9500->9700 mod successful - but slow performance

[Gabbage]

Last week I ordered a Radeon 9500 from Connect3D.com with the intention of soft-modding it to a 9700. The card is a non-Pro model with a 256-bit memory bus and memory in the L-shaped configuration. In other words, it's the perfect candidate for softmodding. When I received the card yesterday I applied the hack by installing the Omega drivers and ran 3dmark2001 and the Radeon 9700 demos from ATI to test for artifacts. Neither gave me any problems. I am therefore positive that the softmod was successful.

However, I have a problem. For some reason (and I can't think of a single explanation for this) my frame rates seem hardly improved over my previous card, a GeForce 2 Pro. My 3dmark scores only increased from 3486 with the GeForce to 7214 with the hacked 9500, and the second demo (Car Chase #2) runs at about 25-30 fps. My frame rates in Unreal Tournament 2003 are simply mediocre (approximately 30-45 fps at 1024x768) and in FIFA 2003, a game that ran perfectly fluidly with AA and AF enabled on a Thunderbird 900 with GF4 Ti4600, my frame rates do not hit anything approaching 60. I have a reasonable system, a Tbird 1.33GHz with 256mb of DDR RAM, so it's hard to imagine that my 9500 is limited by existing hardware. Even if the hack did not work (and I am nearly positive that it did), I think a standard 9500 non-pro should give me better performance. As it is, I have a hacked 9500 that works but provides - by my estimate - GF3 Ti200-level performance.

Anyone care to offer an explanation? Any help or suggestion would be greatly appreciated.

 

[jiffylube1024]

Your platform (Tbird 1.33 GHz, SDR memory) is most likely holding it back. Although the Tbird is still a decent chip, that's an antiquated platform by today's standards with it's PC133 Memory.

Today's platforms have much more system/memory bandwidth, with DDR up to 400 Mhz (200 doubled), And 2+ GHz processors, which can pump those GPU's with data much faster.

 

[Gabbage]

I understand that video performance is directly proportional to CPU speed. However, I do not have anything close to an obsolete system. Agreed, a T-bird 1.33GHz with 256mb of RAM may not be able to draw maximum performance from a Radeon 9700, but it is not so completely out-of-date that I shouldn't still be seeing a major boost in performance.

As I said, I have another system that has a T-bird 900MHz paired with a GF4 Ti4600 and it comprehensively outperforms my current setup. For example, I can run FIFA 2003 on the T-bird 900MHz at 1600x1200 without ever seeing a dip in framerate but I am not even getting decent performance at 1024x768 in the same game with my newer system. Even if I have a CPU bottleneck its influence should not be so pronounced.

 

[Gabbage]

At a friend's recommendation I downloaded 3dmark03 and ran the tests to see how my results compare with his. His system is a P4 2.4GHz with 512mb of RAM and a Radeon 8500. I have broken down our scores here and posted them side-by-side for comparison. The scores on the left are mine.

Game tests (3dmark score: 2490)

-GT1 - Wings of Fury - 61.6 fps / 72.3 fps*
-GT2 - Battle of Proxy - 16.6 fps / 7.5 fps
-GT3 - Troll's Lair - 15.2 fps / 8.5 fps
-GT4 - Mother Nature - 18.4 fps / not supported

Feature tests:

-Fill rate (single-texturing) - 865.7MTexels / 648.6MTexels
-Fill rate (multi-texturing) - 1091.3MTexels / 1907.4 MTexels*
-Vertex shader - 13.0 fps / 9.3 fps
-Pixel Shader 2.0 - 25.1 fps / not supported
-Ragtroll - 11.6 fps / 4.1 fps

As one would expect, my 3dmark score (2490) was much higher than his (1202). Interestingly, however, his multi-texturing fillrate was much higher (1907.4 > 1091.3) though not his single-texturing fillrate (648.6 < 865.7). I am frankly confused by these results because fill-rate is supposed to be a forte of the Radeon 9700.

What do you make of these results? Do they make sense?

 

[modedepe]

Your card might not have modded. A ti4600 would kill a 9500np in games if the 9500 didn't mod. And the reason his 3dmark03 score is so low is because he can't run as many tests as yours can, for the simple reason that yours is a dx9 card and his is a dx8. If the mod had worked your 3dmark03 score should probably be higher, even with the older system, since 03 is a lot less cpu bound than 01.

 

[Ashen Shugar]

Fill rate (multi-texturing) - 1091.3MTexels / 1907.4 MTexels*

There's your problem... The mod did not work. Try reinstalling the drivers. I've had a modded 9500 on my system (P3-1500, 512MB RAM) before, and after applying the softmod, saw the multi-texturing fill rate go from ~1000MTexels to ~2000 MTexels.

The multitexturing fillrate should double after applying the mod, so you could perform that test before and after using the softmod to make sure whether it was applied successfully or not. Hope that helps.

-Parin

 

[McArra]

3dmark 2001 anyway may not increase a lot maore. My PC (P4 1400) coupled with my 9700 Pro wasn't able even to reach 10k. 3dMark score should be around 4000+ with the softmod succesfully made as 3dmark03 is VERY graphic card deppendent.

 

[justagamer]

 

[justagamer]

i suggest downloading the latest rivatuner(v2), and hack the drivers using this program. i remember also once downloading "apparently" hacked drivers, but they did not do a thing....., with this program you can literally see the "4" become a "8" when you have enabled the extra 4 pipelines. also i suggest running a benchmark preferably 2003 as this is dx9, before and after enabling the pipelines (remember to keep v sync off). i have a powercolour 9500np, and have hacked it and it increased performance dramatically (from 2700 to 4200 marks for 3dmark2003)...,
get rivatuner at www.guru3d.com.
let me know if you need any further help