Berek
Regular
You can always infer based on industry sources and other information if you do the appropriate research. Instead of asking the question, you should try and figure it out instead of bashing me without understanding.
NVIDIA GT200 Rumours & Speculation Thread
- Thread starter Arun
- Start date
- Status
You can always infer based on industry sources and other information if you do the appropriate research. Instead of asking the question, you should try and figure it out instead of bashing me without understanding.
Interesting! So just by naming it.. say the R770 instead it can now be INFINITELY FASTER THAN THE RV670! SCORE!
The unfirm info that i have.
Or, believe me, i've done all the research and stuff.
Doesn't look like you did actually.
Are you stupid?
aaronspink
Veteran
Well then you're gonna love this
ok, thats just wrong. Someone seriously needs to take Nvidia's PR out behind the woodshed.
And using that broken PR to brag about the craptastic 9800 GTX? Lets see, we'll charge you ~$350 for something you can already buy for <$200 (MSI 8800 GTS 512 OC at newegg)? Like totally awesome dude!
Aaron Spink
speaking for myself.
But there is a lot of FUD flying around on both sides AFAICT.
I agree, I'm not sure what warranted that comment from Berek?
I think his point is more subtle: codenames can be FUD to confuse everyone. We shouldn't exclude possibilities just because the codename seems to imply otherwise; although I agree that in this case, it doesn't seem very likely be a monolithic single-chip capable of competing with a single-chip GT200. We'll see what happens...
MrWizard6600
Newcomer
bah, finally, Windows Live IDd B3D as spam.
Anyways, I've beein keepin my eye on the RV770 pretty intently over the past couple weeks and it looks like its going to have 480SPs clocked at somewhere around 950MHz (on the highend obviously), 32TMUs, 16ROPs (again, clocked at almost a GHz remember!), a 256bit bus running at an oh so sweet 4GHz effective, thanks to the improvements brought to us by GDDR5. No word on the increase latencies though :S. 512mb is likly to be the SKU but I doubt ATI will hold board partners back from 1024MB. Samsungs GDDR5 fab process might though...
Anyways, the GT200 (Faud was the first to use this name and it just kinda caught on... I don't like it, because A. its faud B. its a new naming scheme), ...where was I, right, the GT200 rumors are still all over the place. I've got links to people claiming a 1024bit memory bus for christs sake (dont hold breath). But yeah, it looks like we're looking at at least a doubling of the SP count. BTW, if the best case scenario for memory holds true (512bit bus, GDDR5), we're looking at upward of 300GB/s of memory throughput on this card! Dude, snap your fingers, your hard drives copied. 300GB/s, thats unfathomable. Thats like a procs L2 speed for shits sake...
What's really baking my noodle is the timeframe. Multiple sources are claiming a G92 shrink, but multiple sources are also claiming a Q3 release for the GT200. It just doesn't fit, Nvidias recently announced Nvision2008 doesn't really clarify things. You can bet it will contain some new product, but which one? Why release a product just to yank it only a few months later? That doesn't even begin to pay the bills Nvidia execs will owe the R&D team for developing this short-lived product. I would also argue the haste of a 9 series start is Nvidia clearing the decks. I think the GT200 might be called something other than Geforce (maybe even Nvision? the "Nvision GTX")
anyways, the best I can put together is that the GT200 will be based on good old GDDR3, it will have a 256bit bus , and its going to have ~200 (192? 256?) SPs. Fuds claiming its going to be built on TSMCs 65nm process, but then, all fuds had to say about the "GT200" has been negative. I donno, its too early to tell. I'm more excited for the RV770 at this point, simply because we have some really confirmed numbers to go on.
Anyways, I've beein keepin my eye on the RV770 pretty intently over the past couple weeks and it looks like its going to have 480SPs clocked at somewhere around 950MHz (on the highend obviously), 32TMUs, 16ROPs (again, clocked at almost a GHz remember!), a 256bit bus running at an oh so sweet 4GHz effective, thanks to the improvements brought to us by GDDR5. No word on the increase latencies though :S. 512mb is likly to be the SKU but I doubt ATI will hold board partners back from 1024MB. Samsungs GDDR5 fab process might though...
Anyways, the GT200 (Faud was the first to use this name and it just kinda caught on... I don't like it, because A. its faud B. its a new naming scheme), ...where was I, right, the GT200 rumors are still all over the place. I've got links to people claiming a 1024bit memory bus for christs sake (dont hold breath). But yeah, it looks like we're looking at at least a doubling of the SP count. BTW, if the best case scenario for memory holds true (512bit bus, GDDR5), we're looking at upward of 300GB/s of memory throughput on this card! Dude, snap your fingers, your hard drives copied. 300GB/s, thats unfathomable. Thats like a procs L2 speed for shits sake...
What's really baking my noodle is the timeframe. Multiple sources are claiming a G92 shrink, but multiple sources are also claiming a Q3 release for the GT200. It just doesn't fit, Nvidias recently announced Nvision2008 doesn't really clarify things. You can bet it will contain some new product, but which one? Why release a product just to yank it only a few months later? That doesn't even begin to pay the bills Nvidia execs will owe the R&D team for developing this short-lived product. I would also argue the haste of a 9 series start is Nvidia clearing the decks. I think the GT200 might be called something other than Geforce (maybe even Nvision? the "Nvision GTX")
anyways, the best I can put together is that the GT200 will be based on good old GDDR3, it will have a 256bit bus , and its going to have ~200 (192? 256?) SPs. Fuds claiming its going to be built on TSMCs 65nm process, but then, all fuds had to say about the "GT200" has been negative. I donno, its too early to tell. I'm more excited for the RV770 at this point, simply because we have some really confirmed numbers to go on.
MrWizard6600
Newcomer
what happened to my post?
Sorry, Windows Live might think B3D is spam, but it looks like B3D also thought you were spam...
We got pretty aggressive moderation to fight spam here so posts with certain characteristics, including links but not only that (can't say what else, sorry), need moderator approval before they are visible to others unless the person has 10+ posts and X+ days on the board. Anyhow, welcome to the forum!
As for GT200, let me confirm that it is indeed the correct codename. Well, not exactly; there are at least 3 (!) codenames which are correct depending on the context. So yeah, confusion ahoy.Regarding G92b/etc. - the way I look at it is that NVIDIA would like to have 55nm shrinks of both G92 and G94 (at the very least, maybe more) for the Back-to-School OEM cycle. Those are OEM deals that'll last for quite a while, so being stuck on 65nm there because they couldn't get 55nm out in time would definitely affect their margins. This is a completely separate issue from GT200 and you can't guess anything about one from the other.
As for NVISION, this is a public event which was planned for some time, so I think the only thing you can conclude from that is NVIDIA really really wants to have GT200 announced either by then or exactly then. Thus, the date must also have been selected conservatively to make that scenario extremely likely; look at it as a likely upper bound. It could be later, but it's much more likely to be earlier. The question is how much; it could be a day, it could be a month, it could even be more for all we know.
Baseless speculation, or something else ? 
I hope they're talking about the ARM-based APX 2500/future derivatives, and not some undercover x86 "thingy".
I hope they're talking about the ARM-based APX 2500/future derivatives, and not some undercover x86 "thingy".
I found that entire article to be really strange. First they claim Nvidia is slowing the pace of innovation yet they believe Nvidia is going to stay just ahead of AMD. So they don't think AMD is capable of and responsible for innovating and pushing the envelope as well?
Quite obvious, me neither. Do you think they are capable at this moment?
I have somewhere a quite old presentation from Jon Peddie research about professional markets and they predicted then that one shouldn't be surprised if NVIDIA will end up competing with Intel at a point in the future and vice versa.
I'm still waiting to see how deep any possible penetration could be in the future. For the time being - and despite the buttload of money Intel is throwing around - I'm still not convinced that Larabee could be a serious threat outside of professional markets for anything graphics. And I'm times less convinced that NVIDIA intends to enter the CPU market in the sense the above article could imply. If someone would tell me that they intend to develop/build something like an entire SoC for small form factor devices that at least I could buy.
I'm still waiting to see how deep any possible penetration could be in the future. For the time being - and despite the buttload of money Intel is throwing around - I'm still not convinced that Larabee could be a serious threat outside of professional markets for anything graphics. And I'm times less convinced that NVIDIA intends to enter the CPU market in the sense the above article could imply. If someone would tell me that they intend to develop/build something like an entire SoC for small form factor devices that at least I could buy.
That's true.
That's why i always use 'i believe' and 'AFAIK' =)
Well it looks like Fuad WAS right about GT200 codename after all =)
But i tend to look a little bit further than codenames and i believe i've presented my reasoning in the post above.
G100, GT200 and D10U?..
Looks like GT200 along with a handful of other GT2xxs' appeared in one of NV's not-so-public tools this week =)
http://www.fudzilla.com/index.php?option=com_content&task=view&id=6702&Itemid=1
Fuds taking a stab in the dark with some numbers regarding memory of this GPU codenamed D10U.
Atleast somethings in the works.
edit - http://en.expreview.com/2008/04/08/geforce-9800gx2-will-be-eoled-in-three-months/
Is it me or this G92b seem like the GT200. Hence the use of the 9900 series moniker. Maybe its more than a mere "die shrink" to 55nm to go toe to toe with the RV770.
Fuds taking a stab in the dark with some numbers regarding memory of this GPU codenamed D10U.
1024MB and 896MB? 512(GTX)/448(GTS)Bit memory interface? Doesnt seem to go in hand to hand with the 55nm "die shrinked" G92b. Q2 release kind of kills current products (GX2) too does it not? or is that why the GX2 has dropped ~$100 dollars in price already and its only been recently launched.
Atleast somethings in the works.
edit - http://en.expreview.com/2008/04/08/geforce-9800gx2-will-be-eoled-in-three-months/
Is it me or this G92b seem like the GT200. Hence the use of the 9900 series moniker. Maybe its more than a mere "die shrink" to 55nm to go toe to toe with the RV770.
Update: According to sources, 9900GX2 do not exist. We are sorry for the mistakes.
I'm not sure what's going on but if this is a FUD campaign they're doing a really good job of it. G92b, G100, GT200, D10-U....I mean come on already ! If G92b is more than just a die shrunk G92 then Nvidia really has gone off the deep end with this naming thing. Even the chinese don't know what's going on and that says something.
Yes. 8 ROPs, 512-bit bus. 1 faulty ROP will give you 896 MB of VRAM on a 448-bit bus for a lower SKU. Looks alright.
I think it's safe to assume that G92b has nothing to do with D10U series.
- Status
