Radeon8500 "high poly count" bug?
- Thread starter BenM
- Start date
OT: Hey Doomtropper,
Look at this Unreal 2 performance test: http://www.anandtech.com/video/showdoc.html?i=1583&p=12
This test is supposed to really stress the GPU with fillrate and polygons.
The GF3Ti200 is side by side with the 8500LE and only 5% slower than a 8500, and today is as low as $120 (see pricewatch).
I have one and I am still tweaking it for best image quality but overall is a very good card (good features, performance, price, value, drivers). I want to keep it for some good time (I hope until XMAS 2003)
Unfortunatelly people will probably have to pay something like $+200 for a GF4Ti4200
Look at this Unreal 2 performance test: http://www.anandtech.com/video/showdoc.html?i=1583&p=12
This test is supposed to really stress the GPU with fillrate and polygons.
The GF3Ti200 is side by side with the 8500LE and only 5% slower than a 8500, and today is as low as $120 (see pricewatch).
I have one and I am still tweaking it for best image quality but overall is a very good card (good features, performance, price, value, drivers). I want to keep it for some good time (I hope until XMAS 2003)
Unfortunatelly people will probably have to pay something like $+200 for a GF4Ti4200
Doomtrooper
Veteran
Pascal,
The Ti200s' are fast cards for the money bit IMO you get a more advanced card with the 8500. Improved PS Support, Truform great image quality.
Here is a example of the higher precision range on the 8500 using diffuse bump mapping.
The other question I have is what CHANGED in the UT builds that changed the results so drastically. Anand refers to the fog bug but your link shows the speed of 8500 increased yet he states the 8500 was slower with the fog issue resolved
http://www.anandtech.com/showdoc.html?i=1580&p=7
The Ti200s' are fast cards for the money bit IMO you get a more advanced card with the 8500. Improved PS Support, Truform great image quality.
Here is a example of the higher precision range on the 8500 using diffuse bump mapping.
The other question I have is what CHANGED in the UT builds that changed the results so drastically. Anand refers to the fog bug but your link shows the speed of 8500 increased yet he states the 8500 was slower with the fog issue resolved
http://www.anandtech.com/showdoc.html?i=1580&p=7
Doomtrooper
Veteran
Not too sure whats going on there, the poly bug that I was seeing was actually a texture bug that was fixed.
This is from TB or Thomas's site:
To: thomas@tommti-systems.com
From: John Carmack <johnc@idsoftware.com>
Subject: Re: Radeon 8500 problem - solved?
At 05:43 AM 2/25/2002 +0100, you wrote:
>Hi!
>
>Sorry, for writing you directly, but I've one question:
>
>Did the latest Radeon 8500 beta driver (winxp 6043) solve the problem, you
>mentioned in you last plan update? It works for GL-Excess and other opengl
>tests. ATi told me, it was some mixed texgen issue.
That was one of the conformance problems that I was working around. There
is a crashing problem they are still working on, and the average
performance is still lower than an NV20.
John Carmack
Maybe I'll email JC too
This is from TB or Thomas's site:
To: thomas@tommti-systems.com
From: John Carmack <johnc@idsoftware.com>
Subject: Re: Radeon 8500 problem - solved?
At 05:43 AM 2/25/2002 +0100, you wrote:
>Hi!
>
>Sorry, for writing you directly, but I've one question:
>
>Did the latest Radeon 8500 beta driver (winxp 6043) solve the problem, you
>mentioned in you last plan update? It works for GL-Excess and other opengl
>tests. ATi told me, it was some mixed texgen issue.
That was one of the conformance problems that I was working around. There
is a crashing problem they are still working on, and the average
performance is still lower than an NV20.
John Carmack
Maybe I'll email JC too
It is a new version (856, the other is 848). Probably the 848 was not optimized. Who knows what will happen with future versions?
Anyway, there is no perfect 3D card, but the 8500LE will soon be alone as the cheapest DX8 card available. This is a good opportunity for expand market share, I dont know about money.
Doomtrooper
Veteran
Yes I noticed the build change, there were different Nvidia drivers used too but my problem is the comments from Anand where he states:
This is just wrong comparing the two tests the newer build shows the 8500 picked up 7 fps over the previous build with the fog fixed.
This is just wrong comparing the two tests the newer build shows the 8500 picked up 7 fps over the previous build with the fog fixed.
well if you assume 10-15% perfromance drop between a 8500LE and a retail (excluding overclocking) then the LE sits at around Ti200 performance levels and wll exceed it quite a bit when you enable 8x or 16x aniso.
But for $50 cheaper, the Ti200 is an excellent card.
I'm sure you woud be pleased with either. If the dual-head, superior DVD etc are not that important, and saving $50 is get the Ti200. Will the store allow you to try both and return the one you dont want??
But for $50 cheaper, the Ti200 is an excellent card.
I'm sure you woud be pleased with either. If the dual-head, superior DVD etc are not that important, and saving $50 is get the Ti200. Will the store allow you to try both and return the one you dont want??
Doomtrooper
Veteran
Competition is always good, Matrox show your cards
Hmm am I right that the Radeon 8500LE has faster Aniso but it's quality is worse? Aniso and quality AA are interesting to me especially as I only have a 15inch monitor and my optimum res is 800x600-1024x768 (def not higher). The Radeon has better AA but it's a lot slower. However Nvidia's AA seems to be around Radeon standard with lowest Aniso enabled.Randell said:
I'm sure I'll be pleased with either too
However since it's not my money (mother's) I have to really consider if the Radeon 8500LE is worth the extra money. One feature which may interest me is the TV-out. Is it true that Radeon 8500 cards tend to have better TV out?Not likely. In any case, I may not be getting them from the same store as I may be able to get on cheaper at one store, and one cheaper at another store.
I still have a short while to decide. I'm waiting for the new Epox KT333 board to arrive here. Hopefully it will be within this week.
technically the radeons aniso is a faster version (rip-mapping) which use bilinear as a base and is poor at angled surfaces.
There are various ananlysis of the difference on differnt web sites. I think digit-lige has a reasonable one. Anyway the majority of the time the conclusion is the aniso looks the same.
The reality is you really have to look hard for it to be worried about it. In the majority of case the aniso does the job its intended to do, ie, keep textures sharp and reduce aliaising into the distance. its mainly wall and floors where you notice the trilinear/bilinear blur anyway, and the aniso clears that up wonderfully.
Having come from V5 FSAA, I'm happier with slighly worse AA and aniso than no aniso at all.
There are various ananlysis of the difference on differnt web sites. I think digit-lige has a reasonable one. Anyway the majority of the time the conclusion is the aniso looks the same.
The reality is you really have to look hard for it to be worried about it. In the majority of case the aniso does the job its intended to do, ie, keep textures sharp and reduce aliaising into the distance. its mainly wall and floors where you notice the trilinear/bilinear blur anyway, and the aniso clears that up wonderfully.
Having come from V5 FSAA, I'm happier with slighly worse AA and aniso than no aniso at all.
Hmm.........
In terms of that tho, GeForce3 Ti200 AA with 64tab (or whatever, the best one) Aniso is faster than Radeon 8500LE AA with no Aniso and I the G3 at these settings seems to be at least as good quality as the Radeon 8500 so I would have to say the G3 is probably the better optimisation in this sense. I suppose The Radeon 8500LE AA with Aniso will be better quality altho I dunno of it's noticeable...
In terms of that tho, GeForce3 Ti200 AA with 64tab (or whatever, the best one) Aniso is faster than Radeon 8500LE AA with no Aniso and I the G3 at these settings seems to be at least as good quality as the Radeon 8500 so I would have to say the G3 is probably the better optimisation in this sense. I suppose The Radeon 8500LE AA with Aniso will be better quality altho I dunno of it's noticeable...
Re: Hmm.........
I doubt that to be honest, where did you see that? Unless you are talking quincunx v 4x, which is not comaprable in terms of IQ.
2xAA plus Aniso on a Ti200 in the majority of cases will not be faster than 8500LE 2x no aniso at 1024x768x32 in most circumstances I can think of.
EDIT:
You're talking about the Geforce 4/Unreal Test review where Anand posted 4xAA scores for the LE in the 2XAA table. ANd conviently left off the LE AA+aniso scores. So no deductions can re reliable made from Anands Unreal Test scores. Especially considering no one else knows how it works etc.
Nil Einne said:
I doubt that to be honest, where did you see that? Unless you are talking quincunx v 4x, which is not comaprable in terms of IQ.
2xAA plus Aniso on a Ti200 in the majority of cases will not be faster than 8500LE 2x no aniso at 1024x768x32 in most circumstances I can think of.
EDIT:
You're talking about the Geforce 4/Unreal Test review where Anand posted 4xAA scores for the LE in the 2XAA table. ANd conviently left off the LE AA+aniso scores. So no deductions can re reliable made from Anands Unreal Test scores. Especially considering no one else knows how it works etc.
well
as an owner of a geforce 3 ti 500 and a radeon retail , i have to say that my radeon beats the geforce by at least 5 fps in all my games which is why i passed the geforce off to my friend... I do believe the radeon 8500 was not a finished part and was rushed out...i doubt it was a big diffrence though.
as an owner of a geforce 3 ti 500 and a radeon retail , i have to say that my radeon beats the geforce by at least 5 fps in all my games which is why i passed the geforce off to my friend... I do believe the radeon 8500 was not a finished part and was rushed out...i doubt it was a big diffrence though.
Tom and Anand
No.... I was talking about Anand and Tom's reviews both of which showed 4xAA (and 2xAA) was a lot faster when comparing the GeForce3 Ti500 (with 64tab) vs. Radeon 8500 (non LE). Considering I will probably be able to overclock my G3 to about Ti500 levels and the Radeon 8500LE to Radeon 8500 levels (probably a bit faster, maybe 275/285) the reviews show the G3 is about double. Quality at G3 4x 64Tab seems to be as good or better then Radeon 8500 4xQ at least in the small images they showed.
Altho we can't be sure, I personally would reckon based on these results, the G3 Ti200 unoverclocked will outperform the Radeon 8500LE. Note that it appears the G3 is at least 60% faster on average in these reviews. Maybe it has changed in the new drivers though?
http://www6.tomshardware.com/graphic/01q4/011123/radeon8500-02.html
P.S. Altho I'm not overclocking my Radeon 8500LE as much as my G3 Ti200, it's not relevant if one can overclock more then the other with no issues.[/url]
No.... I was talking about Anand and Tom's reviews both of which showed 4xAA (and 2xAA) was a lot faster when comparing the GeForce3 Ti500 (with 64tab) vs. Radeon 8500 (non LE). Considering I will probably be able to overclock my G3 to about Ti500 levels and the Radeon 8500LE to Radeon 8500 levels (probably a bit faster, maybe 275/285) the reviews show the G3 is about double. Quality at G3 4x 64Tab seems to be as good or better then Radeon 8500 4xQ at least in the small images they showed.
Altho we can't be sure, I personally would reckon based on these results, the G3 Ti200 unoverclocked will outperform the Radeon 8500LE. Note that it appears the G3 is at least 60% faster on average in these reviews. Maybe it has changed in the new drivers though?
http://www6.tomshardware.com/graphic/01q4/011123/radeon8500-02.html
P.S. Altho I'm not overclocking my Radeon 8500LE as much as my G3 Ti200, it's not relevant if one can overclock more then the other with no issues.[/url]
Nil I can only talk from my own experience, but quite frankly I do not believe Tom's numbers in that section of the review, He is claiming a performance hit of 189fps-->66fps with 2xQuality AA? I dont think any other site would replicate that kind of hit.
Its a long time since I used the 7206 driver Tom used and I know the latest beta drivers have a much better OGL ICD in them, but from I think the 9009 drivers I remember benching roughly as follows for Serious Sam First Edition (I dont have QIIIA installed).
1024x768x32, all High quality seetings, Trufrom enabled models etc on a Radeon 8500LE @ 265/265 on an Athlon XP1600/Epox 8KHA+ mobo, 384meg DDRam, Win98SE.
no Aa, no aniso = ~120fps
2xQuality AA, no aniso = ~100fps
2x Quality AA, 16x aniso = ~60fps.
make from that what you will. I'm not trying to talk you out of the Ti200, I'm just saying some of the 8500 numbers I've seen arent accurate.
Its a long time since I used the 7206 driver Tom used and I know the latest beta drivers have a much better OGL ICD in them, but from I think the 9009 drivers I remember benching roughly as follows for Serious Sam First Edition (I dont have QIIIA installed).
1024x768x32, all High quality seetings, Trufrom enabled models etc on a Radeon 8500LE @ 265/265 on an Athlon XP1600/Epox 8KHA+ mobo, 384meg DDRam, Win98SE.
no Aa, no aniso = ~120fps
2xQuality AA, no aniso = ~100fps
2x Quality AA, 16x aniso = ~60fps.
make from that what you will. I'm not trying to talk you out of the Ti200, I'm just saying some of the 8500 numbers I've seen arent accurate.
I keep meaning to run that comparison for SS:Second Edition, but I'm too busy gaming
For SS that 50% hit for both is the biggest I get, normally its lower because CPU's are the limiting factor in a lot of games. Certainly3DMark only shows 20% to 40% hit across the differnt tests.
Note that the Gf3 2x, 64tab hit is 62% (if you trust those numbers as well).
For SS that 50% hit for both is the biggest I get, normally its lower because CPU's are the limiting factor in a lot of games. Certainly3DMark only shows 20% to 40% hit across the differnt tests.
Note that the Gf3 2x, 64tab hit is 62% (if you trust those numbers as well).
Ooops
Actually Tom's findings did seem a bit excessive which is why I said about 60% not 120% (which is what Tom got). 60% is based on Anandtech's review which I somehow missed out including. Here it is:
http://www.anandtech.com/showdoc.html?i=1562&p=8[/url]
Note however Anand was using lowest quality Aniso (8 tab I suppose) not highest quality. Having said that, I still can't see any quality differences between the Radeon 8500 and Ti500 in the screenshots he showed. Oh an yes, I'm talking about 4x not 2x. Although the G3 Ti500 beats the Radeon 8500 in 2x but not by so much and it is possible that a Ti200 may lose to a Radeon 8500LE in 2x (very unlikely in 4x IMHO assuming Anand's findings are still accurate) and of course, a maximum overclocked Ti200 may also lose to a maximum overclocked Radeon 8500LE at 2x
Actually Tom's findings did seem a bit excessive which is why I said about 60% not 120% (which is what Tom got). 60% is based on Anandtech's review which I somehow missed out including. Here it is:
http://www.anandtech.com/showdoc.html?i=1562&p=8[/url]
Note however Anand was using lowest quality Aniso (8 tab I suppose) not highest quality. Having said that, I still can't see any quality differences between the Radeon 8500 and Ti500 in the screenshots he showed. Oh an yes, I'm talking about 4x not 2x. Although the G3 Ti500 beats the Radeon 8500 in 2x but not by so much and it is possible that a Ti200 may lose to a Radeon 8500LE in 2x (very unlikely in 4x IMHO assuming Anand's findings are still accurate) and of course, a maximum overclocked Ti200 may also lose to a maximum overclocked Radeon 8500LE at 2x
