Gigabyte 570 rev 1&2 SLI

barnik

Honorable
Sep 26, 2012
7
0
10,510
Welcome T'sH comunity!

I am writting to you because I've recently bought second gpu (preparations for Cry3), as written above, gtx570 but to my surprise it happened to be revision 2.0 which PCB is slightly different than 1.0 (identical as reference 570).
Now to the problem - I get extremely large stutering, not sure if it is my PSU (be quiet system power s6 700W - should be enough) OR the differences in PCB schematics.

Can anyone help me pls? Its ridiculous that 4 example BF3 run better on single gpu than on sli.

Edit: I am monitoring gpus with msi afterburner which shows gpu loads drop and top very often, fps readings never affected...
 

Bejusek

Distinguished
It seems that PSU may be hitting its limits.
18A on each +12V rail can be insufficient.

Try something more powerful, 20A for each card at least. I think that the best choice is to buy a PSU with single +12V rail.
 

barnik

Honorable
Sep 26, 2012
7
0
10,510
18A you mean 18 for each cable (8/6 pin) yes? Do I have to buy specific cables with 20a or is it a matter of psu type?
 

Bejusek

Distinguished
Yes, your PSU has four +12V rails, each capable of delivering 18A of current.
18A @ 12V = 216W and GTX570 is likely to consume 210-220W. That's why I think your PSU is not enough.

New cables won't change anything, because this PSU is not capable of delivering more current.

The only thing, apart from testing the platform on higher wattage PSU, is to make sure that each card is exclusively connected on separate +12V rail. I can see that your PSU has only two 6+2 pin pcie connectors, so I guess you use two adapters from molex to pci-e 6pin, right? Make sure these adapters take power from separate rails.
 

barnik

Honorable
Sep 26, 2012
7
0
10,510
Okay, I planned to upgrade psu anyway, non modular cables are terrible to fit in to case. I noticed yesterday that older gpu has one of these metal joints a little melted (joints holding power cables to the pcb - dont know the word) perhaps that is the case?
 

Bejusek

Distinguished
I guess it could be. Is it running fine as a single card?
Extreme stuttering in sli usually points to not enough power delivered to gpus.

Edit:
Have you tried stressing your cards with Furmark? Maybe it will cause the system to shut down and you will know it's the PSU fault.
 

barnik

Honorable
Sep 26, 2012
7
0
10,510
No, furmark is no good, the gpu got failsafe that throttle it down over 300mhz down (i've read its so user couldnt kill it).
Actually it DID run smooth until 2 months ago when i started to get random power kernels playing bf3 and cry2, I had to down clock it 780 to 700 mhz and that fixed the issue. At first i thought it is bf that suck, everyone know it is most buggy game of our times...

E: I thought it might be damaged when started playing on new gpu without older inserted into mobo. Everything run smooth and fine.
 

Bejusek

Distinguished
It seems then that the old gpu is faulty. You said you were monitoring things with afterburner, so I guess it isn't just a temperature issue?

If the problem really is the damaged power connector, it should be easily fixed in any workshop, or if you are handy with soldiering tool, you can fix it yourself. Warranty is over?
 

barnik

Honorable
Sep 26, 2012
7
0
10,510
No, on maximum stress temps are kept below 70 for upper and 50 for lower gpu. Still got the 1,8 year of warranty (3yr) so I will give it a shot and we'll see what happens. Any idea what might caused this?
 

Bejusek

Distinguished
No, not really. Maybe the solder was of poor quality or they missed it on quality check. Surely it is subject to some heat, I don't know how it is designed, but let's say that half of the power flows through this connector. This gives ~10A @ 12V, what isn't a lot of current. To get the solder melted you would need a lot more than that.
 

barnik

Honorable
Sep 26, 2012
7
0
10,510
Yeah I thought so... perhaps OC? I did it once to 900 mhz without raising up voltage (msi Afeterburner got it locked by deafult) maybe this is the cause?
 

Bejusek

Distinguished
I don't think so, without rising core voltage the increase in power draw wouldn't be sufficient to make any damage. It has to be some kind of manufacturing flaw.
 

barnik

Honorable
Sep 26, 2012
7
0
10,510
Thank god I was pretty sure It was done safely :) I'll post as soon as gpu comes back and hopefully workin'

Thanks a lot man!