Posts: 6
Threads: 2
Joined: Dec 2011
This isn't an OclHashcat question exactly...but I know there are lots of experienced users here that have probably faced similar problems. I got my new ATI 5970 yesterday (Newegg Black Friday FTW!). The only other card I have in my case is an ATI 5870...which is quite a few slots over so there is plenty of space between the cards.
After running OclHashcat for about 30 seconds, it quits due to the GPUs hitting the temp threshold...which is about 90 Celsius. I then took the case cover off and put a fan directly toward the card and then it lasted about 120 seconds before hitting the threshold. Is this normal? I know stock fans are crap, but I figured it would work better than that?! The 5870 peaks out at about 70 Celsius.
What are others doing to keep their GPUs cool? I'm trying to avoid water cooling...is that a must for the 5970? Thanks in advance.
Posts: 127
Threads: 16
Joined: Sep 2011
12-02-2011, 12:31 AM
(This post was last modified: 12-02-2011, 12:32 AM by chort.)
Check out
http://hashcat.net/forum/thread-692.html
I need to put that on the wiki today.
PS does it have two fans on the card? You might need to set them independently by either incrementing the display ID or possibly the sensor ID.
Posts: 6
Threads: 2
Joined: Dec 2011
Thanks for the speedy reply. I'm running Windows 7...last I tried, it was a real pain to get Linux running in a triple monitor setup.
I did try maxing out the manual fan speeds last night using the Control Center, but it didn't affect my over heating problem...but I may be having the problem that the fans are actually maxing out at 33%. As far as I can see, there is only one fan...and the rest looks like a heat sink.
How do I change the display/sensor ID? And what exactly does that do? Thanks.
(12-02-2011, 12:31 AM)chort Wrote: Check out http://hashcat.net/forum/thread-692.html
I need to put that on the wiki today.
PS does it have two fans on the card? You might need to set them independently by either incrementing the display ID or possibly the sensor ID.
Posts: 127
Threads: 16
Joined: Sep 2011
I have no idea how it works on Windows; above is for Linux.
It's tough to believe maxing the fan speed wouldn't bring temps down, because on my machine there was an immediate drastic drop in temps. Just keeping the fans at 40% instead of 33% has brought temps down to 64c from 80c. Of course this is all reported through the ATI commands, so who knows what the card is actually doing...
Possibly your fan isn't even running?
Posts: 6
Threads: 2
Joined: Dec 2011
I think I fixed it by putting my 5870 in the primary PCI express slot and the 5970 in a secondary slot. However, now the 5870 is pegging the temp threshold...but not quite as bad. The main problem seems to be with the fans. The fan of the primary card is not adjusting with the temperature properly. For example, now the GPU of the 5870 will be 90 C, but the fan speed is only 45% but ONLY on the card in the primary PCI slot.
I noticed if I bump the temp threshold to 91 C, then the fan will accelerate to 56% when the GPU temp hits 90 C. It seems to be working better, but definitely not optimal...especially since I can't resume if the threshold is hit!
Right now the 5970 is staying in the 70s C...while the 5870 in primary slot is giving me problems. Almost seems to be a bug in the driver.
Actually...my computer just crashed. Fortunately Chrome saved my input in the text box.
The power cable was melted! It was literally melting into the carpet. So does this mean the problem could be due to a small PSU? I think my PSU is 750 watt.
So now I might have to upgrade the PSU. What other gotchas are there? I want to do this right and want to be able to add another video card later if I want. Any good resources for researching this?
Posts: 5,185
Threads: 230
Joined: Apr 2010
moved the thread to general help.
Posts: 649
Threads: 18
Joined: Nov 2010
It means you are drawing too much power from the wall. How many amps is your house rated for? If you have say, 10 amps (9 with sag) and you are pulling 11, it will cause the cabling to overheat and ultimately melt and start a fire.
Posts: 6
Threads: 2
Joined: Dec 2011
I'm away on travel now...I'll check the rating when I get home. I'm sure it's at least 15 amps...the house was just built this year. Is there a fix for this problem? Could it be that my power supply cable is just too thin and flimsy? I know I'm not using the cable that came with the PSU...which may have been thicker. For example, my PS3 power cable is pretty thick as well.
So what do people use for power in their GPU servers with 4+ video cards? Don't they run into similar problems? Thanks for the suggestions...
Posts: 344
Threads: 2
Joined: Aug 2011
For your setup (5970+5870) you need any standard cable rated ~5A. Let's say if you owe FAT PS3 - it's cable will go fine as it's rate is 10/15A (not because it drains so much, but because of security standard)
Posts: 621
Threads: 57
Joined: May 2010
I don't know your cards specifically but overheating is a common problem when you use multiple video cards in a single machine. Usually, the one sitting on top is drawing the heat from the others and overheat. Here's a couple of tips:
1) Put the card with the best cooler on top.
2) Spread your cards to help with the air flow if you can. Don't use contiguous PCI-express slots if you have more than 2.
3) Make sure you have a good air flow in your case.
4) If you have double slot cards, make sure that if your card has a rear exhaust on the second slot (where there are no video connector), that the case metal piece that block the slot is removed.
5) If you can, use water cooling. It makes a HUGE difference.
6) I am not sure that the power supply is a problem but the melting cable is definitely not normal.
