NVIDIA GF100 & Friends speculation

[GZ007]

BTW: How could they build smaller chips with this kind of per/mm^2? They would slower and bigger than a G92b...

Its a whole new architecture, everything can happen(on papers things sound cool). We also dont know how much space is wasted on things like ECC and greatly increased DP.
I remember the ti4800 to be faster sometimes than the super perfect gpu called fx5800.

 

[Alexko]

The numbers are very weird for Crysis WarHead. The GTX 470 scores show that it has almost no performance advantage over the GTX 285, with the same settings in-game (although different systems) as seen below for example:

http://techreport.com/articles.x/17652/8

I won't question the Dirt 2 scores, since those are probably close to reality, given that it's a AMD sponsored game.

What do you mean? PCInlife shows 25.53 FPS for the GTX 470 while the GTX 285 scores 20 FPS, that's a 27.65% improvement.

 

[Razor1]

thats a pretty low increase at such demanding settings again, don't know if they are real, they could be because bandwidth is going to be a pretty important factor at that res and settings.

 

[Razor1]

did anyone catch the title

GTX380, 360

?

 

[Sontin]

I remember the ti4800 to be faster sometimes than the super perfect gpu called fx5800.

Not in <DX9 Games.

 

[WeiT.235]

did anyone catch the title

GTX380, 360

?

The title of the original thread of Asuka said explicitly the gtx470, he changed laterly, maybe for the reason of nvidia.censoring.

The thread seems to be a fighting back for the rumour ' 470 just 5％ faster than 5850' in the chinese forum.

 

[Razor1]

The title of the original thread of Asuka said explicitly the gtx470, he changed laterly, maybe for the reason of nvidia.censoring.

The thread seems to be a fighting back for the rumour ' 470 just 5％ faster than 5850' in the chinese forum.

no its pretty suspect now, especially the unigine benchmark, was told that its not that low on the 470 at those settings.

 

[WeiT.235]

unigine result is bit of salts, as we have been told tesselation is one of most improved function in fermi (by nvidia pr ppt it was about 1.6x rv5870). not sure how he benched the card. but I was told the card most likely be the gtx470, considering rare of full sp units of GF100.

 

[digitalwanderer]

was told that its not that low on the 470 at those settings.

Really? Who told you, who knows?

 

[swaaye]

I remember the ti4800 to be faster sometimes than the super perfect gpu called fx5800.

The FX 5800 Ultra was very competitive with R3x0 in DX8 games, which were prevalent when both cards were being compared. The 5800 actually was faster than the 9700 Pro sometimes in those and OpenGL games. NV30 was terrible once you got into DX9 SM2 territory, of course.

TR's review from back then shows this.
http://techreport.com/articles.x/4966/7

 

[hatter]

The numbers seems plausible. They are in line with what everyone is saying: That GTX480 will end up slightly faster than HD5870 while GTX 470 will be a little faster than HD 5850.

 

[Sontin]

The title of the original thread of Asuka said explicitly the gtx470, he changed laterly, maybe for the reason of nvidia.censoring.

The thread seems to be a fighting back for the rumour ' 470 just 5％ faster than 5850' in the chinese forum.

Sure. Because nVidia doesn't like it when somebody mentions the name...

 

[rjc]

Also, if you believe gf100 to be about 500mm^2, it's not really THAT big, and the smaller chips shouldn't have the problems that they always need to have disabled units (not only because they are smaller but nvidia probably will hopefully fix the design issues by the time they appear...). I don't doubt that nvidia will be able to do some half-fermi chip on 40nm with a similar die size as g92b (but apparently more transistors) which will beat the oldtimer in practical performance (as well as obviously offering more features), something they apparently couldn't do with gt2xx...

The next chip is apparently in process and is slightly larger than the 65nm g92, not the 55nm g92b.

There was a widepread rumor of a 256shader/256bit part, but it appears that for some reason those specs have been cut down.

Personally think they would be having trouble with the memory interface rather than the shader count ie say down to 192bit or something. Can see this in the low data rates of the GT215 and GF100. Need to look at interference and error rates rather than power used.

Tried to look for a site that tested for EMI somehow, but couldnt find anything....a simple high school type radio tuned to the microwave range would probably do, or maybe just a glass of water with a thermometer in it

 

[aaronspink]

Unless of course that gap in the balls means nothing

Which it can. Packages only dictate max die sizes. Its likely that gap only means that they went with outside signal arrays and a dense center P/G array for the balls.

 

[aaronspink]

So you assume that nVidia invested more than twice of the transistor for less than 50% more speed. Even a GTX295 would be faster with less transistors...

BTW: How could they build smaller chips with this kind of per/mm^2? They would slower and bigger than a G92b...

Mr question: Mr Amdahl, Mr Amdahl: Mr question. Doubling resources almost never gets you 2x and in a lot of cases you are happy to get 50%. Its quite easy to screw up. Add to that the whole thing for them is running at a significantly lower frequency than they wanted and...

 

[mapel110]

http://www.abload.de/img/nv-tesselation-benchma0m5f.jpg
Was this picture already posted?

 

[trinibwoy]

Liars, 2x faster is 300%.

 

[rpg.314]

TR's review from back then shows this.
http://techreport.com/articles.x/4966/7

Thanks for posting this link. I moved over to the review page and here are some snippets.

For me, wrapping up this review is a little odd, because it feels more like a post-mortem than a new product review. NVIDIA won't confirm it, but the near-universally-accepted rumor is that very few NV30-based products, either the GeForce FX 5800 Ultra or the plain-jane GeForce FX 5800, will ever see store shelves in North America.

Meanwhile, rumors are flying that the follow-on to NV30, the NV35, is up and running in NVIDIA's labs and coming to market fairly soon.

The NV35 is expected to outperform the current Radeon 9800 Pro without the need for a Dustbuster cooling system.

This would have been a great product had it arrived six months earlier at this same clock speed with lower heat levels, a more reasonable cooler, and lower prices.

This one is particularly illuminating.

Also, I'm not compelled by NVIDIA's talk of "arrays of computational units" and the like to describe the NV30. I'm not convinced they've changed their way of designing GPUs; I think they've mainly just changed their way of talking about them.

 

[rjc]

google engrish transration:


original pic:

Just looking at right hand side - [STRIKE]8[/STRIKE] 10 layer board? Quite a cut down from the GT200 boards, really can cash in here on the lesser sized memory interface. Interesting though from memory the AMD 58xx board has quite a few more layers despite smaller memory interface, remember also they seemed to have quite some trouble with it(required very tight tolerances).

Re figuring die size fom package pin out. From the nvidia site, the GT21x series:
Smaller: GT240m
Close: GT250m
Larger: GTX285m

Sorry

(The GT240m and GTS250m appear to be using the same package).

Edit: Too fast typing - 8 internal + top + bottom = 10 layers

 

[Dave Baumann]

If you go by the right hand side then I'd count 10 layers. And, no, 5800 series are not "quite a few more".