Post your NV40 preview predictions

[Nick Spolec]

Here's my predictions for what we will discover when the NDA is lifted...




It WILL score over 10k on 3DMark03 (weather this is with cheats or not, I cannot say)
It will be hot and noisy
Almost all the article authors will proclaim it as the return of Nvidia.

[SsP45]

I'll add my own input to that.

Will suck a lot of juice
Will be a very big card, and heavy
________
Colorado Dispensaries (http://colorado.dispensaries.org/)

[Natoma]

It'll be fast and it'll be next gen and everyone will argue that it's better than r420. Did I win the prize now? :P

[tEd]

i predict that anand not realized that HQ AF option is broken :wink:

[Crusher]

It will cost more than Russian mail order bride.

[{Sniping}Waste]

i predict that anand not realized that HQ AF option is broken :wink:

Nether will Tom. :shock: :roll:

[digitalwanderer]

I predict it will be a lot longer time from launch to availability than any of the reviews say it will be.


July. 8)

[anaqer]

July. 8)

Aaand there goes another mouthful of precious coffee. :lol:

[nutball]

I predict that there will be massive flame wars over NV40 v R420 IQ, benchmarks, drivers, blah blah blah for the next 18 months until NV5x and R5xx come out.

*sigh*

[anaqer]

I predict that there will be massive flame wars over NV40 v R420 IQ, benchmarks, drivers, blah blah blah for the next 18 months until NV5x and R5xx come out.

*sigh*

And then it stops? For real? Damn, I can't wait for those chips.

[volt]

will win performance wise?

[tEd]

will win performance wise?

most certainly

[jb]

I agree with this:

- Almost all the article authors will proclaim it as the return of Nvidia.

I also think the card will be "pretty good"..no idea if it will be faster than ATI. But think it will be much better than their current line up.

[rwolf]

NV40 will have a loud fan.
It will be really hot and require lots of juice.
It will have low yields and the chip will be expensive to produce.
The performance will be less than the R420 in most situations.
Nvidia will only manufacture 100,000 and then bring out a new card to replace the NV40.
Nvidia will start cheating on benchmarks to make the performance look better.
It will take ten months to get decent performance with the drivers.

[Joe DeFuria]

It WILL score over 10k on 3DMark03 (weather this is with cheats or not, I cannot say)

Agreed. On a side note, I predict that many sites who shunned 3D Mark for the past two years in lock-step with nVidia, will suddenly find a new appreciation for this benchmark.

It will be hot and noisy

Yes, but everyone (reviewers) will comment on how this isn't the final board, and how the shipping product (promises nvidia) will be much quieter, so ignore the noise.

i predict that anand not realized that HQ AF option is broken

Not only that, but will say how much better HQ looks than normal quality. ;) Seriously though, I predict the high quality not working will be "dismissed as a bug, which nVidia is aware of, and promises to fix before release." Though no thought or mention will be made on what performance hit fixing the "bug" will have...

I also predict that everyone will have 'Ultras' to test, but the closest thing we'll have to seeing "what most people might be able to afford" (non-ultra) will be some reviewers trying to underclock the Ultra.

[PaulS]

NV40U will beat R420P, but not by the kind of margin you'd expect from having 4 extra pipes. It'll be noisy, and everyone will comment on the overclocked memory, the heat, and the nightmare power situation.

The R420XT will come out and beat NV40U overall, without the overclocked memory, the heat, or the nightmare power situation.

ATi will win all AA IQ tests, and NV will take the AF tests whenever they get their HQ option fixed.

Neither card will have enough bandwidth to feed all the pipes anyway, particularly NV40. Stupid 8x2 modes.

[Ante P]

Some incorrect (but understandable) assumptions here ;)

[tEd]

Some incorrect (but understandable) assumptions here ;)

ha ha very funny

[Doomtrooper]

Not many

[Nick Spolec]

Some incorrect (but understandable) assumptions here ;)

Really? So does that mean you won't overlook the High Quality AF "bug"?

[Doomtrooper]

Nether will Tom. :shock: :roll:

Make that Borsti, and I-xbit..two sites that have had a field day with angle dependent AF on ATI.

[I.S.T.]

I say it is foolish to make predictions about the NV40 previews unless your name is TheBaron, DaveB, Uttar and probably a few other people on this board. ;)

[Dave Baumann]

25 pages done...

[tEd]

25 pages done...

all right , you can send them to me for spell check 8)

[PaulS]

25 pages done...

Not bad, given that the R300 review was 18 pages, and the NV30 one was 24 pages.

25 pages done...

all right , you can send them to me for spell check 8)

B3D don't do spell checks ;) j/k

[nutball]

25 pages done...

Sounds like another Anandtech special ;)

[LeStoffer]

25 pages done...

Good job Dave, now you might as well upload those first 25 pages so we can start reading! NDA will probably expire as we get to page 26 anyway... :wink:

Seriously, I expect NV40 to be an awesome if somewhat hot running chip. :o

I very much look forward to your preview and I assume it'll be a tech preview without benchmarks?

[I.S.T.]

25 pages done...

Sounds like another Anandtech special ;)


NOOOOOOOOOOOOOOOOOOOOOOOO!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!


Please don't do 50 pages of useless info!

[Kristof]

25 pages done...

Now all together :

WE WANT MORE, WE WANT MORE, ... :twisted:

[{Sniping}Waste]

You finished the reveiw and its 25 pages long? :shock:
This is going to be a good long tech read. The way I like them.

Thanks Dave.

[Dave Baumann]

I said 25 pages done, not that it was finished!

[Nick Spolec]

I'm expecting to see 30+ page monsters, which will try and answer the question of "How many useless benchmarks can you fit into a review?".

[tEd]

I said 25 pages done, not that it was finished!

I think it's good that this time we get some rather deep previews instead of the usual we had the card only for 7hour launch previews

[Kristof]

I said 25 pages done, not that it was finished!

Remember to not neglect Noodle and Nibblet now... oh and Neva... :oops:

And keep the cats away from the keyboard, they are known to press delete at the worst possible moment :?

K-

[PaulS]

I'm expecting to see 30+ page monsters, which will try and answer the question of "How many useless benchmarks can you fit into a review?".

My problem with reviews has never been length - it's been readability. So many places have published graphs which are unreadable. Keep it simple, and provide tables of values. Nothing worse than graphs where you can't tell what's going on :?

[ZenThought]

NV40 will have a loud fan.
It will be really hot and require lots of juice.
It will have low yields and the chip will be expensive to produce.
The performance will be less than the R420 in most situations.
Nvidia will only manufacture 100,000 and then bring out a new card to replace the NV40.
Nvidia will start cheating on benchmarks to make the performance look better.
It will take ten months to get decent performance with the drivers.

During last Cc, Jensen said NV40 was designed for high yield,
so unless he flatly lieing (which would be legally liable), it shouldn't
be too bad.

[Kristof]

During last Cc, Jensen said NV40 was designed for high yield, so unless he flatly lieing (which would be legally liable), it shouldn't be too bad.

"high" could be a very relative term... high compared to other chips of this huge size could still be pretty darn low... 8)

[Joe DeFuria]

During last Cc, Jensen said NV40 was designed for high yield,so unless he flatly lieing (which would be legally liable), it shouldn't be too bad.

During other conference calls JC said that NV30 would be a summer '02 product.

Point: forward looking statements are just that. Forward looking. They can turn out to be incorrect, but that doesn't mean he was purposely lying...

[LeStoffer]

And keep the cats away from the keyboard, they are known to press delete at the worst possible moment :?

K-

... One more reason to upload a 'backup' to the site already! :twisted:

[tazdevl]

NV40U will beat R420P, but not by the kind of margin you'd expect from having 4 extra pipes.

You sure about that? ;)

[Dave Baumann]

I don't know why, but I find this mildly amusing:

We were made aware that when rying to run the Gunmetal benchmark it causes a crash. This is caused by the inability of the benchmark to properly recognize the GeForce6800 Ultra GPU. This benchmark tries to detect the GPUs id; it is impossible for the creator of the Gunmetal benchmark to fix this since the company has gone bankrupt.

[martrox]

I predict DemoCoder will say I'm biased towards ATI.........OOPS, he already did! :wink:

BTW, It will be hot - but not very noisy, will require lots of power, will get incredible benchmark scores - the "optimizations" are still there, and are fully used.

And, it will be a much nicer, more competitive card than the NV3xx........

[Joe DeFuria]

I don't know why, but I find this mildly amusing:

We were made aware that when rying to run the Gunmetal benchmark it causes a crash. This is caused by the inability of the benchmark to properly recognize the GeForce6800 Ultra GPU. This benchmark tries to detect the GPUs id; it is impossible for the creator of the Gunmetal benchmark to fix this since the company has gone bankrupt.

Lol...it would be even funnier if the game itself runs OK, but the benchmark doesn't....is that the case?

That also reminds me...how are the creators of the "bridge-it" game doing...they bankrupt yet? ;)

[Dave Baumann]

Mmmm, I think it will be noisy dependant on which reviews you read.

[Nick Spolec]

We were made aware that when rying to run the Gunmetal benchmark it causes a crash. This is caused by the inability of the benchmark to properly recognize the GeForce6800 Ultra GPU. This benchmark tries to detect the GPUs id; it is impossible for the creator of the Gunmetal benchmark to fix this since the company has gone bankrupt.


Wasn't that basically an Nvidia benchmark?

Oh, right. I mean, "Nvidia favored benchmark".

[martrox]

Mmmm, I think it will be noisy dependant on which reviews you read.

Well......I'm already wrong on this..........hehe, guess that's what I get for giving them (nVidia) the benifit of doubt!

[ZenThought]

During last Cc, Jensen said NV40 was designed for high yield,so unless he flatly lieing (which would be legally liable), it shouldn't be too bad.

During other conference calls JC said that NV30 would be a summer '02 product.

Point: forward looking statements are just that. Forward looking. They can turn out to be incorrect, but that doesn't mean he was purposely lying...

This time it was more specific statement ("Gross Margin will
be approaching historical numbers"). So my guess is that
they have produced enough NV40 in production to estimate yield
properly.

[Psikotiko]

- it will beat X800 pro
- heat problems
- power problems
- New AA

[{Sniping}Waste]

The install papers will say please remove case side panel before powering up your system.

[Kristof]

I don't know why, but I find this mildly amusing:

We were made aware that when rying to run the Gunmetal benchmark it causes a crash. This is caused by the inability of the benchmark to properly recognize the GeForce6800 Ultra GPU. This benchmark tries to detect the GPUs id; it is impossible for the creator of the Gunmetal benchmark to fix this since the company has gone bankrupt.

You could use 3DAnalyze to chang the ID, that might allow the benchmark to run... not that you would want to :roll:

[tEd]

We were made aware that when rying to run the Gunmetal benchmark it causes a crash. This is caused by the inability of the benchmark to properly recognize the GeForce6800 Ultra GPU. This benchmark tries to detect the GPUs id; it is impossible for the creator of the Gunmetal benchmark to fix this since the company has gone bankrupt.


Wasn't that basically an Nvidia benchmark?

Oh, right. I mean, "Nvidia favored benchmark".

it was the *cough* dx9 benchmark *cough* which should show how fast nvidia is in doing dx9 stuff i mean vs2.0 :wink:

[Kristof]

Mmmm, I think it will be noisy dependant on which reviews you read.

The return of MP3 files to indicate how silent the new cards are or aren't ?

[nobie]

I predict it will be much better than NV3x, but still have some IQ issues.

[russo121]

:) :) :) :) :) :)

IMO a lot of people here are taking too much hallucinogenics already..... they seem to see a good card?!

[Nick Spolec]

As long as they aren't smoking from a glass dick.

[Joe DeFuria]

This time it was more specific statement ("Gross Margin will be approaching historical numbers"). So my guess is that
they have produced enough NV40 in production to estimate yield
properly.

That's still a forward looking statement. And it also doesn't say anything about the specific NV40 ultra product.

In any case, what was the full context of the quote? NV40 generation of chips? Gross margin including all products for next quarter?

[dan2097]

We were made aware that when rying to run the Gunmetal benchmark it causes a crash. This is caused by the inability of the benchmark to properly recognize the GeForce6800 Ultra GPU. This benchmark tries to detect the GPUs id; it is impossible for the creator of the Gunmetal benchmark to fix this since the company has gone bankrupt.


Wasn't that basically an Nvidia benchmark?

Oh, right. I mean, "Nvidia favored benchmark".

it was the *cough* dx9 benchmark *cough* which should show how fast nvidia is in doing dx9 stuff i mean vs2.0 :wink:

It doesnt even show that, it only had 1 vs2.0 shader, the gffxs aren't that good at vs2.0 :wink:

I just really hope we get good image qualtity analysis, game benchmarks [H]ard style, and some nice synthetics.

Im really not sure whats going to happen, although these reviews will be interesting as they only have the nv38/r360 to compare to, I dont expect the reviewers will be blind enough to count ATI out of the race.

I think ill just predict that any site that complains about the noise levels will have been assured by Nvidia that the shipping product will be quieter :P

[{Sniping}Waste]

Found this

Also... from Warp2search.net discussion...
Quote:
A certain new card we have caused one of our PCs (an Athlon XP) to auto shutdown after 20/25 minutes gaming. This was due to heat build up in the case, the otherwise OK CPU was overheating, it was 10c+ hotter with this card than with a 5950.. The case had 80mm fans front and back.
If you have an already hot/overclocked computer you may not get on with this certain card (that i cant yet mention of course).
What with this and the need for a warp core to power it i forsee trouble ahead for many purchasers.

[CapsLock]

Judging by these posts NV sure have lost a lot of credibility lately. As well as a few tech sites, though not this one of course.

To all of the above I would like to add the should-be-obvious memory bandwidth handicap. Double the pipes- not anywhere near double the bandwidth. Unless NV has come up with some uber/freakin new techniques, this is going to be the big problem performance wise. At least as far as my completely shallow thinking goes.

Caps

[Hellbinder]

I predict that..

-It will be pleanty Fast.
-Have some improvements (or claimed improvements) to AA and AF.
-be much faster in AF tests comparatively to the last generation.
-It will have greatly improved Shader performance Comparatively to the last generation.

I also Predict that.

-There are still numerous DRivers *optimizations* Specifically in Benchmarks and specific benchmark levels from games

-Certain web sites will be drooling all over themselves to proclaim Nvidia as the new king. Glossing over any negatives they can.

-It will be marginally faster than the R420pro but not in every case.
-Ultimately it will end up losing most benchmarks between 10-20% with some rare wins and loses being bigger that this to the X800XT.

My personal view..

-NV40 is another example of how Nvidia has lost the beat when it comes to engineering a product. They overstuff it with features that no one needs right now and pass up on the opportunity to really make it User Friendly, Quiet, not hot as an oven, heavy as an anvil and requiring more power than a locomotive. Instead they make a product that they have to clock at the brink of artifacting just to deliver on promises.

Elegance is a word that Nvidia seems to have ripped out of its dictionary.

[DemoCoder]

I predict DemoCoder will say I'm biased towards ATI.........OOPS, he already did! :wink:

BTW, It will be hot - but not very noisy, will require lots of power, will get incredible benchmark scores - the "optimizations" are still there, and are fully used.

And, it will be a much nicer, more competitive card than the NV3xx........

I predict you should step off.

[PaulS]

Mmmm, I think it will be noisy dependant on which reviews you read.

I don't know why, but I find this mildly amusing:

We were made aware that when rying to run the Gunmetal benchmark it causes a crash. This is caused by the inability of the benchmark to properly recognize the GeForce6800 Ultra GPU. This benchmark tries to detect the GPUs id; it is impossible for the creator of the Gunmetal benchmark to fix this since the company has gone bankrupt.

God, what's going on :shock: :?

[CapsLock]

-It will be marginally faster than the R420pro but not in every case.
-Ultimately it will end up losing most benchmarks between 10-20% with some rare wins and loses being bigger that this to the X800XT.

And I thought my english was bad.

My personal view..

-NV40 is another example of how Nvidia has lost the beat when it comes to engineering a product. They overstuff it with features that no one needs right now and pass up on the opportunity to really make it User Friendly, Quiet, not hot as an oven, heavy as an anvil and requiring more power than a locomotive. Instead they make a product that they have to clock at the brink of artifacting just to deliver on promises.

Elegance is a word that Nvidia seems to have ripped out of its dictionary.

Sub "efficiency" for elegance and I agree completely.

Caps

edit for superfluous stuff.

[Bjorn]

They overstuff it with features that no one needs right now ....

Just out of curiosity, what features are you talking about ?

[DemoCoder]

-NV40 is another example of how Nvidia has lost the beat when it comes to engineering a product. They overstuff it with features that no one needs right now

Wasn't the R300 also an example of a card overstuffed with features that no one needed and hasn't needed for the last 2 years? The R300 supported way more DX9 features than the NV30, not just PS2.0, but FP cube maps, MRT, HDR, etc. And who needed PS2.0 at all? How much more "elegant" and power friendly their existing chip could have been if they simply made it an 8-pipe R200 with multisample gamma corrected AA and left out the other "unneeded" baggage.

Right about the time the R500 comes out, we'll see you flip about how ShaderModel 4.0 is the best thing since sliced bread and jumping for joy to high heaven about it.

You went from "ATI is going to OWN Nvidia in performance" to "well, it will be close" to "doesn't matter anyway, cause it's hot and noisy and that invalidates everything" Denial is not a river in Egypt.

As for Elegance, I'd reserve that word for PowerVR. Both ATI and NVidia are "brute force" chips: mega pipelines, mega buses, mega ram.

[Hellbinder]

-It will be marginally faster than the R420pro but not in every case.
-Ultimately it will end up losing most benchmarks between 10-20% with some rare wins and loses being bigger that this to the X800XT.

And I thought my english was bad.

My personal view..

-NV40 is another example of how Nvidia has lost the beat when it comes to engineering a product. They overstuff it with features that no one needs right now and pass up on the opportunity to really make it User Friendly, Quiet, not hot as an oven, heavy as an anvil and requiring more power than a locomotive. Instead they make a product that they have to clock at the brink of artifacting just to deliver on promises.

Elegance is a word that Nvidia seems to have ripped out of its dictionary.

Sub "efficiency" for elegance and I agree completely.

Caps

edit for superfluous stuff.
Ok so i am really from Mars not earth... And occasionally i typw stuff that sounds really really dumb.

[PaulS]

And occasionally i typw stuff that sounds really really dumb.

You just know that's going to come back and bite you in the ass :D

[ninelven]

<ot>
I thought R500 was SM3.0...
</ot>

[DemoCoder]

<ot>
I thought R500 was SM3.0...
</ot>

I've seen speculation it will support whatever DX-Next does. In anycase, doesn't matter. Whatever the R500 supports is what HB will deem needed and elegant, irrespective of the time delay and level of developer support before it gets used. Stuffed with features? Naw, ATI only puts *exactly* what is optimal and useful into their cards.

[Hellbinder]

-NV40 is another example of how Nvidia has lost the beat when it comes to engineering a product. They overstuff it with features that no one needs right now

Wasn't the R300 also an example of a card overstuffed with features that no one needed and hasn't needed for the last 2 years? The R300 supported way more DX9 features than the NV30, not just PS2.0, but FP cube maps, MRT, HDR, etc. And who needed PS2.0 at all? How much more "elegant" and power friendly their existing chip could have been if they simply made it an 8-pipe R200 with multisample gamma corrected AA and left out the other "unneeded" baggage.

Right about the time the R500 comes out, we'll see you flip about how ShaderModel 4.0 is the best thing since sliced bread and jumping for joy to high heaven about it.

You went from "ATI is going to OWN Nvidia in performance" to "well, it will be close" to "doesn't matter anyway, cause it's hot and noisy and that invalidates everything" Denial is not a river in Egypt.
Well the R300 was no more overstuffed with features than the Nv30 was. I often wonder how it is you actually process information.

-R300 = meets DX9 spec and delivers a single slot/Quiet well performing solution with massive performance increase and improved AA.

-Nv30 =exceeds DX9 specs, has shader issues, has pipeline issues, needs lots of Driver hacks, takes 2 slots, is very Hot not improved AA.

There is an obvious difference betwen the two above cards and anyone with an OPEN MIND can see the difference in Two seconds.

-R420 = meets the Standards for games available TODAY, has greatly improved performance improved AA is not overly Hot, Power hungry or needing multiple slots.

-Nv40 = Beyond the needs of Todays games, Features that will not be used for Quite some time. is HOT HEAVY NOISE and LARGE.

How many people have to say that loud, hot, and large are not desirable factors for a Video card for you to get the picture? Performance is indeed important but so is elegant design and functionality.

IF Card X is very close to Card Y in performance and usable features and is a far more elegant design. That will be important to most people. Wether you choose to admit it or not. Further When Card Z is faster than card Y accross the board and is still more elegant in design.. You figure it out.

[Brian3DGPU]

I'll also bet that the the 16 pipe ATI card will require two molex connectors. DaveB...can you email me at brian.evans@3dgpu.com

[CapsLock]

And occasionally i typw stuff that sounds really really dumb.

You just know that's going to come back and bite you in the ass :D

LOL, Bail, HB, Bail out now!!!!!

Seriously, easy sniping aside, what about the bandwidth prob?

8*500=4000
16*400=6400
60% difference

500 mem - 550 mem = 10% difference

Now thats a problem.

Caps

[Hellbinder]

The above is not to say that the Nv40 is not Fast and has cool features.

It is simply a case of we dont need a 220m transistor card that is hot as an oven and heavy as an anvil. I *personally* think Nvidia should have kept their PS 2.0+ system made improvements to it, increased the pipelines, imporived the AA and AF and just made a really solid card that is an elegant design.

But again, thats just me.

[ninelven]

Features that will not be used for Quite some time

Just curious... what are these features and how can you know they won't be used for "quite some time" ?

[Psikotiko]

8*500=4000


8?

[anaqer]

-R420 = meets the Standards for games available TODAY, has greatly improved performance improved AA is not overly Hot, Power hungry or needing multiple slots.

Improved AA wrt ATi is a "yeah, whatever" thing to most ( please signal if you think otherwise ).
Now AF on the other hand... well, what can you say about AF?

[surfhurleydude]

Well the R300 was no more overstuffed with features than the Nv30 was. I often wonder how it is you actually process information.

-R300 = meets DX9 spec and delivers a single slot/Quiet well performing solution with massive performance increase and improved AA.

-Nv30 =exceeds DX9 specs, has shader issues, has pipeline issues, needs lots of Driver hacks, takes 2 slots, is very Hot not improved AA.

There is an obvious difference betwen the two above cards and anyone with an OPEN MIND can see the difference in Two seconds.

-R420 = meets the Standards for games available TODAY, has greatly improved performance improved AA is not overly Hot, Power hungry or needing multiple slots.

-Nv40 = Beyond the needs of Todays games, Features that will not be used for Quite some time. is HOT HEAVY NOISE and LARGE.

How many people have to say that loud, hot, and large are not desirable factors for a Video card for you to get the picture? Performance is indeed important but so is elegant design and functionality.

IF Card X is very close to Card Y in performance and usable features and is a far more elegant design. That will be important to most people. Wether you choose to admit it or not. Further When Card Z is faster than card Y accross the board and is still more elegant in design.. You figure it out.

I have to say that you are definitely twisting things around from the way they actually are...

Truth - There are only a handful of DX 9 games out right now, almost two years after the release of the R3xx... If you do not consider that "exceeding the standards for the games at the release of the card" I don't know what does.
Subtruth - The R3xx certainly does everything well enough though ;)

Truth - NV40 uses PS 3.0. This is definitely not in the same vein as the NV30's "exceeding DX9 spec" classifcation. Because should I decide to purchase an NV40, there will be a game on DAY ONE that I have that will support PS3.0, Far Cry.
Subtruth - I'm sure you've read by now that PS 3.0 is very easily to incorporate into games with PS2.0 already coded into the game, and PS 3.0 compatibility is expected to catch on pretty fast. PS 3.0 in Half Life 2? Maybe!

Truth - In reality you could make the case saying that both the NV40 and R420 will "EXCEED standards for games TODAY" because upon looking at my cd wallet full of games, only one of them actually uses DX 9 in any sort of meaningful way... And looking at release dates for upcoming games, the vast majority of them are DX 8 titles as well.


On a side note though, your claim of elegance being more important than performance is actually a very valid one, and it is something I personally hold in higher esteem than performance.

[CapsLock]

8 or 4X2, whatever.

Caps

[surfhurleydude]

oops double post... sorry :(

[Brian3DGPU]

As far as useless features...same old chicken and egg argument that has been around since 16 vs. 32 bit. Fact is, devs need something to develop on to make the games to support it. That and unlike a lot of us, tons of people don't upgrade their cards twice a year, or even once every two years...

[Doomtrooper]

http://www.webhostingtalk.com/images/smilies/popc1.gif

[martrox]

I'll also bet that the the 16 pipe ATI card will require two molex connectors. DaveB...can you email me at brian.evans@3dgpu.com

Brian's calling Dave out......FIGHT!!! FIGHT!!!!......

:wink:
(j/k - really)

[Hellbinder]

Features that will not be used for Quite some time

Just curious... what are these features and how can you know they won't be used for "quite some time" ?
Obviously Nvidia is going to get some devs to use PS3.0 stuff in the not to distant future and that is Cool. But.. Even Mark Reign is saying that the bulk of really cool Shader stuff is still 18+ months away.

[noko]

What is expected is this is Nvidia's come back card to take the crown back. Now what crown is beyond me. Now will the many more transitors offset the faster and less transitor R420 series, I don't think so. Still the NV40 will kick some butt over ATI a.k.a DoomIII and games that spin from that game engine plus ps3.0 stuff, which FarCry might be able to show.

Overall though I think ATI will deliver a significant blow to Nvidia with their X800 XT card. The NV40 era I predict is when Nvidia actually starts to lose money, ATI just has an overall superior line up. The old Nvidia lines NV3x series repackage for PCI express is utterly inferior to ATI's R420 line up coming out. ATI allowance for failed chips to be used on a lesser product - - That is just plain smart as far as I am concerned.

Nvidia market share I predict will shrink rapidly is the bottom line.

[Miksu]

Well, I'm just waiting for B3D's, Hardocp's and NordicHardware's previews. Maybe that way I can get a real picture what the card can do as I predict that most of the preview-sites still give us benchmarks from Quake3 run without aa/aniso. And that really isn't the info I'm looking for.

My predictions about the card itself has already been stated in this thread: Fast, hot etc.

[DemoCoder]

-Nv30 =exceeds DX9 specs, has shader issues, has pipeline issues, needs lots of Driver hacks, takes 2 slots, is very Hot not improved AA.


It does not exceed DX9 specs. In totality, it lacks many DX9 specs that R300 supported. It exceeded the spec in just a few areas with respect to # of instructions supported, and a couple of extra instructions. ATI also exceeds the specs. Their HW is capable than more than what is exposed in DX9, especially if you include F-Buffer, which no real-time game is going to use. (has any game used a shader approaching 96 instructions?)

How many people have to say that loud, hot, and large are not desirable factors for a Video card for you to get the picture? Performance is indeed important but so is elegant design and functionality.

I don't care what people say on this board. This board is a biased sample on both sides. The market will speak. NV3x was a loud and hot card that *did not deliver features and DX9 performance*, so the heat and noise were not justified. Personally, I will buy the best performing card for the money. I think the majority of the market will buy the mid-range cards, which will be the 12-pipe R420 vs the 12-pipe NV4x, that's where the true comparison of noise and heat will lie. This X800XT vs NV40U contest is just a PR mindshare contest. Vast majority of products sold won't be either of those.

I don't mind buying something hot and loud *IF* I get something for it.

[martrox]

I don't mind buying something hot and loud *IF* I get something for it.

And you will get........something hot and loud! :wink:

[DemoCoder]

Features that will not be used for Quite some time

Just curious... what are these features and how can you know they won't be used for "quite some time" ?
Obviously Nvidia is going to get some devs to use PS3.0 stuff in the not to distant future and that is Cool. But.. Even Mark Reign is saying that the bulk of really cool Shader stuff is still 18+ months away.

Sounds about the time HL2 will finally ship. Team Fortress 2 and Duke Nukem Forever will ship about 3 years after that. :)

[Brian3DGPU]

I predict that most of the preview-sites still give us benchmarks from Quake3 run without aa/aniso. And that really isn't the info I'm looking for.

Oh, come on...a few sites will , but not the majority. Also realize that many smaller sites do not have staff that are full-time. We do the best we can in the time given.

[Hellbinder]

Hey, I just want to say one more time that I think NV40 is an ok product from early information. I only think that Nvidia is focusing on the Wrong things right now thats all. I am not suggesting that people in general should not buy it. All the cards comming out are going to be plently fast for gaming over the next year. ITs going to come down to other factors. Like Design elegance, total pakcage, Actual in game IQ stuff like that. I suppose performance in specific games that people are really looking forward to is a big issue as well. Doom-III people will perhaps want the Nv40 etc...

[ZenThought]

What is expected is this is Nvidia's come back card to take the crown back. Now what crown is beyond me. Now will the many more transitors offset the faster and less transitor R420 series, I don't think so. Still the NV40 will kick some butt over ATI a.k.a DoomIII and games that spin from that game engine plus ps3.0 stuff, which FarCry might be able to show.

Overall though I think ATI will deliver a significant blow to Nvidia with their X800 XT card. The NV40 era I predict is when Nvidia actually starts to lose money, ATI just has an overall superior line up. The old Nvidia lines NV3x series repackage for PCI express is utterly inferior to ATI's R420 line up coming out. ATI allowance for failed chips to be used on a lesser product - - That is just plain smart as far as I am concerned.

Nvidia market share I predict will shrink rapidly is the bottom line.

X800Xt is panic response to 6800 Ultra. It will have very high price
and low yield.

[Tim Murray]

I guess we can't comment if we already know the answers to all of these...

damn, where's the popcorn smiley. :)

[Joe DeFuria]

X800Xt is panic response to 6800 Ultra. It will have very high price
and low yield.

Neither the X800XT and 6800 are "panic" responses, as much as they are (as DC said) PR Marketing tools and mindshare devices.

Both will have very high price and be low yield.

The "real" high end products (the ones with more realistic prices and therefore more producible volume) will be the X800 Pro and the 6800 non-Ultra.

[anaqer]

X800Xt is panic response to 6800 Ultra. It will have very high price and low yield.

And you know this for sure because..? :roll:

[surfhurleydude]

Hey, I just want to say one more time that I think NV40 is an ok product from early information. I only think that Nvidia is focusing on the Wrong things right now thats all. I am not suggesting that people in general should not buy it. All the cards comming out are going to be plently fast for gaming over the next year. ITs going to come down to other factors. Like Design elegance, total pakcage, Actual in game IQ stuff like that. I suppose performance in specific games that people are really looking forward to is a big issue as well. Doom-III people will perhaps want the Nv40 etc...

I appreciate you trying to save face and stick up for the NV40, but spare us the BS... Design elegance, total package, and Actual in game IQ has been what purchasing has been about EVER SINCE the NV30 fiasco. But nevertheless, it still hasn't been what nVidia has really focused on :(

[noko]

I would also say price. I think ATI designs allows for a lower cost overall which means even more manufacturers producing ATI chipped cards. Nvidia much higher power and heat issues must hurt OEM sells in the end.

[ZenThought]

X800Xt is panic response to 6800 Ultra. It will have very high price
and low yield.

Neither the X800XT and 6800 are "panic" responses, as much as they are (as DC said) PR Marketing tools and mindshare devices.

Both will have very high price and be low yield.

The "real" high end products (the ones with more realistic prices and therefore more producible volume) will be the X800 Pro and the 6800 non-Ultra.

Then, why did they leak information about having 16 pipes and release
dates?
They saw that 6800 spec and realized they need to response even
thought it will cost them yields.

[Doomtrooper]

ATI could have easily buried the competition if they felt they needed to release what I would call a 'brute force' approach last year. If people feel this current Nv40 is ready for primetime on the market (meaning hot, noisy and power hungry ), then I see no difference with that mentality with the card below. I would wager this card here would outperform some of the next generation stuff, but last year....and why stop at two, lets go 4 VPU's :)

How far do you want to go.



http://www.forum-hardtecs4u.com/3dcenter/images/2004/03-23_pic6.jpg

[Joe DeFuria]

Then, why did they leak information about having 16 pipes and release
dates?

Um, because it's exactly what I said? A PR Battle?

[Dave Baumann]

Then, why did they leak information about having 16 pipes and release
dates?

If they have a part capable of 16 pipes, do you think they would have never realesed it in that configuration?

[noko]

They saw that 6800 spec and realized they need to response even
thought it will cost them yields.

It won't cost them any yield since if the chip doesn't make it at 16 pipes it may at 12 or 8 unless the chip totally fails. This doesn't effect yields at all. ATI designed the chip to be 16 pipes anyways, why not take advantage of those that make it, ATI just being smart again.

[Hellbinder]

Hey, I just want to say one more time that I think NV40 is an ok product from early information. I only think that Nvidia is focusing on the Wrong things right now thats all. I am not suggesting that people in general should not buy it. All the cards comming out are going to be plently fast for gaming over the next year. ITs going to come down to other factors. Like Design elegance, total pakcage, Actual in game IQ stuff like that. I suppose performance in specific games that people are really looking forward to is a big issue as well. Doom-III people will perhaps want the Nv40 etc...

I appreciate you trying to save face and stick up for the NV40, but spare us the BS... Design elegance, total package, and Actual in game IQ has been what purchasing has been about EVER SINCE the NV30 fiasco. But nevertheless, it still hasn't been what nVidia has really focused on :(
Wow.. you guys are harsh... ;)

Ok.. the Nv40 totally SUCKS spoiled Eggs and it will lose more benchmarks than Janet Jackson loses Clothes on national TV. Its IQ will suck, and be inferior to ATi's LAST GENERATION products. It weighs 22 pounds per square inch, runs at 2,000 degrees, and will Blow out the inner ear of the average person in under 2.2 seconds. Nvidia will prop the lame duck up with Bully tactics, Lies, Evil PR practices, Driver hacks and *special agreements* with specific developers to cripple everyone elses hardware on their software.

Whew.. I feel so much better now :lol:

[ClyssaN]

X800Xt is panic response to 6800 Ultra. It will have very high price
and low yield.

Neither the X800XT and 6800 are "panic" responses, as much as they are (as DC said) PR Marketing tools and mindshare devices.

Both will have very high price and be low yield.

The "real" high end products (the ones with more realistic prices and therefore more producible volume) will be the X800 Pro and the 6800 non-Ultra.

Then, why did they leak information about having 16 pipes and release
dates?
They saw that 6800 spec and realized they need to response even
thought it will cost them yields.

So if The Inq had posted the ATI 16 pipes specs before the nvidia 16 pipes specs, that means NVIDIA response was a panic one ?

[CapsLock]

I guess we can't comment if we already know the answers to all of these...

Ok Baron Smarty Pants... theoretically speaking of course, how would you deal with the drastic change in available bandwidth per pipe NV is facing? Does going X1 TMU somehow improve efficiency? Do they have amazing new compression tech? Much higher than leaked mem speed? Magic Dawn pixie dust?

Caps

And do not be coy! We have ways of making you talk. ie.: find an electrical socket near your current location. Stick a fork in till you decide to co-operate. :shock:

edit: added reference since people are posting so darn fast

[Joe DeFuria]

Ok Baron Smarty Pants... theoretically speaking of course, how would you deal with the drastic change in available bandwidth per pipe NV is facing?

Easy....encourage review sites to benchmark in 16 bit mode. ;)

[ZenThought]

They saw that 6800 spec and realized they need to response even
thought it will cost them yields.

It won't cost them any yield since if the chip doesn't make it at 16 pipes it may at 12 or 8 unless the chip totally fails. This doesn't effect yields at all. ATI designed the chip to be 16 pipes anyways, why not take advantage of those that make it, ATI just being smart again.

16 pipes was having yield problem which is not unexpected with
that many transistors on low-k. So 12 pipes was their original plan but
when they saw 16 pipe specs, they release 16 parts.

[surfhurleydude]

Easy....encourage review sites to benchmark in 16 bit mode. ;)

Yeah, I knew that 3dfx tech would be good for something. :D

[noko]

Hmnm, I just thought of something. . . What if the yields of the R420 with 16 pipes is much higher then expected? I actually think that will be the case since ATI seems more conservative. Will that mean those extra capable chips being stuck in X800 Pro's (I think so :D) or prices across the line go down?

[Stryyder]

They saw that 6800 spec and realized they need to response even
thought it will cost them yields.

It won't cost them any yield since if the chip doesn't make it at 16 pipes it may at 12 or 8 unless the chip totally fails. This doesn't effect yields at all. ATI designed the chip to be 16 pipes anyways, why not take advantage of those that make it, ATI just being smart again.

16 pipes was having yield problem which is not unexpected with
that many transistors on low-k. So 12 pipes was their original plan but
when they saw 16 pipe specs, they release 16 parts.

Zen is smoking the glass pipe.

[Doomtrooper]

Ahhh this whole launch thing is just hilarious...nothing but mind games :lol:

[{Sniping}Waste]

Caps

And do not be coy! We have ways of making you talk. ie.: find an electrical socket near your current location. Stick a fork in till you decide to co-operate.


I about spit my Coke all over my screen reading that. Please put a warning up before posting stuff like that. :shock:

[Moose]

I predict that the Nv40 will fast as hell with awesome new features for developers. It will be proclaimed as the return of the king of all video cards (at least for a couple of weeks or so), especially by Nvidia friendly sites. Ps3.0 will be the big new feature that we absolutely must have even though a couple generations of cards will come and go before Ps 3.0 is actually useable in games.

Futuremark will once again be relevant to Nvidia and will now feature tests for PS3.0 at least as tech demos. Cheats will slowly disappear and or reappear in direct proportion to how far ahead or behind Nvidia is in benchmarks.

All this will come at a cost though. Around $500 for the card and $60 or more for the required power supply, possibly even more for larger cases and more fans.

I also predict that the heat and noise of the card will turn off many including computer companies (Dell etc) who will have a big problem getting this thing installed even in their high end computers.

Lastly I will predict that I am going to wait for ATI's line up before I even think about this one. From what I've seen so far, as long as ATI can stay in the ballpark speedwise they may have won this round again with a cooler more customer oriented card.

Oh, and of course the ATI/Nvidia battle will rage on and on until one company goes out of business or a third company comes along to knock one of them out of the battle. :)

[CapsLock]

Caps

And do not be coy! We have ways of making you talk. ie.: find an electrical socket near your current location. Stick a fork in till you decide to co-operate.


I about spit my Coke all over my screen reading that. Please put a warning up before posting stuff like that. :shock:

A thousand apologies good sir. Too much coke is bad for you anyway. :wink:

Caps

[noko]

16 pipes was having yield problem which is not unexpected with
that many transistors on low-k. So 12 pipes was their original plan but
when they saw 16 pipe specs, they release 16 parts.

That sounds like conjecture. I believe they gave themselves targets to reach and options depending on a number of factors (a.k.a ram available, ram speeds, yields, performance of part etc.) The 16 pipe chip was designed and already tested and may have been pushed up in the release date due to competition in the high end. Still it is the same chip in the end and everything was in place to support it. Panic - - I think not. Actually I think Nvidia is the one panicking now with their slower 400 mhz beast compared to a 500mhz R420 with overdrive (hmmm, 550mhz ;) )

[Doomtrooper]

Does the trend continue for scalable VPUs from ATI :?: :D

[jvd]

Does the trend continue for scalable VPUs from ATI :?: :D

dual r 420s on a card ?
that be sweet

[Dave Baumann]

Now, that would need dual molex's!

(It'd also be pretty stupid).

[Doomtrooper]

I seriousally doubt it would happen but, I'm waiting to see the reaction from the internet community and review sites from the Nv40. If they feel this approach is acceptable...why not...double the AA modes :D

[Doomtrooper]

Now, that would need dual molex's!

(It'd also be pretty stupid).

Huh....there would be buyers though :lol:

[noko]

Multiple cards in PCI express slots should be possible with 32 pipes in action ;). Only thing is that you will probably need two 16x slots :(.

[Dave Baumann]

Gah. All this pipeline talk will be so irrelevent soon.

[Miksu]

Oh, come on...a few sites will , but not the majority. Also realize that many smaller sites do not have staff that are full-time. We do the best we can in the time given.

Didn't mean to say that every other hardware-site is beneath those three I mentioned. Those are just the ones I've personally found to cover the things I like to know. There can be and there are other sites too, but I haven't "found" them yet.

[CapsLock]

Ok Baron Smarty Pants... theoretically speaking of course, how would you deal with the drastic change in available bandwidth per pipe NV is facing?

Easy....encourage review sites to benchmark in 16 bit mode. ;)

Actually that might be a good way of testing bandwidth limitations, wouldn't it?

Caps

[CapsLock]

Gah. All this pipeline talk will be so irrelevent soon.

Hey Dave, I'm trying my best to get em to switch.

Caps

[noko]

Gah. All this pipeline talk will be so irrelevent soon.

Well it is easy for people to relate, you know more inches the better, the more pipe the better ;).

[Joe DeFuria]

Ok Baron Smarty Pants... theoretically speaking of course, how would you deal with the drastic change in available bandwidth per pipe NV is facing?

Easy....encourage review sites to benchmark in 16 bit mode. ;)

Actually that might be a good way of testing bandwidth limitations, wouldn't it?

Caps

Yes, it's one way to test for that. But as we know, nvidia frowns heavily upon any synthetic type test, and I'll be damned if anyone is going to buy an NV40 to play games in 16 bit mode. ;)

[sonix666]

Damn Dave, just one more day. How late will the review be posted in GMT?

[Kristof]

Gah. All this pipeline talk will be so irrelevent soon.

Naah will be plenty of talk about heatpipes... :lol:

[Stryyder]

Gah. All this pipeline talk will be so irrelevent soon.

Dave stop teasing us.

[DemoCoder]

All this will come at a cost though. Around $500 for the card and $60 or more for the required power supply, possibly even more for larger cases and more fans.

I also predict that the heat and noise of the card will turn off many including computer companies (Dell etc) who will have a big problem getting this thing installed even in their high end computers.


Dell would only think of installing this card in their gamer systems (e.g. Alienware competitor), ditto for Alienware. Gateway and Dell will install mid-range cards in their regular desktops, if that. Remember, this is Dell you're talking about, the company that sells Dimension desktops by the boatload that don't even have AGP slots!

[Waltar]

Gah. All this pipeline talk will be so irrelevent soon.

Dave the nda for the 6800 goes off tonight at midnight right?

[martrox]

Midnight your time or mine......? :wink:

[Tim Murray]

Pipelines are irrelevant because both of the cards are 8x2 very often...

[CapsLock]

Yes, it's one way to test for that. But as we know, nvidia frowns heavily upon any synthetic type test, and I'll be damned if anyone is going to buy an NV40 to play games in 16 bit mode. ;)

Well of course I meant as a means of exploring the design. Actually I'm surprised that low end cards aren't tested more with 16 bit. That Steam survey shows that its still a popular speed enhancing option.

Caps

[PaulS]

Gah. All this pipeline talk will be so irrelevent soon.

Dave the nda for the 6800 goes off tonight at midnight right?

Depends what timezone. It launches in the evening of the 13th.

Pipelines are irrelevant because both of the cards are 8x2 very often...

Damn bandwidth (I also edited it into my original speculation 7 pages ago a little while back).

[LeStoffer]

Pipelines are irrelevant because both of the cards are 8x2 very often...

Oh God no, please not the old 2 x TMU per pipeline or 'but it can apply two textures per clock!' debate again?! :shock: :wink:

[Tim Murray]

No card this generation is Nx2. Bandwidth, dearies, bandwidth.

[Hellbinder]

Pipelines are irrelevant because both of the cards are 8x2 very often...
Riiiight... :lol:

Perhaps you are talking about the Nv40 and Nv40 ultra as "Both cards" ?

[Quitch]

I predict this round will be little different from the last one. nVidia will release a card, it will be proclaimed a winner because it's from nVidia and out-performs last-gen tech. I wouldn't be at all surprised if PS 3.0 is the same to nVidia as PS 2.0 was last time around.

Unless the nv40 is radically different from the nv3x, I really don't see nVidia winning this round as all ATI have to do is more of the same. No, the onus should be on nVidia to prove they have a capable card, and all these optimistic people have obviously already blanked the last 12-18 months.

[DemoCoder]

I love that quote. It really crystalizes some of the attitudes on this board. :)

[John Reynolds]

I love that quote. It really crystalizes some of the attitudes on this board. :)

Narcissist! :twisted:

[digitalwanderer]

The more I'm hearing getting closer to release, the more I think the nV40 is going to be a good card; but I'm more convinced than ever that nVidia is going to over-promise/over-hype it somehow. :(

[PaulS]

The more I'm hearing getting closer to release, the more I think the nV40 is going to be a good card; but I'm more convinced than ever that nVidia is going to over-promise/over-hype it somehow. :(

Yet it occured to me earlier that - if you just went by what's on their main website - NVIDIA barely mention anything at all. It's bizarre. No cryptic flash, no slogans. Nothing. Zip.

[Waltar]

Perhaps you are talking about the Nv40 and Nv40 ultra as "Both cards" ?

You are literally the Chalnoth of ATi fans.

[anaqer]

Perhaps you are talking about the Nv40 and Nv40 ultra as "Both cards" ?

You are literally the Chalnoth of ATi fans.

Or Chalnoth is the Hellbinder of NV fans.
It goes both ways you know. :wink:


On the relative silence from nVidia - yup, somehow reminds me of the hush-hush Prescott launch... in many ways, actually.
Let's hope it ends up better though. :?

[John Reynolds]

Guys, let's stop the calling of each other out. Mmmkay?

[digitalwanderer]

Yet it occured to me earlier that - if you just went by what's on their main website - NVIDIA barely mention anything at all. It's bizarre. No cryptic flash, no slogans. Nothing. Zip.
Yeah, but it's not like they're really gonna start to sell the new cards anytime soon. ;)

Isn't tomorrow's "launch" pretty much a geeky enthusiasts/industry people thing just to let the specs be known?

[ghola]

Gah. All this pipeline talk will be so irrelevent soon.

Well it is easy for people to relate, you know more inches the better, the more pipe the better ;).

WTF...? Wrong forum buddy........

[PaulS]

Yeah, but it's not like they're really gonna start to sell the new cards anytime soon. ;)

Sure, but think back to NV30.

Isn't tomorrow's "launch" pretty much a geeky enthusiasts/industry people thing just to let the specs be known?

Yes, but NVIDIA knows it's all going to come out, and it's the start of the whole thing - and yet no mention on their front page. It's not like it's some secret event than no information will come from.

[digitalwanderer]

Yes, but NVIDIA knows it's all going to come out, and it's the start of the whole thing - and yet no mention on their front page. It's not like it's some secret event than no information will come from.

Just noticed this over at EB, try:

digitallllll..go to nvidia.com..then click on the link to nzone....
when the site is loaded...up to the left u see the eye of dawn looking at u...click on it...and ull knw if tomorow is the launch of a new card or not :lol:

EDITED BITS: It just takes you to here (http://www.nzone.com/object/nzone_geforcelanparty_home.html), a PR release for their LAN party tomorrow.

[PaulS]

Just noticed this over at EB, try:

digitallllll..go to nvidia.com..then click on the link to nzone....
when the site is loaded...up to the left u see the eye of dawn looking at u...click on it...and ull knw if tomorow is the launch of a new card or not :lol:

EDITED BITS: It just takes you to here (http://www.nzone.com/object/nzone_geforcelanparty_home.html), a PR release for their LAN party tomorrow.

Yep, but that's what, four clicks in? Compare that to ATi.com :D

Sorry to go on, it just seems strange to me.

[digitalwanderer]

Yep, but that's what, four clicks in? Compare that to ATi.com :D

No argument, that IS weird and very un-nVidia like IMHO. Good observation.

Sorry to go on, it just seems strange to me.

You kidding? This is the stuff that rampant speculation the night before a big release is made of! ;) :lol:

[thatdude90210]

The hype comes in the form of whispering into people's ears, the people who are likely to leak the info. You know who, the people who were insisting early that it's 16pipes, over 12k in 3dmark03, ps3.0, etc.

[T2k]

I love that quote. It really crystalizes some of the attitudes on this board. :)

I agree.
First of all its author's...

[DemoCoder]

Put up or shut up, find a "gloating" post of mine.

[T2k]

Put up or shut up, find a "gloating" post of mine.

Put up or shut up? Where do you live, in The Bronx?
:roll:

FYI: it's not about gloating, don't try to derail the subject.
The quoted post was made by you after my comment on HB's fun on one of Chalnoth's funny statements - which clearly showed me your intentions to turn around the story 180 degrees... :roll: whatever, I'm not interested in these stupid accusations, neither pro or contra.

PS: BTW it wouldn't take too much time to find one of your hundreds of NV-defending posts...

[John Reynolds]

I agree.
First of all its author's...

Cough:

Guys, let's stop the calling of each other out. Mmmkay?

I don't make posts like the above to practice my typing skills.

[T2k]

I agree.
First of all its author's...

Cough:

Guys, let's stop the calling of each other out. Mmmkay?

I don't make posts like the above to practice my typing skills.

OK, I've just noticed yours. :lol:

[digitalwanderer]

Put up or shut up, find a "gloating" post of mine.

DemoCoder please don't feed the trolls, you're better than that. And T2K, please refrain yourself. If y'all want to fight, may I humbly suggest PMs?

Please don't get this thread closed you two, some of us have some pretty boring Monday nights ahead of us and just might need this distraction. ;)

EDITED BITS: "two" for "too", gads I'm an idiot! :roll:

[Bouncing Zabaglione Bros.]

Yep, but that's what, four clicks in? Compare that to ATi.com :D

Sorry to go on, it just seems strange to me.

Could it be an indication that this will be a paper launch and Nvidia will have nothing to actually ship for months? After all, you wouldn't advertise if you have no product to sell.

[tEd]

Put up or shut up, find a "gloating" post of mine.

DemoCoder please don't feed the trolls, you're better than that. And T2K, please refrain yourself. If y'all want to fight, may I humbly suggest PMs?

Please don't get this thread closed you two, some of us have some pretty boring Monday nights ahead of us and just might need this distraction. ;)

EDITED BITS: "two" for "too", gads I'm an idiot! :roll:

i have a job for you dig if you are bored:
http://www.beyond3d.com/forum/viewtopic.php?t=11445

[digitalwanderer]

i have a job for you dig if you are bored:
http://www.beyond3d.com/forum/viewtopic.php?t=11445
Can't right now, I'm just killing time waiting 'til it's time to give the kids a bath and tuck 'em in...then I'm really not sure if I'm going to try that or not depending on how much time I have before me wife gets off work. (She HATES coming home to me playing with her 'puter. :roll: )

Sorry for the OT, back to the thread. :)

[gb25]

Yep, but that's what, four clicks in? Compare that to ATi.com :D

Sorry to go on, it just seems strange to me.

Could it be an indication that this will be a paper launch and Nvidia will have nothing to actually ship for months? After all, you wouldn't advertise if you have no product to sell.

That didn't stop nvidia from hyping the nv30. Remember the "Are you ready?" campaign they started well before the nv30 was actually released.

[digitalwanderer]

That didn't stop nvidia from hyping the nv30. Remember the "Are you ready?" campaign they started well before the nv30 was actually released.
Hard as it is for me to even contemplate, but mebbe they learned from their past mistake? :shock:

[Dave Baumann]

There is a number of things they have learnt from their past mistakes, certainly from a press perspective.

[digitalwanderer]

There is a number of things they have learnt from their past mistakes, certainly from a press perspective.

Is it any better/easier dealing with their PR people this time around? (Can you talk about that?)

[DemoCoder]

That didn't stop nvidia from hyping the nv30. Remember the "Are you ready?" campaign they started well before the nv30 was actually released.

Apparently ATI learned something from NVidia. *cough* "Latest Secret Weapon"

[Bouncing Zabaglione Bros.]

That didn't stop nvidia from hyping the nv30. Remember the "Are you ready?" campaign they started well before the nv30 was actually released.

That's my point. After last time, many people won't stand for a constantly delayed release date. Anything in the NV40 launch that echoes the NV30 launch problems will be bad news for Nvidia's already tarnished reputation.

[demalion]

The commentary on PS 2.0 being useless for its lifetime strikes me as rather nonsensical. :-? The minimums it brought to the table have been used to offer more advanced effects in quite a few places, and this seems rather remarkably obvious. Just because it hasn't been baseline for engine development doesn't change that.

This makes about as much sense as comparing PS 3.0 to prior shader models instead of looking at how it is distinct from them. It doesn't directly compare to PS 1.4 for being proposed as having a lack of adoption except in how it doesn't make sense as a baseline immediately after release. Its advantages over PS 2.0 (for real-time usage on hardware in the near future) seem to be of lesser significance than for PS 1.4 in its time of release, primarily delivering some ease of implementation and possible efficiency improvements that might or might not be delivered by the hardware implementation. This lesser significance compared to what PS 2.0 offers is largely in relation to tools that facilitate handling the difference more easily and there being less critical "doors to be opened" compared to PS 2.0 than for PS 1.4 compared to PS 1.3, both of which highlight a positive difference from the time of PS 1.4 as well...

As a result of this is, I don't see why in the world developers won't take advantage of those tools and make PS 3.0 a target right away, to go along with PS 2.0 becoming more of a baseline. The discussions on this matter seem IHV-centric and end up ignoring how things already depart greatly from the NV3x and its PS 2.x, the most comparable recent example: the NV40 seems to deliver a full set of useful pixel shading features, so the hardware offering the "beyond PS 2.0" this time does not resemble the NV3x and its "beyond PS 2.0" in this important way, even if the marketing goes overboard with it again.

PS 3.0 is already less of a purely "checklist" feature for developers than the NV30 and NV35 offered. The marketing and technical hardware aspects of the name do not preclude each other, one can be bad or good without dictating the other must be as well.

[hoom]

My guesses: (like it matters :roll: )

Locally it'll be the 15th before I get to see any info...

NV40 will be crowned king by most sites/magazines for the simple fact that it is the first next generation card to be tested.
Unless there are already some r420 cards in the hands reviewers that is...

If the buzzer is actually there for sound cancellation & it actually works & the review boards were sent out with the sticker on it, then only reviewers who bothered to read all the documentation/got in touch with nv/? will have taken the sticker off & discovered that their card is not so noisy after all.

Nobody will mention that the fan design is totally wrong :cry:

There's likely to be much mention of 'two slot cooler' (which I actually don't mind since the only expansion card I have is my 56k & its right down the bottom).
But if ATI can get away with single slot (even cut down from 9800xt :!: ) then thats surely better.

Much 'OMFG'/'well its the price you pay for performance' about dual molex.
Again, if ATI can pull off similar performance with a single molex, it makes nv look bad.

PS2.0 performance will be vastly better than nv3x but given that nv3x had very poor ps2.0 performance, that doesn't necessarily mean too much.
I think nv40 will beat 9800xt on some really shader heavy tests by a surprisingly (or not depending on bias) small margin.

3dmark03 won't be back in a big way (give or take no driver cheats).
Especially when (?) r420 winds up doing 10000+ too.
3dmark04 is surely not far away, hopefully with scoring ps2.0 tests & demonstration ps3.0.
I hope it will 'feel' better than 03.
Certainly 03 won't be getting a ps3.0 test added.

Ps3.0 will be widely used in games fairly early (Farcry already with no hardware out yet :shock: ) but mostly only in the form of ps2.0 compiled to ps3.0 (much as many complex ps1.x can be recompiled as fairly simple/more efficient ps2.0).
NV will play this up big time regardless of whether it actually equates to a performance/image quality gain.

Much will be made of fp blending, fp textures, mrt etc until (I suspect) r420 shows up with everything but dynamic branching (& fp32?), I mean, heck r300 had many of these supposedly revolutionary new things & I don't think ATI have been sitting on their ass, dynamic branching or no.

There will be usage of ATI demos to show that nv40 can actually use things like mrt & fp textures :twisted:

[demalion]

That didn't stop nvidia from hyping the nv30. Remember the "Are you ready?" campaign they started well before the nv30 was actually released.

Apparently ATI learned something from NVidia. *cough* "Latest Secret Weapon"

Actually, I think ATI had the same low key efforts for prior launches...didn't they do something like this for the R300 launch? Well, so far it seems "same old" for ATI for me, though I suppose there could be a change to go with the apparent thing they did learn from nVidia as far as making the launch possibly "sexier"...anything associated with the lips they showed has to be sexier than a chimp :P, even if it doesn't go to the same (lack of) limits as Dawn.

nVidia's "Are you Ready?" was significantly higher volume, and I think more accurately nVidia learned something (good) from nVidia by its absence this time.

[BRiT]

My question is, will Nvidia remove the cheats for the NV30 to make the NV40 look that much better (perhaps without telling anyone this is what they did)? ie: "In this test the NV40 is 700% faster than the NV30" even if in reality it would only be 200% if the NV30 had the cheats.

[PaulS]

My question is, will Nvidia remove the cheats for the NV30 to make the NV40 look that much better (perhaps without telling anyone this is what they did)? ie: "In this test the NV40 is 700% faster than the NV30" even if in reality it would only be 200% if the NV30 had the cheats.

Whilst it's a good tactic in theory, you also a) lose any potential NV3x sales, b) You reveal yourself to be a cheater (as if anyone doubted), and c) you lose even more of your reputation.

[demalion]

1) The NV3x will still live for quite a while. I'm not even sure the NV17/18(?) will be dead! :shock:
2) The rumors of the heat and power issues with the card seem to indicate to me that it won't be necessary, especially with picking and choosing benchmarks and the possibility of new titles and graphics options that simply won't run well on the NV3x (i.e., they might stop adding cheats and, unfortunately, optimizations).

[Geeforcer]

Yep, but that's what, four clicks in? Compare that to ATi.com :D

Sorry to go on, it just seems strange to me.

Could it be an indication that this will be a paper launch and Nvidia will have nothing to actually ship for months? After all, you wouldn't advertise if you have no product to sell.

Question: Is there anything Nvidia can (or cannot) do that you wouldn't interpret as either the tell-tale sigh of an insidious plan or the foreshadowing of their imminent doom?

[digitalwanderer]

Question: Is there anything Nvidia can (or cannot) do that you wouldn't interpret as either the tell-tale sigh of an insidious plan or the foreshadowing of their imminent doom?
Lemme think on that one and get back to you, it's a toughy! :|

[radar1200gs]

Hey, I just want to say one more time that I think NV40 is an ok product from early information. I only think that Nvidia is focusing on the Wrong things right now thats all. I am not suggesting that people in general should not buy it. All the cards comming out are going to be plently fast for gaming over the next year. ITs going to come down to other factors. Like Design elegance, total pakcage, Actual in game IQ stuff like that. I suppose performance in specific games that people are really looking forward to is a big issue as well. Doom-III people will perhaps want the Nv40 etc...

Just about every card nVidia has ever released from the TNT-1 onwards has been bigger, hotter, heavier, more "brute force" (less efficient than the competition) and none of that has ever stopped consumers from purchasing them, as their competitors will sadly attest.

[Bahadir]

All i can say is the Geforce6800U will beat the X800Pro in almost all areas and will take on the performance crown although it will be very hot, noisy and suck alot of juice, hence 450W min PS would be required. :shock: The X800Pro would be alot quieter and would draw less current from the PS.

But, when the X800XT will be launched in a couple of months, the performance gap will be much narrower with the XT leading the various benchmarks in some areas...

That's just a prediction :P

[Joe DeFuria]

Just about every card nVidia has ever released from the TNT-1 onwards has been bigger, hotter, heavier, more "brute force" (less efficient than the competition) and none of that has ever stopped consumers from purchasing them, as their competitors will sadly attest.

That is, until the R300/NV3X generation.

Bigger, hotter, more power hungry is OK...unless your competitor is meeting or beating your product in performance while it's smaller, cooler, and less power hungry....then you're in trouble, as nVidia would attest.

[Nick Spolec]

No argument, that IS weird and very un-nVidia like IMHO. Good observation.



Oh, give them a chance, will ya?.......to show their normal behavior.

If the NV40 isn't ready for mass-market for another 2 months or more, I can guarentee you Nvidia will flood the web with propaganda.

[digitalwanderer]

No argument, that IS weird and very un-nVidia like IMHO. Good observation.



Oh, give them a chance, will ya?.......to show their normal behavior.

If the NV40 isn't ready for mass-market for another 2 months or more, I can guarentee you Nvidia will flood the web with propaganda.
Hey, I'm counting on that! :lol:

I just don't see the point in antagonizing anyone un-necessarily yet by reminding 'em, everyone seems so darn sensitive lately. :roll:

;)

(Sorry, sorry...prelaunch jitters. I'll calm down sometime later tomorrow.)

[reever]

Does the trend continue for scalable VPUs from ATI :?: :D

How much more do you need? They already have 4gpu solutions, granted that's on 2 pcb's, and then we have SGI systems with 32 R300's in it. Isn't the r300 capable of 256+ gpu's anyways?

[DemoCoder]

Looking at the leaked (possibly bogus) prices for the NV40 ($299 for NV40, $399 for NV40U), I have to wonder if NVidia is saving something for the $499 pricepoint. ATI's X800XT "leaked" prices (again, possibly bogus) is $499. I'm thinking in June they will announce something new, perhaps a PCI-E version of the NV40 @ > 450Mhz, with 1.2Ghz or higher RAM.

The is possibly the most exciting 3D launch we've had in a while. Both vendors are launching nearly simultaneously, launching extremely interesting parts, and they are racing neck and neck to upstage the other. Hell, if the $399 NV40U rumor is true, it sounds like Nvidia is selling this one at a loss, but whom am I to argue if companies want to slash prices and offer uber deals to get my patronage? :)

[mikechai]

That prices couldn't be true. Just too good to be true.
________
EXTREME VAPORIZER REVIEW (http://vaporizer.org/reviews/extreme)

[Psikotiko]

little OT.....

http://www.hardocp.com/


ATI Radeon R420 Roadmap:
Secret double agents in Zimbabwe have turned over classified ATI documents. The below specifications outline a whole host of information about the upcoming flagship VPU from ATI.

First we have the names that will be attached to the VPUs that will scale across three levels of performance with the same VPU. The "X800XT" and the "X880XT" being the "same" card but with a different interface, with the latter being a PCI-Express X16 card. You will see that the PCI-Express core is labeled "R423" as it is in fact a native PCI-Express part and not the same CPU as the R420 with an adaptive bridge to allow the VPU to work on both interfaces. (It is yet to be seen if there is a performance impact from using a bridged VPU/GPU.) Notice as the clocks and memory speeds scale across the product line, so does the pixel pipelines that are turned on or off to further define performance levels.

Last but certainly not least are the projected release dates and MSRP prices in US dollars.

This information did not come from ATI but a source that has proven to be correct in the past. I would categorize this information as "solid but unconfirmed".


http://www.hardocp.com/images/news/1081819095xDxNGnqwG1_1_1_l.gif

If it was already posted........sorry.

[radar1200gs]

Just about every card nVidia has ever released from the TNT-1 onwards has been bigger, hotter, heavier, more "brute force" (less efficient than the competition) and none of that has ever stopped consumers from purchasing them, as their competitors will sadly attest.

That is, until the R300/NV3X generation.

Bigger, hotter, more power hungry is OK...unless your competitor is meeting or beating your product in performance while it's smaller, cooler, and less power hungry....then you're in trouble, as nVidia would attest.
The problems for NV30 were time to market and quantities available (not helped by out of control PR for an unvailable product).

[Geeforcer]

I personally tend to scratch my head at any purported "official IHV roadmap" that denotes the amount of transistors as two numbers with a dash.

Edit: What you say you say?

[Joe DeFuria]

little OT.....

http://www.hardocp.com/



Looks like their source is Dark Crow / gzeasy, who posted an indentical chart yesterday or the day before. Dave has already commented that "certain aspects" of that are wrong. Don't know what though. (I personally have a hard time believing in 600 Mhz clock speeds for the XTs...)

[Joe DeFuria]

The problems for NV30 were time to market and quantities available (not helped by out of control PR for an unvailable product).

...becuase they couldn't make it bigger, hotter and consume enough power to actually beat a R300 that was launched months earlier.

[DemoCoder]

Just about two years ago, the R300 Pro was released, with 19.2Gb/s of bandwidth and 2.6Gpix/s (or op/s) peak fillrate. Now in June, we're looking at (if it is believed), a card with 38.4Gb/s bandwidth, and 9.6Gpix/s (or op/s) 2x the bandwidth, and 3.5x the fillrate in 2 years.

Exciting times we live in. CPU evolution (except for what I read about CELL) is practically boring. Where's my 6Ghz pentium?

[radar1200gs]

The problems for NV30 were time to market and quantities available (not helped by out of control PR for an unvailable product).

...becuase they couldn't make it bigger, hotter and consume enough power to actually beat a R300 that was launched months earlier.
If it had launched on time, the R300 wouldn't have mattered. Of course TSMC stuffed that up nicely. They won't be getting the chance with NV40.

[demalion]

...

The is possibly the most exciting 3D launch we've had in a while. Both vendors are launching nearly simultaneously, launching extremely interesting parts, and they are racing neck and neck to upstage the other.
...

This seems a pretty safe prediction, I agree. ;)

[Joe DeFuria]

Just about two years ago, the R300 Pro was released, with 19.2Gb/s of bandwidth and 2.6Gpix/s (or op/s) peak fillrate. Now in June, we're looking at (if it is believed), a card with 38.4Gb/s bandwidth, and 9.6Gpix/s (or op/s) 2x the bandwidth, and 3.5x the fillrate in 2 years.

Exciting times we live in. CPU evolution (except for what I read about CELL) is practically boring. Where's my 6Ghz pentium?

My thoughts exactly. It seems that Intel / AMD have been stuck in the 3 Gz (or 3000+) rut for what, a year?

[Joe DeFuria]

If it had launched on time, the R300 wouldn't have mattered. Of course TSMC stuffed that up nicely.

No, if NV30 launched on time, then R300 wouldv'e launched 6 months earlier, of course. (It's easy to play the "but if" game, eh?) And TSMC doesn't seem to be stuffing up anything from ATI....

[DemoCoder]

AMD really annoys me. If you look at their CPU's per clock performance, they could OWN Intel if they just released a real 3Ghz chip instead of a 2.4Ghz one. I'd like to see how a 3.4Ghz FX53 compares to a 3.4 P4EE :)
Anyway, for a while, CPU's were rapidly advancing, then, as you said, they seem to be stuck in 2-3Ghz land.

[reever]

Exciting times we live in. CPU evolution (except for what I read about CELL) is practically boring. Where's my 6Ghz pentium?

Wait until leakage becomes a big problem with GPU's once we get smaller

[Nick Spolec]

If it had launched on time, the R300 wouldn't have mattered. Of course TSMC stuffed that up nicely.

If the NV30 launched on time, the R300 would have still been faster, quieter, better.

[Ardrid]

AMD really annoys me. If you look at their CPU's per clock performance, they could OWN Intel if they just released a real 3Ghz chip instead of a 2.4Ghz one. I'd like to see how a 3.4Ghz FX53 compares to a 3.4 P4EE :)
Anyway, for a while, CPU's were rapidly advancing, then, as you said, they seem to be stuck in 2-3Ghz land.

Easier said than done. Because AMD has such a high IPC, they can't simply ratchet the clock speeds up like Intel can. They have to go slow. In any event, why go faster when you're already doing more with less?

[DemoCoder]

Yeah yeah, but I can dream, can't I? Anyway, both Intel and AMD seem stuck in a rutt. Maybe IBM/Toshiba/Sony et al can shake them up.

[Joe DeFuria]

AMD really annoys me. If you look at their CPU's per clock performance, they could OWN Intel if they just released a real 3Ghz chip instead of a 2.4Ghz one. I'd like to see how a 3.4Ghz FX53 compares to a 3.4 P4EE :)
Anyway, for a while, CPU's were rapidly advancing, then, as you said, they seem to be stuck in 2-3Ghz land.

I still get a chuckle when I think back to my days of just getting interested in 3D acceleration. The days of the Gaming Glint (Creative Labs 3D Blaster) and nVidia NV-1 (Diamond Edge 3D). Something called AGP was supposed to mean the end of on-board graphics memory, and MS Talisman was going to be more or less "required" for high performance graphics, because, (paraphrasing), "memory bandwidth just isn't progressing. We won't be seeing the ceiling of 10 GB /sec needed for high quality rendering any time soon."

[T2k]

AMD really annoys me. If you look at their CPU's per clock performance, they could OWN Intel if they just released a real 3Ghz chip instead of a 2.4Ghz one. I'd like to see how a 3.4Ghz FX53 compares to a 3.4 P4EE :)


DC: I have a very direct line to AMD, let me just put this way what I know.
Are you suggesting they're holding back their faster/stronger products because no need for rush or what?

If so, you are very very far from the truth.

[Ardrid]

Yeah, you can certainly dream :) But I definitely agree, clock speeds have slowed down tremendously. AMD's still got a little left in them before they run into the problems Intel's experiencing, but both companies are going to hit a wall, sooner rather than later. Multiple cores and high IPC are definitely the way of the future.

[Joe DeFuria]

Exciting times we live in. CPU evolution (except for what I read about CELL) is practically boring. Where's my 6Ghz pentium?

Wait until leakage becomes a big problem with GPU's once we get smaller

My daughter had a leakage problem...but changing diaper brands fixed that right up...

[T2k]

Multiple cores and high IPC are definitely the way of the future.

And future is pretty close, believe me... ;)

[T2k]

BTW this 'clock = performance' stupidity gained its popularity by Intel. They did start this clock race several years ago - and now?
The 'clockspeed' company forced to choose the AMD-way to differ its products... LOL1 I think it's really funny... :lol:

Edit: typos

[digitalwanderer]

If it had launched on time, the R300 wouldn't have mattered.
If it had launched on time the R300 would have still came out and spanked it in speed, image quality, and support. 8)

But at least it would have been on time.

Gads, I sure hope that SE info is wrong. 128bit memory bit-width? Eeew!

[T2k]

If it had launched on time, the R300 wouldn't have mattered.
If it had launched on time the R300 would have still came out and spanked it in speed, image quality, and support. 8)

But at least it would have been on time.

Gads, I sure hope that SE info is wrong. 128bit memory bit-width? Eeew!

I don't see what's the problem there: SE is targeted for middle-budget wallets and they don't play where I or you do (1600 or 1900 w/ 4-6xAA+8-16xQAF).
With 450MHz chip/400MHz memory clock and next-generation pipeline architecture and everythin g else this chip should be fine for 1024 and 4xAA...

If you want more, pay more - it's capitalism, Dig.

[Ardrid]

Multiple cores and high IPC are definitely the way of the future.

And future is pretty close, believe me... ;)

Yes, indeed it is. 2005 can't come any quicker :)

[Joe DeFuria]

I don't see what's the problem there: SE is targeted for middle-budget wallets and they don't play where I or you do (1600 or 1900 w/ 4-6xAA+8-16xQAF).
With 450MHz chip/400MHz memory clock and next-generation pipeline architecture and everythin g else this chip should be fine for 1024 and 4xAA...

If you want more, pay more - it's capitalism, Dig.

The problem would be if that chart's right... it's $299 MSRP for a card that may not perform as well as a card you can get today for $200. (9800).

$299 is not a "middle budget price."

[digitalwanderer]

I don't see what's the problem there: SE is targeted for middle-budget wallets and they don't play where I or you do (1600 or 1900 w/ 4-6xAA+8-16xQAF).
With 450MHz chip/400MHz memory clock and next-generation pipeline architecture and everythin g else this chip should be fine for 1024 and 4xAA...

If you want more, pay more - it's capitalism, Dig.
Hrrrm. If it happens that way, I'll argue the point and we can just agree to disagree for now...fair? :|

[g__day]

The wait is almost over, its 1pm 13th in Sydney, I awoke to a thread of 3 pages and given family distractions of school holidays it got to 11 pages before I got thru it all!

I wonder what benchmarks will be pushed most heavily by NVidia? I tend to think FM and Aquamark 3 featuring predominantly amongst others. I am less interested in knowing 1,000 FPS in Quake 3 is now possible...

[hoom]

I'd like to see how a 3.4Ghz FX53 compares to a 3.4 P4EE
Me too!
But the AMD needs to be running a 500mhz r300 vs a stock nv30/p4 too :twisted:

If x800xt is 600mhz with 16 pipes (I am highly sceptical) then even if nv manages a 500mhz 6800, they gonners.
2 competing 16 pipe chips at 500mhz would be a very interesting fight though 8)

edit: 3:15pm 13th in nz

[ninelven]

If you want more, pay more - it's capitalism, Dig.

And capitalism works because of competition :).

[DemoCoder]

If x800xt is 600mhz with 16 pipes (I am highly sceptical) then even if nv manages a 500mhz 6800, they gonners.
2 competing 16 pipe chips at 500mhz would be a very interesting fight though 8)


Not neccessarily, because of bandwidth restrictions, you will need heavily ALU shader bound games to show off such a big difference (500Mhz vs 600Mhz) However, I think the reverse of your argument is true: NV40 needs more Mhz to beat the R420. Due to the complexity of NV's pipeline, I expect it will run per-clock slower, even allowing for shaders to be well optimized by the drivers. So I think NV would need a 650Mhz NV40 to beat a 600Mhz R420.

(except on Doom3 and any game with heavy use of shadows)

[T2k]

I don't see what's the problem there: SE is targeted for middle-budget wallets and they don't play where I or you do (1600 or 1900 w/ 4-6xAA+8-16xQAF).
With 450MHz chip/400MHz memory clock and next-generation pipeline architecture and everythin g else this chip should be fine for 1024 and 4xAA...

If you want more, pay more - it's capitalism, Dig.

The problem would be if that chart's right... it's $299 MSRP for a card that may not perform as well as a card you can get today for $200. (9800).

$299 is not a "middle budget price."

Yay, Joe, you know very well: a) these are some leaks or plain rumours b) MSRP doesn't mean anything - that's (middle-budget) probably means ~$200 IMO.

[Doomtrooper]

Not neccessarily, because of bandwidth restrictions, you will need heavily ALU shader bound games to show off such a big difference (500Mhz vs 600Mhz) However, I think the reverse of your argument is true: NV40 needs more Mhz to beat the R420. Due to the complexity of NV's pipeline, I expect it will run per-clock slower, even allowing for shaders to be well optimized by the drivers. So I think NV would need a 650Mhz NV40 to beat a 600Mhz R420.

(except on Doom3 and any game with heavy use of shadows)

Doom 3 is capped @60 fps :wink:

[ninelven]

I think I recall someone hinting that nvidia kinda ditched the deep pipeline and went for something with a bit more ppc than the nv3x. However its still probably slower than ati's per clock.

[Pete]

The problem would be if that chart's right... it's $299 MSRP for a card that may not perform as well as a card you can get today for $200. (9800).

$299 is not a "middle budget price."
[H]'s chart shows the $300 X800 with 800MHz memory, which means it's basically got more bandwidth than the 9800XT right off the bat. Along with core tweaks and a higher core clock, it should do very well, no? Pity it's only supposed to show up in June.

[T2k]

Not neccessarily, because of bandwidth restrictions, you will need heavily ALU shader bound games to show off such a big difference (500Mhz vs 600Mhz) However, I think the reverse of your argument is true: NV40 needs more Mhz to beat the R420. Due to the complexity of NV's pipeline, I expect it will run per-clock slower, even allowing for shaders to be well optimized by the drivers. So I think NV would need a 650Mhz NV40 to beat a 600Mhz R420.

(except on Doom3 and any game with heavy use of shadows)

Doom 3 is capped @60 fps :wink:

Still: if you have, say, 8xfps average or 6x-7xfps average, the latter will hit 5x as lowest fps.

[Nite_Hawk]

If it had launched on time, the R300 wouldn't have mattered.
If it had launched on time the R300 would have still came out and spanked it in speed, image quality, and support. 8)

But at least it would have been on time.

Gads, I sure hope that SE info is wrong. 128bit memory bit-width? Eeew!

Not to mention that its only clocked at 400MHz DDR... That's just under 2/3rds the memoy bandwidth of a 9700pro!

Nite_Hawk

[Nite_Hawk]

The problem would be if that chart's right... it's $299 MSRP for a card that may not perform as well as a card you can get today for $200. (9800).

$299 is not a "middle budget price."
[H]'s chart shows the $300 X800 with 800MHz memory, which means it's basically got more bandwidth than the 9800XT right off the bat. Along with core tweaks and a higher core clock, it should do very well, no? Pity it's only supposed to show up in June.

Er, I'm assuming they mean 400MHz DDR = "800" MHz right? on a 128bit bus that isn't going to be anywhere close to a 9800XT...

Nite_Hawk

[Joe DeFuria]

The problem would be if that chart's right... it's $299 MSRP for a card that may not perform as well as a card you can get today for $200. (9800).

$299 is not a "middle budget price."
[H]'s chart shows the $300 X800 with 800MHz memory, which means it's basically got more bandwidth than the 9800XT right off the bat.

No, H's chart shows the $300 X800 SE with 400 Mhz, 128 bit memory, which means it would only have about 60% of the memory bandwidth of the 9800 XT, which has 375 Mhz, 256 bit wide ram. As shown on the chart, the X800SE would only have 75% of the bandwidth of the 9700 non pro..

I really hope that chart is wrong...either the price, or the memory bus.

[T2k]

The problem would be if that chart's right... it's $299 MSRP for a card that may not perform as well as a card you can get today for $200. (9800).

$299 is not a "middle budget price."
[H]'s chart shows the $300 X800 with 800MHz memory, which means it's basically got more bandwidth than the 9800XT right off the bat.

No, H's chart shows the $3800 X800 SE with 400 Mhz, 128 bit memory,

OFF
Wow, that would be REALLY expensive, I agree. :lol :D :D

PS: OK, it's late enough for this, isn't it?

Edit: I know, it's lame but almost midnight, my wife is sleeping... should I watch TV? :)

[Joe DeFuria]

No, H's chart shows the $3800 X800 SE with 400 Mhz, 128 bit memory,

OFF
Wow, that would REALLY expensive, I agree. :lol :D :D

PS: OK, it's late enough for this, isn't it?

Doh! (I'll go back and fix...)

[digitalwanderer]

So, we're all planning on waiting up for the next 36-48 hours or so chatting about what we're gonna find out?

Cool, I'm in. 8)

[gunblade]

The problem would be if that chart's right... it's $299 MSRP for a card that may not perform as well as a card you can get today for $200. (9800).

$299 is not a "middle budget price."
[H]'s chart shows the $300 X800 with 800MHz memory, which means it's basically got more bandwidth than the 9800XT right off the bat.

No, H's chart shows the $3800 X800 SE with 400 Mhz, 128 bit memory, which means it would only have about 60% of the memory bandwidth of the 9800 XT, which has 375 Mhz, 256 bit wide ram. As shown on the chart, the X800SE would only have 75% of the bandwidth of the 9700 non pro..

I really hope that chart is wrong...either the price, or the memory bus.

Besides, there is almost no way a DDR memory to scale to 800Mhz, so it should most certainly indicate 400Mhz.

The SE also is listed as 8 pipe only. All these seem to indicate the R420 SE will only beat the radeon 9800 in ultra-shader-heavy game. Everything else will be totally owned by radeon 9800.

[Ardrid]

And the clock ticks...

[T2k]

So, we're all planning on waiting up for the next 36-48 hours or so chatting about what we're gonna find out?

Cool, I'm in. 8)

Sorry, I'll pass this. It'd be nice but I have to work tomorrow - I already took off today (Easter, church, etc ;))

[T2k]

And the clock ticks...

And nothing happens... :P

[nelg]

Lets waste time by guessing who will have there review up first. My guess is Anandtech. It is easy to be first when you don’t let details stand in the way. :lol:

[Geo]

Lets waste time by guessing who will have there review up first. My guess is Anandtech. It is easy to be first when you don’t let details stand in the way. :lol:

Nah, it'll be some German or French site jumping the gun. Or Czech, or something like that.

[T2k]

Lets waste time by guessing who will have there review up first. My guess is Anandtech. It is easy to be first when you don’t let details stand in the way. :lol:

Nah, it'll be some German or French site jumping the gun. Or Czech, or something like that.

Obviously some Far-Eastern or European site.
Think about time zones... ;)

[mikechai]

Lets waste time by guessing who will have there review up first. My guess is Anandtech. It is easy to be first when you don???t let details stand in the way. :lol:

Anandtech's reviews were always later than others...
________
BABY MAC (http://babimac.com/)

[digitalwanderer]

For some reason I'm thinking it'll be Tom's that jumps the guns, but that's just a gut guess.

[Ardrid]

Whoever it is, make sure it gets posted here ASAP :D

[JD]

It's most likely that both ihvs will have some functionality slow. I think ati has problems with flow control in shaders. It's not like we can say all the features are 2x faster across the board. It would be nice if everything worked out but that's not realistic. For example, my gf2 has a hw bug that only surfaced when I used cubemaps with filtering. I hope ihvs would disclose this info to devs before they're done pulling their hair out :)

So to me it's just another board with potential headaches. Not looking forward to that that much. Yeah, call me pessimist :)

[T2k]

For some reason I'm thinking it'll be Tom's that jumps the guns, but that's just a gut guess.

:twisted: Let me gues what makes you think that - he has better connections for this IHV, right? :twisted:

[digitalwanderer]

:twisted: Let me gues what makes you think that - he has better connections for this IHV, right? :twisted:
Kind of, I figure him and Anand's actually both got about an equal chance of feeling they could post it up an hour or two early "due to confusion over the time difference" without too much backlash from nVidia since they're on very good working terms with 'em.

But I give the nod to Tom's just 'cause I think he's got a bit more nuggers for this kind of thing, although Anand's has a better chance of truly releasing it accidently early thru a screw-up. :lol:

[Farid]

Nah, it'll be some German or French site jumping the gun. Or Czech, or something like that.

For the french sites you can go watch http://www.phoenixjp.net/news/fr/ it contains RSS from all (at least the more importants) french hardware sites
There's also an US version http://www.phoenixjp.net/news/us/

[digitalwanderer]

There's also an US version http://www.phoenixjp.net/news/us/

OMG! I love the link, thank you very much for it. (Sorry, it's new to me and it's a good new one!)

[Geo]

:twisted: Let me gues what makes you think that - he has better connections for this IHV, right? :twisted:
Kind of, I figure him and Anand's actually both got about an equal chance of feeling they could post it up an hour or two early "due to confusion over the time difference" without too much backlash from nVidia since they're on very good working terms with 'em.

But I give the nod to Tom's just 'cause I think he's got a bit more nuggers for this kind of thing, although Anand's has a better chance of truly releasing it accidently early thru a screw-up. :lol:

Make it http://www.tomshardware.de and we'll split the difference. . .

[Farid]

There's also an US version http://www.phoenixjp.net/news/us/

OMG! I love the link, thank you very much for it. (Sorry, it's new to me and it's a good new one!)

:wink: i'm glad that link is somekind usefull to you.

I spend everyday a few time watching the news of all the majors sites on one look , it help with the classical "i've read it all before" when you go to one site to another.

[T2k]

G'nite everybody. 8)

[Pete]

[H]'s chart shows the $300 X800 with 800MHz memory, which means it's basically got more bandwidth than the 9800XT right off the bat. Along with core tweaks and a higher core clock, it should do very well, no? Pity it's only supposed to show up in June.

Er, I'm assuming they mean 400MHz DDR = "800" MHz right? on a 128bit bus that isn't going to be anywhere close to a 9800XT...
OK, no more internet bickering for me until the NDA lifts. I, uh, was high on Sudafed at the time.

But a 128-bit bus for an R420 seems kind of stupid, no? Is this another 9500-type stopgap solution? Why would that even be necessary with R360 still around? It seems more likely that it'll be 256-bit across the line, no?

[webmedic]

This time it was more specific statement ("Gross Margin will
be approaching historical numbers"). So my guess is that
they have produced enough NV40 in production to estimate yield properly.

this is assuming they can fool as many people into bying one as they did the nv3x cards. There is no gross margin on parts you cant sell.

[webmedic]

Found this

Also... from Warp2search.net discussion...
Quote:
A certain new card we have caused one of our PCs (an Athlon XP) to auto shutdown after 20/25 minutes gaming. This was due to heat build up in the case, the otherwise OK CPU was overheating, it was 10c+ hotter with this card than with a 5950.. The case had 80mm fans front and back.
If you have an already hot/overclocked computer you may not get on with this certain card (that i cant yet mention of course).
What with this and the need for a warp core to power it i forsee trouble ahead for many purchasers.

:shock: :shock: :shock: :lol: :lol: :lol:

[Evildeus]

My predictions:
- We will see the previews tomorrow
- B3D preview will be 25+ pages 8)
- Nv will use some water demos
- that's all folks :!:

[webmedic]

Perhaps you are talking about the Nv40 and Nv40 ultra as "Both cards" ?

You are literally the Chalnoth of ATi fans.

more like hellbinder is the anti-chalnoth.

The only person that understands chalnoth is himself on the other hand hellbinder is very understandable.

[LeStoffer]

While waiting, I predict that John Carmack will be happy!


The future is in floating point framebuffers. One of the most noticeable
thing this will get you without fundamental algorithm changes is the ability
to use a correct display gamma ramp without destroying the dark color
precision.


Floating point framebuffers and complex fragment shaders will also allow much better volumetric effects, like volumetric illumination of fogged areas with shadows and additive/subtractive eddy currents.

(Carmacks plan update on DoomIII, game design and NV30 vs R300 last year)

[Evildeus]

http://www.hardocp.com/images/news/1081819095xDxNGnqwG1_1_2_l.gif
http://www.hardocp.com/