Pc keeps shutting down when running 4-way cfx in any benchmark or demanding game

Toggle sidebar Toggle sidebar
R

radeson

Member
Dec 20, 2014
174
0
41
Hello,

I just installed a new system, new motherboard, CPU, Ram. I am having problems with the computer shutting down any time i run high intensity stress tests or benchmarks. I literally just put all this together friday so its kind of fresh. I just finished installing windows 8.1 64 bit. I updated it fully. I have all the latest drivers installed for the MOBO, GPU's etc.

I start up furmark or firestrike from 3dmark and it gets like 10 secs to a min into either one and shuts down cold. Then it starts back up on it's own. I think it might be my PSU. But maybe not. Its pulling 1300-1400 watts at the wall when this happens.

I only have the XMP profile set for the RAM @ 2800mhz which pushes the strap to 125 and the clock up to 3.5ghz

I am running:

i7 5960x; ASUS Rampage V Extreme; 16gb Corsair Dominator Platinum 2800 Mhz; (2800Mhz XMP profile set in BIOS and no other tweaks) 4x Sapphire Radeon R9 290 Reference; 1600w "G-Series" LEPA G1600 PSU; 4x Samsung 850 pro SSD(RAID 0); 1x Seagate 1tb HDD; Corsair H100i CPU cooler; Custom water loop through 3 of the 4 R9's (4th one on bottom) Temps never over 55 celcius anywhere except the 4th GPU gets like 60-70 max with fan speed @ 60%
 
TurtleCrusher

TurtleCrusher

Lifer
Apr 20, 2008
10,067
990
126
1400w is really, really pushing it. Peak usage should be 75% or less, and sustained 70% of rated power.

My opinion though.
 
V

.vodka

Golden Member
Dec 5, 2014
1,203
1,538
136
Taking a card out will be a pain since they're watercooled, undervolt all four (reference 947/1250 clocks usually allow -50mV) to shave some watts off each card to confirm you're running close to what your PSU can supply here.
 
M

Micrornd

Golden Member
Mar 2, 2013
1,368
232
106
I doubt that it is the power supply, as the Lepa G1600 is capable of pulling 2000w at the wall and supplying 1600w continuously to the PC, as long as you don't overload the individual rails. ;)

It sounds more like your loads are unbalanced on the rails causing a shutdown.
(Which is what it should do with an overloaded rail, and then it should reset once the overload is gone. Sound familiar?)

1st make sure each 290 card is on a separate rail (12v3, 12v4, 12v5, 12v6)
The MB and RAM is supplied by the 12v1 rail and the CPU is (predominantly) powered by the 12v2 rail

This leaves you with about 12A+ to spare on the 12v4 and 12v6 rails (after powering a 290 on each).
HDD, SSDs, and the H100i can go on the 12v4 rail and the water loop pump(s) on the 12v6 rail.
(If for some unknown reason your water loop needs more than 12A @ 12v, you'll have to add auxiliary power supply for it.)

Try that, I've run quad 7970s, with dual xeons, dual h110s, and 8 SSDs that way without problems using that Lepa G1600.
BTW - I think you'll find anything beyond 2x - 290's or 290x's is only worth it for e-peen and benchmarks, but doesn't really help in gameplay enough to matter ;)
 
Last edited:
flexy

flexy

Diamond Member
Sep 28, 2001
8,464
155
106
You, sir, just won the price for best 1st world problem of the century:

A 1600W (!) PSU but it's not sufficient for your rig. /mind blown


Otherwise what Micrornd said, try to check how your loads are balanced and whether a rail could be overloaded.
 
Last edited:
daveybrat

daveybrat

Elite Member
Super Moderator
Jan 31, 2000
5,822
1,035
126
It sounds like an easy to solve problem. If it truly is the power supply, just run 2 or 3 of the cards and see if the symptoms persist. If they don't, then it is a power issue.
 
Jacky60

Jacky60

Golden Member
Jan 3, 2010
1,123
0
0
It's your PSU, my sig rig did the same with both Enermax 1350W and EVGA 1300W on 4 way xfrire (2x295x2) but has vever once done it since upgrading to EVGA 1600W and I have a much leaner system than you do with 4770K not 5960X.
 
Last edited:
D

DownTheSky

Senior member
Apr 7, 2013
800
167
116
It's your PSU. Undervolt and underclock the GPUs to 800-900 Mhz and whatever voltage is stable and the power consumption should go down significantly. Your cards should have Dual-Bios feature. Modify your bios to whatever you found stable and flash one of them. Do this for all cards. Should take an evening but it's worth it in the long run.
 
Last edited:
R

radeson

Member
Dec 20, 2014
174
0
41
Thank you everyone for the replies. Last night i tried to switch off my 4th card with the convenient little pci switches on this board. Well, I have to say that i didnt get it to shut down one time.

I have been playing with some OC stuff. Im still kind of a noob at that and I dont want to push it too high. I have the voltage set to AUTO and the strap at 125 for the XMP ram profile, Then i pushed the multiplier up to 32 from 28 and its running just shy of 4.1Ghz and it runs stable with Prime on small fft's never going over 60c. The RAM is at like 2792 i think, not a solid 2800......
I would like to get a stable 4.4-4.5. I will be upgrading from the H100i to a more complex custom water cooling loop soon.

As for the Power/Crossfire Problem.... The comp hasn't shut down since I switched off the 4th card. Which thankfully the 3 top cards are imported from my former AMD system, and the only ones that have water blocks. The 4th has stock cooler. I would rather not under-volt the cards as I have never done any bios flashing on gpu's before and I'm a little hesitant. I would love to learn how to do it, but I would much rather have a system capable of running the cards stock.

I am open to any sound advice.......
 
R

radeson

Member
Dec 20, 2014
174
0
41
Also do the Rails run in columns??? I have Each GPU on their own cable which has 2x 6+2 connectors... but i have the majority of the connectors organized to the right side of the PSU opposit of the CPU/MOBO connectors.....????
 
Y

YBS1

Golden Member
May 14, 2000
1,945
129
106
modular_panel_scheme_small.jpg


Yes, it's entirely possible your system can shut down that 1600W supply depending upon clocks/voltage. Mine was shutting down a 1300W with just two Classifieds when heavily overclocked and the cpu was at 5.1GHz. I'm now running a SeaSonic Platinum 1200W for aesthetic/cabling reasons after my Lepa G1600 died (and was replaced under warranty), but if I were to heavily bench I would need to run the cards off the Lepa. An 8 core cpu and 4 R290s are probably pushing the boundaries of that psu if heavily overclocked.
 
Last edited:
R

radeson

Member
Dec 20, 2014
174
0
41
Ahhhhh..... So two of my cards are on rail 5! I will switch it over and update later.... Thank you!
 
R

radeson

Member
Dec 20, 2014
174
0
41
So I have the GPU's hooked up to 3,4,6 on the bottom, and the middle 5 rail in the bottom. I have the 4 pin motherboard hooked up to the other 5 on the bottom right. In going to start up right now....
 
R

radeson

Member
Dec 20, 2014
174
0
41
So far no shut downs with all 4 running!!! I just scored a 26080 on firestrike from 3dmark!!!! im like 1200 points from the top 100 list!!!
 
TurtleCrusher

TurtleCrusher

Lifer
Apr 20, 2008
10,067
990
126
High-end GPUs throw out any room for power consumption rationale. It's hilarious to see people quibble about a 40w peak power, 5w idle spread on desktop CPUs then suggest slapping in a 1-3 250w+ GPUs.
 
R

radeson

Member
Dec 20, 2014
174
0
41
Jacky60

Jacky60

Golden Member
Jan 3, 2010
1,123
0
0
gradoman said:
Glad you figured it out!

I've always wanted to try a 3+ GPU system, but it seems like once you go with more than two, there's only trouble and a lot of power consumption.
Click to expand...

Well yes and no, You do need more power for obvious reasons and because such rigs are unusual driver teams tend to fix any 3+card issues last but the gains can make it worth it.
 
Last edited:
R

radeson

Member
Dec 20, 2014
174
0
41
It's actually pretty quiet when I'm only running 3 way cfx with the 4th air cooled card turned off... The cpu doesn't go above 55c at 4.4ghz 1.32v
 
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom