X-bit labs

Started by: Wingless | Date 08/24/07

Comments: 44 | Last Comment:  12/06/07

[1-20 | 21-27]

1. I suck...I bought a 2900XT for DX9 games =(

This test really is a wake up call because it shows that Nvidia's 8800 series that was praised and touted as the worlds first DX10 cards runs DX10 terribly. The only saving grace are the large amounts of memory that keep the minimum frame rates high. My 2900XT 512mb runs DX10 slightly better but not good enough for my tastes. It will probably run Crysis fine as long as the driver engineers SPECIFICALLY code the drivers for that game. Its annoying having to wait for driver updates just to run specific games well. Bioshock is an exception to that rule because it ran between 60 and 120fps at all times at 1680x1050 before the Bioshock driver patch was released on the Cat 7.8s. I can't wait to see Bioshock DX10 performance tests on these cards.

2. I don't why, but xbitlabs team have chosen old ForceWare drivers.
With 162.18 the situation is different:
http://www.anandtech.com/video/showdoc.aspx?i=3029&p=2

Making a comprehensive article is not a one-day process. Nvidia GeForce 8 graphics cards were tested using the latest driver available at the time of testing. Further performance checks with drivers that became available later indicated that there are no performance improvements for the GeForce 8800 there compared to version 158.24.

The article was posted on 8/22/07 the 160s were posted the end of JUNE, not July, and deffinitely not Aug.

correction
The 16x drivers were posted in July for xp. However it's still almost a month.

Nvidia Corp. published Nvidia ForceWare 162.22 drivers for Microsoft Windows Vista 32-bit operating system on the 26th of July, 2007.

Making a comprehensive article is not a one-day process. Nvidia GeForce 8 graphics cards were tested using the latest driver available at the time of testing. Further performance checks with drivers that became available later indicated that there are no performance improvements for the GeForce 8800 there compared to version 158.24.

3. One more example in CoH DX10:
http://www.firingsquad.com/hardware/diamond_radeon_2900_xt_1gb/default.asp
Take a notice, that in anandtech, xbitlabs and firingsquad the same Catalyst version is used.

4. I find this interesting indeed. Still, even though the ATI cards provide better performance for DX10 games it dosn't seem like a great selling point due to their relatively low speed. The way I see it if your interested in a mainstream card you are better off with a ATI X1950 or NVIDIA 7900 card. By the time games come out that you need DX10 to play their will be a more acceptable graphics solution out.

I dislike the 2600 series for being rather slow in current games, but ATI may be vindicated by proving that these cards will be better for DX10 games than thier 8600 rivals. Thing is your just comparing slow and really slow at that point ^_^

5. Now what need is mid/low end multi-GPU combinations (SLi/Crossfire) benchmarks against 2900/8800 single series.

Forthcoming articles will bring new benchmark results.

However, considering that current mainstream solutions are hardly even 2 times slower compared to the high-end products, we do not think two of such cards will deliver competitive performance when compared to a single-chip high-end graphics card.

6. totally agree with your post #4

That is why i'm thinking of Upgrading my X1950pro 512 to a X1950XT 512mb To hold me over for alittle longer. This way I can skip the current highend cards. Maybe check out the refresh products if they offer better performance in DX games. If not I wait for the next gen which should be summer 08. I think by that time I will be ready to run vista as my primary os as SP1 will also be out at the time. I think sticking to XP right now and the X1950 or 7900 series is a good idea.

The performance of these Highend cards in DX10 is pitaful, and they aren't getting my money until I see more performance and less drivers issues on both sides.

7. I have been telling people to skip this generation of cards. Wait.....wait.....wait....wait....I have a 7950GT and will prob continue to use that for another year.

8. Dude you really should have through in Bioshock.

9. I only have an AMD 4800X2 with an 8800GTX and all of my benchmarks were higher than yours in Lost Planet. I always run at 1920x1200 and I always had over 30fps with 4xAA. What gives?

It's really possibile:
http://www.hardocp.com/article.html?art=MTM3MSwsLGhlbnRodXNpYXN0
All midrange DX10 cards SUCKS!
P.S. these all tests done with old ForceWare is worthless.

We do not comment on third party's benchmark results, as they are often obtained with different in-game and driver settings.

10. only have an AMD 4800X2 with an 8800GTX and all of my benchmarks were higher than yours in Lost Planet. I always run at 1920x1200 and I always had over 30fps with 4xAA. What gives?
[Posted by: DRen72 | Date: 08/24/07]


YEAHH RIGHT keep on dreaming

anyway these high end card sck cause they are not even future proof

Correct on that one yasin. No use buying any of them. Mid range is much better.

11. Its not only the drivers and cards at fault but MOSTLY programming on the game side. DX10 is a new API and developers of the current games you have listed have yet to optimize the code for it, it is more their fault then the fault of the hardware developers. Please stop blaming ATI and Nvidia, get a clue...please!

12. What kind of comparison is this with forceware drivers over 2 months old vs the latest ATI beta?

kind of smells like fish to me!

13. " ... the Radeon HD 2900 XT can challenge the much more expensive GeForce 8800 GTX." You make that statement, yet you do not even point out, much less explain why, the HD 2900 XT 512MB shows better performance than the HD 2900 XT 1GB; or that the 8800 GTX showed better than the 1GB part in the first few results.

I think there is something wrong with your testing ... do not know what it is, but something is not right. Therefore, the whole test should be ignored, and is not useful in reaching a decision on how to spend one's money. I don't lean towards either brand reviewed as I own both, and a reviewer needs to be totally unbiased. That may not be the case here.

The quality of this review is not up to the normally achieved excellent quality that XBit Labs is known for.

We cannot explain driver issues of AMD and Nvidia, as we are writing reviews, not programming drivers.

Currently there are a lot of talks regarding performance differences between the GeForce 8800 GTS 320MB and 640MB as well as Radeon HD 2900 XT 512MB and 1GB. Nobody, including officials from the developer companies can explain performance differences officially.

We outlined that "Performance of both ATI Radeon HD 2000 and Nvidia GeForce 8 is fully dependent on drivers, therefore, it is nearly unpredictable which graphics card will eventually become better DirectX 10 performer."

14. In half of CoJ tests HD2900 XT 1Gb (740/2000) is slower than HD2900 XT 512Mb(740/1650) - nonsense.

15. Did you disable ATI optimizations too (i.e. Catalyst AI)? From what I've read, nVidia's default filtering quality is actually better than ATI's... AF on the HD 2000 series is no better than X1000 series.

Its pretty sad when reviewers dont understand what there doing isnt it. Cat AI does introduce texture filtering optimizations yet Xbit doesnt disable it but disables Nvidias. And yes Nvidias default AF is better then ATIs High Quality. Cat AI on my x1900xtx shimmers in Source engine games, Farcry, ANNO 1701 to mention a few, turning Cat AI off and the shimmer is gone. I dont see why Cat AI would act any differently on the 2900 xt. Either bench with Optimizations disable for both or bench at default. Pull your head out of the sand Xbit.

What are you guys talking about?
I have both cards and NVIDIA is much worst, the textures are all blurry with the NVIDIA.
XBitlabs is doing the right thing in disabling the Nvidia cheats.
In fact if they want to keep the NVIDIA cheats enabled just go to Ati Control Center and lower the Texture and MIP level one value and they are both comparable.

You've got a GeForce 8 or a GeForce 7? I had a GeForce 7900 GTX in my system, but I upgraded to an X1950 XTX because it had better filtering quality. G7x filtering quality was quite frankly awful and I used to run with optimizations disabled (high quality) on that card. However, it's one area that Nvidia improved with GeForce 8, as there is now angle-independent filtering at default quality - it's a shame ATI didn't see the need to improve its filtering quality with R600 because it has been an area where ATI has been fairly dominant in the last few years.

16. The Call of Jaruz benchmarks are completely useless. If you do some reading on its development, you see that it was heavily optimized for ATI hardware.
Second, CoH DX10 benchmarks are useless because the DX10 version is just a patch and the games engine itself is based on dx9 code. Its not a dx10 game.
That leaves us with Lost Planet, a true dx10 game where ATI gets their ass handed to them.

ATi is not going to even survive if they crank out more shitty architecture while Intel and nVIDIA are stomping all over the place. RIght now, AMD's mediocre performance really gives meaning to Intel's and nVIDIA's slogans.

"We started on our DX10 journey with Lost Planet. In that evaluation we found that there were no major visual differences between DX9 and DX10."

There goes your true dx10 game...

17. "Call of Jaruz was heavily optimized for ATI hardware."
"That leaves us with Lost Planet, a true dx10 game where ATI gets their ass handed to them."

And the Lost Planet was paid by NVIDIA didnt you read the news?
Ati didnt even know that game was comming.
And Lost Planet was ported from the XBox.
Get your info right or dont even post.

All I can say is "bring on the BioShock, Crysis, ET:QW, HL2E2, TF2, Portal, Hellgate: London, World in Conflict and UT3 benchmarks!"

None of the games benchmarked here are true DX10 games, and none are particularly compelling. CoH was compelling (and still is to an extent) when it came out, but that was almost a year ago.

18. Where is 8600GT?

Due to poor performance of mainstream and entry-level DirectX 10-compatible graphics cards, we decided not to test cut-down solutions and restrict the testing with the "base" offerings.

19. I'm not going to play soon DX-10 games. I prefer DX-9, high FPS and high res.. with a DX-10 card like the 8800 GTS !

20. Very disappointing. Maybe the performance of true DX10 titles is better? both BioShock and WiC perform decently in Vista with DX10, much better than the trio you guys tested anyway. this article was lacking the depth we usually see with X-bitlabs articles. what gives?

[1-20 | 21-27]