IBM T40p FireGL9000 Benchmark

[mocca]

I just got a new IBM T40p and run some benchmarks on it. Unfortunately, this unit have to go back to IBM. There are few problems with the built quality of this unit which I will not get into the details but suffice to say that it make me feel uncomfortable with IBM QC now. Anyway, I ran some 3DM2k1SE (330) with the following results. Note that the default GPU/Mem for the FireGL is 252/200. I cannot publish the results since the Futuremark say that my CPU change the speed during the tests which is normal for the P-M 1.6GHz. BTW, I used the default FireGL driver from IBM.

At Default (everything stock): 7247
O/C to GPU/Mem (270/218): 7909
O/C to GPU/Mem (274/218): 7952
O/C to GPU/Mem (277/218): 7966

The GPU/Mem of 277/218 seem to be the max o/c I can do without artifacts. For comparison, with ATI M9 32MB (also IBM T40 with P-M 1.6GHz), the default score is about 4900 with max o/c score of about 6000 (this is at GPU/Mem 279/234).

Mocca

Update: I got the maximum score of 8143 (details below). This is with GPU/Mem of 286/218 with no artifacts.

--------------------------------------------------------
Program Version 3DMark2001 SE
Resolution 1024x768 32bit
Texture Format Compressed
FSAA Disabled
Z-Buffer Depth 24bit
Frame Buffer Double
Rendering Pipeline D3D Pure Hardware T&L

Detailed Test Results
3DMark Score 8143 3D marks

Game 1 Car Chase - Low Detail 105.8 FPS
Game 1 Car Chase - High Detail 55.3 FPS
Game 2 Dragothic - Low Detail 121.4 FPS
Game 2 Dragothic - High Detail 67.2 FPS
Game 3 Lobby - Low Detail 132.8 FPS
Game 3 Lobby - High Detail 63.7 FPS
Game 4 Nature 40.9 FPS

Fill Rate (Single-Texturing) 576.8 MTexels/s
Fill Rate (Multi-Texturing) 1128.6 MTexels/s

High Polygon Count (1 light) 22.0 MTriangels/s
High Polygon Count (8 lights) 5.2 MTriangels/s

Environment Bump Mapping 141.6 FPS
DOT3 Bump Mapping 87.7 FPS

Vertex Shader 75.7 FPS
Pixel Shader 114.8 FPS
Advanced Pixel Shader 84.2 FPS

Point Sprite 16.1 MSprites/s

------------------------------------------------------------

 

[esaleris]

Hey, what do you use to overclock, man? I haven't found one that works correctly with the Mobility FireGL yet...

 

[mocca]

Try Rage3d Tweaker from here. Work fine for me with ATI M6, M7, and M9 (including the FireGL9000). Don't o/c too high though. O/c your gpu too high will damage the gfx chip and for the disclaimer purpose: I'm not responsible for any damage that might occur when you o/c your graphic card. Use the program at your own risks. Good luck.

Mocca

 

[BeauJangles]

I didn't think the FireGL was good for gaming?

 

[mocca]

The FireGL9000 have the same core as the regular 128bit M9 but with OpenGL CAD certified drivers and tweaks. It should be as fast as 128bit m9 in games and run stable and efficient with openGL certified programs like ProE or AutoCAD.

Mocca

 

[ES135]

Mocca or anyone -

With regard to the review this website did on the T40/P, Dell 800 etc, it showed that
the Fire GL took major hits in the MM and Gaming test ie compared to the D800 with
the Nvidia card.

Yet, when you look at these GL benchmarks and if its true that the GL will do everything
that the radeon M9( 64/128) will do, then does this mean that if you stack up the Dell's
D800 (Nvidia card) to that of the Radeon M9 with 64/128, the Nvidia blows it away?

Although in the test they used Dell's Nivida which had 64/128 and the GL's 64/128,
in effect your also comparing the nvidia to a radeon m9 with 64/128 if its true that the
core of the M9 is in the GL. Unless the Nvidia has drivers tweaked that the GL did not.

If so, then does the M9 come with these tweaks automatically where the GL does not?
If yes, how can we get the GL up to speed? Does this make any sense?

Happy T40P owner still in 30 day reveiw period.

 

[mocca]

Originally posted by: ES135
Mocca or anyone -

With regard to the review this website did on the T40/P, Dell 800 etc, it showed that
the Fire GL took major hits in the MM and Gaming test ie compared to the D800 with
the Nvidia card.

Yet, when you look at these GL benchmarks and if its true that the GL will do everything
that the radeon M9( 64/128) will do, then does this mean that if you stack up the Dell's
D800 (Nvidia card) to that of the Radeon M9 with 64/128, the Nvidia blows it away?

Although in the test they used Dell's Nivida which had 64/128 and the GL's 64/128,
in effect your also comparing the nvidia to a radeon m9 with 64/128 if its true that the
core of the M9 is in the GL. Unless the Nvidia has drivers tweaked that the GL did not.

If so, then does the M9 come with these tweaks automatically where the GL does not?
If yes, how can we get the GL up to speed? Does this make any sense?

Happy T40P owner still in 30 day reveiw period.

The GF4Go 4200 is faster than the M9 even with 128bit. The FireGL9000 is the same core as regular M9 128bit and hence should have similar performance but with better OpenGL support. The problem with Anand benchmark is that he didn't really do extensive benchmark between the ATI M9 and GF4GO 4200. We need more benchmarks (i.e., with Quake3, 3DM2k1SE, etc.) to see the overall comparison. But at the end, there is no deny that GF4GO 4200 is indeed faster than M9 (if it is not faster, then there is no point for nVidia to release the GF4GO 4200, 6months after they realize the GF4Go 440 cannot compete with M9 128bit, to compete with M9). IMO, GF4Go4200 requires a lot of power and also produce more heat than M9. Therefore, it is not a really good candidate for a thin-and-light laptop such as T40. Look at D800 size and their short battery run time. The D800 have 9cell battery and still have shorter battery run time than the similar T40 by as much as 2-3 hrs. The LCD size do contribute to the short battery size but another major power drain is the GF4Go 4200.

Mocca

 

[mocca]

FYI: the best GF4Go 4200 3DM2k1SE score is around 10000 (just 10-15% faster than the best scores from M9, 8900). The average scores of GF4Go4200 with P4m 2.5GHz (a little bit faster than P-M 1.6GHz) is about 9200 compared to the average scores of M9 128bit of 7500 (about 20% faster).

Mocca