***Official*** NV40 Benches (Updated as they go live) ANANDTECH Review Added

[Corporate Thug]

Originally posted by: Alkali

Battlefield Vietnam

1280 1024 4x/8af GF6800U 52fps -->>73fps at 0xfsaa? (IE 24fps increase) (49% increase in speed)
1024 768 0x/8af 9800XT 49fps

COD

1600 1200 4x/16af 6800u 147fps -->> 200+ at 1280x1024 surely... (IE 100fps increase) (107% increase in speed)
1280 1024 4x/16af 9800xt 107fps

UT2K4

1280 1024 4x/16af 6800u 46fps -->> 65fps at 0xfsaa? IE 20fps increase (44% increase in speed)
1280 1024 0x/8af 9800xt 45fps

Flight Sim 04

1280 1024 4x/16af 6800u 45fos -->> 54fps at 2xfsaa? IE 20fps increase (60% increase in speed)
1280 1024 2x/8af 9800xt 34fps

Lets use some good old-fasioned guesswork... to try and get a picture here.
Lets assume a 2xfsaa setting gives the 6800 a 20% hit in fps, and a 4xfsaa setting gives a 40% hit in fps...



Seems to me those percentage increases are good....

So basically, the 6800 Ultra seems to be between 50 and 100% faster than current top end cards according to my guestimates based on the pics.[/quote]


if you looks at anands 9800XT review, you will realize that 20% for 2x FSAA and 40% for 4x FSAA are both gross overestimates

 

[reever]

Originally posted by: GTaudiophile
Wish Kyle could comment on these...

Be aware that not all results represent "apples to apples" comparisons. We try to find the highest playable resolution, anti-aliasing and anisotropic filtering settings for each card in each game. Keep that in mind as you look at the graphs and find that each card may be set at a different resolution, AA and AF level. What we strive to illustrate to our readers is consistency in performance and which cards provide the highest level of gameplay performance.

underclocked gpu core

You don't even know the concrete specs, so how can you make this comment? They had to solder on more power regulation to the board, I think this is very close to production speeds

 

[Luthien]

Yep, with the PS requirement disspointing benches two slot requirement the NV40 may be the end of Nvidea. Keep your triggor finger on the sell button if you own their stock. Based on these benches I can certainly see ATI winning this round also. But, if COD is the norm after driver updates or something then Nvidea wont be in trouble but I am having serious doubts now. Hopefully some sense will be made out of all this one way or the other in the next day or two.

 

[ponyo]

I never bought into Nvidia hype about being up to 8x fast. I thought we would be lucky to see around 30-50% improvement from 9800 XT/5950 Ultra. It's still early and we just got selected benchmarks. We still don't have the whole picture and don't know how much more the drivers will improve. Given the reported price, I don't think it's that bad.

I don't see R420 doing much better. We shall see.

 

[flexy]

i will refrain from giving further comments on these gifs, because i dont understand 'em. Leads to more confusion which is exactly the opposite a benchmark is supposed to do.

Period.

NV40 numbers look nice...every comparison 4xAA vs 0xAA and different resolutions and these graphs are utter BS.
I am a ATI fan....but in this case i just shut my mouth because these graphs are WORTHLESS.

Edit:

Even same AA settings comparisons are worthless IF the AA algorithms are different....only an in-depth review about the AA methods with image quality comparison and THEN the fps is worth something...and i know from experience that you often cant compare NV AA and ATI AA.
All what counts is HOW IT LOOKS on the screen in real life..and then do a comparison and measure FPS <--- only very few people to benchamrks that way and one of the few which usually does good ones is on anandtech. I dont give a **** about some fancy named AA technique if (in reality) it just sucks and (de facto) is not comparable on the other card. (like 4xAA which looks like exactly the 2xAA on the other card....which is only misleading of consumers and introduces fake/wrong benhmarks if someone does a quick/sloppy review !!!)

 

[Maximilian]

I think we should all just wait for anands benches as they will be accurate, besides nvidia arent stupid... they wouldnt release a card thats supposed to be the first in a new generation of cards if it was only a tiny bit better than the previous generation.

 

[daction]

Those benchmarks are so obviously fake, gimme a break people wait until 9am tomorrow!

 

[BFG10K]

16 pipes and SM 3.0 is looking a definite. Also FP32 at full speed, assuming it's not marketing speak, is very good news for nVidia.

But 16xAF? Woohoo! Finally nVidia, you got with the program.

 

[GTaudiophile]

Originally posted by: daction
Those benchmarks are so obviously fake, gimme a break people wait until 9am tomorrow!

Seems like someone went to A LOT of trouble...

 

[Corporate Thug]

Originally posted by: daction
Those benchmarks are so obviously fake, gimme a break people wait until 9am tomorrow!

can u explain how you can tell they are "obviously" fake?

 

[Jojo7]

Everyone needs to take a deep breath, sit back and relax.

Any self respecting hardware site wouldn't compare video cards in such skewed resolutions/AA/AF settings.

It's like apples and oranges.

 

[GTaudiophile]

Originally posted by: Jojo7
Everyone needs to take a deep breath, sit back and relax.

Any self respecting hardware site wouldn't compare video cards in such skewed resolutions/AA/AF settings.

It's like apples and oranges.

Has someone taken a look at HardOCP's other recent reviews? Did they do something similar?

 

[GTaudiophile]

this answers my own question...

Now how many review sites do this?

 

[Corporate Thug]

Originally posted by: GTaudiophile

Originally posted by: Jojo7
Everyone needs to take a deep breath, sit back and relax.

Any self respecting hardware site wouldn't compare video cards in such skewed resolutions/AA/AF settings.

It's like apples and oranges.

Has someone taken a look at HardOCP's other recent reviews? Did they do something similar?

yes, in their very last review (BFG 5900XT OC) they used different AF at times and even rarely different resolutions

 

[Luthien]

I bet the reviewer based the comparisons on how closely the quality is when compared and that is why the difference in setting. I cannot see any difference in quality and would need it pointed out to me. There is no clear winner in the screen shots showed based on quality IMO so maybe that is why the different specs again. If that is true then the 6800U with settings the same regardless of image quality may be faster by whatever percent but again that means it sucks. If ATI can show it is faster or almost as fast but looks better and when the 6800U is force to use settings that compare in quality ATI shows it spanks the 6800U again Nvidea is in a world of hurt. I do believe that this review is probably the result of that kind of comparison. SELL YOUR NVIDEA STOCK

 

[Jojo7]

Originally posted by: GTaudiophile
Has someone taken a look at HardOCP's other recent reviews? Did they do something similar?

Heh. You make a good point. I just went back and looked at the MSI FX5900 review done on March 23 and this "skewed" situation seems to apply here as well.

I never read any reviews by hardocp anyway. I don't like the layout/news style posting on their site.

 

[Leon]

I bet the reviewer based the comparisons on how closely the quality is when compared and that is why the difference in setting.

From GT Link

Remember, the goal is to find the highest level graphic settings that are still playable on each card.

 

[flexy]

so its not really a 'comparison' - but rather a test what has to be disabled to keep it 'playable' (meaning above 30fps).

He's cranking down settings 'til the game staus above 30fps and (in his opinion) is "playable"

We all got the concept wrong then - its NOT (!) a comparison of performance using the same settings...i dont think the idea is bad per se, but personally i'd rather like the old-fashioned approach and just have PLAIN and EASY comparison w/ same resos and same settings.....!H! is confusing us there

 

[apoppin]

Originally posted by: GTaudiophile
Wonder how many ATi engineers are pouring over these pics...

None.

rumour has it that the ati crew got ahold of an nV40 sample . . . that's why they UNlocked the rest of their x800pipes . . .

I won't spend much time with a puzzle when the entire pciture will be clearer in a few hours.

 

[RussianSensation]

Originally posted by: nick1985
why is the 6800 running 4*AA while the XT isnt????

Sometimes HardOCP has a tendency to compare cards with different resolutions to try to prove that 1 card is a lot better since it can run at higher res while having AA and AF turned on or higher settings of those

That has been their way of benchmarking videocards forever, which is why i dont like their reviews too much. Either way, even after taking this into account, if this is true i am dissappointed to say the least.

 

[flexy]

uhhh..apopping.....gimme a break with "ATI unlocking the secret hidden pipes for our pleasure"

We all know these pipes were there from the start and (IMHO) there is not ONE logical reason why they (OTHERWISE) should have kept 4 pipes secretley hidden and locked. Just doesnt make any sense to me...

 

[RussianSensation]

Originally posted by: klah
http://mbnet.fi/elixir/NV40/10817474486qLMOmeutS_2_1_l.gif

400/1100. Is this the Ultra or non-Ultra?

Those are NON-ultra specs.

 

[mrbios]

After doing some looking around on the ard|Forums, it looks like those are in fact legit ard|OCP benchmarks. Although there is no "official" comment from the FrgMstr himself, some people who do talk with him have basically confirmed it.

 

[Blastman]

The last few graphic card reviews at Hardocp have used different settings on the cards.

Basically the idea is to find the max settings that are ??playable?? on the cards. A faster card can achieve better IQ at a ? ?playable? ? setting so they run the faster card with higher settings to show it?s more capable.

Hardocp started that type of review a couple of months ago with a 5700U/9600XT review. Showing that the 9600XT could run higher IQ settings a lot of the time and still match the 5700U in speed. While I liked the novelty for one review, there was a thread at Beyond3d about the reviews, and I voiced that I much preferred running the settings the same -- it is easier to gage the speed difference between the cards at the same settings --- and it would be obvious to the user that the faster card could run higher IQ settings.

 

[SneakyStuff]

The core is only 400 MHz?