Bitcoin Forum

4x7950 gigabyte stock clocks(1000/1250)
gigabyte 970a ds3 mb
2x4gb ram
sempron 145
lepa g1000 1000w psu
i have two 1->1  risers and 2 1->16 standard risers
12.8 drivers
2.8 sdk
cgminer 3.1.0

Everyhing worked fine when there were 3 gpu(2x 1->1powered raiser and one in 16x slot), after adding 4th one(2x 1->1 powered,1->16 in 1x slot, 1-> in 16x slot) rig just turns off after about 15 minutes(sometimes more sometimes less), temps are ok i use autofan with target temp 75 and they all run at about 73 all the time while mining bitcoin. I can't turn it on for a while.I used it for a while with lcd plugged in and don't remember it happening(maybe it was to short time). Psu is 4 rail and i tried to split is as equal as i can:
1:mobo+one card(not modular so has to be i guess)
2:one card, impossible to plug anythin else
3:one card+ssd+powered riser
4:one card+powered riser

only way to determine the rails was a photo of pcb in some review so not sure it really is split that way.
Looks like something with psu for me but how can it be to small for this setup?

You're drawing too much from the power supply and it's shutting down.  If I recall correctly, gigabyte cards draw 250 watts each at 1000mhz... (might be 225w, can't remember, but either way.. you're up there.)

Looked a bit in internet, in every test i found 7950 was below 200w, even if it was 220 it should work well since the psu is solid acordning to reviews and tests.

Use a kill a watt. You might be surprised how much power they use.


The 7950's TDW is 200W at stock clocks.

From my own experience, this seems accurate.  I would also recommend you take one card out or turn it off before trying to continue.  My experience also included me continuing to try different configurations to get it working and blew out an old PSU.  Luckily the mobo or cards weren't damaged but they certainly could be.

You don't want to be pushing your PSU anywhere near the max - most people will say it's safe to run it at 70-80%.

Well, when I did the same... I found 200 watt for the 800mhz card... so, assuming voltage was the same, 1000mhz is 25% higher than 800mhz, 200x 1.25 = 250watts.

So.. uhm, yeah.. might even be a bit higher ;)

I have 2 Sapphire 7950's that are at 1000/1250 and they draw 250-270 W.


I dunno if your using Windows or Linux.  But if your using windows and the computer just turns off with no BSOD and no Windows shut down unexpectedly error thing than its the power supply.

thanks guys, guess i'll just run 3 now and get killawat thing to know for sure.

Plus you have to add in everything else.


The powered risers will draw 75W each if Im not mistaken.  Plus you got the rest of the system and the SSD.

well, if it lasts 15 minutes, then it suggests to me that you could probably get it working with ubuntu on a usb stick, take out one of the RAM sticks, underclock/undervolt the CPU (and RAM)

turn off all the stuff you don't need to use, like the onboard audio and serial ports and what not

Maybe I just haven't heard of it but is Lepa a decent brand?


I only see 1 review on Newegg, and 1 review on Amazon of this PSU.  Even though its 80 plus gold rated, running it near 100% capacity especially 24/7 isnt probably going to last long.



http://en.wikipedia.org/wiki/80_Plus (http://en.wikipedia.org/wiki/80_Plus)

Also look at what 80 plus gold actually means.  At 100% load its only putting out 87% at best.  So 870W.


So even if your cards were only putting out 200W then your still too much power.

You're failing to understand how PSUs and their power/efficiency ratings work.  Some Gold rated PSUs like Seasonic made ones are well over 90% efficient at full load.

If  PSU is only 87% efficient, it means it will draw 1.0/0.87 * 1000W (in this case) if it was maxed out (assuming it can run the full 1000w - cheap PSUs can't).  87% efficiency doesn't mean it can only put out 87% of maximum power.

Lepa is a decent brand (part of Enermax's budget line).

This is the safest route, it might suck to lose 25% of your hashes right now, but that's waaaaay better than losing 100% of your hashes until you have a PSU replacement and/or more equipment!

it probably shuts off due to increase in ambient temp

depending on how many peripherals are attached to it and possibly moving it to a cooler location, it probably wouldnt even be stressing it too much.. or just do all that and drop the voltage on a card (* if you can on 7950's dunno)

temp is not an issue i think, i put it outside of my open setup so it sucks completely fresh air from 30cm fan. Looks like lack of power. If i were to buy a replacement should i go for single rail or it doesn't matter?

single rail is just easier to manage

and probably easier to sell on ebay too

You could buy a second PSU (even a 500w one), and use the second PSU to run the 4th card. You will need to "jump" the PSU to turn it on:

http://forum.overclock3d.net/showthread.php?t=394

...you wouldn't happen to have some power saving feature enabled that automatically powers down your computer when it's not in use would you?

This is what sent me mad for a day and nearly resulted in an RMA of the GPU! Check it!

it went all night on stock clocks mining bitcoin, i switched to litecoin now and it turned off after getting few shares. Litecoin intensity is 19 and bitcoin 16(any higher and cgminer shows some very high hashrates but doesn't mine anything and gpu usage is like 37% some other random amount).

no power saving

i tried with some old 400w nonapme psu(connected one 8pin to one card and one 6 pin to other) everything freezes on start of cgminer.

bitcoin mining over intensity 9 is wrong

Most likely a power issue as mentioned.  If you put a Kill-a-watt on it and start mining you'll notice the power draws goes up until the temps reach a steady state.  Probably when the rig shuts off it is overextending itself.  Either take a card out, undervolt or downclock one card to reduce the load.  Maybe downclock the hottest card.

Or underclock the CPU, that's what I did. If it isn't your main pc you could even disable one of the cores reducing draw even more!

flash your cards to F43 bios
It will lower voltage to 1.09... and save you around 40W per card.
Here is the tutorial:
https://bitcointalk.org/index.php?topic=197200.0

Mining litecoin with 3 gpus pulls about 1k from wall(bitcoin is about 900w), something must be wrong?
Beside gpus i have cpu, ram, ssd, usb wireless.
Solid psu if it wokred for 10 minutes with one more gpu i guess :)

Would prefer not to flash bios due to warranty.

Yep, you're just out of power. Mine the real coin at lower power and be happy.

Still won't be able to use 4 gpus and i have 4, i'll try to return the psu and just get a bigger one and then figure out max performance.

Flashed to f43, looks a lot better now. Had some problem with no video on booting, only could run 2 cards at once and not in any combination... but did this finally. Clocks are 1000/1250 tho not 900/1250 but voltage is 1.09, autofan works fine at about 500rpm less to keep target, hash is more or less the same - 600 on litecoin. Most important thing, it takes 800w from wall so significant drop. Maybe i can get the 4th gpu going, pretty sure it'll be fine for bitcoin :)

edit: power draw for bitcoin at 1100/750 770w about 560mh/s fluctuates a lot unlike previous bios where it was very steady in cgminer, also there is difference between cards and before there wasn't bu i guess it's a lot bettter now :)


edit 2: Could some moderator please move it to mining support, i wated to start it there but something went wrong.

I got the 4th card in but cgminer starts with it being at 500/100 even when i specified engine and memclock for each card seperately, when i disconnected on of proprly working cards this one starts to work fine(1100/750). Can it be that it lacks power and acts this way(before it just turned off so dunno)? Its connected to a rail with mb(2x pci-e not powered, proc, ram and other stuff), non modular so can't really change much besides powering 2 more risers. Also i can't change engine and memclock in cgminer it's fixed ant this 500/100 or 500/150 for litecoin.

Same thing happened to me.  2x "Gigabite" 7970 with 750 Watts OCZ power supply.

Shut down after both cards came online to mine with GUI miner.  Upgraded to a 950 Watt PSU, problem solved.

See my thread https://bitcointalk.org/index.php?topic=205406.0 (https://bitcointalk.org/index.php?topic=205406.0).



I am was having power draw questions too.  I think that mining LTC on top of OC ing the cards makes you draw alot of power.

Mining bitcoin for an hour with 4 gpus at 1100/1250 1020w at most from wall so easily within psu. Powered one more raiser what allowed for more even split between rails. Looks like i'm getting a bit less hashes, stable 570 with 3 gpus, now 550 at most and a lot of time is like 500 or 520. Also cannot underclock memory with cgminer, setting gpu works ok but memory always stays at 1250.

For anyone who wants to do this i'd definately recommend single rail psu.


edit: weird gpu-z shows 1250memclock and afterburner shows 500(what i typed in --gpu-memclock) when i open it while mining.

edit: so it failed anyway, only after 3 hours not 15 minutes, while pulling 1060 from wall, 4 rail psu sucks, especially that 2 rails are not modular(atx+2xgpu,2gpu!!!), will just get 1300w single rail psu.

You must limit your memory underclock to no more than 150 less than the engine speed (driver limitation). Best way to do this with cgminer is to add --gpu-memdiff -150

Could also be an issue with chipset over heating!!!  :-[ can you post your motherboard model?

I had this issue as well and thought it to be bunky PSU at first...(I was in the setup stage for this rig trying out 2x7950 on an 850 watt PSU) I tried at least 30 different configurations with PSU/GPU/Motherboard combo's until I actually took a good look at the board and found that two of my 7950's were sitting right above the motherboard chipset.

Plain and simple there is inefficient heat dissipation for chipset heatsinks on a lot of motherboards, whenever my miner shutdown randomly and unexpectedly, I checked the heat on this (and burned my hand quite a few times!!)

My conclusion on random insta-shutdowns is it typically boils down to 3 things
-Kernal Panic due to driver failures
-Insuffiecient Power Draw from PSU to the motherboard when Cards are running 100%
-Component overheating/damage (chipset or in rare cases CPU)


Hopefully this helps!!!

Even a junk 750 watt power supply should be able to handle two 7970's...

have a single core CPU or disable the other cores, undervolt all that stuff (memory, cpu, fsb, etc), ubuntu on a usb, use some energy efficiency lcd monitor with a kvm switch

hmm, do ps2 mice/keyboards have less of a power draw than usb?   does disabling USB slots in bios save you a few watts from motherboard?

i never mine litecoins, but, yeah, gpu's overheating wouldnt surprise me (cough not being able to lower memory to nothing).   sometimes you'll get cards with really crappy thermal paste job that'll have some sensor +20oC  higher than 'main'