Topic: 3 x Sapphire 7950 Rig Crash - High Temperatures - Reward for Help (Read 4952 times)

100% fan. if that doesnt decrease temps, decrease volts (and frequency).

this is strange... i would try to use the card on factory setting (no overclocking)...
Otherwise everything looks good. Maybe 3 cards one over another is a bad idea use some risers.

Are you living in the desert at 40C ?

My main rig is in a Haf 932 with only TWO 7950s and it struggles sometimes. Underclock and undervolt your ram, and whack as many fans as you can in there.

Did you try putting the fan speed at max for all 3 cards?
Also use small anti-static foam blocks to space the cards further apart. (an addional cm/half inch could double the airflow).

Could be the VRM/Mem temp that is getting too high.

What also worked for my card(s) is removing the shrouds, especially when you have casefans blowing directly onto the cards. All that plastic/metal just blocks the airflow.

I'm all for the initial statement about the cards heating the others. Usually the top card will be fine but the second and third cards are not as open to the air. I have a full-sized room fan blowing on my cards for now and the top card is keeping the bottom card at 80-84C. Top card sits at 75C. Without that top card running the bottom runs at 75C because that top card is dumping its heat right on the backside of it.

I don't understand.

I have nearly the same set up - I get 600 mhash btc each card and 560 khash (still tweaking here) on ltc. high 70s temps, all in case, no extra fans.

What settings are you using?

So the PSU was a problem after all?

Next thing: Cooling. You squeeze three factory overclocked high-end GPUs tight to each other. They will heat and/or choke each other. Either go for excessive cooling (highly optimized airflow or water blocks) or use extenders. Which you choose depends on the chassis options or financial situation.

Best of luck!

Hello,

I fixed the PSU crash by changeing the PSU with a 1200W model.

But now i have other issues, AMD driver chases o stock settings when i start the 3 video cards in guiminer-scrypt. , if i try to undervolt it crahes.

At one try i got all of them working and stable at 81 degeres, but at the next tryout huge temps and driver crash.

Any ideeas?

use pcie extenders, bring the cards out of the case onto a rail, separate them, and have a big fan blowing on them.

these cards were not not meant to be right next to each other running at full capacity 24hrs a day.

estimated cost for this solution: ~30$

money it could save you: cost of 3x cards, time lost mining

best of luck. i don't think this method could fail in terms of cooling and is easy to do.

i know you likely want to keep things contained in the case, but it may not be viable.

I don't think that PSU is up to the task at only 850 Watts especially with 3 cards.

Also, try setting the bandwidths on the PCI channels to x8.

What fans are you using in the case?  Airflow can make a huge difference.  If you are using low end fans that do not move enough air, or the flow is not efficient enough, then that also would affect things. Sometimes you have to shell out a bit and get some higher end fans that displace more air to ensure you are getting maximum airflow and cooling.

3 cards is gonna get hot. Try de-clocking.

I'm guessing the cards are humping eachother? Even ATX cases really aren't designed to handle 3 cards unless the computer is in a freezer...

working on it..

I know this sounds weird but sometimes the cards like to be burned in a little.

Take the 2 extra cards out of the rig.  Run each card 1 at a time sitting in the 1st slot.  Start it low and work your way up. If you're using GUIMiner use very conservative settings, don't use -f0 right off the bat for example.  Make sure each card can handle the settings you will be using for all 3 cards.  If the card can mine 30 min with no problem it's OK.

This will help you narrow down if one card is bad (maybe has bad VRMs).  If you're curious you can also try to find out the stable max for each card.  Bump up the core clock speed 10MHz every minute until it freezes or locks up, then back off by 20 to be safe.

If all 3 cards can run up to the 560MH/s speed you quoted, now try to hook them all up again paying attention to spread out power on each rail.  HCP-850 has 4 rails each providing 40Amps.  Try to connect 1 card to each rail if possible.  If you need molex-->PCIe for the 3rd card try putting it on different spots.

Try the above stuff and come back to us.

next step I guess is to check how your powersupply is routing the juice. Did it come with 6 6pin conectors? if its modular it will have "rails". make sure its split up right. if you had to add 6pin with a molex adapter see try and split that so again your not pulling to much off one side of the powersupply.

The PSU is at 92% efficiency, i saw a lot of forumers with this cofig and since i got the started for a little while i dont think here is the issue.


They are

Make sure when you start your 2nd and 3rd miner that in the drop down menu the right gpu is selected.

Ah ok. I'm trying to figure out if you have successfully run all 3 GPUs in 100% before the crashing? 850W sounds a little weak for 3 Turbo (factory overclocked) GPUs... Most PSUs usually don't perform with stability near their max load. Trying to eliminate the usual caveats here...

The case has 3 extra fans pointing exactly on the GPUs, i dont think is a cooling problem, since on some settings the temperature allmost leveled on a constat level when the 3rd GPU starts then the thing goes down.

It might sound simple but the cooling may be inadequate. Sapphire cards tend to be cheap and cheap cards tend to go hot in extreme scenarios (like running three of them besides each other). I recommend some extra chassi fans directed towards the GPU cards. Have you considered this?

Best luck!

Hello all,

Unfortunately i am facing some real problems with my new rig, this is the setup i just received:

CPU: AMD - Procesor AMD Athlon II X2 Dual Core 250 (C3) (BOX)
MotherBoard: GIGABYTE 990FXA-UD5, AMD 990FX + SB950, AM3+, DDR III, USB 3.0, PCI-E 16x, SATA III
PSU: Antec HCP-850, 850W, Modulara, Vent. 135mm, Efficiency 92% (80 PLUS GOLD)
Ram: Corsair Vengeance DDR3, 2x4GB, 1600Mhz (Dual Channel)
Case: CoolerMaster - CoolerMaster HAF 932 Advanced
Hdd: some old suff i found at home
Video Card:3 x Sapphire Radeon HD 7950 Boost, 3GB, GDDR5, 384bit, DVI, HDMI, Mini-DisplayPort, PCI-E 3.0

On a fresh windows i have installed the following:

MSI Afterburner
AMD SDK
AMD Drivers 12.8
GUIMiner

Start 1 #1: I run the miner on the first to GPU, good hashrates ±560, when i start mining the 3rd GPU-system crash/shut down

I thought i had to undervolt or something and restarted the computer. I run cmd and enter: startx GPU_MAX_ALLOC_PERCENTAGE 100
and enter various settings in afterburner that i found on the forum, for one setting the rig laster a little bit with all the GPUs started (although the pool did not show the same hashrate) but when i closed one GPU and started it again same thing happened.

I also watch the GPU temperatures, it reached 90c on some settings - on other settings the GPU temperature was quite ok (maxing 81 for one GPU the others at about 60) but the GPU was reaching 99% constantly (on other settings the was to change in load)

I really have no ideea what to do...i am just looking at an expensive piece of computer.

I am offering a 10% out of the first month reward to the person that can help me with this one + 25$ via paypal.