Anyone else with 2xSLI 970 FTW noticing any asynchronous data between the cards? I speak mainly in regards to the power usage (and its limit report) and the running voltage (and its limit report).
 
The attached screenshot is the MSI Afterburner data regarding two Fire Strike Ultra runs. One on stock and one with an overclock (+87mv, 110 power, +100 core, +250 memory).
If you take a look at the power and voltage data, you'll notice a few things:
- The "limit" will pass from 0 to 1 (to report that it has been reached) in moments where said limit hasn't been reached (this is mostly noticeable in GPU 1 - left column).
- The data isn't synched between the two cards, meaning voltage and power usage isn't synched.
- GPU2 never touched 110 power usag e, but you can its limits chart showing a few instances where the limit had been reached.
 
What I seem to notice so far, apart from the asynchronous data:
- GPU 1 limit charts report limits reached for voltage and power usage before the limits are reached. It gives the idea that despite the power limit being set to 110, the limit chart flags the limit being reached the moment the card passes the 100 power usage mark. Same goes for voltage.
- GPU 2 seems to report voltage and power usage limits more accurately, but occasionally also flags limits before they are reached.
- While benching on air, I keep fan speeds at 100% at all times. I noticed that GPU 1 usually runs between 10ºC to 20ºC higher than GPU 2, something expected I think, since they're in PCI-E slots 1 and 3.
- The cards rarelly (if ever) run at the same voltage, despite being the exact same model, having a similar ASIC (around 68%) and having their serial numbers 7 numbers apart.
 
Currently, I'm not sure if I should install the watercooling blocks on them and try to adjust the BIOS to avoid having the variable speeds and see if both voltage and power usage stabilize on the same values, or, if there's actually something wrong with one of the cards and installing the watercooling block would just give me double work for an RMA. =)
 
Is everyone running SLI experiencing the same? Is this asynchronous data normal?

Yeah, before running them SLI, I tested them individually, and the temps where better than the hotter card in SLI, this should indeed be normal. The only reason I didn't install the waterblocks yet, is really because first I want to be sure the cards are A-OK. Both of them have coil whine on menus, but didn't notice it yet in games; though I'll be able to judge better once they're on water.
 
One thing I forgot to mention (but I think this is normal); while testing both cards individually, either of them could do +100 core and +500 memory. However, SLI is not all that stable with those values. I'm hopping the WC blocks and a BIOS to fix the fluctuating clocks and voltage help stabilizing those values in SLI. =)
 
I look forward to your findings!

 Awww, there's always tomorrow! Good luck tackling the UPS dude. =)
 
And here's another one, before I call it a day. This time, it's a Fire Strike bench (stock settings, and result here: http://www.3dmark.com/3dm/4686084?):

Apart from the usual asynchronous volt/power limits reported between the two cards; another curious fact is GPU1 reporting the power voltage limit below it (the stock limit is 1.212v), and while being below GPU2 voltage (1.187v vs. 1.193v). GPU2 shows its higher accuracy again, not flagging a limit where no limit has been reached.
 
Regarding power limit; during this round, even though GPU2 reported it less times, some of the limits reports where when GPU2 power was at 94%. GPU1 reported more power limits, but the earliest reported was at 96%. Earlier in Unigine Valley though, GPU1 was the one triggering the limits 6% earlier than it should.
 
Note: I'm still running drivers 344.48, as 344.60 introduced a lot of stuttering, and therefore I won't be bothered with 340.65, or any other BS game enhancement driver.

Hi,
 
I experienced the same issues like you. So I installed my waterblocks (same like you) as well and use a modded BIOS with raised powerlimit and unlocked voltage up to 1,275V now. With it the cards run stable and sync @ 1531/4005 MHz. Powerlimit is at 245 W and they still throttle in benches but not in games ;-)
 
EDIT: ASIC is 72 % and 73,1 % and my serials are in a line.

Hi,
 
with the modification I did on the BIOS (made by myselft) the cards run sync and at mx clock and voltage without the need of any OC software right out of the box. I'll attach it if you're intersted. As you have double BIOS it's worth a try ;-).
 
Good luck...
 
https://www.dropbox.com/s.../evga_ftw_mod.zip?dl=0

Nice to hear that it worked for you. Of course the temps are higher due to the increased voltage. For 24/7 safety I can't guarantee... I raised the total powerlimit to 250W. My cards throttle at about 95-98 % (about 240 W) anyway and to reach that I let the BIOS draw 80 W from the PCIe slot. With letting it at standard (66 W and 75 W max) they throttle at 215 W. 240 W is quite a lot for a 4 phase design. It would be nice to get an official statement from EVGA. Without the mod the cards do not clock like they should with the "FTW" mark in my opinion...
 
Greets