Author: Aaron Gordon
Date: 06:26:45 06/27/03
Go up one level in this thread
On June 27, 2003 at 05:19:08, Tom Kerrigan wrote: >On June 27, 2003 at 04:44:40, Aaron Gordon wrote: > >>On June 27, 2003 at 04:06:00, Tom Kerrigan wrote: >> >>>On June 27, 2003 at 02:25:54, Aaron Gordon wrote: >>> >>>>On June 27, 2003 at 01:58:45, Peter Stayne wrote: >>>> >>>>>On June 27, 2003 at 01:42:20, Aaron Gordon wrote: >>>>> >>>>>>On June 27, 2003 at 01:27:19, Peter Stayne wrote: >>>>>> >>>>>>>>In my tests my Athlon XP 2.5GHz beats a P4-3.38GHz, it's close but it does. >>>>>>> >>>>>>I think Carmack is partial to Intel, because he hasn't in all of the Quake3 >>>>>>patches done a quick fix in his code, making SSE work in Quake3. Quake3 treats >>>>>>the Athlon like a K6 and tries to use the non-existant/broken 3DNow! code. Also, >>>>>>he wouldn't allow AMD to make an official patch out of my DLLs. Why? One can >>>>>>only wonder... >>>>> >>>>>The last part of your remark here certainly was a business decision. Those DLL's >>>>>were made by a 3rd party. First off, id would have to negotiate ownership rights >>>>>for the programmer's code. Secondly, even if he were to give them to id for >>>>>free, the moment you include them in an official patch, you're bound to support >>>>>it. This requires that id test them thoroughly, analyze them for issues that >>>>>would need to be addressed. Also, we can't know if it was a 'quick fix' or not, >>>>>perhaps it remained broken because it wasn't a quick fix. Thing is, with Q3, >>>>>most cpu's of the day pulled >60fps with a good GPU, whereas there were quite a >>>>>few GPU's that weren't pulling their weight, which is why I think he focused >>>>>more on GPU performance. He has remained faithfully honest and reliable in the >>>>>nVidia vs ATI wars. >>>>> >>>> >>>>The source code wasn't modified in any way, only compiled with profiling with >>>>the Intel C compiler. So all of the original code remained as such, and wasn't >>>>changed. Even so Carmack still did not allow it. >>>> >>>>I'm pretty sure getting the SSE support enabled on the Athlon XP/MP is no big >>>>deal, it's there for the P3 and P4 (P4 gets detected as a P3, which has SSE, and >>>>that gets enabled). However ID could just change the cpu id detection (very >>>>simple) to detect an Athlon, and then use the pre-existing (and working) SSE >>>>code. >>> >>>Just because Carmack doesn't do something to help AMD doesn't mean he's partial >>>to Intel. >>> >>>-Tom >> >>If you want to better sell your product why not optimize for both AMD and Intel >>cpus? Also, if he isn't partial to Intel why isn't he doing the few simple >>things I explained earlier to fix it? He knows its used as a 3D gaming >>benchmark. >> >>Also, you may say, "He doesn't want to waste the time to blah blah etc etc".. >>well, I did that. AMD and I got to the point to where all he had to do was say >>'yes' or 'no'.. and he chose to say no. He didn't even have to support the DLLs >>in any way, shape or form. Why'd he say no? Who knows (you know what I think >>though), it wasn't like his code was modified.. and he wouldn't have to support >>it in any way. Out of the thousands of people that have downloaded it I have not >>received *1* email of crashing due to the dlls. Him denying all AMD related >>tweaks and not fixing his code sure has a hard time making me think otherwise... > >He knows it's used as a benchmark, but from what I know of him, he could care >less. Q3A is more interesting as a graphics card/memory bandwidth benchmark, he >probably sees CPUs as inconsequential, esp. considering that current CPUs are >putting out hundreds of FPS. Besides, he was probably busy with Doom 3 by the >time you approached him about optimizing Q3A, so that's another reason why he >could probably care less. So you gave him the choice of improving something that >he didn't care about and didn't think mattered at the cost of something >potentially going wrong (however remote the chances) and he'd have to waste time >sorting out. I can understand his decision. > >If the guy isn't partial to ATI or NVidia, which make a much bigger difference >in the work he does, what makes you think he would sell out to Intel somehow? >Besides, weren't his engines one of the first to make use of 3DNow? If you >search for Carmack and 3DNow, the 2nd link is him endorsing it on AMD's web >site. Another thing to think about is that, IIRC, Q3A was written, tested, and >released long before any current Intel hardware so it's not like he could have >optimized anything for, e.g., the P4. > >-Tom Not if the 3DNow! code is completely broken. You won't even gain 1/100th of 1 fps, or anything. Test a K6-266 vs a K6-2/266 (only difference is 3DNow on one chip, the other no 3dnow).. They will run the same, and infact the K6-2/266 was 0.1 fps slower. Try the same test on a Katmai P3-450MHz vs a P2-450MHz.. only difference is SSE on one of the chips. You'll see a significant boost in fps. Also, Quake 1 & 2 never supported 3dnow.. AMD did make a Quake2 patch for 3dnow but that was for Voodoo2 cards only...
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.