Valve Half Life 2 Presentation and Benchmarks @ Beyond3D

Dom · Sep 16, 2003, 06:04 AM

While ATI’s “Shader Day” was a very useful educational experience, inevitably it will mainly be remembered for the comments from Valve's presentation by Gabe Newell concerning NVIDIA’s products and how they perform with Half Life 2. Which is a shame in some ways. However, the strong way in which Valve has spoken out is more or less a direct result of the strength of ATI’s hardware shader capabilities.

First Valve were keen to stress that Half Life 2 was written in such a fashion to make effective use of a wide range of DirectX9 functionality – in fact in many places they started with DX9 and deconstructed the operations for DX8 paths. Because of this they go on to suggest that the performance trends displayed under Half Life will be more or less the case with other games that make effective use of DirectX9.

Read More...

BWX · Sep 17, 2003, 05:00 PM

"One of the issues with 5800 is that it had more integer (“DX8”) pixel shader units than it did floating point (“DX9”) units, but with 5900 these integer units are said to have been removed and replaced with float units. If the 5900 pipeline is fully float in the first place then there shouldn’t be any inherent performance issues in running PS1.4 shaders in favour of PS2.0. It’s also slightly ironic that NVIDIA bemoaned the use of PS1.4 shaders in 3DMark03’s Mother Nature test, citing that few games utilise it, and yet are suggesting that it could be a more correct choice here – there seems to be a level of inconsistency in the messages that are being sent from NVIDIA in this regard."

http://www.beyond3d.com/misc/hl2/index.php?p=5

What gets me is, I think we would be playing this game already if Valve just wrote the code for DX9 hardware, and left out any specific code/ architecture path.

I don't think they should have made all the concessions they did for Nvidia's "problem" and let Nvidia deal with it w/ drivers and or new "fixed" hardware. After all, DX9 is here, and it's been such a long wait for a full DX9 game.. It's not fair to ATI to go to all these time consuming and expensive optimizations for HL2, when ATI just runs it on the default DX9 code, like ALL GPUs should be doing in 2003, especially if you paid over 400 DOLLARS for them..

The paragraph I posted from the article also is VERY interesting, and really makes me wonder what Nvidia was thinking when they designed the NV35, surely they new this day would come. I just don't understand why they would make hardware that need "special optimizations" to run DX9 apps.. I just don't get it.

Sep 16, 2003, 06:04 AM	#1
Dom DriverHeaven Extreme Member Join Date: Jun 2002 Posts: 12,940 Rep Power: 0	Valve Half Life 2 Presentation and Benchmarks @ Beyond3D While ATI’s “Shader Day” was a very useful educational experience, inevitably it will mainly be remembered for the comments from Valve's presentation by Gabe Newell concerning NVIDIA’s products and how they perform with Half Life 2. Which is a shame in some ways. However, the strong way in which Valve has spoken out is more or less a direct result of the strength of ATI’s hardware shader capabilities. First Valve were keen to stress that Half Life 2 was written in such a fashion to make effective use of a wide range of DirectX9 functionality – in fact in many places they started with DX9 and deconstructed the operations for DX8 paths. Because of this they go on to suggest that the performance trends displayed under Half Life will be more or less the case with other games that make effective use of DirectX9. Read More...

Sep 17, 2003, 05:00 PM	#2
BWX unplugged Join Date: Nov 2002 Location: USNY Posts: 19,669 Rep Power: 110 System Specs	"One of the issues with 5800 is that it had more integer (“DX8”) pixel shader units than it did floating point (“DX9”) units, but with 5900 these integer units are said to have been removed and replaced with float units. If the 5900 pipeline is fully float in the first place then there shouldn’t be any inherent performance issues in running PS1.4 shaders in favour of PS2.0. It’s also slightly ironic that NVIDIA bemoaned the use of PS1.4 shaders in 3DMark03’s Mother Nature test, citing that few games utilise it, and yet are suggesting that it could be a more correct choice here – there seems to be a level of inconsistency in the messages that are being sent from NVIDIA in this regard." http://www.beyond3d.com/misc/hl2/index.php?p=5 What gets me is, I think we would be playing this game already if Valve just wrote the code for DX9 hardware, and left out any specific code/ architecture path. I don't think they should have made all the concessions they did for Nvidia's "problem" and let Nvidia deal with it w/ drivers and or new "fixed" hardware. After all, DX9 is here, and it's been such a long wait for a full DX9 game.. It's not fair to ATI to go to all these time consuming and expensive optimizations for HL2, when ATI just runs it on the default DX9 code, like ALL GPUs should be doing in 2003, especially if you paid over 400 DOLLARS for them.. The paragraph I posted from the article also is VERY interesting, and really makes me wonder what Nvidia was thinking when they designed the NV35, surely they new this day would come. I just don't understand why they would make hardware that need "special optimizations" to run DX9 apps.. I just don't get it. __________________