9262 SLI Upgrade Surprise

I added a second 8800m GTX card about a week ago, and I would have posted my general impressions and benchmark results at that time had I not encountered some strange results. My 3Dmark06 results were OK but not great (I expected to see a 50% increase in SM2/3 scores rather than the 45% I actually got), and the system ran smoothly and rendered well in all games.

When I began benchmarking games, however, I began to see something peculiar. The percent FPS difference between non-SLI and SLI was GREATER at lower resolutions than at higher resolutions, just the reverse of what would be expected! The biggest offender was Crysis which only increased 30%!

After a few days of thinking, I began to believe that this result might be a typical symptom of a failed, or missing, bridge cable, so this morning I opened the machine up for a look. The cable was connected, but I had installed it backwards!!! I flipped the cable, buttoned things up, and now I'm not embarrassed to post the benefit I realized from the second card.

3DMark06 SM2/3 scores went up 52%. Crysis benchmark scores at 1900x1200 high settings went up by 70%, and (this was a surprise to me-must have been the decrease in sparks ) GPU temps ran 5 degrees cooler. Here are some scores with the 174.82 driver (174.31 runs a little faster, but I prefer the later one):

Non-SLI

3DMark Score 9801
SM 2.0 Score 4352
SM 3.0 Score 4177
CPU Score 2689

SLI

3DMark Score 13367
SM 2.0 Score 6297
SM 3.0 Score 6634
CPU Score 2700

Crysis at 1900x1200 High Quality

Non-SLI

!TimeDemo Run 1 Finished.
Play Time: 104.91s, Average FPS: 19.06
Min FPS: 13.97 at frame 151, Max FPS: 21.99 at frame 1013
Average Tri/Sec: -17567498, Tri/Frame: -921510
Recorded/Played Tris ratio: -0.99

SLI

!TimeDemo Run 1 Finished.
Play Time: 62.02s, Average FPS: 32.25
Min FPS: 9.93 at frame 147, Max FPS: 38.67 at frame 83
Average Tri/Sec: -22977790, Tri/Frame: -712561
Recorded/Played Tris ratio: -1.29

 

glad it didn't fry the cards

 

hehe-you are lucky!

 

I'm so jealous. A 3D Mark of over 13,000...that so sweet.

 

I guess I am going to get the second card now for my 9262. Those are great results.

Could you post a pic of the correct cable installation?

 

In order to take a picture of the cable I would have to take the cards out again, and I'm reluctant to do that for fear of screwing something else up. However, if you look at step nine of the following guide you will see how the cable should be oriented: http://pcmicroworks.com/Drivers/Black-Hawk/XR5_SLI_Upgrade.pdf

Note that the procedure for the Sager is different from this guide, and much easier. With the Sager, cards can be removed/installed without removing the heat sink, and the upgrade card will arrive already enclosed in its heat sink.

Just as a heads up: My kit came without the three screws required for the second fan. I "borrowed" three identical screws from the back cover to finish the job, and XoticPC promptly sent me the three I was missing.

Good luck...

 

WOW, Temps were 5C cooler?!?!? Great setup BTW

 

Wait 4 the 9800MGTX. With SLI u can get 19XXX 3dMarks......

 

Will the D901C require a mainboard update to be compatible with the 9800m GTX or is there no official information out on it yet?

 

There is no official word yet.

But if Nvidia releases the 9800M GTX with the same low level config/requirements for the MXM-IV slot, then there should be no motherboard update required.

 

How do you know this?

 

Because the 9800GTX is out for the desktop and thats what he is looking at.
Mobile cards never do as good.

8800GTX in sli on a desktop can score 18k+ and we only see what? 13k?

 

mobile 8800M GTX (G92) is almost equivalent to desktop 8800 GTS (G92)

so the score is about right.

 

^are you sure?