Whats with the 12 pipe cards?

[Gamingphreek]

In recent video articals, more specifically Doom III and HL 2, we have shown the Mid and highend scales to shift drastically in the favour of one company or another (not so much in HL2). In DIII we see the 12 pipe 6800NU beating or equaling performance of the X800XT(PE) in many tests. In HL2 we see the 12Pipe X800Pro beating the 6800U a fair amount of times.

What is happening here? I know heavy optimizations are part of the reason however why are 16 pipe cards getting beaten by 12 pipe cards.

As a sidenote: In a recent X-Bit labs review we see the 6800NU getting beaten regularly by the 5950U, how in the hell is that happening aside from rendering in DX8.1 mode.

-Kevin

 

[HeroOfPellinor]

Originally posted by: Gamingphreek
As a sidenote: In a recent X-Bit labs review we see the 6800NU getting beaten regularly by the 5950U, how in the hell is that happening aside from rendering in DX8.1 mode.

-Kevin

HL2 and D3 are both very poor benchmark examples as each is specifically optimized for a certain card.

As for any previous generation card beating the 6800, every time I've seen that occur it's with high res and lots of AF and AA and all the cards involved were generating unplayable FPS with the 9800XT or 5950 being 2-3 FPS higher, when they're all in the teens, which amounts to a whopping 15% in some cases. But you may be referring to a test I haven't seen yet.

 

[Genx87]

It really depends on the game engine and what it is doing. In Doom3 carmack is using alot of stenciling which favors Nvidia hardware bigtime. Also it is opengl and Nvidia has a track record of having much better opengl drivers.

In HL2 it appears fillrate is king and the ATI cards with their high clocks reign supreme. I believe the X800 Pro has slightly higher fillrate than the 6800GT due to its higher clocks.

Then you have memory speed. The 5950U has more memory bandwidth than the 6800NU. So if the test requires bandwidth then the 5950U can beat the 6800NU.

 

[Gamingphreek]

Oh damn i completely forgot about memory bandwidth and stuff. However shouldn't those 16 pipes count for something there otherwise it seems as though all of them are not being filled like they should be.

-Kevin

 

[jiffylube1024]

D3 was designed using many features that Nvidia incorporated in hardware. The way Nvidia handles shadows and overdraw, for example, is more fit for D3 than the way ATI does them.

Also, D3 is an OpenGL game, which is Nvidia's forte, and ATI's achilles heel.

HL2 was designed as a DX 9.0 game, so it runs great on ATI's DX9 cards (9700 and up), which were always strong DX 9 performers.

D3 is Nvidia's showcase game and HL2 is ATI's, and performance is on the side of the company who is pushing the game. This can't be just a coincidence. I'm not saying either company is cheating, but it's obvious they designed their respective games to run better a certain way, using algorithms and rendering methods more favourable to one hardware featureset over another.

 

[eastvillager]

the game wasn't designed to shine using nvidia features, the current nvidia GPU was designed to provide the features id told everybody they were using in doom3.



ati knew what id was doing too, but coming up with a new core wasn't their plan for this round of the graphic wars.

 

[apoppin]

Originally posted by: HeroOfPellinor

Originally posted by: Gamingphreek
As a sidenote: In a recent X-Bit labs review we see the 6800NU getting beaten regularly by the 5950U, how in the hell is that happening aside from rendering in DX8.1 mode.

-Kevin

HL2 and D3 are both very poor benchmark examples as each is specifically optimized for a certain card.

As for any previous generation card beating the 6800, every time I've seen that occur it's with high res and lots of AF and AA and all the cards involved were generating unplayable FPS with the 9800XT or 5950 being 2-3 FPS higher, when they're all in the teens, which amounts to a whopping 15% in some cases. But you may be referring to a test I haven't seen yet.

Try THIS 10x7 one where the 9800xt kicks the 6800Standard's ass

Generally speaking all graphics cards performed well in the video stress test. Even ATI?s previous-generation mainstream graphics cards perform pretty well in 1024x768. What you should keep in mind when comparing results of the GeForce FX-series to the rest is that they worked in DirectX 8.1 mode with aggressive optimizations of texture filtering, whereas others worked in DirectX 9.0 mode.

or 16x12 w/aa+AF

or this one in all 3 benchs

The RADEON 9800 XT appeared to be better performer than the GeForce 6800, which is a good sign for this slightly outdated warrior. Fortunately for their owners, the RADEON 9500- and RADEON 9600-series graphics cards also perform pretty well in 1280x1024 and 1024x768 resolutions.



Even though the GeForce FX 5950 Ultra scores inline with the GeForce 6800, with respect to its rendering methods such victory cannot be considered as correct. The same applies to the rest GeForce FX products in the comparison as well.

So, "that's 'why'".

some cards use different render pathways, some cards like the 6800 are "limited" in other ways - while GENERALLY faster - lose in SPECIFIC instances.

(edit: of course some games are OPTIMIZED for one brand or another, also - like CS)

 

[Genx87]

You know what is funny about the above posters links? It shows a 5950 faster than both the 6800 and 9800XT.

 

[apoppin]

Originally posted by: Genx87
You know what is funny about the above posters links? It shows a 5950 faster than both the 6800 and 9800XT.

why don't you READ the article to find out "why"

:roll:

nothing "funny" except your ridiculing post.

(hint: why don't you read ALL of my post - see what's in "quotes" - and figure it out for yourself?)

 

[KevinH]

All this talk about ATI being better at HL2...what reviews/previews are you guys basing this off of? The Stress test reviews seem to show the difference to be marginal at most.

-Edit- Never mind found the FS review. Guess ATI does do very well in it.

 

[DAPUNISHER]

Originally posted by: KevinH
All this talk about ATI being better at HL2...what reviews/previews are you guys basing this off of? The Stress test reviews seem to show the difference to be marginal at most.

That's leads me to ask; With the HUMUS tweaks now available and increasing performance of ATi cards, does it make the same true for D3???

 

[KevinH]

I can't speak for the X lines of cards because I'm still chugging away with my 9700 pro but it did yield about a 5 fps increase in timedemo at 1024x768 HQ (from 35 to 40). Maybe it was a placebo effect but D3 certainly felt smoother after the tweaks.

 

[gururu]

HUMUS tests

as for myself I got a 20% boost on a 9700pro. KevinH, it only works well if you use medium settings and force AF through control panel.

 

[KevinH]

Originally posted by: gururu
HUMUS tests

as for myself I got a 20% boost on a 9700pro. KevinH, it only works well if you use medium settings and force AF through control panel.

Really? Interesting. By CP you mean the ATI control panel right? What AF lvl should I set?

 

[gururu]

Originally posted by: KevinH

Originally posted by: gururu
HUMUS tests

as for myself I got a 20% boost on a 9700pro. KevinH, it only works well if you use medium settings and force AF through control panel.

Really? Interesting. By CP you mean the ATI control panel right? What AF lvl should I set?

i was using 16x quality AF. no AA.