R300 the fastest for DoomIII, John Carmack Speaks Again

Status
Not open for further replies.
OpenGL Guy, seeing as you work at ATi, surely you know that ATi included a special "hook" for Quake 3 that changed the MIP-map levels for demo runs? I'm not sure why you're denying it, as it was proven by a few sites.

I don't mean to drag this thread down, but the rose-tinted glasses are getting too numerous, speculative thread or not. 8)

BTW, no one has an 8500 and is willing to do a few minutes' worth of benchmarking? I'm not sure I've seen a review (recent or otherwise) compare aniso between a GF3/4 and an 8500 with the former explicitly set to bilinear. I'm more interested given ATi's recent fps boost with their latest drivers.
 
Did you make sure that the texture and geometry quality were set to high?
Click to expand...

Yeapper. And ensured the desktop was set to 32bit color as well. Thats why I posted them as they were the same settings. Just wish I had a PC closer to your specs to make it more "fair".

Bambers I have the newest dirvers but I upgraded my PC to a AMD 1800+ so its running almost at x2 the speed of his. I had those numbers on-line and posted them as they were the same settings he had on hardware that was close. I figued if it was not close, but maybe in the ball park (if not then in the parking lot of that ball park). I also posted them from work so I could only use data I had..
 
Pete said:
OpenGL Guy, seeing as you work at ATi, surely you know that ATi included a special "hook" for Quake 3 that changed the MIP-map levels for demo runs? I'm not sure why you're denying it, as it was proven by a few sites.

I don't mean to drag this thread down, but the rose-tinted glasses are getting too numerous, speculative thread or not. 8)

BTW, no one has an 8500 and is willing to do a few minutes' worth of benchmarking? I'm not sure I've seen a review (recent or otherwise) compare aniso between a GF3/4 and an 8500 with the former explicitly set to bilinear. I'm more interested given ATi's recent fps boost with their latest drivers.
Click to expand...

Chalnoth can you tell me which version of q3 and which demo you used? Then I can get some 8500 results.

I have got a fairly similar PC (950 tbird @ 1073 and it will be no problem for me to reduce it to chalnoths speed).
 
Joe DeFuria said:
But if you're to assume that those products were not released because of marketing reasons and not "physical" reasons, then we can assume the same could happen to NV30. Ala: GeForce4 Ti Ultra can be "competitive" with R300, especially on todays "benchmarks." NV30 is soooo advanced, that it would blow R300 out of the water. So why not wait until the next product cycle to release NV30? Why should nVidia launch it now?
Click to expand...

Yeah, and why release NV30 at all? Think about all the R&D they can save by just reiterating the NV20. So when ATI launches R400, nVidia can just counter with GeForce5 XP30000, at a mind-blowing 600 MHz clockspeed! Who really cares if ATI supports DX10 and OGL2.0? Everyone will buy GF5 as long it gives 500+ FPS in Q3!
 
Bambers said:
Chalnoth can you tell me which version of q3 and which demo you used? Then I can get some 8500 results.

I have got a fairly similar PC (950 tbird @ 1073 and it will be no problem for me to reduce it to chalnoths speed).
Click to expand...

I used the Quake3 demo (and demo001). I don't have the full version (I actually nabbed it from a friend once...it was fairly quickly deleted...).

Btw, my board is running at 266MHz FSB.

Also, don't forget that all settings were at their maximum, with the exception of TC and trilinear filtering.
 
Some new comments from ol' JC at 3Dchipset...

AbRASiON : John, Quick one I'll ask straight off the bat. A lot of people love speculating hardware (as do I) and I'm reading that ATI was the supplier of the gfx chip that ran Doom at E3, can you please confirm if it is either an un-announced product from ATI or The 8500 card that's already out? No one has really been specific if it's "new ati" or current! :(

Carmack : It is not an 8500. The 8500 is consistently slower than the GF4 at Doom, except when you turn on anisotropic texturing, which the 8500 is significantly faster at.
Click to expand...

My good friend who is very close to the developer/programmer side of things has shot me an email that shows that indeed the Doom III demo was powered by the ATi R300 next generation hardware. I'm not going to quote the exact email as it could lead my friend in some trouble. Here is the key elements:

The card that Carmack was using was the R300. Some of the engineers were onsite and they worked through the night to get it working just in time for the show. Doom III has been running at E3 since 9am this morning on R300 hardware. The ID team seems very satisfied with the system and the ATI team seems quite happy as well. The system was advertised as running on ATI "Next generation HW".
Click to expand...

http://www.3dchipset.com/news.shtml#newsitem1022701034,96010,
 
Results for demo001 on 8500 64Mb @ 275/275 (Retail specs). 933Mhz tbird ABIT KT7A

Max detail except compression.

1600x1200x32:

Bilinear: 99.0fps
Trilinear: 96.5fps
2xaniso: 95.1fps
4xaniso: 93.5fps
8xaniso: 92.3fps
16xaniso: 91.6fps

1024x768x32 16xaniso in all

no FSAA: 127.8fps
2xP: 93.3fps
2xQ: 80.0fps
3xP: 66.3fps
4xP: 45.0fps
4xQ: 34.9fps
5xP: 30.1fps
6xP: 20.0fps << :LOL:

Tried 320x240 min detail and got 143fps.
 
Is your card overclocked? Mine isn't.

Still, it does verify what we knew all along...though now the GF4 is a bit closer with max aniso on (Of course, it's way ahead with both aniso & FSAA enabled, which is what I always use).

Oh, and btw, I ran a little timedemo at 320x200 w/ min details. Result:
156.0 The main reason for this score being higher is that I have an nForce. Remember that the nForce has an integrated DASP, and so its advantage disappears once you insert an Athlon XP.
 
Its exactly the same as the retails (3.6ns ram etc). I got it back in early nov so I assume ATI were short of OEMs (soon to be called LEs) and just used boards that would normally pass for retails and lowered their voltages to stop overclocking. I used the pencil mod on it to take it to the retail standard volts. I'll bumb them up a bit more once I get some decent cooling on the core.
 
Geeforcer said:
Some of the engineers were onsite and they worked through the night to get it working just in time for the show.
Click to expand...

So Kristof was right...
Click to expand...

So, because they worked on it, it shows they were "hacking/cheating/whatever"? You don't know the whole story, so maybe you should keep quiet.
 
Chalnoth said:
Is your card overclocked? Mine isn't.

Still, it does verify what we knew all along...though now the GF4 is a bit closer with max aniso on (Of course, it's way ahead with both aniso & FSAA enabled, which is what I always use).

Oh, and btw, I ran a little timedemo at 320x200 w/ min details. Result:
156.0 The main reason for this score being higher is that I have an nForce. Remember that the nForce has an integrated DASP, and so its advantage disappears once you insert an Athlon XP.
Click to expand...

I would say for a 8500 LE that is very close to your scores, even with FSAA, its only when 4X FSAA is when the 8500 starts to lose significantly, not bad for Supersampling.not to mention the 8500 is set to 16X Ansitropic. I also know that the default lod on some of Nvidia drivers are not set as agressive, as seen on Rage3D with RTCW shots..it took tweaking of the lod to -1.3-1.5 ...LOD affects frames quite a bit.

http://63.231.44.178/board/showthread.php?s=&threadid=33611472

compare.jpg
 
OpenGL guy,

So, because they worked on it, it shows they were "hacking/cheating/whatever"? You don't know the whole story, so maybe you should keep quiet.
Click to expand...

Why are you so hung up about "hacking/cheating/whatever"? Nobody is suggesting cheating or anything - how can you cheat on an unreleased board? Kristof was just saying that in a situation like this is fairly normal for a company to pull out all the stops and get the application running by whatever means necessary - the quote clearly say that this is what ATi have done, and quite rightly; why construe that as anyone implying 'cheating'?
 
Pete said:
OpenGL Guy, seeing as you work at ATi, surely you know that ATi included a special "hook" for Quake 3 that changed the MIP-map levels for demo runs? I'm not sure why you're denying it, as it was proven by a few sites.
Click to expand...

If you are talking about the Quake/Quack issue, I am pretty sure I've commented on this before. Compare the benchmark results before the bug was fixed to after the bug was fixed. I think you'll see they are pretty much the same. In fact, I didn't even buy my 8500 until I knew the problem was resolved.

I don't mean to drag this thread down, but the rose-tinted glasses are getting too numerous, speculative thread or not. 8)
Click to expand...

I don't have rose-colored glasses on, but it sounds like some people remember every bit of dirt no matter what.
 
Hmmm I would say that with the past Quake 3 reference that has been in Radeon drivers since Day 1, even the Radeon 1..anyone mentions anything like hack, cheat to a ATI coder and fur starts to fly :LOL:
 
Doomtrooper said:
Hmmm I would say that with the past Quake 3 reference that has been in Radeon drivers since Day 1, even the Radeon 1..anyone mentions anything like hack, cheat to a ATI coder and fur starts to fly
Click to expand...

Why don't you go look for nvidia's cheats? Some aren't that hard to find.
 
OpenGL guy said:
Doomtrooper said:
Hmmm I would say that with the past Quake 3 reference that has been in Radeon drivers since Day 1, even the Radeon 1..anyone mentions anything like hack, cheat to a ATI coder and fur starts to fly
Click to expand...

Why don't you go look for nvidia's cheats? Some aren't that hard to find.
Click to expand...

I don't need to look, I know they opitmize to..all companies cheat...I never thought the Quake 3 thing was a big deal and you read my post wrong...I was basically saying some websites made a big deal about the Quake 3 reference on the 8500 debut...a Mr. Kyle Benett little knowing that reference was in there since his own websites Radeon DDR review, where they made reference to the better IQ on the Radeon vs. the Geforce 2....How Ironic ;)

Opengl guy, I don't think Kristof is saying ATI was cheating or meant anything like that..at least I don't read that hence my sensetivity post about the ATI hacks ..it was supposed to be a joke..you know :LOL:

Smile, relax looks like the R300 is gonna smoke, and according to nvnews front page link NVIDIA President and CEO Jen-Hsun Huang said the Nv30 is releasing in August..thats only three Months away so Nvidia MUST have Nv30 boards but JC chose ATI on what they saw.. thats good :p
 
OpenGL guy said:
Geeforcer said:
Some of the engineers were onsite and they worked through the night to get it working just in time for the show.
Click to expand...

So Kristof was right...
Click to expand...

So, because they worked on it, it shows they were "hacking/cheating/whatever"? You don't know the whole story, so maybe you should keep quiet.
Click to expand...

Actually, I was referring to his "burning midnight oil" comment. No need to be rude.
 
Status
Not open for further replies.
Back
Top