Radeon 9800 Pro preview
- Thread starter Ante P
- Start date
Again (don't people even read the thread, let alone the reviews
) our review used the 9800 drivers and clocked the 9800 down to 9700 PRO speeds and showed pipeline improvements in some cases.That's interesting to see. I wonder how it will play out in the long run?
DaveBaumann said:Again (don't people even read the thread, let alone the reviews
I'm talking about using the hacked 9800 drivers on a 9700! Geez.
Errrr, so am I!!!
Slides said:
You don't need to "hack" any drivers. They R300 and R350 both use the exact same files and configuration. Actually the driver inf handles the 9800 just like it handles the R300. They don't even have separate settings in the inf: they're both called R300.
The 9800 is about 5-10% faster when using the same clockspeeds AND the same drivers.
Then what is this all about:
http://warp2search.net/article.php?sid=11098&mode=thread&order=0
Test System:
Athlon XP 1800+
Radeon 9700 Pro
Windows XP SP1
Results:
Fablemark [4%]
3DMark 03 | 1024x768 Default | [3-5%]
3DMark 03 | 1024x768 16x AF | [up to 20%]
3DMark 03 | 1024x768 2x FSAA | [up to 10%]
3DMark 03 | 1024x768 4x FSAA | [up to 10%]
3DMark 2001 SE | 1600x1200 | [Pixel Shader 1.1 40%; generally faster]
3DMark 2001 SE | 1600x1200 16x AF | [generally faster]
ATI Island Demo 1 [partially 19% faster]
Shader Mark 1.7 | DX8 | [generally faster]
Shader Mark 1.7 | DX9 | [even more faster]
Shader Mark 1.7 | DX9 PS2.0 | [up to 23% faster]
Z-Reject Overdraw Tester 1.6 | [up to 4% faster]
http://warp2search.net/article.php?sid=11098&mode=thread&order=0
Test System:
Athlon XP 1800+
Radeon 9700 Pro
Windows XP SP1
Results:
Fablemark [4%]
3DMark 03 | 1024x768 Default | [3-5%]
3DMark 03 | 1024x768 16x AF | [up to 20%]
3DMark 03 | 1024x768 2x FSAA | [up to 10%]
3DMark 03 | 1024x768 4x FSAA | [up to 10%]
3DMark 2001 SE | 1600x1200 | [Pixel Shader 1.1 40%; generally faster]
3DMark 2001 SE | 1600x1200 16x AF | [generally faster]
ATI Island Demo 1 [partially 19% faster]
Shader Mark 1.7 | DX8 | [generally faster]
Shader Mark 1.7 | DX9 | [even more faster]
Shader Mark 1.7 | DX9 PS2.0 | [up to 23% faster]
Z-Reject Overdraw Tester 1.6 | [up to 4% faster]
Lets take this S-L-O-W shall we....
You see the drivers that came with the Radeon 9800, Yes? I took those drivers and forced an install on to a Radeon 9700 PRO - OK? So, all the benchmarks you see in that review use the driver for the 9800, OK?
Now (this is where thing get tricky....) Not only did I use the same drivers, but I also down clocked the 9800 to 9700 PRO levels to see what improvments there were, OK?
You see the drivers that came with the Radeon 9800, Yes? I took those drivers and forced an install on to a Radeon 9700 PRO - OK? So, all the benchmarks you see in that review use the driver for the 9800, OK?
Now (this is where thing get tricky....) Not only did I use the same drivers, but I also down clocked the 9800 to 9700 PRO levels to see what improvments there were, OK?
And let me make take this even SLOWER for you.
The meager performance difference between the 9700pro and the 9800pro at the same clock speed and using the same drivers means that the 9800pro is essentially an overclocked 9700pro with some enhancements.
Meaning, that I (not you or anyone else) will get a much better deal with a 9700 using the 9800 drivers. Get it?
So before you get all bitchy again, this is my personal opinion on what I should buy.
The meager performance difference between the 9700pro and the 9800pro at the same clock speed and using the same drivers means that the 9800pro is essentially an overclocked 9700pro with some enhancements.
Meaning, that I (not you or anyone else) will get a much better deal with a 9700 using the 9800 drivers. Get it?
So before you get all bitchy again, this is my personal opinion on what I should buy.
DaveBaumann said:Lets take this S-L-O-W shall we....
You see the drivers that came with the Radeon 9800, Yes? I took those drivers and forced an install on to a Radeon 9700 PRO - OK? So, all the benchmarks you see in that review use the driver for the 9800, OK?
Now (this is where thing get tricky....) Not only did I use the same drivers, but I also down clocked the 9800 to 9700 PRO levels to see what improvments there were, OK?
Dave seems as if Slides is referring to what Thomas found out. He's made the new drivers somehow use a different path in the 6307 drivers.
Thomas did not modify the inf but the actual drivers files. Apparently this increased performance quite a bit. Enough to rule out that the 9800 is really any more efficient at anything.
"Note that the number in brackets shows the approximate performance difference to standard R300 path."
Since ATi's memory controller is "programmable" perhaps this "optimized" stuff is actually referring to how software optimizations?
Dave, the issue is that the ATI driver internally seems to lower the R300 scores in order to show a bigger R350 advantage. Look carefully at this link :
http://www.tommti-systems.de/main-Dateien/reviews/catalyst/catalyst.html
and you'll understand the trick...
Edit : me too slow ! Ante beat me and his explanation is better
http://www.tommti-systems.de/main-Dateien/reviews/catalyst/catalyst.html
and you'll understand the trick...
Edit : me too slow ! Ante beat me and his explanation is better
I'm not saying that isn't that case - its just that you didn't exactly seem clear what the results in the test were actualy representing.
We all knew that R350 was an improved R300, thats not in dipute - its just a refresh product. And if you actually note the conclusion of the review I point out that its hardly worth the upgrade for current 9700 PRO users.
aaXiom3D said:
I dunno if I'd phrase it like "lower the scores" rather they don't apply the same optimizations.
I sure hope they frikkin do.
I'd be pretty mad if I found out that ATi are cheating me on performance just to make their latest product look good.
This also makes large parts of previews obsolete. I'm gonna see if Thomas wants to share the trick, that way I can redo the benches to see if all the differences I saw was just becuase the driver was using different paths...
I also e-mailed ATi and asked them to comment this real quickly, people probably won't be too happy when they hear about this.
I didn't mod th inf either, but the 9700 PRO was installed as 'Radeon 9800 SERIES' which would clearly indicate to me that it was operating on the same 'path' as a 9800.
