SLI temperature Differential on cards.

[rower30]

I’ve noticed about a 4-5C difference between GPU#1 (monitor card) and GPU #2 (slave card) in SLI with my EvGA 7800GT cards. I don’t know if this differential was always the case because I never looked at the Nview utility until I mounted some new coolers that were selected to be quiet more than “cool”. After installation I checked the general temps to make sure that thing were not WAY of line and within reasonable expectations. They were at 47C on VPU#2 and 52C on CPU#1. But the spread seemed odd.

Has anyone seen this slight non-linearity in GPU temps at all? I’m going to switch the cards around and see if the temps follow the card, or the slot. It may be that the “master” card just does a little more work and runs hotter. If it follows the slot, it seems to say the card is either working harder, or the slot position doesn’t cool as well (not sure which). If it follows the card, I may have goofed on the heat sink installation or…one GPU just isn’t as good a bin as the other. I don’t want to tear a card apart and start over if I can avoid it. I don’t think that 5C will kill me as long as it stays under 65C, anyway.

The coolers? I used the Arctic-Coolers Accelero 1 coolers and they seem to mount and work just fine as far as quiet goes, they are nearly silent. Now if my BFG 650W power supply fan was near that level I’d be set, it’s LOUD! The 52C GPU temp on card #1 and 47C on card #2 after 3Dmark05 seems reasonable, too. I’m just curious what other SLI set-ups see on temperature differentials.

EDITED
I switched the cards between the two slots, and the temperature differential stayed the same. So, it isn’t the GPU’s being different or the mounting of the cooler being different. It is either the upper slot cools more poorly, or the master card (monitor card) is being worked harder.

At idle it’s GPU #1 @ 48C and GPU @42C with the cards switched places, which is the same as it was before. The cards are the same, the slot is different for cooling or GPU loading reasons.

 

[delanooch]

Yes this is normal. I was thinking of adding one of those Vantec slot fans between my cards but the 2- 120mm blowing on my cards from the side panel seem to do the trick.

 

[Herr_Fritz]

I've noticed the same thing on my rig and I attributed it at the time to the master card being above the slave. As the heat from the lower card rises and heats the upper slightly. Though I have no idea if this has any basis in fact.

Fritz

 

[rower30]

I don't think that the "heat" from the lower card is the issue. The coolers I run are too good to show that much differential between the card temps. I see 65C on "master" and 52C on the "slave" card in extrended game play.

I'm not knocking the temps, mind you. I've seen X1900XT's tested at almost 90C which is insane.

My guess is that the GPU's are load "balanced" on just the 3D rendering, and the master card still compiles all the rendered data into a composite or digital signal to your monitor not to mention, one card has to keep track of who's on first between the two cards. So the master card is still working harder than the slave card which must be chuncking rendering only.

Does anybody know the block diagram for SLI mode 3D set-ups? This would answer the question as to why one card runs 12C hotter than the other under extended game play, and cooling can't possibly account for the difference.