Hello all,
I haven't been able to get a straight answer on max safe vrm temperatures for 5970s. Some say stay under 100 and some other say under 120 degree, for 24/7 usage. I want to get max out of my cards.

If you have a good answer, please tell me so that i can tweak accordingly.

Officially the VRM limit is 120, above that and it gets into unhappy territory, the thermal cut out may or may not operate, not a good idea to find out. Its not instant Armagedeon at 120, but its dodgey territory, and could well fail at any time over 120. You void your warrenty if it fails over 120 (and they have circuitry which tells Vendors that happened).

Ideal temp is below 100, which is somewhat anechdotal and popularist comment based on the highest (practical)safe temps for GPUs (not VRMs) of 100, VRMs naturally operate at higher temps than GPUs. That 100 figure however is not any kind of mandatory requirement, only the 120 top limit is. Operating at VRMs 100-110 is fine, no dramas. Above 110 should be avoided only in so far as if some unexpected issue occurs, there is a "buffer" before hitting 120.

Also bare in mind that operating 7 x 24 places a big stress on the card, and whilst it can operate at higher temperatures/gpu/memory settings, its a bad idea to keep them at the very top edge, it will degrade card life significantly.

Bottom line - If you kept the VRMs at 105 to 110 on 7 x 24, thats about as far as you should push it to preserve long card life and sleep at night (!). IMHO, at £500 a card its not worth the grief to push the VRMs to 110-115 on 7 x 24, as some do just for the sake of 2/3 seconds. The choice as they say, is yours :)

Regards
Zy

Thank you Zydor. My VRMs are currently at 96 and i will push them 30-40 mhz more. Just below 105 degree.

Thank you Zydor. My VRMs are currently at 96 and i will push them 30-40 mhz more. Just below 105 degree.

Actually, oc them to 850 which is the default 5870 clock speed and the vrm is 110 degree. I wont push it further.

My 5970 is currently at 941MHz core clock and 1000MHz memory.

Via GPU-Z my VRMs are around 95°C. I have altered the fan profile via MSI Afterburner and the GPU's are at 66°C & 76°C, but the fan is at 80% and noisy.

Maybe you can push yours a little harder?

My 5970 is currently at 941MHz core clock and 1000MHz memory.

Via GPU-Z my VRMs are around 95°C. I have altered the fan profile via MSI Afterburner and the GPU's are at 66°C & 76°C, but the fan is at 80% and noisy.

Maybe you can push yours a little harder?

I have 2x5970 running crossfirex and there are no slots between them. Therefore the first gpus vrm is this high (110). Otherwise, i would oc them more.

By CCC i suggest you to reduce the memory clock to lowest official value which is 500 mhz, if you only use the 5970 for milkyway. It will save you heat and power

I have had two of my 5970's running @ 950 core 1000mhz memory and I am seeing temps@ VDDC 109,114,114 and it's been running for days 24/7.
My fan is set at 100% (I don't mind,lol) and GPU temps are @ 66,67,84,85
I have cards waiting if she blows so will let you know but as of yet for a long time I have been running with these numbers and oh yeah 1200mV.
Oh, I do adjust one core to 910(or any little different number) to keep it running smooth (it wants to oscuillate back n forth 2d,3d@950 for some reason)and I have seen sustained temps around VDDC 120 and so far she is still crunching.
All of these things are kind of nebulous from my experience as each of my 6 cards act a bit different, but I haven't had anything fail yet except a fan!
You can see the results by graphing my current challenge performance these last few days.

I have had two of my 5970's running @ 950 core 1000mhz memory and I am seeing temps@ VDDC 109,114,114 and it's been running for days 24/7.
My fan is set at 100% (I don't mind,lol) and GPU temps are @ 66,67,84,85
I have cards waiting if she blows so will let you know but as of yet for a long time I have been running with these numbers and oh yeah 1200mV.
Oh, I do adjust one core to 910(or any little different number) to keep it running smooth (it wants to oscuillate back n forth 2d,3d@950 for some reason)and I have seen sustained temps around VDDC 120 and so far she is still crunching.
You can see the results by graphing my current challenge performance these last few days.

Would you state the working environment? My hosts work at a room cooled to continuous 25C. How did you oc those 2x5970 to 950 core?

Also i checked your host and despite all that high clock, your wus take long time to complete. Are you sure running only milkyway@home?

I am running 2 wu's on each core and I have found all these settings to be a bit nebulous as all my cards act differently. I'm far from any expert, I just experiment and check the results and I haven't had anything fail yet except 1 fan.
My room has an open window letting in -30C airflow.
That helps, yes.

The 5970 has really crappy VRM cooling, which is why I got 2 5870's.

I am running 2 wu's on each core and I have found all these settings to be a bit nebulous as all my cards act differently. I'm far from any expert, I just experiment and check the results and I haven't had anything fail yet except 1 fan.
My room has an open window letting in -30C airflow.
That helps, yes.

-30C airflow!!:) Are you living in north pole? That is one favorable temperature for oc:)

Oh Boy big crazy!
I was doing too many things at once! The airflow into my room is only 0C !!
Still helps but yeah that would've been like having phase change on it!
But I am able to maintain the 950 cores with no apparent problems.
Now whether they are all the way optimized, well I'm not at all sure.

My 5970 is currently at 941MHz core clock and 1000MHz memory.

Via GPU-Z my VRMs are around 95°C. I have altered the fan profile via MSI Afterburner and the GPU's are at 66°C & 76°C, but the fan is at 80% and noisy.

Maybe you can push yours a little harder?

I have 2x5970 running crossfirex and there are no slots between them. Therefore the first gpus vrm is this high (110). Otherwise, i would oc them more.

By CCC i suggest you to reduce the memory clock to lowest official value which is 500 mhz, if you only use the 5970 for milkyway. It will save you heat and power

When MW is down it does Collatz so the memory is higher than I would normally run it purely for MW.

I hear ya on the no room between the 2*cards.

It will depend a lot on individual mobo layouts whether ot not this helps those with closely packed cards (and use crossfire not dummy plugs), but usually works just about anywhere as even high end cards will work just fine in a x8 slot if thats all there is.

ATI cards come with a 7cm strap as standard for Crossfire, you can get a 10cm strap which allows the cards to be separated by a substantial proportion, usually 3 slots worth. I have custom coolers on each of my 5970s (These Coolers) , and the 10cm strap still allows very good card separation on the mobo, greatly increasing airflow around the cards.

Worth a shot - they are usually around £2-3 on EBay, stay clear of the `£10-12 offerings .....sharks...

There have been allegations by some vendors that the 10cm strap cause minor frame rate loss. That may be so, dont know, but I have never had any issue with it, nor have others I know who use the 10cm straps. Even if it did cause minor frame rate loss, when compared to the output of 2 x 5970s its hardly an issue.

(Those coolers are awsome, they really do take off 40c from GPU temp as they claim - they do them now for many card types - if you are not used to fiddling with hardware, I strongly recommend you get them fitted by your local PC shop)

Regards
Zy