- Jan 29, 2005
- 5,202
- 216
- 106
It's something I've personally noticed, mind you, perhaps many of you actually knew it, maybe despite not having actually read about it black on white (maybe you just "heard about it").
This subject has potential for a long discussion, and quite frankly I'm not sure where to begin with.
Well, let's start somewhere shall we ?
Simply said, nVidia's GeForce 7 series' top-end product, the 7800 GTX (as I type this), has 8 Vertex Pipelines. And so do the X1K series from ATi.
So, 8 VP's on each sides.
Now, let's take a look at those articles from XBit Labs:
http://www.xbitlabs.com/articles/video/display/radeon-x1800_17.html
http://www.xbitlabs.com/articles/video/display/asus-en7800gtx_22.html
The first article is the X1K review. And the last is a review on ASUS' 7800 GTX "TOP" version.
As you can see, I brought you guys on the page where the GPU's are tested in WarHammer 40,000: Dawn Of War. In both articles that is.
In the first paragraph, of the first article, we can read: Dawn of War is a classical real-time strategy, it doesn?t have any of those graphics ?crotchets? the previous game, Perimeter, boasts. However, there can still be quite many military units on the screen at the same time, which imposes pretty heavy workload on the vertex processors.
And in the single paragraph, of the second article, we can read: Warhammer 40000: Dawn of War is a geometrically heavy application. It?s easy to reckon that 100 models in a scene, each made up of 2000 polygons, give a total of 200,000 polygons ? and this wouldn?t be the most complex scene possible! The winning of this test comes naturally to NVIDIA?s solutions with their UltraShadow technology since the game makes use of stencil shadows. Note that the ASUS Extreme N7800GTX TOP is not much faster than the GeForce 7800 GTX ? less than by 10% in the hardest operational mode.
Now, look at the benchmarks themselves.
Notice the minimum frames-per-second ?
At 1024 x 768
nVidia: 45+
ATi: 25+
And here's where it gets more difficult for ATi even ...
At 1024 x 768 still, but add 4x FSAA and 16x AF
nVidia: 45+
ATi: 15+
Now, what do you think ? nVidia's performance hit was actually none, still on the minimum frames-per-second side. While ATi suffered from another 10+ f-p-s drop.
Now, what I find a bit strange is the following ...
In the first article, at 1024 x 768 (without AA nor AF), ATi's X850 XT PE
gets 10+ more frames than ATi's X1800 XT ! How come ?! Especially when you remember that the X1800 XT version tested there as most places on the web was the 512 MB one, and the X850 XT PE is 256 MB (even if the Memory amount didn't count in for any sort of performance gain nor loss, it's still a technical fact not to forget about, as I personally believe, especially when you're in face of a "professional" review via benchmarks).
NVidia has UltraShadow II, and that single feature seems (well, not seems, but actually does) help, according to the benchmarks (and not only XBit Labs'), in terms of "shadowing through Vertex processing needs", since, as confirmed, that game, WarHammer 40,000: Dawn Of War, is heavy on the Vertex workload. Thus "requires" good Vertex processing performance, or it's going to get hit in the face.
I also believe that a good game "tester" for Vertex processing power would be Rome: Total War, where literally 15,000+ troops on the battle field can duke it out in real-time (in custom games).
Why am I writing all this actually ... well, because, honestly, I am still very interested by ATi's X1800 XT, for my second up-coming system. One system with ATi GPU, and the other with nVidia GPU, as I am not a "fanboy", or so I do believe not to be one. But because I like to benchmark as well, and over-clock, and, well ... you know, tweak / play around with the hardware products I buy.
I am a little bit mixed-up about ATi's Vertex processing though, especially when you compare it with the last generation, the X800 series.
Isn't it a technical "logic" that a new generation should be better compared to the previous one, in all departments, or at least in most of them ?
People were saying that ATi might have made a mistake not to make its XT version at least with 24 Pipelines, and instead decided to stay with 16.
But personally, the more reviews I see, and the more I believe that ATi should have made more Vertex Pipelines instead, since the Pixel Pipelines' sheer number doesn't seem to "affect" ATi's X1K performance directly related to it anyway.
Any thoughts ? Facts ?
I'd appreciate.
Thanks for your time.
This subject has potential for a long discussion, and quite frankly I'm not sure where to begin with.
Well, let's start somewhere shall we ?
Simply said, nVidia's GeForce 7 series' top-end product, the 7800 GTX (as I type this), has 8 Vertex Pipelines. And so do the X1K series from ATi.
So, 8 VP's on each sides.
Now, let's take a look at those articles from XBit Labs:
http://www.xbitlabs.com/articles/video/display/radeon-x1800_17.html
http://www.xbitlabs.com/articles/video/display/asus-en7800gtx_22.html
The first article is the X1K review. And the last is a review on ASUS' 7800 GTX "TOP" version.
As you can see, I brought you guys on the page where the GPU's are tested in WarHammer 40,000: Dawn Of War. In both articles that is.
In the first paragraph, of the first article, we can read: Dawn of War is a classical real-time strategy, it doesn?t have any of those graphics ?crotchets? the previous game, Perimeter, boasts. However, there can still be quite many military units on the screen at the same time, which imposes pretty heavy workload on the vertex processors.
And in the single paragraph, of the second article, we can read: Warhammer 40000: Dawn of War is a geometrically heavy application. It?s easy to reckon that 100 models in a scene, each made up of 2000 polygons, give a total of 200,000 polygons ? and this wouldn?t be the most complex scene possible! The winning of this test comes naturally to NVIDIA?s solutions with their UltraShadow technology since the game makes use of stencil shadows. Note that the ASUS Extreme N7800GTX TOP is not much faster than the GeForce 7800 GTX ? less than by 10% in the hardest operational mode.
Now, look at the benchmarks themselves.
Notice the minimum frames-per-second ?
At 1024 x 768
nVidia: 45+
ATi: 25+
And here's where it gets more difficult for ATi even ...
At 1024 x 768 still, but add 4x FSAA and 16x AF
nVidia: 45+
ATi: 15+
Now, what do you think ? nVidia's performance hit was actually none, still on the minimum frames-per-second side. While ATi suffered from another 10+ f-p-s drop.
Now, what I find a bit strange is the following ...
In the first article, at 1024 x 768 (without AA nor AF), ATi's X850 XT PE
gets 10+ more frames than ATi's X1800 XT ! How come ?! Especially when you remember that the X1800 XT version tested there as most places on the web was the 512 MB one, and the X850 XT PE is 256 MB (even if the Memory amount didn't count in for any sort of performance gain nor loss, it's still a technical fact not to forget about, as I personally believe, especially when you're in face of a "professional" review via benchmarks).
NVidia has UltraShadow II, and that single feature seems (well, not seems, but actually does) help, according to the benchmarks (and not only XBit Labs'), in terms of "shadowing through Vertex processing needs", since, as confirmed, that game, WarHammer 40,000: Dawn Of War, is heavy on the Vertex workload. Thus "requires" good Vertex processing performance, or it's going to get hit in the face.
I also believe that a good game "tester" for Vertex processing power would be Rome: Total War, where literally 15,000+ troops on the battle field can duke it out in real-time (in custom games).
Why am I writing all this actually ... well, because, honestly, I am still very interested by ATi's X1800 XT, for my second up-coming system. One system with ATi GPU, and the other with nVidia GPU, as I am not a "fanboy", or so I do believe not to be one. But because I like to benchmark as well, and over-clock, and, well ... you know, tweak / play around with the hardware products I buy.
I am a little bit mixed-up about ATi's Vertex processing though, especially when you compare it with the last generation, the X800 series.
Isn't it a technical "logic" that a new generation should be better compared to the previous one, in all departments, or at least in most of them ?
People were saying that ATi might have made a mistake not to make its XT version at least with 24 Pipelines, and instead decided to stay with 16.
But personally, the more reviews I see, and the more I believe that ATi should have made more Vertex Pipelines instead, since the Pixel Pipelines' sheer number doesn't seem to "affect" ATi's X1K performance directly related to it anyway.
Any thoughts ? Facts ?
I'd appreciate.
Thanks for your time.