First (real) GMA X3000/G965 Benchmarks

I can't seem to connect to that site to download those drivers.
However I'll go ahead and try to test the 14.21 drivers, I'm currently at 14.10 (I haven't really had a chance to test the card because I've been too busy to play games).

BTW, my driver info for GMA950 shows 14.10.4670. According to Intel, that's actually 14.24 for GMA950.

You can see from the Readme file. It says:
Graphics: 14.10.4670
HDMI Audio: 5.10.0.1014

(Not sure why they show audio driver version as it should be a seperate driver, not that important for us anyways)

14.21 is actually 6.14.10.4624.

I suggest you test the older versions of 3D Mark like 2001 or 2003. The scores aren't important now as the drivers aren't complete, and if you can't run 06, run 05 or 03. Check out what the fillrate for the card is on the test though, it should be same across all 3Dmarks(since its measuring the same thing).

Hope that helps.
 
Well, I guess I already downgraded my drivers...haha. Oh well.

Here are the scores, with 5 tests
CPU1 - Red Valley 0.484 FPS CPU Tests
CPU2 - Red Valley 0.756 FPS CPU Tests
HDR1 - Canyon Flight 0.000 N/A Not supported HDR/SM3.0 Graphics Tests
HDR2 - Deep Freeze 0.000 N/A Not supported HDR/SM3.0 Graphics Tests
Fill Rate - Single-Texturing 382.661 MTexels/s Feature Tests
Fill Rate - Multi-Texturing 1182.539 MTexels/s Feature Tests

Pixel Shader 15.862 FPS Feature Tests
Vertex Shader - Simple 13.398 MVertices/s Feature Tests
Vertex Shader - Complex 5.143 MVertices/s Feature Tests
Shader Particles (SM3.0) 0.000 N/A Not supported Feature Tests
Perlin Noise (SM3.0) 0.000 N/A Not supported Feature Tests
8 Triangles 0.010 MTriangles/s Batch Size Tests
32 Triangles 0.041 MTriangles/s Batch Size Tests
128 Triangles 0.163 MTriangles/s Batch Size Tests
512 Triangles 0.641 MTriangles/s Batch Size Tests
2048 Triangles 2.450 MTriangles/s Batch Size Tests
32768 Triangles 12.044 MTriangles/s Batch Size Tests

CPU score was 1511
 
So I am guessing you are on 14.21(6.14.10.4624)?? Or is it 14.24(6.14.10.4670)?? Mind if you test Quake 3 Arena Demo on 1024x768 high quality??
 
Fill Rate - Single-Texturing 382.661 MTexels/s Feature Tests
Fill Rate - Multi-Texturing 1182.539 MTexels/s Feature Tests

My system:

Celeron D 2.53GHz
Intel GMA 950
1GB single DDR2-533
WD 360

I got ~1500MTexels/s in both Single and Multi-Texturing, which is close to theoretical maximum of
1.6GTexels/s. That's equal to 400MHz clock with 4 pixel pipeline and 1 texture units per pipeline.

G965, according to your tests, sounds like there are 3 texture units per pipeline. The fill rate is rather low, as even with 1 pixel pipeline it should achieve close to 667MTexels/s in single texturing.
There could be two reasons:
1. Immature drivers don't expose the full performance
2. Intel doesn't use zone rendering on G965, since zone renderers achieve close to 100% maximum fillrate.
 
I was testing with the older drivers 14.21 (14.10.4642) And yes, it's 42.
My RAM is 512 DDR2-533
OS: WinXP Home
I did do it at 1024 res.
As far as the Quake III goes, I'll have to wait a few days...I'm rather busy for at least the next day or two.
 
I appreciate what you are doing TraumaPill. Now if I can get my hands on a G965 board that would be sweet.
 
My system:

Celeron D 2.53GHz
Intel GMA 950
1GB single DDR2-533
WD 360

I got ~1500MTexels/s in both Single and Multi-Texturing, which is close to theoretical maximum of
1.6GTexels/s. That's equal to 400MHz clock with 4 pixel pipeline and 1 texture units per pipeline.

G965, according to your tests, sounds like there are 3 texture units per pipeline. The fill rate is rather low, as even with 1 pixel pipeline it should achieve close to 667MTexels/s in single texturing.
There could be two reasons:
1. Immature drivers don't expose the full performance
2. Intel doesn't use zone rendering on G965, since zone renderers achieve close to 100% maximum fillrate.
3DMark can't and doesn't measure single-textured fillrate. It displays something with that label, but that should be completely ignored if you want to keep your sanity.

Fun debate on the topic:
http://www.beyond3d.com/forum/showthread.php?p=705730#post705730
 
3DMark can't and doesn't measure single-textured fillrate. It displays something with that label, but that should be completely ignored if you want to keep your sanity.

Fun debate on the topic:
http://www.beyond3d.com/forum/showth...730#post705730

So according to what they are talking about the reason single texturing in GMA 950 is so close to theoretical fillrate is its not bandwidth limited?? Or does it have to do with something else??

Shared 4.2GB/s bandwidth with CPU is enough for GMA 950?? Interesting.
 
G965, according to your tests, sounds like there are 3 texture units per pipeline.
Why? The score sounds perfectly fine if there were 2 texture units overall.

zeckensack said:
3DMark can't and doesn't measure single-textured fillrate. It displays something with that label, but that should be completely ignored if you want to keep your sanity.
Well, it's not really measuring single-textured fillrate, but the score is low nonetheless. Some quick math tells me it's a score I would expect with about 3GB/s bandwidth. Maybe ok if that's a single-channel DDR-533 configuration, but otherwise that score is really low.
 
Well, it's not really measuring single-textured fillrate, but the score is low nonetheless. Some quick math tells me it's a score I would expect with about 3GB/s bandwidth. Maybe ok if that's a single-channel DDR-533 configuration, but otherwise that score is really low.
It does sound like he's using single channel DDR2-533

I was testing with the older drivers 14.21 (14.10.4642) And yes, it's 42.
My RAM is 512 DDR2-533
OS: WinXP Home
I did do it at 1024 res.
As far as the Quake III goes, I'll have to wait a few days...I'm rather busy for at least the next day or two.

Single channel can do around 3.5-3.8GB/s on DDR2-533 measured I think with sandra buffered.
 
Why? The score sounds perfectly fine if there were 2 texture units overall.

Because Intel's previous gen graphics reach very close to theoretical in single and multi. I get 1.5Texels/s out of max 1.6GTexels/s with my GMA950, using only single channel DDR2-533.

Either it tells its unoptimized drivers, or this gen is more bandwidth limited, leading to another possible conclusion: It may not use zone rendering technology(at least for now).
 
Yes, I only have single channel DDR2-533. I also apparently downgraded my drivers due to Intel's ridiculous driver numbering and the fact that they don't have the latest drivers posted on their website. Intel seriously needs to get their shit together. Anyone have any ideas why 3DMark would've crashed though? I'll go ahead and try to see if I can get an older version, and I think I may try the QIII demo test here in a bit.
 
Yes, I only have single channel DDR2-533. I also apparently downgraded my drivers due to Intel's ridiculous driver numbering and the fact that they don't have the latest drivers posted on their website. Intel seriously needs to get their shit together. Anyone have any ideas why 3DMark would've crashed though? I'll go ahead and try to see if I can get an older version, and I think I may try the QIII demo test here in a bit.

You gotta go from Support & Downloads section up at the tab not from your mainboard page. The motherboard page aren't updated as often. Then you go to Browse Product at the left, then go find Graphics section, and you can see G965, etc. You can then see the latest is 14.24.
 
Results of the Quake III Arena benchmarks for the GMA965:
With the old drivers, demo001 gets 75.9 fps
but I had it set at 640x480.
So for the new drivers, it actually got worse...
74.2 fps at 640x480
and 52.2 fps at 1024x768
I think my low allocation of video memory may be hindering performance though.
 
Run 3dmark2001se on it so I can compare it to the score I got with a GMA950.
 
Performance of GMA X3000 on windows drivers, version 14.25

http://www.forum-3dcenter.org/vbulletin/showthread.php?t=321049&page=4

Google translation:
G965: “a diagram driverâ€￾ approved yet by Intel. “supports the numerous hardware featuresâ€￾. In the comparison to the old driver the values are improved around approximately 10-15%. Only Doom 3 jumps from 4 fps to 10 fps. Benchmarks in 1024x768.

E6400, 2x 1 GiB 6400-555.

Asrock Conroe945G-DVI / Asus P5B-VM
3dMark 2003 Patch 350: 2101 / 1760
3dMark 2005 V 1.2.0: 717 / 878
UT2004 Botmatch dm-ranking: 41 / 37
Quake 3 high detail: 139 / 107
Doom 3 medium: 7 / 10
Windows Vista Grafik für Aero: 3,2 / 3,9

Performance Testing by Phoronix on the Q965(GMA 3000):
http://www.phoronix.com/scan.php?page=article&item=571&num=1

Performance is rather good on the Open Source Linux drivers on UT2003. It outperforms ATI X300SE.

Personal Blog which compares GMA X3000 on Linux on UMark: http://deadmoo.com/articles/2006/09/28/intels-new-onboard-video-benchmarked
http://deadmoo.com/articles/2006/09/28/intels-new-onboard-video-benchmarked

Could we have something here?? :D.
 
Back
Top