CS: Source benches

[Brandon]

I noticed the thread linking to the original VST scores we posted, I don't have time to go through it all, but I think now that we have the CS: Source results online, that's a better resource to go by:

http://firingsquad.com/hardware/counter_strike_source/

I'm now going back and re-running scores with the FX cards forced to run the DX9 path with VST, CS, and a HDR demo I made up real quick in dust. I don't know the command to force the FX mode (mixed mode) from ATI Shader Day last year. Valve may have decided to remove it entirely.

[Pete]

Good stuff, Brandon. Those are some heady resolutions, there. :)

I'm looking forward to the mid-range review. The mixed-mode setting is also interesting, as Valve's latest CSS CSV seems to show they still apply a "DX81" rather than "DX80" or just DX9 mode to the 5800 and 5900. It may be that nV switched detection from the game to its drivers, though, as it seems to have happened with Doom 3.

[trinibwoy]

Now we just need 5 other sites to come out with completely different results :lol:

Very nice for the 6800 though if FS's results are indeed accurate.

[PatrickL]

On the extreme high-end side, we also included eVGA’s e-GeForce 6800 Ultra Extreme Edition. For the Extreme Editions, eVGA handpicks the fastest 6800 Ultra cards and overclocks them by default. All cards are validated and guaranteed to run at 450MHz core/1.2GHz memory, which is an improvement of 50MHz on the graphics core and 100MHz (effective) on the memory. This makes these boards some of the fastest GeForce 6800 Ultras on the planet.
The only downside is that these cards are ultra rare. To date, eVGA has only sold 60 cards

I am not really happy when people benchmark cards like that. I bet more of theses cards were sent out to reviewers than sold, and should be more seen as part of the marketing budget than anything else.

[trinibwoy]

I am not really happy when people benchmark cards like that. I bet more of theses cards were sent out to reviewers than sold, and should be more seen as part of the marketing budget than anything else.

Agreed but it doesn't matter though since the XT beats both comfortably - it's not like the Ultra Extreme pulled a coup - and it didn't affect the results at the lower segments where it matters more.

[PatrickL]

Yes it is not the result the problem but the way things are going that does not seem right in my book.

[^eMpTy^]

I am not really happy when people benchmark cards like that. I bet more of theses cards were sent out to reviewers than sold, and should be more seen as part of the marketing budget than anything else.

Agreed but it doesn't matter though since the XT beats both comfortably - it's not like the Ultra Extreme pulled a coup - and it didn't affect the results at the lower segments where it matters more.

Yeah the EE is bullshit...it's fun to see in benchmarks...but really doesn't belong there...more of a novelty than anything

How do you figure the XT beats both comfortably? I'm looking at the numbers now and the only place I see the XT winning by more than 2fps is 2048x1536...hardly what I would call comfortable...looks pretty damned even to me...

You ARE talking about the firingsquad numbers right?

[jvd]

I am not really happy when people benchmark cards like that. I bet more of theses cards were sent out to reviewers than sold, and should be more seen as part of the marketing budget than anything else.

Agreed but it doesn't matter though since the XT beats both comfortably - it's not like the Ultra Extreme pulled a coup - and it didn't affect the results at the lower segments where it matters more.

Yeah the EE is bullshit...it's fun to see in benchmarks...but really doesn't belong there...more of a novelty than anything

How do you figure the XT beats both comfortably? I'm looking at the numbers now and the only place I see the XT winning by more than 2fps is 2048x1536...hardly what I would call comfortable...looks pretty damned even to me...

You ARE talking about the firingsquad numbers right?

2048x1536x32 trilinear
6800ultra extreme 75.8
6800ultra 69.4
radeon xt pe 77.9
2.1 frames faster than the extreme and 8.5 than the ultra

2048x1536x32 4x af
ultra extreme 67.5
ultra 60.6
xt6 pe 76
8.5 faster than the extreme and 15.4 frames persecond more than the ultra. The pro is even almost 3 frames faster

2048x1536x32 16xaf
extreme 62.3
ultra 56.2
xt pe 72.5
10.2 faster than the extreme and 16.3 faster than the ultra. Once again the pro is faster than the ultra .

2048x1536x32 4x aa 8xaf

ultra extreme 43.8
ultra 39.5
xtpe 57.3
13.5 faster than the extreme and 17.8 faster than the ultra. The pro is once again faster than the ultra and is with in 1 frame of the ultra extreme

Now in all other reses you can see the geforces close or faster than the radeons

Whats really interesting is the max settings . It shows full well that the x800tpe has power to spare as even at 6x aa/16xaf 1600x1200 its with in two frames of the 6800ultra which is only running with 4x

though the link posted here is not representive of half life 2

according to the article

Keep in mind that we’re pumping significantly fewer shaders and bump/normal maps than in the video stress test, which could make it more of a shader performance test than an indicator of performance with the final release of Half-Life 2.

[Moose]

The 6800Gt is looking like a great bang for buck card. :)

[jvd]

The 6800Gt is looking like a great bang for buck card. :) not as good as the x800pro vivo :-)

[trinibwoy]

The 6800Gt is looking like a great bang for buck card. :) not as good as the x800pro vivo :-)

You do realize that 99% of people who buy a vivo are going to be running it stock right?

[^eMpTy^]

The 6800Gt is looking like a great bang for buck card. :) not as good as the x800pro vivo :-)

You do realize that 99% of people who buy a vivo are going to be running it stock right?

The GT stock easily beats the Pro vivo stock...and for less money...if you're a hardcore type looking to mod to 16 pipes and overclock...the vivo starts to look better...but I don't think many people are looking to void warranties on $400 hardware, I know I'm sure not...

[jvd]

The 6800Gt is looking like a great bang for buck card. :) not as good as the x800pro vivo :-)

You do realize that 99% of people who buy a vivo are going to be running it stock right?

The GT stock easily beats the Pro vivo stock...and for less money...if you're a hardcore type looking to mod to 16 pipes and overclock...the vivo starts to look better...but I don't think many people are looking to void warranties on $400 hardware, I know I'm sure not...

I disagree with that statment .

The gt does not easily beat a pro stock . Unless we are talking about doom3.

But with half life 2 as the stress tests show the x800pro will sometimes beat a 6800ultra.

[trinibwoy]

But with half life 2 as the stress tests show the x800pro will sometimes beat a 6800ultra.

Hanners' findings (http://www.elitebastards.com/page.php?pageid=6336&head=1&comments=1) strongly disagree. This half-life stress test whatever has to be the most inconsistent benchmark I've ever seen.

[jvd]

But with half life 2 as the stress tests show the x800pro will sometimes beat a 6800ultra.

Hanners' findings (http://www.elitebastards.com/page.php?pageid=6336&head=1&comments=1) strongly disagree. This half-life stress test whatever has to be the most inconsistent benchmark I've ever seen.

all his findings show is that if you put a 400$ video card with a 70$ cpu your not going to get the best performance .

vr zone shows it diffrent http://www.vr-zone.com/?i=1181&s=8


here we the gt run out of steam and the x800pro out pacing the 6800ultra and almost outing the 6800ultra extreme !seehttp://firingsquad.com/hardware/counter_strike_source/page5.asp


http://firingsquad.com/hardware/counter_strike_source/page4.asp and here we see it beat the ultra in all but one setting .


Fring squad even shows that 6x is playable .


To say that hte 6800gt easily beats the x800pro in any benchmark but doom3 is wrong.

Both cards trade blows really well at stock speeds .

[Bjorn]

2048x1536x32 trilinear
6800ultra extreme 75.8
6800ultra 69.4
radeon xt pe 77.9
2.1 frames faster than the extreme and 8.5 than the ultra

2048x1536x32 4x af
...


The 2048*1536 res results doesn't seem valid at all compared to the 1600*1200. As in, probably a driver problem. And the authors mentions this in the article also:


At 2048x1536 driver limitations are likely playing a role in holding back the GT and Ultra cards.

[jvd]

then i guess nvidia doesn't have the golden drivers any more eh ?

[Bjorn]

then i guess nvidia doesn't have the golden drivers any more eh ?

How many people play games at 2048*1536 ?

I would consider those number to be pretty unimportant and i guess Nvidia does too. Especially these days with people moving to LCD's. (It'll take a while before 2048*1536 res LCD's are mainstream)

[FUDie]

then i guess nvidia doesn't have the golden drivers any more eh ?
How many people play games at 2048*1536 ?

I would consider those number to be pretty unimportant and i guess Nvidia does too.
That's the lamest excuse I have ever heard. If you're going to be an NVIDIA apologist (a far cry from the old Rendition days, eh?) then at least think of some good excuses.

To answer your question, a lot less people will be gaming at 2048*1536 on NVIDIA cards than ATI cards :lol:... at least in HL2/CS:S.

-FUDie

[euan]

then i guess nvidia doesn't have the golden drivers any more eh ?

Their only golden, after the driver has a compete new rendering path implemented internally for the specific application. ;)

[Bjorn]

That's the lamest excuse I have ever heard. If you're going to be an NVIDIA apologist (a far cry from the old Rendition days, eh?) then at least think of some good excuses.


Well, we just have to disagree. I don't really believe that NVidia ignores that resolution but i still think my point is valid, benchmarks in a res that hardly anybody is using and that for some reason (bug or non optimized path in the drivers) doesn't match the ones in the other resolutions isn't something that i would care about. + that we're talking about a beta game.

[Dave Baumann]

It may not be driver issues that are holding back that resolution - it could be the case that it exceeds the onchip ZCULL mechanism, meaning its only operating on part of the screen or disabled entirely (I don't know what NVIDIA do in that situation).

[Bjorn]

It may not be driver issues that are holding back that resolution - it could be the case that it exceeds the onchip ZCULL mechanism, meaning its only operating on part of the screen or disabled entirely (I don't know what NVIDIA do in that situation).

Interesting, that should show in other games also then i guess. Though benchmarks in that resolution aren't that common.

[Rodx]

I want to see some overclocked x800 pro scores.
Dave, if you'll be reviewing the pro, please include overclocked benches in HL2 and CS:S!

[WaltC]

Interesting, that should show in other games also then i guess. Though benchmarks in that resolution aren't that common.

Shows the problem with benchmarks, doesn't it?....;)

If you really want to look for average resolutions used by most people, I think you'd be looking at 1024x768, tops (especially true for people limited by LCD resolution caps.) For many people, a benchmark run at 1600x1200 is no more reflective of practicality than is a benchmark run at 2048x1536. The reason to bench at such higher resolutions, though, is to push the rendering hardware to the point where the software is limited by the gpu instead of the cpu, so as to get a better picture of the graphics hardware itself. The higher the resolution, the harder the push on the graphics hardware. If 3d acceleration was disabled in the hardware at resolutions above 16x12, then you'd certainly have a point...;)

[Bjorn]

The reason to bench at such higher resolutions, though, is to push the rendering hardware to the point where the software is limited by the gpu instead of the cpu, so as to get a better picture of the graphics hardware itself.

That's true. Benchmarks in 2000*.. of course gives us some information. But is performance in that resolution critical when comparing the NV40 and R420 ? IMO, no. Others may disagree.

[trinibwoy]

Including a particular resolution in a benchmark article should be dependent on the playability of the game at that resolution. Since the XT demonstrated that it can handle it then I see no reason to diminish it's importance - granted that 16x12 is a much more practical resolution than 20x15.

[jvd]

I really want to see 6x fsaa and 8x fsaa peformance compared more often.

They give close to similar image quality .

4x is getting boring. I used to play all games with 4x on my 9700pro.

[RickCain]

I want to see some overclocked x800 pro scores.
Dave, if you'll be reviewing the pro, please include overclocked benches in HL2 and CS:S!

Here you go! X800 Pro (560 core 545 mem w/ XP SP1 and CAT 4.8)

(A64 3500+ 1gig RAM)

1024 x 768 0x AA 8x AF 138.75fps (world)
1024 x 768 0x AA 16x AF 133.73fps (world)

1024 x 768 0x AA 8x AF 115.96fps (reflect all)
1024 x 768 0x AA 16x AF 115.37fps (reflect all)

1024 x 768 4x AA 8x AF 122.76fps (world)
1024 x 768 4x AA 16x AF 118.18fps (world)

1024 x 768 4x AA 8x AF 110.11fps (reflect all)
1024 x 768 4x AA 16x AF 106.35fps (reflect all)

[trinibwoy]

I really want to see 6x fsaa and 8x fsaa peformance compared more often.

Only reason I can see for this is if you get a kick out of seeing ati beating nvidia in benchmarks. 8xSSAA will rarely be playable in any widely used benchmark.

[T2k]

I really want to see 6x fsaa and 8x fsaa peformance compared more often.

They give close to similar image quality .

4x is getting boring. I used to play all games with 4x on my 9700pro.

Bottom config, 1920x1200 w/ 6xAA+16xQAF is ~5xfps.

[jvd]

I really want to see 6x fsaa and 8x fsaa peformance compared more often.

They give close to similar image quality .

4x is getting boring. I used to play all games with 4x on my 9700pro.

Bottom config, 1920x1200 w/ 6xAA+16xQAF is ~5xfps.

I know i have one too. I'm saying in reviews

Only reason I can see for this is if you get a kick out of seeing ati beating nvidia in benchmarks. 8xSSAA will rarely be playable in any widely used benchmark

that or i want to see which card is capable of displaying the best image at the best speed.

Not my fault nvidia is using such a demanding mode.

alot of people don't play at 1600x1200 and at the lower res both cards should have enough steam to enable it

[Bjorn]

I really want to see 6x fsaa and 8x fsaa peformance compared more often.

They give close to similar image quality .

4x is getting boring. I used to play all games with 4x on my 9700pro.

Close to the same quality at edges perhaps but 8XFSAA will also help with texture aliasing if i'm not mistaken. Nothing wrong with benchmarking them though, as long as you're not trying to comparing them as they were apples to apples.

[Degrader]

How is it possible that the Radeon 9800 XT is much faster than a Radeon 9800 Pro 256MB? The clockspeed of the 9800 XT is not more than 8% faster? And both cards have 256MB VRAM. The only thing may be the different BIOS (more optimized :?) of the 9800 XT, I think.

Link:
http://www.firingsquad.com/hardware/half-life2_vst/page7.asp

[Bjorn]

How is it possible that the Radeon 9800 XT is much faster than a Radeon 9800 Pro 256MB? The clockspeed of the 9800 XT is not more than 8% faster? And both cards have 256MB VRAM. The only thing may be the different BIOS (more optimized :?) of the 9800 XT, I think.

Link:
http://www.firingsquad.com/hardware/half-life2_vst/page7.asp

Some people have mentioned that different runs can result in differences of about 15% which probably explains a lot of peculiar results in these benchmarks.

[Degrader]

How is it possible that the Radeon 9800 XT is much faster than a Radeon 9800 Pro 256MB? The clockspeed of the 9800 XT is not more than 8% faster? And both cards have 256MB VRAM. The only thing may be the different BIOS (more optimized :?) of the 9800 XT, I think.

Link:
http://www.firingsquad.com/hardware/half-life2_vst/page7.asp

Some people have mentioned that different runs can result in differences of about 15% which probably explains a lot of peculiar results in these benchmarks.

But the 9800 XT is always significantly faster, so I don't think that might be the problem. But I just saw some people saying that the problem is that the 9800 Pro 256MB runs too hot, due to it's DDRII memory. That might be a good reason, I think.

[^eMpTy^]

But with half life 2 as the stress tests show the x800pro will sometimes beat a 6800ultra.

Hanners' findings (http://www.elitebastards.com/page.php?pageid=6336&head=1&comments=1) strongly disagree. This half-life stress test whatever has to be the most inconsistent benchmark I've ever seen.

all his findings show is that if you put a 400$ video card with a 70$ cpu your not going to get the best performance .

vr zone shows it diffrent http://www.vr-zone.com/?i=1181&s=8


here we the gt run out of steam and the x800pro out pacing the 6800ultra and almost outing the 6800ultra extreme !seehttp://firingsquad.com/hardware/counter_strike_source/page5.asp


http://firingsquad.com/hardware/counter_strike_source/page4.asp and here we see it beat the ultra in all but one setting .


Fring squad even shows that 6x is playable .


To say that hte 6800gt easily beats the x800pro in any benchmark but doom3 is wrong.

Both cards trade blows really well at stock speeds .

Maybe you should read the conclusion of the reviews you link too instead of just picking the one page out of the whole thing that supports your point.

With the exception of 2048x1536, the GeForce 6800 GT sweeps the X800 PRO in 4xAA/8xAF and 4xAA/16xAF testing. At 2048x1536 driver limitations are likely playing a role in holding back the GT and Ultra cards.

The Pro competes well but loses to the GT in every single benchmark I've seen, including the firingsquad review you linked too and the xbitlabs benchmarks...not to mention it kicks the shit out of the Pro in OpenGL games...

If you want to recommend the vivo, I can understand that, but to question whether the GT is faster than the Pro is just plain silly...ALL benchmark data points to that not being true...nearly without exceptiong...

[gordon]

There were quite a few people who would normally say the x800 Pro only beats the 6800GT when AF is turned on. Sorta the other way around in that source review since the 6800GT only matches or beats the x800 Pro when AF is turned on.

[jvd]

I really want to see 6x fsaa and 8x fsaa peformance compared more often.

They give close to similar image quality .

4x is getting boring. I used to play all games with 4x on my 9700pro.

Close to the same quality at edges perhaps but 8XFSAA will also help with texture aliasing if i'm not mistaken. Nothing wrong with benchmarking them though, as long as you're not trying to comparing them as they were apples to apples.

Well they are both the top of the image quality mark. So why not ?

For over a year we had ati's 4x fsaa against nvidia's 4x fsaa although the 4x from nvidia was inferior to the nvidia mode.

The Pro competes well but loses to the GT in every single benchmark I've seen, including the firingsquad review you linked too and the xbitlabs benchmarks...not to mention it kicks the shit out of the Pro in OpenGL games...


I don't read conclusions . THe reviewers normaly let bias into the end views.

ALso there are plenty of other benchmarks from sites showing the x800pro faster than the 6800gt .

[Bjorn]

Well they are both the top of the image quality mark. So why not ?

For over a year we had ati's 4x fsaa against nvidia's 4x fsaa although the 4x from nvidia was inferior to the nvidia mode.


And i think every review mentioned that it had higher quality also. But in this case, you also need to show that it also does something about texture aliasing which is harder to show in pictures. Makes it to much of a problem imo.

[jvd]

Well they are both the top of the image quality mark. So why not ?

For over a year we had ati's 4x fsaa against nvidia's 4x fsaa although the 4x from nvidia was inferior to the nvidia mode.


And i think every review mentioned that it had higher quality also. But in this case, you also need to show that it also does something about texture aliasing which is harder to show in pictures. Makes it to much of a problem imo.

The major problems i see in many reviews is the pimping of nvidia's sm3.0 .

Esp in the reviews .

Also the lack of temporal aa testing is another major problem with reviews.

They should write up about temporal and explain it and give personal opinons on it .

Also the lack of higher lvl aa testing is a problem.

[Bjorn]

...
Also the lack of higher lvl aa testing is a problem.

I agree. Especially since Ati's 6X FSAA is playable in a lot of games these days. Hell, even Nvidias 8xS is playable in a lot of games. But most reviews concenrate on the settings that are comparable between the IHV's. Which will be 4X FSAA and so forth. Perhaps better then to do Beyond 3D style of reviews and concentrate on the cards actual features more then how it compares to the competition. Both styles are necessary but there's to many of the "which card is faster at ... res and 4X FSAA..." type of reviews.

[^eMpTy^]

I really want to see 6x fsaa and 8x fsaa peformance compared more often.

They give close to similar image quality .

4x is getting boring. I used to play all games with 4x on my 9700pro.

Close to the same quality at edges perhaps but 8XFSAA will also help with texture aliasing if i'm not mistaken. Nothing wrong with benchmarking them though, as long as you're not trying to comparing them as they were apples to apples.

Well they are both the top of the image quality mark. So why not ?

For over a year we had ati's 4x fsaa against nvidia's 4x fsaa although the 4x from nvidia was inferior to the nvidia mode.

The Pro competes well but loses to the GT in every single benchmark I've seen, including the firingsquad review you linked too and the xbitlabs benchmarks...not to mention it kicks the shit out of the Pro in OpenGL games...


I don't read conclusions . THe reviewers normaly let bias into the end views.

ALso there are plenty of other benchmarks from sites showing the x800pro faster than the 6800gt .

Right because you're not biased?

And why don't you link me to the one or two situations in another benchmark where the Pro wins...or was that the only place you could find?

:roll:

[T2k]

Khm, interesting thing: WinXP x64 Edition b1218, A64 3200+ CS:S Beta in 1280x1024 w/ 4xAA +8xAF and R9700 Pro produced some ~19.xxfps but REALLY nice water, waves etc...

And Quadro 3000 ONLY in 1024 w/ same filters, ForceWare 61.76 (AMD64) and 17.xxfps and NO WATER AT ALL.

Weird, isn't it? :shock:

[jvd]

Empty i may be bias but i'm not writing reviews. I also don't sign up on forums to spread fud .

I can post you to driverheaven and all thier benchmarks show the x800pro wining .

http://www.anandtech.com/video/showdoc.aspx?i=2102&p=5

here it is beating the 6800gt by as much as 5 fps

here again beating it by 3fps http://www.anandtech.com/video/showdoc.aspx?i=2102&p=6

You can even follow it further and see it beating the 6800gt in nvidia recorded benchmarks with out the use of instancing and sm2.0b .



I was going to link u to the xbit labs reviews with sm2.0b enabled . But i can't get to the site

[jimmyjames123]

Selectively quoting again, huh? :D

For the nth time (for people like jvd, blastman, etc): DriverHeaven typically benches the NV cards with all optimizations off, while the X800 cards have tri and aniso filtering optimizations enabled at all times. Same deal with Xbit Labs.

There is really little doubt from looking at the reviews in general that the 6800GT is slightly faster overall than the X800Pro, with or without AA/AF (up to 4xAA obviously). The arguably more advanced feature set, and ability to overclock past Ultra speeds without needing modification, make the 6800GT is a pretty fine and logical choice. That's why ATI is planning on releasing a $399 16 pipeline part. ;)

[Cynips]

DriverHeaven typically benches the NV cards with all optimizations off, while the X800 cards have tri and aniso filtering optimizations enabled at all times. Same deal with Xbit Labs.
Don't know about that, but this time DH explicitly ran with NV optimizations on. Several of the people behind DH actually have NV cards in their personal setups, so I don't give your allegations much credit.

[dan2097]

Same deal with Xbit Labs.


Not any more. I cant remember when from but xbitlabs has been benchmarking the 6800s with both anisotropic filtering and trilinear filtering optimizations on, see the comments about their half life 2 stress test performance review when their site comes back up.

I personally feel this invalidates their half life 2 results (possibly why the 6800GT significantly beats the X800 pro) as the ATI cards do not use texture stage optimizations when set to application preference whilst enabling anisotropic filtering optimizations on the 6800s forces texture stage optimizations on both control panel and application af

[T2k]

Selectively quoting again, huh? :D

For the nth time (for people like jvd, blastman, etc): DriverHeaven typically benches the NV cards with all optimizations off, while the X800 cards have tri and aniso filtering optimizations enabled at all times. Same deal with Xbit Labs.


I smell some BS here... :?:

[RickCain]

Empty i may be bias but i'm not writing reviews. I also don't sign up on forums to spread fud .

I'm sorry... I think this is my fault.

[^eMpTy^]

Empty i may be bias but i'm not writing reviews. I also don't sign up on forums to spread fud .

I can post you to driverheaven and all thier benchmarks show the x800pro wining .

http://www.anandtech.com/video/showdoc.aspx?i=2102&p=5

here it is beating the 6800gt by as much as 5 fps

here again beating it by 3fps http://www.anandtech.com/video/showdoc.aspx?i=2102&p=6

You can even follow it further and see it beating the 6800gt in nvidia recorded benchmarks with out the use of instancing and sm2.0b .



I was going to link u to the xbit labs reviews with sm2.0b enabled . But i can't get to the site

So by stating the general accepted rationale that the GT is faster than the pro I'm stating FUD???

And look at your benchmarks...July 02, 2004...with new drivers the GT is ahead and you know it...

[jvd]

So by stating the general accepted rationale that the GT is faster than the pro I'm stating FUD???

And look at your benchmarks...July 02, 2004...with new drivers the GT is ahead and you know it...

No yu know why you spread fud.


In those tests the gt had the benfit of new drivers. Ati had 2 new leaked sets and one new offical driver at the time of those benchmarks none of which were used. The 6800s had the newest leak set . Not to mention the fact that ati didn't have sm2.0b or instancing support at that time

Depending on the day or driver set the 6800gt or the x800pro is ahead.

THe only exception is doom3. Its the only thing the 6800gt is continuely faster than the x800pro in

[^eMpTy^]

So by stating the general accepted rationale that the GT is faster than the pro I'm stating FUD???

And look at your benchmarks...July 02, 2004...with new drivers the GT is ahead and you know it...

No yu know why you spread fud.


In those tests the gt had the benfit of new drivers. Ati had 2 new leaked sets and one new offical driver at the time of those benchmarks none of which were used. The 6800s had the newest leak set . Not to mention the fact that ati didn't have sm2.0b or instancing support at that time

Depending on the day or driver set the 6800gt or the x800pro is ahead.

THe only exception is doom3. Its the only thing the 6800gt is continuely faster than the x800pro in

lol...and there are even more Nvidia drivers coming out making the GT even faster...whatever man, if you don't want to admit ithe GT is faster, that's fine, you're entitled to your delusions...

[RickCain]

lol...and there are even more Nvidia drivers coming out making the GT even faster...whatever man, if you don't want to admit ithe GT is faster, that's fine, you're entitled to your delusions...

We will see if NV is resorting back to their bad habits. Based on the increase I am reading, either they did something real good with their drivers or they will be distroyed by many websites.

[jvd]

lol...and there are even more Nvidia drivers coming out making the GT even faster...whatever man, if you don't want to admit ithe GT is faster, that's fine, you're entitled to your delusions...

We will see if NV is resorting back to their bad habits. Based on the increase I am reading, either they did something real good with their drivers or they will be distroyed by many websites.

and he makes it sound like we are never going to get new ati drivers.

He makes me laugh alot

[oddfellow]

I don't know the command to force the FX mode (mixed mode) from ATI Shader Day last year. Valve may have decided to remove it entirely.

I sent you a message via your website last night regarding forcing ps2.0 on nv3x hardware.

I don't trust those web based message forms, so if you didn't receive it let me know. I'd be interested in seeing some tests done on your hdr dust demo when forcing the "stdshader_hdr_dx9.dll" on nv3x hardware.

[mongoled]

Glad to see some members here posting abt the results seen on R3xx cards. I posted this over @ rage3d only to see it get bogged down with peeps who seem more interested in arguing abt the same things over and over again. Hopefully some R3xx uses would care to comment on my results and the differences certain websites are showing compared to these.

thanxs for providing those results in the manner u have, that has saved me alot of time. I have to say that the firing squad results with the 9800XT do seem strange to me. My results are more consistant with those of VR-Zone testing, one thing I have been suprised with is the performance penalty antialiasing is having with my gfx card. Their is a substantial hit going to 2xAA, after 2xAA, the penalty is alot less. This result im seeing is mirroring the results Hanners has shown in his article.

But why the initial massive hit in enabling AA? And why are Firing Squads 9800XT results so scewed? There is no way a stock 9800XT 256MB card is going to outperform a 128MB card clocked @ 480/400.

Here are my results, each time a benchmark was run and a setting changed the app was exited re-started and then the stress test was run.

Stats:

9800Pro128>XT (R360) 480/400
k7 @ 235x11.5 (2700mhz)
1GB ram

----------------0xAA_8xAF----2xAA_8xAF----4xAA_8xAF
1024x768-------90.35----------53.48------------51.40
1280x1024------58.30----------32.33------------31.19
1600x1200------46.20----------27.99------------24.48


FiringSquad - A64 3800+ 1GB 4.8 - 9800XT
1024x768------------------------------------------77.4
1280x1024-----------------------------------------57.7
1600x1200-----------------------------------------42.3

VR-Zone - A64 3800+ 1GB 4.8 - 9800XT
1024x768------------------------------------------40.78
1280x1024-----------------------------------------32.92
1600x1200-----------------------------------------27.98


Hope these results help some peeps out who havnt got the stress test available. Results from a real-life user who just wants a rig ready to play HL2 when its released

mong

[Skinner]

Well I share my results of the Source VST with both X800XT and 6800U.


X800XT cat 4.8

1280x1024x32 x16 AF Controlpanel, waterreflect on "all"----82 fps ?

1280x1024x32 x16 AF ingame, waterreflect on "all"---------103 fps

GF6800U

65.62
1280x1024x32 x16 AF Controlpanel, waterreflect on "all"----87 fps

66.00
1280x1024x32 x16 AF Controlpanel, waterreflect on "all"---100 fps

note
The GF6800u cut's faster through the rotating shader plates then the X800XT with driver 66.00.

[Lezmaka]

Khm, interesting thing: WinXP x64 Edition b1218, A64 3200+ CS:S Beta in 1280x1024 w/ 4xAA +8xAF and R9700 Pro produced some ~19.xxfps but REALLY nice water, waves etc...

And Quadro 3000 ONLY in 1024 w/ same filters, ForceWare 61.76 (AMD64) and 17.xxfps and NO WATER AT ALL.

Weird, isn't it? :shock:

I don't see how that's weird at all considering the Quadro FX 3000 is using an NV35GL chip. And in addition to that, it's conceivable that, since it's a workstation-class card, it doesn't use the same driver path/optimizations as a regular NV35-based GFFX 5900 card.

[Blastman]

There is really little doubt from looking at the reviews in general that the 6800GT is slightly faster overall than the X800Pro, with or without AA/AF (up to 4xAA obviously). The arguably more advanced feature set, and ability to overclock past Ultra speeds without needing modification, make the 6800GT is a pretty fine and logical choice. That's why ATI is planning on releasing a $399 16 pipeline part. ;)


If the performance difference is only “slight” -- either way (not overclocking), then other factors are going to weight in. So I hardly see the GT as the … “logical” … choice for everyone when the Pro has things in it’s favor too. Like a 6AA that’s much more usable than the 8AA of the GT. The X800Pro at 6AA will be twice as fast as the 6800GT at 8AA. Also, the X800’s Trylinear is better than the Brilinear of the GT it’s usually being benched against. The Pro also has lower power consumption. etc.

[DarthFrog]

Also, the X800’s Trylinear is better than the Brilinear of the GT it’s usually being benched against.

How so? It is difficult to determine or compare filtering quality because the ATI driver switches to trilinear if you use coloured mips and ATI's sample optimizations (which NVIDIA are currently adding to their driver as well or already have) are even more difficult to pin down.

Two more advantages of the X800 that I factored into my decision (even though it went against the ATI card in the end) are temporal AA and the ability to force triple-buffering in OpenGL games.

[hovz]

uhhhh digit life had a filtering comparison that used methods other than colored mip maps and atis trylinear is far ahead of nvidias bri.

[Skinner]

I take it you guys follow the progress of both IHV's driverdevelopment right? I can assure you both optimizations are more orr less comparable, with small fluctuations in different sets.

The performance of the 6800(U) is increasing in each driverupdate, so it's not that relevant to point to old reviews.

With the latest, availabe 66.00 driver the 6800U is on par in the Source VST and I cann't see any quality tadeoffs.

[kyleb]

How so? It is difficult to determine or compare filtering quality because the ATI driver switches to trilinear if you use coloured mips...

yet with briliniear is easy to see the degradation from triliniear in game without colored mips while with trylinear it is not; where is the difficulty in that?