Stuttering 4way SLI?

[Gamer]

Just installed a 4way SLI GTX680 on the Maximus V Extreme.

Came from 3way SLI where my Unigene score was 5500 stock on 3770k at 4500mhz.

 

In 4way my score is 4250.

Noticed some heavy stuttering during the bench.

I think this is not normal?

 

1200watt and 1000watt PSU.

Solutions?

 

Installed 3D01 on same system, max fps is 180.

http://www.hwbot.org/submission/2304105_zupernico_3dmark2001_se_4x_geforce_gtx_680_18026_marks

 

Looks normal according to this result, can't be right no?

Edited September 5, 2012 by Gamer

[Massman]

Rampage V Extreme?

 

During pre-testing MVE and 4x 680 Sam and me noticed something similar. In 3DMark11, GT2 and GT3 would throttle and physics would hang the first second making us lose 1000pts in that subtest.

[K404]

Is that "the way it is?"

 

BIOS? BCLK? PCI-E gen? System memory settings?

 

 

http://hwbot.org/submission/2304106_zupernico_3dmark03_4x_geforce_gtx_680_28960_marks

 

Board/BIOS or GPU drivers IMO

Edited September 1, 2012 by K404

[TaPaKaH]

(talking about Z77 M5E)

We had the same only going below certain temperature (-80 or so), above that all was well.

Fiddling with PCIe Gen settings got us issue-free .. but only for one run, after that not even elmor could figure it out.

[subaruwrc]

so low 3d perf maybe not bios related. I've tested an mve some days ago and it gave me a bad 3d performance ( 230fps 3d05 gt3 at 6.3ghz cpu with gtx 580 .. ) low score in 3d06 etc. I had some serious problems with board not giving signal to the monitor after an hour or two under subzero.. strange

[Gamer]

Same trouble on X79-UD7 OC board.

Must be driver related I think.

[Massman]

Spoke with KP about this yesterday - seems to be driver-related indeed.

 

Do you get the issue if you stay below 5GHz CPU frequency?

[rsnubje]

Is this on every driver? I can remember when I tested 2x GTX 690 I had a lot of stuttering with some drivers, but the least/none with 304.48, or maybe 304.79. Don't know about 306.02. These are beta's tho, not sure if you want/can (to) use these for your WR's

[K404]

I'm surprised a problem as big as this was allowed to leave the lab

 

To make £1400+ of GPUs perform like £250 is NOT COOL.

[Massman]

I don't think there's anyone in Nvidia doing any validation testing for 4way sli. There's not much people using it anyway...

[Gamer]

Spoke with KP about this yesterday - seems to be driver-related indeed.

 

Do you get the issue if you stay below 5GHz CPU frequency?

 

Yes, 4500mhz same issue.

Official drivers used, no beta yet.

[K404]

I don't think there's anyone in Nvidia doing any validation testing for 4way sli. There's not much people using it anyway...

 

Chicken and egg. Does no-one use it because it sucks, or does it suck because no-one uses it?

 

nVidia advertise this. Every time they drop the ball on something that costs a user so much money, a user goes over to ATI.

 

 

Dear nVidia. I would like a job validating your tri-SLI and Quad-SLI driver performance.

[Gunslinger]

I'm sitting on 4x 680 Classifieds but am feeling pretty unmotivated to bench them at the moment.

[Massman]

Last week Sam and me tested the M5E with 4x 680 and we noticed some weird performance issues when increasing the CPU frequency. I've attached some screenshots:

 

- 27647: reference run with SB-E at 5.2G

- 25499: ivy run at 5G

- 26855: ivy run at 6G

 

There's two performance issues:

 

- GT2/GT3

- Physics

 

For GT2/GT3, for some reason they are lower at 5900 than they are at 5000. Cards were clocked the same and there was no crash or anything. I've seen the same on GT1 in a different run: the FPS is just lower from the start. 4way was always enabled, though.

 

Physics at 6G should be around 16k, so I'm losing about 1K. The reason is that when the test starts, it'll hang for the first second before actually starting. When 4way is disabled (so single gpu), the problem is solved. Didn't have any issues with 4way and physics subtest when running on air and also had it working once at 6G when switching from gen3 pcie to gen2 pcie in bios. That trick only worked once, though. Later I had the same issue and switching pcie gen didn't fix it anymore.

 

We reinstalled the drivers too, but it didn't fix the problem.