my predictions were spot-on as pretty much always. And I'm only using my common sense, since I have no insider info whatsoever, just lots of experince in the SW/HW industry.
This delay was another christmass present for nV, they can happily try stuff out and develop new tech while ATI has to struggle with getting stuff done. That's a priceless advantage every CEO or investor drools over in his wet dreams.
I'd be very surprised if the (high-end gfx) ATI part of AMD still exists in 2009/2010 timeframe. Unless some miracle happens and R650 turns out to be a real killer, but I'll believe it when I see it on the shelves and not before that.
Oh, and please stop parroting that "OMG, much lower prices for R6xx parts" fairy tale, it's not even funny anymore. Wake up and turn your brains on, really. First, nV doesn't even need to compete with prices, having the lead they have now and they could go even lower than ATI easily without making losses, if so needed.
EDIT: and before someone calls me nV fan, I'm not. But I'm a huge fan of their management, I'd give them a paycheck with blank space for filling in their wished payment for themselves if I was the "big boss".
Their managment isn't stellar from what I see, they just have an inept competitor. 8600GTS, a poor part really, is just another new example, along with their driver messes and poor SLI support and on and on. If Nvidia is raking in the cash and growing bigger and bigger in employee numbers as we're led to believe, these driver problems are rather baffling, unless they're just not run all that efficiently. Nvidia imo has been focused on profits rather than performance lately far too much imo, they're just lucky they dont have a competitor to make them pay. You can bet Intel will.
The high end of Nvidia wont be around in 2009/10 either I wager. The one company that's going to be left is Intel, not Nvidia. Intel's process shrink lead continues to grow, they are playing around on 45nm while Nvidia isn't on 80 yet. If Intel had a GPU it would be three times as big as Nvidia's. And you can bet Intel engineers will make it count.
This is all OT, but since the whole AMD goes out of business OT constantly infects this thread anyway, why not.
They have a stellar management because they`re getting the big money in nV`s pocket and have had spotless execution save for NV30. This is looking at it from an economics perspective, not internet-tech junkie that gets a hard on for extra precision bits, sooper dooper AA, unified shaders etc. Their management does what it`s meant to do in a great way, IMHO.
This is looking at it from an economics perspective, not internet-tech junkie that gets a hard on for extra precision bits, sooper dooper AA, unified shaders etc.
Yeah, that's exactly how I'm looking at it above.
While I have no doubt the R600XT 512MB will be a very competitive part indeed I'm begining to wonder whether it's going to be the only competetive part they are going to get out of the R6x0 series. :lol:
Their managment isn't stellar from what I see, they just have an inept competitor. 8600GTS, a poor part really, is just another new example, along with their driver messes and poor SLI support and on and on.
How is the 8600 a poor part? It's slightly slower than the X1950Pro and has about 1/3 as many transistors and costs less than 30% to produce.
As for drivers troubles, let's first see what ATI brings to the table, shall we? Everyone and his grandpa seems to believe that ATI drivers for the R6xx series will be perfect out of the box - I call nonsense.
The one company that's going to be left is Intel, not Nvidia. Intel's process shrink lead continues to grow, they are playing around on 45nm while Nvidia isn't on 80 yet. If Intel had a GPU it would be three times as big as Nvidia's. And you can bet Intel engineers will make it count.
This song I've been hearing since 1997 or so, over and over with each new gen of gfx chips. "Intel will crush them all, they're just soooo much biggerer, OMG!" :lol:
Again, I'll believe it when I see it.
I think if the XT is going to be in a lower price bracket and not a whole lot faster the the G80, then that leaves AMD with a gap at the top end, it terms of both performance and addressing that market segment. I think they'll be eager to get a 65nm XTX with 1 gb of DDR4 and higher clocks as soon as possible.
I wonder if midrange has been delayed because AMD are using some of that fab capacity for R650?
Oh I'm sure they want the an xtx or r650 out now as soon as possible, just hope the the 65nm node isn't giving problems (I think we know by now at 80nm ATi did have some kind of technical issues, what ever it may be, perfromance or manufacturing), but they are sacrificing alot with the mid range delay, if they are going to release numbers and not release these cards nV's apparrent weakness might become a strength.
It's slightly slower than the X1950Pro and has about 1/3 as many transistors and costs less than 30% to produce.
LOL, it's actually around 12% smaller in transistor count.
Sound_Card
19-Apr-2007, 15:06
How is the 8600 a poor part? It's slightly slower than the X1950Pro and has about 1/3 as many transistors and costs less than 30% to produce.
Again, I'll believe it when I see it.
Slightly slower? Even more slower when you put on some AA/AF. :razz:
Might want to check your transistor count too. The 8600GTS has 298m transistors while the x1950pro has 330m. :wink:
Silent_Buddha
19-Apr-2007, 15:11
They have a stellar management because they`re getting the big money in nV`s pocket and have had spotless execution save for NV30. This is looking at it from an economics perspective, not internet-tech junkie that gets a hard on for extra precision bits, sooper dooper AA, unified shaders etc. Their management does what it`s meant to do in a great way, IMHO.
And if rumors are to be believed Nvidia Managements inability to get decent stable Vista Drivers is costing them millions from OEMs jumping ship to ATI.
As someone else said, Nvidia's management right not isn't so much the fact that they are good, it's that ATI's has been pretty dismal the past couple years.
Had ATI launched R520 and R600 "on time" then Nvidia wouldn't be looking nearly as good right now. Which is entirely ATI's fault and has nothing to do with how well (or not well) Nvidia is managed.
In fact, I'd argue that Nvidia's management this past year has been far worse than it has been in previous years. Driver issues is just the most noticeable effect of this. And I'm going to guess that it's mostly because they've become lazy expecting ATI to do their job for them by constantly missing target dates for their products.
It's sort of like when you see a skinny girl standing next to a fat girl at the club. The skinny chick always looks hot, but that's only because the fat chic looks so blah. :wink: :lol: :smile: Now, stand that exact same skinny girl next to a really hot girl that doesn't need a fat girl to make her look good and suddenly that skinny girl is looking pretty nasty.
Heck, if early reports are to be believed, ATI (X1K series) may actually have across the board better OGL performance under Vista than Nvidia with the new 7.4 drivers. Something that would have been unheard of when Nvidia was run well.
Also, I'd have to agree that 65nm R650 might be delayed by delays in RV610 and RV630. As it appears, according to some rumors that ATI is targetting 100 million shipments for those chips for this year. It also appears that ATI is having greater than expected demand for RV610 and RV630 from OEMs which might help explain the delays. Which means capacity at 65nm might be a little tight IF true.
Regards,
SB
Really gotta bury the hatchet with huge OEM BS from AMD for thier midrange products, Dell just picked up the 8600's for thier computers. As I said before, I didn't think Dell picked up AMD midrange, there were no hints of this.
INKster
19-Apr-2007, 15:25
Slightly slower? Even more slower when you put on some AA/AF. :razz:
Might want to check your transistor count too. The 8600GTS has 298m transistors while the x1950pro has 330m. :wink:
Actually, it's 289. DX10 mandatory support for FP32 precision, SM 4.0, the new VP2/BSP's, etc, consume a large part of this.
Also, the tests prove that the G84's performance scales better when the software requires complex shading.
Might not be enough to beat X1950 Pro, but it is an interesting trend nevertheless.
trinibwoy
19-Apr-2007, 15:34
All these 8600GTS comparisons will be for nought once RV630 hit. That's where the true comparison lies because all the last generation parts are not hindered by DX10/VP logic. If HD 2600XT is significantly faster then the 8600GTS will look poor in comparison and I'd be confused if it isn't given the considerable cooling it seems to require.
LOL, it's actually around 12% smaller in transistor count.
Really? You surely mean only the _active_ transistors? We may as well go with die area, if it fits you better. Or power consumption and cooling as well.
But ok, I exaggerated with the transistor count, I give you that ;)
ATI is targetting 100 million shipments for those chips for this year
One hundred million chips? *boggle*
Get out! :lol: :lol: :lol:
We may as well go with die area, if it fits you better. Or power consumption and cooling as well.
Why not, it's on smaller process so the die is er.. smaller, the real life power consumption (http://techreport.com/reviews/2007q2/geforce-8600/index.x?pg=10) seems to be pretty much the same as with X1950Pro.:wink:
As it's clocked so high it does better in situations needing shader performance, most of the time it's slightly slower than X1950Pro quite often dropping to GT level or below, in some rare situations it seem to go all the way down to previous generation 128-bit card levels.
So on and all there's nothing really outstanding about it, the word is it's not actually flying out of the shelves which is not surprising considering the price/perf. ratio.
But it's not really that bad a card just the price is definetly wrong.
vertex_shader
19-Apr-2007, 16:06
One hundred million chips? *boggle*
Get out! :lol: :lol: :lol:
Its not impossible Link (http://img.alibaba.com/photo/11123391/Cassava_Chips.jpg) :twisted:
trinibwoy
19-Apr-2007, 16:06
It's sort of like when you see a skinny girl standing next to a fat girl at the club. The skinny chick always looks hot, but that's only because the fat chic looks so blah. :wink: :lol: :smile:
Heh, I guess you've never seen a cute fat girl standing next to an ugly skinny one :razz:
And I agree with razor, attempts to criticize what Nvidia is doing while using imaginary numbers for what ATI could be doing aren't really worth much. I'm sure ATI wishes that the world runs on "ifs" but that's not how it is.
jamis: I call that pretty impressive considering the laughably low count of shader units in the 8600 (although they're clocked so high).
Add the same improvements to the current 8800 and there is your R600 killer already (not that I think they'll be doing anything like that, but for the sake of argument).
Where I was going is the margins. Thus nV could easily sell 8600 for half the price without losing money on them, while ATI can do nothing like that with any of their offerings.
Add the same improvements to the current 8800 and there is your R600 killer already (not that I think they'll be doing anything like that, but for the sake of argument).
Where I was going is the margins. Thus nV could easily sell 8600 for half the price without losing money on them, while ATI can do nothing like that with any of their offerings.
Nah, it's just that pretty much everything is decoulpled so they can play around with the amounts of different units to get optimal perf/transistor ratio for every price point. That kind of configurationd wouldn't make much sense in the high end.
I have no doubt they can reduce the price and it will need to come down for these things to start selling properly. Afterall looking at the specs of G84/G80 there's a huge gap between them which can be filled with not only 1 but 2 different SKU's.
I dunno, but that first segment struck me first as doing something like bicubic filtering using bilinear lookups - which is pretty easy actually (I believe there was even a Gems 2 chapter on it). Also since it can be done quite efficiently in shader code, that theory would seem to contradict the later paragraph.
Still the first quoted paragraph could definitely be referring to something like higher-order filtering although the patent-speak may be confusing the issue.
In bed last night I convinced myself it must be something to do with anisotropic filtering, haunted by the horizontal and vertical ratios:
[0032] Step 204 is generating new Sfract, Tfract values that will be placed in the look-up table for that particular sub-texel alignment. The look-up table values are based on the horizontal ratios and the vertical ratios of the initial filter weights. In one exemplary embodiment, FIG. 3 illustrates a plurality of weights. Step 206 is generating the scale factors using the horizontal ratios and the vertical ratios. In one embodiment, the horizontal ratio can be determined based on the fraction of a first weight and a second weight in a horizontal relationship, such as the relation of W00 and W01 of FIG. 3. W00/W10=W01/W11=horizontal ratio Equation 2:
[0033] The vertical ratio is further computed based on the spatial relationship between the various weights in a vertical relationship, such as the quotient or fraction of W00 and W10 of FIG. 3. W11/W12=W21/W22=vertical ratio Equation 3:
[0034] The horizontal ratio is applicable to all relative weights in a horizontal relationship and the vertical ratio is relative to all weights in a corresponding vertical relationship. Moreover, this ratio relationship exists for all filters that are separable.
Rummaging some more (I only briefly skimmed it last night):
[0041] As recognized by one having ordinary skill in the art, the computations and techniques for data transformation of the present invention, including the above-noted equations, provide an equivalent actual result as the prior art computational techniques. For example, the resultant computation of Equation 7 produces the same result as the prior art computation of summing the multiplications of texels and weights (e.g. W00*p00+W10*p10+W01*p00+W11*p11). However, the present invention provides for improved computation through the effective utilization of existing computation resources and the manipulation of computational terms, such as the generation of Equation 7 above.
the bolded part makes me think that this is just a faster way of performing each "loop" of anisotropic filtering, where each loop consists theoretically of 16 samples.
I'm hoping the texture filtering gurus will show up at some point...
Jawed
icecold1983
19-Apr-2007, 16:21
As for drivers troubles, let's first see what ATI brings to the table, shall we? Everyone and his grandpa seems to believe that ATI drivers for the R6xx series will be perfect out of the box - I call nonsense.
you would have to agree tho, its hard to imagine how the drivers could be worse. 6 months after launch we have an xp driver thats barely better than the beta launch drivers.
Evildeus
19-Apr-2007, 16:32
Slightly slower? Even more slower when you put on some AA/AF. :razz:
Might want to check your transistor count too. The 8600GTS has 298m transistors while the x1950pro has 330m. :wink:
Well i don't know, but considering all the test on HFR we have this summary:
(AF activated)
1680*1050 => 100,0 X1950 Pro vs 104,4 8600 GTS
1680*1050 + AA 4x => 97,2 8600 GTS vs 100,0 X1950 Pro
http://forum.hardware.fr/hfr/Hardware/2D-3D/topic-geforce-g86-sujet_753192_13.htm
http://www.hardware.fr/articles/664-4/nvidia-geforce-8600-gts.html
More or less a tie i would say
silent_guy
19-Apr-2007, 17:14
Might want to check your transistor count too. The 8600GTS has 298m transistors while the x1950pro has 330m. :wink:
It's a good thing then that this crappy management doesn't care one bit about the number of transistors. :wink:
230mm2 for R570 without DX10 vs 169mm2 for a G84 with... Let's just hope for AMD that the R600 architecture is radically more efficient than the R500 serious.
Well, 44 vector units (PS+VS) require much more silicon than 32 scalar ones. :wink:
Voltron
19-Apr-2007, 20:39
It's a good thing then that this crappy management doesn't care one bit about the number of transistors. :wink:
230mm2 for R570 without DX10 vs 169mm2 for a G84 with... Let's just hope for AMD that the R600 architecture is radically more efficient than the R500 serious.
Though weirdly Beyond 3D lists the R570 dimensions as 13.8 x 13.8 and die size of 230mm2. If it is indeed 13.8 x 13.8 then the die size should be 190mm2.
http://www.beyond3d.com/resources/chip/119
So which info is correct?
Not that this negates the point that G84 is more efficient.
Sound_Card
19-Apr-2007, 20:49
It's a good thing then that this crappy management doesn't care one bit about the number of transistors. :wink:
230mm2 for R570 without DX10 vs 169mm2 for a G84 with... Let's just hope for AMD that the R600 architecture is radically more efficient than the R500 serious.
I was pointing out that G84 was not 1/3 of the transistors of x1950pro....
G84 better be more effiecient than R570 per transistor since it is a next generation arc... if not then that would be a completly different thread would it not?:wink:
jimmyjames123
19-Apr-2007, 21:02
You can't really talk about efficiency per transistor when comparing two different cards with vastly different feature sets (ie. one with DirectX 10 compatibility and advanced video processing engine, vs one without).
I sincerely severely doubt that. I'm quite confident that nV already could start churning out G9x (or whatver it will be called) in a matter of weeks if needed. Same thing I told numerous times since R520 debacle, obviously my predictions were spot-on as pretty much always. And I'm only using my common sense, since I have no insider info whatsoever, just lots of experince in the SW/HW industry.
What makes you think nvidia is capable of pulling some super performing part from thin air? We know the R600 is on par with the GTX, that leaves a die shrink for both companies as an option and I don't think nvidia has anything beyond a higher clocked "Ultra" or another 7950 like part, the latter of which wouldn't exactly be in the same price bracket at any rate. This is a repeat of what happened with the last generation. The X1800 XT was late but on par with the 7800 GTX and ATI quickly followed up with an improved X1900 XT, which was on par or better than nvidia's refresh, the 7900 GTX.
Oh, and guys please stop parroting that "OMG, much lower prices for R6xx parts" fairy tale, it's not even funny anymore. Wake up and turn your brains on, really. First, nV doesn't even need to compete with prices, having the lead they have now and they could go even lower than ATI easily without making losses, if so needed.
It's the best way to challenge nvidia, a company who cares about the money above all else. Believe me, most people will go for the better card, especially when that card has equal performance for more than $100 less.
vertex_shader
19-Apr-2007, 22:24
Sound and physics are the buzzwords for AMD's R600
R600 combines graphics with audio and physics process in a whole another way, and those 64 vect5D shader pipelines yield in 320 scalar units total, directly comparable to Nvidia's 128 scalar ones. Now, 128 scalar units at 1.35 GHz is still a tad more efficient that ATi's own 320 at 740 MHz, but ATi's pipeline is more complex than it was previously thought.
Link (http://www.theinq.com/default.aspx?article=39042)
Though weirdly Beyond 3D lists the R570 dimensions as 13.8 x 13.8 and die size of 230mm2. If it is indeed 13.8 x 13.8 then the die size should be 190mm2.
http://www.beyond3d.com/resources/chip/119
So which info is correct?
Not that this negates the point that G84 is more efficient.
13.8 x 16.7
Twinkie
19-Apr-2007, 22:52
What makes you think nvidia is capable of pulling some super performing part from thin air? We know the R600 is on par with the GTX, that leaves a die shrink for both companies as an option and I don't think nvidia has anything beyond a higher clocked "Ultra" or another 7950 like part, the latter of which wouldn't exactly be in the same price bracket at any rate. This is a repeat of what happened with the last generation. The X1800 XT was late but on par with the 7800 GTX and ATI quickly followed up with an improved X1900 XT, which was on par or better than nvidia's refresh, the 7900 GTX.
Dont be so quick to judge. IMHO, this isnt what it was like last gen. Why? because ATi is 7 months late, and their R600 part at launch is only on par with the GTX (trades blows through benchmarks according to recent rumours) thats been around forever now.
Also, nVIDIA can pretty much go for an optical die shrink going to 80nm, while ATi has to design from ground up (ala the R650) if their moving to the 65nm process. (Since R600 is already on 80nm)
So your suggesting the R650 is going to replace the R600 in 3 months?
"We don't do soft launches" --Henri Richard re R6xx family launch
Also said 10 products launching.
And "second half of May", which is a bit more puzzling if they all launch on the same day, or just the last of them launch second half of May, or what.
Link (http://www.theinq.com/default.aspx?article=39042)
Pay attention to the "Sound Physics" part. New buzzword? ;)
Bouncing Zabaglione Bros.
19-Apr-2007, 23:12
So your suggesting the R650 is going to replace the R600 in 3 months?
I think R650 will come in above the R600 in the XTX spot, rather than replace the XT. It might launch soon because AMD needs to fill that space. They might be waiting to see what response Nvidia has to the R600.
Arnold Beckenbauer
19-Apr-2007, 23:14
Link (http://www.theinq.com/default.aspx?article=39042)
R600 combines graphics with audio and physics process in a whole another way, and those 64 vect5D shader pipelines yield in 320 scalar units total, directly comparable to Nvidia's 128 scalar ones. Now, 128 scalar units at 1.35 GHz is still a tad more efficient that ATi's own 320 at 740 MHz, but ATi's pipeline is more complex than it was previously thought.
Ouch, my brain hurts...
Dear Inquirer, R600's 5D ALUs (Vec4+1 and not Vec5) are not just scalar, they are superscalar. You know, it's like a man <-> Superman.
Well, forget it.
Is it possible, that R600's Vec4-+1 ALUs are split in 2:2:1?
trinibwoy
19-Apr-2007, 23:32
It's the best way to challenge nvidia, a company who cares about the money above all else. Believe me, most people will go for the better card, especially when that card has equal performance for more than $100 less.
Ahahahahhahaha!! Whew....that was a good laugh. :lol: :lol:
3dilettante
19-Apr-2007, 23:35
If R600 is truly capable of being repurposed for sound and physics, it would need some very robust threading control. I doubt people are going to buy 3 R600s to keep each workload from trashing the rest.
If R600 is truly capable of being repurposed for sound and physics, it would need some very robust threading control. I doubt people are going to buy 3 R600s to keep each workload from trashing the rest.
Threading control is one thing I expect R600 to be pretty good at. Xenos supports 8 concurrent contexts. Couple that with the bevy of virtual-memory related patent documents and it looks like a slam-dunk to me.
Still, for absolute performance, I'm sure more R6xx GPUs will be required. And, anyway, won't it be years before we see it in use in games?
Unless AMD gets OpenAL running on R600? Would that work?
Jawed
Cuthalu
20-Apr-2007, 02:46
2600xt doesn't look like a better performer than 8600gts. :/
3dmark numbers: http://www.hardspell.com/english/doc/showcont.asp?news_id=414&pageid=517
Cuthalu
20-Apr-2007, 02:47
Edit: wrong place.
R300King!
20-Apr-2007, 03:40
2600xt doesn't look like a better performer than 8600gts. :/
3dmark numbers: http://www.hardspell.com/english/doc/showcont.asp?news_id=414&pageid=517
Here (http://forums.whirlpool.net.au/forum-replies-archive.cfm/715719.html) the 06 numbers look very close.
8600GTS
======
E6600
250gb 16mb cache Sata 3gb/sec
2gb 6400 Gskill ram
3dMark06...
Main Test Results
3DMark Score 5542 3DMarks
SM 2.0 Score 2284 Marks
SM 3.0 Score 2036 Marks
CPU Score 2605 Marks
Detailed Test Results
Graphics Tests
1 - Return to Proxycon 17.534 FPS
2 - Firefly Forest 20.532 FPS
CPU Tests
CPU1 - Red Valley 0.834 FPS
CPU2 - Red Valley 1.302 FPS
HDR Tests
1 - Canyon Flight (SM 3.0) 18.063 FPS
2 - Deep Freeze (SM 3.0) 22.653 FPS
The "leaked" HD2600XT (http://www.hardspell.com/english/doc/showcont.asp?news_id=414&pageid=517)
=================
Radeon HD2600XT(3DMark06 and game benchmark)
Computer hardware configuration:[display card:RV630]CPU[Core 2 Duo E6600]
HDR Tests
1 - Canyon Flight (SM 3.0) 18.299 FPS
2 - Deep Freeze (SM 3.0) 23.799 FPS
http://www.hardspell.com/pic/pic.asp?picurl=http://www.hardspell.com/pic/2007/4/20/81b5d6ff-b4c9-43f7-991e-525105c48315.jpg
They look very similar in 3DMark06 but we all know the 8600GTS tanks in games compared to it's 06 marks.
overclocked_enthusiasm
20-Apr-2007, 04:00
From the AMD CC I got the following:
1. R600 family will be a hard launch (as Geo said).
2. The whole stack will be launched at once.
3. "Market leading performance" was the answer to which was the top part XT or XTX at launch. While this was NOT asked verbatim nor answered that way I assume the 2900 XT is the top launch part...no surprise there.
4. Barcelona 2 core and 4 core parts delivered in Q3.
5. New 65 nm mobile architecture to appear in "first half 2008".
Looks like the launch in Tunis will be a hard launch of R600, RD790 and a soft launch of Barcelona desktop parts...or a hard launch of the Barcelona server part? Either way, you can bet your bottom dollar they will use Barcelona as the CPU in the display rig(s).
Geeforcer
20-Apr-2007, 04:12
Barcelona IS the server version of K10.
BTW, someone need to explain the rumors of RV6xx delay and "hard launch" AMD line, one is obviously incorrect. At one point, I was thinking that maybe RV610 or RV630 was delayed while the other RV was on-target. But, 10-product launch would seem to indicate both are a go.
Dont be so quick to judge. IMHO, this isnt what it was like last gen. Why? because ATi is 7 months late, and their R600 part at launch is only on par with the GTX (trades blows through benchmarks according to recent rumours) thats been around forever now.
Yes, and that is exactly what happened with the last generation, including the R520s greater feature list over the G70. I fail to see how the current situation is any different.
Also, nVIDIA can pretty much go for an optical die shrink going to 80nm, while ATi has to design from ground up (ala the R650) if their moving to the 65nm process. (Since R600 is already on 80nm)
So your suggesting the R650 is going to replace the R600 in 3 months?
Shrinking the G80 to 80nm will not give it that much headroom and it's almost a given the R650 will be 65nm giving it a much bigger advantage over the R600.
Ahahahahhahaha!! Whew....that was a good laugh.
Are you being facetious? I can't tell. If not all you have to do is look at how long nvidia has held onto the $550+ GTX price and has not released any other cards since the launch of the 8800 besides a GTS with less ram. They love to milk prices and always have since the fall of 3dfx.
Twinkie
20-Apr-2007, 04:32
Yes, and that is exactly what happened with the last generation, including the R520s greater feature list over the G70. I fail to see how the current situation is any different.
Shrinking the G80 to 80nm will not give it that much headroom and it's almost a given the R650 will be 65nm giving it a much bigger advantage over the R600.
R520 was only 3 months late to my knowledge. It was actually faster across the board as well.
But current situation points to a 7month delay, with a part that only performs on par with the GTX. Not sure what kind of features it might have but from IQ perspective, both are feature rich in that department. Also GTX still has alot of headroom from the driver side of things so i dont see the R600 outperforming G80 later on (like the R520 vs G70), unless R600 is really strong on DX10 apps.
And how do you know that? (the bolded part above)
Unless you work for nVIDIA.. :lol:
R520 was nearly four months late and basically traded blows with the G70, it did have more features however, just like now. So it seems to me that the only difference is the R600 is around three or four months later, granted which does hurt.
It's been seven months since the launch of the 8800 GTX, at this point I would be surprised to see much of a performance boost from driver updates. Besides, nvidia's primary problem is stability within Vista, which is a priority for them above all else.
As far as 80nm not giving much headroom, let's call it a hunch.
Silent_Buddha
20-Apr-2007, 04:57
Barcelona IS the server version of K10.
BTW, someone need to explain the rumors of RV6xx delay and "hard launch" AMD line, one is obviously incorrect. At one point, I was thinking that maybe RV610 or RV630 was delayed while the other RV was on-target. But, 10-product launch would seem to indicate both are a go.
It's possible they may be fudging "hard launch" a little with availability in OEM systems purchaseable by the consumer on launch day, but not available in retail as an add in card until 2-4 weeks later.
*shrug* All any of us has to go on is rumors. ATI guy says 10 products will hard launch. Internet rumors says at least 2 of the products are delayed.
Pick your poison with who you believe. Myself, I'll just wait and see.
Regards,
SB
trinibwoy
20-Apr-2007, 05:33
Are you being facetious? I can't tell.
Yes I was, sorry. I just find it amusing that you think Nvidia as a company is unique in its focus on money.
including the R520s greater feature list over the G70.
That's not how I remember it...
Yes I was, sorry. I just find it amusing that you think Nvidia as a company is unique in its focus on money.
Oh no, I recognise nvidia is in no part unique in that respect, just more so than ATI (although maybe not so much now with AMD).
:smile:
Oh no, I recognise nvidia is in no part unique in that respect, just more so than ATI (although maybe not so much now with AMD).
:smile:
Is that whey they like to rename some of their previous gen cards with the new gen naming scheme ?
I'm sorry, but i think statements like this are laughable. The 600 XT will be priced according to it's performance/features and how desperate AMD is to gain back market share/goodwill.
Pretty old article though, Nov 9. I'm not sure how much faster the newer drivers are in 3dmark-06, but they should be faster after 6 months.
Brand loyalty is a bitch sometimes. Must be exhausting.. :lol:
Promise us you won't jump off a bridge if g80 actually beats R600 in a few benchies. ;)
Who cares! :lol:
icecold1983
20-Apr-2007, 06:51
R520 was nearly four months late and basically traded blows with the G70, it did have more features however, just like now. So it seems to me that the only difference is the R600 is around three or four months later, granted which does hurt.
It's been seven months since the launch of the 8800 GTX, at this point I would be surprised to see much of a performance boost from driver updates. Besides, nvidia's primary problem is stability within Vista, which is a priority for them above all else.
As far as 80nm not giving much headroom, let's call it a hunch.
u could be right, but its fair to mention that besides 3d mark and that gpu dot test or whatever its called, there have been virtually no increases in performance in any games on windows xp since the earliest launch driver.
The difficult task for ATI was to create/make R600 - it takes a-lot of time, but R650 should be a-lot easier. So logically can be calculated that R600 might have short life before it will be replaced by R650.
Is that whey they like to rename some of their previous gen cards with the new gen naming scheme ?
I'm sorry, but i think statements like this are laughable. The 600 XT will be priced according to it's performance/features and how desperate AMD is to gain back market share/goodwill.
You're entitled to your own opinion, however wrong it may be. :twisted:
Russell
20-Apr-2007, 07:26
Barcelona IS the server version of K10.
BTW, someone need to explain the rumors of RV6xx delay and "hard launch" AMD line, one is obviously incorrect. At one point, I was thinking that maybe RV610 or RV630 was delayed while the other RV was on-target. But, 10-product launch would seem to indicate both are a go.
The way I read into it is that AMD doesn't think it's a soft launch if they launch some of the products and the others appear a couple weeks later.
R520 was only 3 months late to my knowledge. It was actually faster across the board as well.
But current situation points to a 7month delay
Actually, R600 was "technically" only delayed once and will be about 2 months late. Of course, this assumes you go by official delays. R520 had more than one unofficial delay that resulted in it being more than 3 months late (which may be how long it was officially late, I'm not really sure).
As for the 80nm question, it's only an optical shrink. It's not an entirely new process, so it's not going to give the same results as a full process shrink. There will be improvements, and nVidia will milk every last ounce of performance out of it as incredibly well as nVidia does, but it won't be nearly as much as can be potentially gained from the shrink to 65nm. I bolded one word for a reason.
EDIT: Oh, one more thing,
Dailytech (http://www.dailytech.com/AMD+Posts+611+Million+Net+Loss+in+Q1/article6988.htm) on the AMD CC: "On a positive note, AMD reported $197 million USD in revenue from its graphics division in Q1 2007. This represented a 19 percent gain from Q4 2006."
So much for those assuming that ATI is losing money due to this delay. Revenue is actually UP since G80's release, in a slow time of year. Considering how late it is in the last-gen's product cycle, yields are surely good and I think it's safe to say that profits are up as well. I wonder if any of this revenue comes from pre-sales of R600. Frankly, I doubt it. Not that early.
Geeforcer
20-Apr-2007, 07:32
The way I read into it is that AMD doesn't think it's a soft launch if they launch some of the products and the others appear a couple weeks later.
So to paraphrase: "We don't do soft launches - we just redefine what a soft launch is." Maybe that's why they were so eager for "broad technology rollout" and "family" launch - that way you can sill paper-launch half the products in the "family" but claim that the whole thing was "hard".
Russell
20-Apr-2007, 07:35
So to paraphrase: "We don't do soft launches - we just redefine what a soft launch is." Maybe that's why they were so eager for "broad technology rollout" and "family" launch - that way you can sill paper-launch half the products in the "family" but claim that the whole thing was "hard".
I love how you paraphrased it :)
But hey, I wouldn't put it past them. Talk on all fronts indicates that the delay was due to drivers. They've been talking out of their asses from the moment they made that decision.
Geeforcer
20-Apr-2007, 07:41
Dailytech (http://www.dailytech.com/AMD+Posts+611+Million+Net+Loss+in+Q1/article6988.htm) on the AMD CC: "On a positive note, AMD reported $197 million USD in revenue from its graphics division in Q1 2007. This represented a 19 percent gain from Q4 2006."
So much for those assuming that ATI is losing money due to this delay. Revenue is actually UP since G80's release, in a slow time of year. Considering how late it is in the last-gen's product cycle, yields are surely good and I think it's safe to say that profits are up as well. I wonder if any of this revenue comes from pre-sales of R600. Frankly, I doubt it. Not that early.
Unfortunately, Dailytech is missing a huge point: The Q4 results for graphics division only included sales from October 25 and on. I am not a rocket surgeon, but $197 million in 3 months is not really better than $166 in two... in fact :runaway:
silent_guy
20-Apr-2007, 07:51
G84 better be more efficient than R570 per transistor since it is a next generation arc... if not then that would be a completely different thread would it not?:wink:
One doesn't necessarily have to follow from the other: The additional price for DX10-only features is far from clear, but it will definitely bring down the efficiency/mm2 for DX9 applications.
To be honest, the current AMD rumors aren't very promising: they were at a huge disadvantage during previous generation, so they already had some catching up to do. Right now, it doesn't feel like that's going to happen. So once again they'll be at a disadvantage wrt pricing power. (That doesn't really matter for the consumer, but it won't make their bean counters happy.)
Also GTX still has alot of headroom from the driver side of things so i dont see the R600 outperforming G80 later on (like the R520 vs G70), unless R600 is really strong on DX10 apps.
You can't dismiss the possibility that it will be indeed ATI's drivers that will put them ahead of 8800 GTX ;)
Remember the OpenGL 4xAA 30% performance increase for X1800 series?
silent_guy
20-Apr-2007, 08:06
I think R650 will come in above the R600 in the XTX spot, rather than replace the XT. It might launch soon because AMD needs to fill that space. They might be waiting to see what response Nvidia has to the R600.
Depends on how you define 'soon'.
Say ATI had planned R600 for January. In that case, the earliest possible upgrade would have to be planned for, say, October, following the standard refresh time lines.
If the delay of R600 hasn't delayed other projects, then October is still an option, I guess. But there's no way you can reasonably accelerate chip schedules, even if they need to fill that space. There are no corners to cut like you can with software.
It may be different if R600 was planned to be released much earlier than January...
Bouncing Zabaglione Bros.
20-Apr-2007, 08:19
Depends on how you define 'soon'.
Say ATI had planned R600 for January. In that case, the earliest possible upgrade would have to be planned for, say, October, following the standard refresh time lines.
If the delay of R600 hasn't delayed other projects, then October is still an option, I guess. But there's no way you can reasonably accelerate chip schedules, even if they need to fill that space. There are no corners to cut like you can with software.
You never know until you see the schedule and how far along they are. Suppose they allowed for three respins of R650 in the schedule, but it turns out they only needed two? That's a fair bit of wiggle room there that could easily take a couple of months off the schedule.
AMD is going to want to get that 65 nm XTX out as soon as possible to address that market segment they've conspicuously left empty above the XT pricing. If there's anything they can do to bring that product forward, they will, especially if they are waiting on Nvidia's response to R600.
It may be different if R600 was planned to be released much earlier than January...
IIRC, wasn't R600 originally in the frame for last October/November time?
DemoCoder
20-Apr-2007, 08:20
Corporations which aren't in the business of making money will soon be out of the business. I don't call companies which are forced into price reductions and low margins charitable, I call them companies whose management failed. ATI is not offering its cards cheap because it doesn't care about making money, or because it's magnaminous, but because its execution has been managed worse than nVidia.
I mean, for the love of god, this is a publically traded company we're talking about, ATI, now AMD, whose stockholders, investors are demanding returns, whose employees' futures are dependant upon continued profits, where enormous numbers of people are affected when the company suffers lower margins. NVidia isn't anymore "in love" with margins or profits than any other Fortune 500 company. They just happen to have a management that knows how to connect desire with reality.
The idea that they're somehow more evil, greedy, or rapacious is absurd.
silent_guy
20-Apr-2007, 08:44
Suppose they allowed for three respins of R650 in the schedule, but it turns out they only needed two? That's a fair bit of wiggle room there that could easily take a couple of months off the schedule.
I understand you're just giving an example here, but just FYI: any chip leader who comes up with a 3-spin project plan will be sent back on sight by his manager. It's an accepted practice pretty much everywhere to take into account 1 full spin and 1 metal spin. You don't plan for more. And usually, you don't need more.
AMD is going to want to get that 65 nm XTX out as soon as possible to address that market segment they've conspicuously left empty above the XT pricing. If there's anything they can do to bring that product forward, they will, especially if they are waiting on Nvidia's response to R600.
Yes, but all project plans (in all engineering disciplines) are made with optimistic best case schedules. It's a fundamental law of Nature! More than once, I've seen desperate managers promise bonuses to accelerate chip schedules. It doesn't work: there are almost no corners to cut. If they originally planned for October, there's no way they can reign it in by, say, 2 months.
IIRC, wasn't R600 originally in the frame for last October/November time?
Possible. But if their 65nm part is ready in July,why bring out a 80nm part? (And once again piss off its buyers, like with R520?)
Corporations which aren't in the business of making money will soon be out of the business. I don't call companies which are forced into price reductions and low margins charitable, I call them companies whose management failed. ATI is not offering its cards cheap because it doesn't care about making money, or because it's magnaminous, but because its execution has been managed worse than nVidia.
I mean, for the love of god, this is a publically traded company we're talking about, ATI, now AMD, whose stockholders, investors are demanding returns, whose employees' futures are dependant upon continued profits, where enormous numbers of people are affected when the company suffers lower margins. NVidia isn't anymore "in love" with margins or profits than any other Fortune 500 company. They just happen to have a management that knows how to connect desire with reality.
The idea that they're somehow more evil, greedy, or rapacious is absurd.
Some people desire money more than others. Whether those people are shareholders or running the company is another matter. You are correct, most organizations exist to make money, however there are also grades of morality and greed, it's not simply black and white.
This is getting off topic though.
Russell
20-Apr-2007, 09:15
Unfortunately, Dailytech is missing a huge point: The Q4 results for graphics division only included sales from October 25 and on. I am not a rocket surgeon, but $197 million in 3 months is not really better than $166 in two... in fact :runaway:
Ick. Noted. :/
DemoCoder
20-Apr-2007, 09:33
Some people desire money more than others. Whether those people are shareholders or running the company is another matter. You are correct, most organizations exist to make money, however there are also grades of morality and greed, it's not simply black and white.
So in your theory, ATI's margins have been lower not because they had higher internal costs due to various factors like lower yields/bigger chips, and inability to price their chips higher, but because ATI's management are less greedy and more moral, and therefore, they took actions inline with those values which lowered margins, such as---???
I mean, your comments might seem reasonable if nVidia's margins were high because they employed child labor/sweat shop labor, dumped wasted into local rivers, and basicallly did whatever they could to lower costs by doing things inefficiency. But nVidia's high margins in fact, seem to derive from efficiency and sales demand, and not inefficiency and aggressive anti-labor tactics. Let's face it, nVidia's margins, especially on the G70, stem from aggressive focus on optimizing the architecture to minimize production costs.
I mean, this isn't a case of your local cobbler shoe store vs asian Nike sweatshots.
This might be offtopic, but I don't think one can just let your original comments slide without some rebuttal.
Bouncing Zabaglione Bros.
20-Apr-2007, 09:49
Possible. But if their 65nm part is ready in July,why bring out a 80nm part? (And once again piss off its buyers, like with R520?)
But they are not bringing out that part. R600 is 80 nm and dropped into a lower price segment. It's different from the R650 at a higher price and performance point. A new part would have been out in six months anyway, so if R600 is late and R650 is early/on time...
nicolasb
20-Apr-2007, 10:06
Depends on how you define 'soon'.
Say ATI had planned R600 for January. In that case, the earliest possible upgrade would have to be planned for, say, October, following the standard refresh time lines.Why not say they had R600 planned for November? That was the original rumour.
Also, it's not out of character for graphics chip companies to alternate two different types of "new" product. The first is a whole new "architecture", while the second is a "speed-bump" release, which is architecturally similar but with performance tweaks. So, for example, GeForce 6xxx to GeForce 78xx was a significant architectural change. 78xx to 79xx was a "speed-bump" change.
Given that R650 would be a "speed-bump" release, architecturally similar to R600 but on a smaller process so as to get a better clockspeed and lower power consumption, I don't think it's at all unreasonable to suppose that the gap between the two might be only six months, with the next major architectural revision (R700) planned to follow 12 months after R600. It would also seem plausible that R650 would follow on fairly close behind RV6xx - again, it wouldn't be out of character for a company to produce a high-end part, followed by mid-range ones on a smaller process, followed closely by a die-shrink of the high performance part as soon as the experience of the mid-range chips has ironed out all the bugs inherent in the new process.
I wouldn't be surprised if R650 hits in September. Conceivably even late August (3 months after R600 XT). Though I think people who are expecting it in June are going to be dissappointed!
Silent_Buddha
20-Apr-2007, 12:46
Unfortunately, Dailytech is missing a huge point: The Q4 results for graphics division only included sales from October 25 and on. I am not a rocket surgeon, but $197 million in 3 months is not really better than $166 in two... in fact :runaway:
Of course that's oversimplying things as you are also comparing the busiest quarter (Q4) against the one of the slowest quarters (Q1 and Q2).
Also, considering the ATI brand had no response to Nvidia's G80, makes it all the more impressive.
Or it could just mean the loss of the high end market had virtually no impact on sales of ATI branded midrange/low end chips, where approximately 90% of revenue is made.
Without being able to compare it to ATI's Q4 2005 results and Q1 2006 results, it's hard to say how well or not well they did.
As regards R600/R650 compared to R520/R570. It's only really similar in the fact that both of the first parts are delayed. R650 doesn't appear to be as much of a major design change as R570 was in comparison to the parent chip.
I'd imagine that unless there were some major additions (IE - the increased shaders in R570 vs R520) I'd imagine that R650 will face a much easier shrink.
Regards,
SB
SB, I think your 8 is broken...
Of course that's oversimplying things as you are also comparing the busiest quarter (Q4) against the one of the slowest quarters (Q1 and Q2).
Also, considering the ATI brand had no response to Nvidia's G80, makes it all the more impressive.
Or it could just mean the loss of the high end market had virtually no impact on sales of ATI branded midrange/low end chips, where approximately 90% of revenue is made.
Without being able to compare it to ATI's Q4 2005 results and Q1 2006 results, it's hard to say how well or not well they did.
As regards R600/R650 compared to R520/R570. It's only really similar in the fact that both of the first parts are delayed. R650 doesn't appear to be as much of a major design change as R570 was in comparison to the parent chip.
I'd imagine that unless there were some major additions (IE - the increased shaders in R570 vs R520) I'd imagine that R650 will face a much easier shrink.
Regards,
SB
Over simplyfing? Hmm look at nV's quarters the same time, and you can see where ATi dropped the ball a few times, the entire year or year and half they have been fumbling all over. Its like a high school football team pitted against a seasoned NFL championship team. ATi's/AMD's price drops are just to slow the bleading we have seen this tactic used many times with many different companies. Its nothing new, make sure you don't lose marketshare in bad times, its very hard to gain back, you have to rebuild consumer confidence once that confidence is gone, its hard to convinice consumers we are back on top again.
Silent_Buddha
20-Apr-2007, 13:40
I don't dispute that, ATI's execution has been rather dismal. You'll even see I posted that same view a few pages back.
However, that doesn't negate the fact that with AMD losing buckets of money by having to cut margins to maintain marketshare against Intel, their ATI division isn't suffering the same fate...yet.
I'm neither a fan of Nvidia nor a fan of ATI. I currently have a preference for ATI as their hardware is currently working better with Vista vs. Nvidia's hardware. My friends with Nvidia motherboards and 8800 GTX's are having an rather amazing amount of stability problems. However, as I've said before if Nvidia's Vista 64 drivers were stable, I'd be using an 8800 GTX right now rather than waiting to see what ATI is coming out with.
I'm a fan of one thing and one thing only, hardware that does what I need it to do. And until the 8800 GTX came out, ATI just had flat out better IQ. Unfortunately, even though Nvidia has both better speed AND IQ right now, it's completely and totally useless to me.
So...IF HD 2900 XT is at least as fast as the 8800 GTX...and IF it's actually launching at 400 USD...and IF the Vista 64 drivers for it are as stable as current Catalyst drivers for my X1800XT...then yes, I will be "fan" of it. Just like I was a fan of the GF4 4600-TI. Like I was a fan of the V5 5500. A fan of the Tseng ET-4000.
Just because I think (IE - my opinion) Nvidia's managerial performance right now is much worse than it has been in the past, doesn't mean I like ATI better. After all, it's only because ATI's management has been unable to execute consistently that Nvidia's management continues to look good.
The only people I see as obvious fans here are ones that just automatically discount everything as false if it doesn't promote their company as being better. And considering I only signed up at B3D because I felt it was one of the more impartial boards around, that should say something about how many fans I think are actually here. :)
Regards,
SB
trinibwoy
20-Apr-2007, 15:01
Stability.....the new IQ! :smile:
Stability.....
Obviously of no interest to someone who parks their PC on the desk with the side left off...
Jawed
trinibwoy
20-Apr-2007, 15:15
I've actually neglected my desktop for some time now.....probably misses me something awful. Though I'm not sure what an open case has to do with stability in the first place or how my comment implied that I don't value stability :???:
overclocked_enthusiasm
20-Apr-2007, 16:05
http://www.beyond3d.com/content/news/195
"AMD today reported a net loss of $611M, primarily due to an operating loss of more than $320M in the Computing Solutions (AMD CPUs and ATI chipsets) group while former-ATI operations (GPUs and consumer DTV, consoles, etc) also incurred losses of about $40M."
"GPU revenue in Q1 is down by 20-25%."
So much for any R600 units being shipped or sold to OEMs in Q1...it simply did NOT happen. I highlight this financial information to rebutt the notion that R600 has shipped to OEMs...not to get OT.
EDIT* I added the (stuff in parenthesis) to highlight the different business units AMD now reports.
I've actually neglected my desktop for some time now.....probably misses me something awful. Though I'm not sure what an open case has to do with stability in the first place or how my comment implied that I don't value stability :???:
Enthusiasts are an epitome of out of the box thinking, hard disks run perfectly fine outside your box, so do external cooling apparatus, volt meters and other overclocking tools. people that value that extra fps in general don't care about a lock-up more or less in their quest for the edges of durability of a component..
I mean.. haven't we all?
What makes you think nvidia is capable of pulling some super performing part from thin air?
As said, they had lots of time to tinker around with stuff and try different approaches while ATI had to struggle getting stuff done. That's just a logical conclusion, I may as well be totally wrong.
So much for any R600 units being shipped or sold to OEMs in Q1...it simply did NOT happen. I highlight this financial information to rebutt the notion that R600 has shipped to OEMs...not to get OT.
Are you fighting ghosts? who ever said ATI shipped R600 to OEMs?
overclocked_enthusiasm
20-Apr-2007, 16:22
Are you fighting ghosts? who ever said ATI shipped R600 to OEMs?
Hopefully not...I should have said R6xx or RV6xx parts. It was only 1 or 2 posters who were on this line of thinking and they have likely changed their minds already on this topic. I'll drop it as fighting ghosts is not my speciality.
If not all you have to do is look at how long nvidia has held onto the $550+ GTX price and has not released any other cards since the launch of the 8800 besides a GTS with less ram. They love to milk prices and always have since the fall of 3dfx.
Now please show me one single company which wouldn't do the same... :lol:
If you can make more money, you will. They'd be stupid to do otherwise.
That's not how I remember it...
He meant HDR+AA I guess.
trinibwoy
20-Apr-2007, 16:30
Nvidia hasn't held onto the $550 GTX price. The market has. No competition = increased demand = higher price.
Stability.....the new IQ! :smile:
LMAO, pitty I can't rep you for the "laugh of the week"! :lol:
nicolasb
20-Apr-2007, 16:45
Though I'm not sure what an open case has to do with stability in the first placeWell obviously if you take one side off the case it makes the case unbalanced and more likely to tip over. :yes:
silent_guy
20-Apr-2007, 16:46
Given that R650 would be a "speed-bump" release, architecturally similar to R600 but on a smaller process so as to get a better clockspeed and lower power consumption, I don't think it's at all unreasonable to suppose that the gap between the two might be only six months, with the next major architectural revision (R700) planned to follow 12 months after R600.
Is there one example where refresh followed a new flagship less than 9 months later? (R580 doesn't obviously doesn't count.)
I don't remember there is, but I didn't look for it... But didn't ATI say some time ago they are working towards longer cycles?
icecold1983
20-Apr-2007, 17:00
r350
INKster
20-Apr-2007, 17:05
Nv35.
r350
pretty much the same chip wasn't it?
According to B3D's chip tables, it had the same stuff, same process, everything. Just different clocks.
http://www.beyond3d.com/resources/chip/14
http://www.beyond3d.com/resources/chip/15
icecold1983
20-Apr-2007, 17:47
most refreshes are the same
Well obviously if you take one side off the case it makes the case unbalanced and more likely to tip over. :yes:
You mean it "worsens the stability"? ;)
flippin_waffles
20-Apr-2007, 19:59
Question, what would be the 10 products they will have available at launch as they said in there Q1 report? The rumours seem to be all over the map lately. If there is a R600, RV630, and RV610 then how would those 10 products align within those catagories? Anyone have a guess? The most i've heard of is 3 sku's for each chip.
Question, what would be the 10 products they will have available at launch as they said in there Q1 report? The rumours seem to be all over the map lately. If there is a R600, RV630, and RV610 then how would those 10 products align within those catagories? Anyone have a guess? The most i've heard of is 3 sku's for each chip.
10 products = GPU (5 rv610/rv630/r600 of 2 SKUs each for RV610/RV630 ?) + MB Chipset (3 enthusiast, main-stream, integrated gpu) + CPU (2 ?)
vertex_shader
20-Apr-2007, 20:18
Question, what would be the 10 products they will have available at launch as they said in there Q1 report? The rumours seem to be all over the map lately. If there is a R600, RV630, and RV610 then how would those 10 products align within those catagories? Anyone have a guess? The most i've heard of is 3 sku's for each chip.
Rv610 le/pro/xt, rv630 pro/xt, r600xt, r600 lasso version, r600 stream computing version, r600/rv630 firegl version, this is 10 :smile:
vertex_shader
20-Apr-2007, 20:20
10 products = GPU (5 rv610/rv630/r600 of 2 SKUs each for RV610/RV630 ?) + MB Chipset (3 enthusiast, main-stream, integrated gpu) + CPU (2 ?)
In the 2nd half of May the R600 GPU will be introduced. AMD did not say what the product name will be. They did say that there will be a full line up from top to bottom with 10 products that are all DirectX 10 capable.
:wink:
How are CPU and MB Chipsets not DX10 capable? :wink: They can both run DX10 software.
vertex_shader
20-Apr-2007, 20:30
How are CPU and MB Chipsets not DX10 capable? :wink: They can both run DX10 software.
Everyone know this, this is why i think he mean 10x dx10 VGA, or he talk in PR language :smile:
Everyone know this, this is why i think he mean 10x dx10 VGA, or he talk in PR language :smile:
I think it's all PR language. Why would they be straight shooters now when they weren't with the reasons for delaying the R600 launch?
flippin_waffles
20-Apr-2007, 20:43
Hmm yeah I never thought about chipsets. I thought they weren't releasing those until towards of the end of the year though? I've no idea. But FireGL sounds pretty reasonable to me and some for of stream computing solution.
Maybe 2 sku for R600, 3 each for RV630 and RV610, 1 FireGL and 1 Stream compute. :lol: Feels like i'm making a shopping list! :lol: Too bad I can't afford it. :(
Sound_Card
20-Apr-2007, 21:41
I bet RD790 will launch along R600.:razz:
If I'm not mistaken the rd790 is a AM2+ socket that would mean AMD would have to roll out new CPU's along with that, so probably not going to launch with the r600.
silent_guy
21-Apr-2007, 07:25
r350
According to this (http://www.gamepc.com/labs/view_content.asp?id=r9800pro&page=1&cookie%5Ftest=1), it was introduced 9 months after R300.
Nv35.
NV30 was launched in November 18th, 2002 (and was already WAY later than expected). NV35 was launched in May 2003.
Granted, NV30 wasn't available until much later, but my point is that ATI/NVidia don't deliberately plan successors to be less than 9 months away from each other.
If G80 was as much a surprise to ATI as it was to us, then there was no reason for ATI to schedule an R600 successor less than 9 months after its release. Anyway, if indeed R600 was initially planned for november, then a successor in July is a possibility. Just don't hold your breath. :wink:
Anon Lamer
21-Apr-2007, 10:18
If I'm not mistaken the rd790 is a AM2+ socket that would mean AMD would have to roll out new CPU's along with that, so probably not going to launch with the r600.
Yes and no...AM2+ can take regular AM2 cpus.
flippin_waffles
21-Apr-2007, 11:21
I bet RD790 will launch along R600.:razz:
I hope so Sound_Card! :)
Wow, 100 pages...and still so much time left.
What are the plans for this thread? Let it continue on until R600 actually gets released, or is there going to be another addition to the hall of fame? :razz:
Wow, 100 pages...and still so much time left.
What are the plans for this thread? Let it continue on until R600 actually gets released, or is there going to be another addition to the hall of fame? :razz:
It will run until all pages combined reach 600 (which will be somewhere in the middle of may...)
Twinkie
21-Apr-2007, 22:38
It will run until all pages combined reach 600 (which will be somewhere in the middle of may...)
600 pages..!!
Ok people put your backs into it. We got a LONG way to go :twisted:
nutball
21-Apr-2007, 22:42
600 pages..!!
Add up the thread views. Over 664,000 in the three threads. More views than R600 will sell in its lifetime (probably) :)
Techno+
22-Apr-2007, 15:31
Add up the thread views. Over 664,000 in the three threads. More views than R600 will sell in its lifetime (probably) :)
AMD should reward each beyond3der an r600? :grin:
Megadrive1988
22-Apr-2007, 19:50
NV35 was more of a leap over NV30 than R350 was over R300.
NV35 got a 256-bit bus and more math power, over NV30.
R350 only got some tweaks and something called an F-buffer (no not framebuffer)
that said, the original R300 was still superior to NV35 in almost every way.
Megadrive1988
22-Apr-2007, 19:58
Why not say they had R600 planned for November? That was the original rumour.
Also, it's not out of character for graphics chip companies to alternate two different types of "new" product. The first is a whole new "architecture", while the second is a "speed-bump" release, which is architecturally similar but with performance tweaks. So, for example, GeForce 6xxx to GeForce 78xx was a significant architectural change. 78xx to 79xx was a "speed-bump" change.
Given that R650 would be a "speed-bump" release, architecturally similar to R600 but on a smaller process so as to get a better clockspeed and lower power consumption, I don't think it's at all unreasonable to suppose that the gap between the two might be only six months, with the next major architectural revision (R700) planned to follow 12 months after R600. It would also seem plausible that R650 would follow on fairly close behind RV6xx - again, it wouldn't be out of character for a company to produce a high-end part, followed by mid-range ones on a smaller process, followed closely by a die-shrink of the high performance part as soon as the experience of the mid-range chips has ironed out all the bugs inherent in the new process.
I wouldn't be surprised if R650 hits in September. Conceivably even late August (3 months after R600 XT). Though I think people who are expecting it in June are going to be dissappointed!
typically there are new architectures
(i.e. NV20, NV30, R300, NV40, G80, R600)
there are speed bumps
(i.e. GF2 Ultra, GF3 Ti 500, Radeon 9800, GF 7900 )
and there are refreshes with major architectural improvements & performance gains
(i.e. NV25, R420, R520, R580, G70)
Geeforcer
22-Apr-2007, 20:26
typically there are new architectures
(i.e. NV20, NV30, R300, NV40, G80, R600)
there are speed bumps
(i.e. GF2 Ultra, GF3 Ti 500, Radeon 9800, GF 7900 )
and there are refreshes with major architectural improvements & performance gains
(i.e. NV25, R420, R520, R580, G70)
While R420 was clearly an R300 derivative, I'd say R5xx were a new architecture.
MrBelmontvedere
23-Apr-2007, 03:21
so it's the ATI Radeon HD 2900 XT now? these names are getting way too long :shock:
Indeed they are, I propose a new name :
Radeon 2
It's simple, effective, and states "We're ready for round two.". Now if they could just get off their asses and bring us another R300 :razz:
vBulletin® v3.8.4, Copyright ©2000-2010, Jelsoft Enterprises Ltd.