Supposed NV40 numbers (it's packin' some serious heat).

[Flyermax2k3]

Originally posted by: lordtyranus2
Judging by this benchmark

http://www.xbitlabs.com/articles/video/display/ati-nvidia-roundup_13.html

We see that an overclock of the 9800 Pro from the standard 380/340 to a 420/360 setup yields a 3 FPS increase in speed, or a 9% speed increase.

So clearly, at even those settings, overclocking the GPU does increase performance, so it is potentially conceivable that putting in an even faster GPU could yield even better numbers. In other works, the game is clearly not CPU limited at those settings.

But as for the 72 FPS figure, that seems to be complete BS. If nothing else, I can NOT believe for a second that this card is 100% better than the current best on the market. 30%? Great. 50-60%? Stretching it but conceivable. But 100%? No way.

One must first understand what the term "CPU-lmited" means before making a judgement as to whether or not an application is, indeed, CPU-limited.
CPU-limited does not mean that the CPU is the only factor which determines an app's performance, just that it determines an app's performance moreso than any other factors. An app can be CPU-limited and still receive increases in performance from other factors such as FSB speed, GPU speed, memory timing, etc.
If we're going to talk about the UT botmatch being CPU-limited, then there's no point in providing benches of any other apps or even UT if it's not a botmatch. Compare apples to apples and then we'll talk

 

[Pete]

The 9700P was 100% faster than GF4Ti 4600 @ high res AA+AF, so why is it so hard to believe the NV40 would be just as big of a leap?

Acanthus, analysis seems to indicate those purported NV40 screenshots are running 4xMSAA with a 9-tap blur filter, apparently the same thing that's been available in nV cards for a while. Although Vegetto claimed RGMS, the screenshots don't seem to entirely agree. Check Hyp-X's pics in the middle of this page.

 

[Acanthus]

Originally posted by: chsh1ca

Originally posted by: Acanthus
And i thought AT was ATi biased. Wow.

Hahaha. AT is one of the saner places around IMO.

They dont even take into account that not only is he using an OCed A-64. We are dealling with a 16pipe card, clocked HIGHER than the NV35, with much faster memory.

Actually, no, it wasn't, the game tests were run on a Dual Opteron box as mentioned in this quote (Page 5, The Baron):

Okay, I got some clarification on the CPU speed/botmatch controversy. The system specs were not accurate for Painkiller and UT2004--for these, a dual Opteron machine was used (workstation, probably). The rest of the scores were done on the more reasonable machine posted before. Remember, the card was only in the tester's possession for two days, so I imagine he took it between home and work.

They then go on to claim that the NV40s AA+AF scores in botmatch are impossible. On a card with much higher fillrate, on an A-64 @ 2400mhz. Then they go on to talk about SSAA and MSAA, which the NV40 USES NEITHER ONE (they use a new "superior" AA implementation). Even if the guy is lieing, his numbers are very possible.

With AA/AF turned off, botmatch appears to scale with the CPU more than the GPU, though it is definitely capable of being limited by either at any given time.

In the end, we're going to have to wait until early April before we see some benches. If they are launching with a huge lanparty on the 13th, cards should be sent out to reviewers about a week before, right?

I would imagine if its that close to readyness, AT either has it now, or its on its way.

 

[Acanthus]

Originally posted by: Pete
The 9700P was 100% faster than GF4Ti 4600 @ high res AA+AF, so why is it so hard to believe the NV40 would be just as big of a leap?

Acanthus, analysis seems to indicate those purported NV40 screenshots are running 4xMSAA with a 9-tap blur filter, apparently the same thing that's been available in nV cards for a while. Although Vegetto claimed RGMS, the screenshots don't seem to entirely agree. Check Hyp-X's pics in the middle of this page.

So that means he's lieing. It means nothing about NV40
nV is claiming up to SEVEN TIMES as fast as the FX5950 in certain games. (doom3 was specifically named).

 

[BFG10K]

Check the link out, there is no way that bench is close to CPU limited at 16x12 w/AA+AF.

Where does it say that bench is a botmatch (genuine question)? In any case with full AA and AF I'd agree that even a botmatch would probably be GPU limited.

Finally, the proof for once and for all that Unreal Botmatch is ONLY CPU LIMITED AT =< 800X600,and proof that unreal botmatch is ALSO GPU LIMITED at 1600x1024

Barefeats is the worst benchmarker on the web and his results have been shot to pieces each time they're released. Not only that but using Macs to determine dynamics in 3D games is just silly. A 2 GHz G5 for example, is slower than a 2 GHz Northwood in UT2003.

Is anyone else surprised at how bitter this ostensibly fun discovery has been contested?

The numbers and claims don't irritate me that much, it's posters (usually nVidiots) who continue to hammer that they're correct and ignore all evidence to the contrary.

I mean look at what was promised with the NV30: nVidiots hyped it up yet it turned out to be one of the biggest flops in the history of 3D cards. Even nVidia themselves admitted it was a mistake and it should've never been released, yet if you listened to the hype at the time you would've thought it was a 9700 Pro killer.

 

[stardust]

Originally posted by: Pete
Considering the overwhelming number of redundant threads on when the NV40 will be available, I'm surprised no one has linked to the only thread I know of posting both screenshots AND framerates of nVidia's salvation.

Obviously those numbers are uncorroborated and second-hand, but Anand and others have mentioned incredible performance jumps before, and those numbers are pretty impressive (particularly the 3DM2K1 fillrate, but 2x 5950U/9800XT @ 16x12 ain't too shabby ). They still leave some room for R420, though.

Without reading the whole of this post, the screenshots @ Beyond3D with the comparison between the "next gen card" and This, it seems nvidia has managed to cover the whole screen with a "blurring" mask that makes AA'ed edges even softer and more rounded. I discovered that when looking at the weapon of the 2 screenshots, the detail of the gun is crisper on the ATI card but the lines on the front part of the scope are rounder on the next-gen nvidia card due to the blurr effect.

This is just what i noticed, don't hold me accountable for any wrong assumptions please.

 

[BenSkywalker]

The 9700P was 100% faster than GF4Ti 4600 @ high res AA+AF, so why is it so hard to believe the NV40 would be just as big of a leap?

Actually, the huge failure NV30 is frequently 100% faster or more at high res with AA+AF versus the Ti4600; and that was a board people were slamming for being too slow regularly. I think a lot of people have got the R300 performance level stuck in their head as it has been the upper echelon for 18 months now, somewhat forgetting what a generational leap can look like

Where does it say that bench is a botmatch (genuine question)? In any case with full AA and AF I'd agree that even a botmatch would probably be GPU limited.

He posted the XBit numbers to cross reference and stated the flyby numbers were ~130FPS.

The numbers and claims don't irritate me that much, it's posters (usually nVidiots) who continue to hammer that they're correct and ignore all evidence to the contrary.

Did you read the thread @B3D? It was a bunch of fanatics calling the guy an idiot because nothing could possibly be faster then the R9800Pro that nVidia would ever build. They insisted that the bench was CPU limited @37FPS despite the fact that the R9800Pro was pushing roughly 100FPS at the lowest settings used in the same bench. I don't really see anyone stating that the benches are accurate outside of those that claimed to have witnessed them, just people pointing out that the fanatics are exhibiting a rather huge amount of idiocy in the way they are trying to disprove the benches.

 

[Pete]

Acanthus,

So that means he's lieing. It means nothing about NV40
He may just be repeating info without knowing much about it. That's my impression, anyway, but it doesn't nec'ly discount his numbers.

nV is claiming up to SEVEN TIMES as fast as the FX5950 in certain games. (doom3 was specifically named).
Uh, yeah, I have a bridge for sale. Interested? ;P

stardust, yes, I linked to where Hyp-X seemed to show the NV40 screenshots were 4xOGMS with a 9-tap blur filter (and Digital Vibrance). The overall shot still looks great, though.

Ben,

Actually, the huge failure NV30 is frequently 100% faster or more at high res with AA+AF versus the Ti4600
I was going to mention that, but didn't feel like typing the extra words, and didn't think anyone would notice. But I'd imagine ppl refer to 9700P simply because the darn thing showed up about half a year before the 5800U.

 

[lordtyranus2]

Did you read the thread @B3D? It was a bunch of fanatics calling the guy an idiot because nothing could possibly be faster then the R9800Pro that nVidia would ever build. They insisted that the bench was CPU limited @37FPS despite the fact that the R9800Pro was pushing roughly 100FPS at the lowest settings used in the same bench. I don't really see anyone stating that the benches are accurate outside of those that claimed to have witnessed them, just people pointing out that the fanatics are exhibiting a rather huge amount of idiocy in the way they are trying to disprove the benches

It has nothing to do with Nvidia. I'd be just as skeptical if an ATI card more than doubled the 9800 XT.

I think a lot of people have got the R300 performance level stuck in their head as it has been the upper echelon for 18 months now, somewhat forgetting what a generational leap can look like

Perhaps. But I think labeling them as biased is rather unnecessary.

 

[Acanthus]

Originally posted by: Pete
Acanthus,

So that means he's lieing. It means nothing about NV40
He may just be repeating info without knowing much about it. That's my impression, anyway, but it doesn't nec'ly discount his numbers.

nV is claiming up to SEVEN TIMES as fast as the FX5950 in certain games. (doom3 was specifically named).
Uh, yeah, I have a bridge for sale. Interested? ;P

stardust, yes, I linked to where Hyp-X seemed to show the NV40 screenshots were 4xOGMS with a 9-tap blur filter (and Digital Vibrance). The overall shot still looks great, though.

Ben,

Actually, the huge failure NV30 is frequently 100% faster or more at high res with AA+AF versus the Ti4600
I was going to mention that, but didn't feel like typing the extra words, and didn't think anyone would notice. But I'd imagine ppl refer to 9700P simply because the darn thing showed up about half a year before the 5800U.

NVIDIA is comparing their next gen part to current in a specific game at specific settings and says "up to" 7 times as fast. This would obviously be on a very high end system. But calling NVIDIA flat out liars is just asinine pete.

 

[BFG10K]

He posted the XBit numbers to cross reference and stated the flyby numbers were ~130FPS.

Yes but he also said he ran a botmatch so I fail to see how the X-Bit benchmarks are relevant.

Besides, to my knowledge UT2004 doesn't even have a built-in benchmark so he must've created one himself and again, I fail to see how that relates to the X-Bit numbers.

 

[caz67]

Drool..I cant wait

 

[BenSkywalker]

It has nothing to do with Nvidia. I'd be just as skeptical if an ATI card more than doubled the 9800 XT.

So you are a new enthusiast to the vid card market then. If you had followed it for some time you would have seen that a 100% bump in performance for new generations is not unheard of by any means. As has already been pointed out, the current gen cores when new exceeded the twice as fast mark over their predecessors. So did the V2 when it hit the scene, as did the GFDDR when it hit the market. Seeing a new gen twice as fast as the former is certainly no shock.

For that matter, I'm trying to think if we have ever seen such glacial advancement as we have over the last 18 months in the vid card market. For the gens that failed to match up with the doubling of performance, they were relatively rapid fire. In the last 18 months we have seen top tier parts improve their performance by ~15%-20%, at this point we should be expecting an enormous performance leap.

Perhaps. But I think labeling them as biased is rather unnecessary.

Saying they are biased is akin to say a McLaren F1 is sort of fast. The posters in that thread in particular and the B3D forums in general are extremely biased, and that is far from some big discovery. Wait until the new round of boards hits and watch them declare whatever element ATi wins at the most important one this generation, no matter what it is. If they lose to nVidia on all fronts then they will be adamant that nVidia is cheating their way to the top, even if the margin was in excess of 100%(not that that is going to come close to happening) they would still come up with some excuse as to why ATi is better. With the strong rumors of ATi's new core lacking PS 3.0 support they have already turned a complete 180 on that front saying the only thing that matters this gen is PS 2.0 performance(despite their hailing of the second coming of the crucial PS 2.0 performance of the R300 as it was paramount and all developers were chomping at the bit to advance to the new feature set).

BFG-

Besides, to my knowledge UT2004 doesn't even have a built-in benchmark so he must've created one himself and again, I fail to see how that relates to the X-Bit numbers.

Then you should easily realize how utterly moronic the fanatics were being stating the bench had to be CPU limited. They are doing better with their spin now, the 36K on 2K1 should have been an easy indicator that something was all screwed up- much better PR angle for their masters as it is one that is not utterly preposterous. There is no defending their stance here, they are blindly rabid fanatics on the attack because someone dared state that nVidia could make something better then the euphoric R9800XT- and they failed to let rational thought guide them for long enough to notice the really screwed up number and instead jumped all over the one they knew nothing about.

BTW- The XBit numers were also from a botmatch bench.

 

[Pete]

Originally posted by: Acanthus
But calling NVIDIA flat out liars is just asinine pete.

Can I start laughing now, or should I wait until the reviews are out? Seriously, that 7x figure is as much of a joke as the NV30 pre-launch hype was. Do you expect NV40 to offer 7x the performance of NV38 with only .5-2x the transistors and only 20% more bandwidth (assuming 600MHz DDR3)? I'm a fan of miracles, too, but 7x is likely a highly-specific number that I'm almost sure we won't see in almost any game, even in DX9 HL2 compared to the relatively weak NV38. But 7x faster in Doom 3, a game which the NV3x series seems to have been optimized for, seems astonishing.

Yes, I suppose my comment was asinine. After all, it's not like nV has put out misleading info before, right?