Author

Topic: HD5970 temperature vs fan speed (Read 3988 times)

newbie
Activity: 41
Merit: 0
May 26, 2011, 04:56:41 AM
#13
Well i literally just upgraded yesterday form a 5870 to a 5970, and all i did, was push the voltages to 1175, then set it to 850/1200(same as the 5870) and its now been running for the last 21 hours perfectly. I still need to install sdk 2.1 as im only averaging about 620 for the card, whereas i think i should be getting about 700mh/s for those clocks. I have a fan blowing on the card and its sitting at a constant 90*C which is fine for these cards. What i say, is that leave the fan on auto, if it get too hot, it WILL ramp up its speed and your temps will drop. No need to worry about a card at 90*c or under. especially with something like a 5970, which they know runs hot and have made the cards with better components so that it can handle it.  Im going to try clock it higher, with increasing the fan speed and decreasing the mem rate.

I'm getting 624mhash with stock clocks. I even dropped voltage to 0.975v hoping to get some temperature improvements. Didn't help much, temps still between 89-93C depending on time of day, fan around 53% (auto)

So, check your miner settings or switch miners/sdk, you should get more mhash from that clock.

I'm reluctant to overclock due to it running too hot (don't have a spare fan right now) but overcloking to 735 (just a 10mhz push) makes the card run at 640mhash.
newbie
Activity: 5
Merit: 0
May 26, 2011, 01:39:01 AM
#12
Well i literally just upgraded yesterday form a 5870 to a 5970, and all i did, was push the voltages to 1175, then set it to 850/1200(same as the 5870) and its now been running for the last 21 hours perfectly. I still need to install sdk 2.1 as im only averaging about 620 for the card, whereas i think i should be getting about 700mh/s for those clocks. I have a fan blowing on the card and its sitting at a constant 90*C which is fine for these cards. What i say, is that leave the fan on auto, if it get too hot, it WILL ramp up its speed and your temps will drop. No need to worry about a card at 90*c or under. especially with something like a 5970, which they know runs hot and have made the cards with better components so that it can handle it.  Im going to try clock it higher, with increasing the fan speed and decreasing the mem rate.
hero member
Activity: 531
Merit: 505
May 16, 2011, 04:23:23 PM
#11
I run two 5780 cards at about 75-85 C. I prefer less noise over temperature. The cards are stable around 80C.

Moreover, I think that "lowering GPU lifetime using hot temperature" is kind of urban legend, partly invented by manufacturers to prevent folks from overclocking (thus requiring them to buy better HW for more money).  Do you really know a case when the GPU failed from overheating in 2 years frame?

There is a thermal throttling built-in into cards BIOS, so unless its activated, it seems the card is fine with that temperature.

Not talking about the fact that these cards may be obsolete for mining in half a year from now.
sr. member
Activity: 254
Merit: 250
May 16, 2011, 11:49:07 AM
#10
With lower temp you mitigate also the electromigration process and so increasing your card lifetime. This is why i dont overvolt my gpu card.
http://en.wikipedia.org/wiki/Electromigration
legendary
Activity: 2058
Merit: 1452
May 16, 2011, 10:52:32 AM
#9
Another thing you can do to keep the temps down is underclock the memory to 300MHz. With Phoenix this won't reduce your hashrate at all. I'm not sure how to do this on Linux other than flashing the BIOS though.
that actually boosts hash rate :p
full member
Activity: 219
Merit: 120
May 16, 2011, 06:02:46 AM
#8
Be careful when overvolting 5970s. The stock cooler doesn't provide sufficient cooling for the VRMs if you up the voltage too much. This will cause the card to downclock and reduce your hashrate. Personally I wouldn't use more than 1.1v for the core voltage.

Another thing you can do to keep the temps down is underclock the memory to 300MHz. With Phoenix this won't reduce your hashrate at all. I'm not sure how to do this on Linux other than flashing the BIOS though.
hero member
Activity: 560
Merit: 517
May 15, 2011, 08:35:12 PM
#7
Quote
How did you get to 980Mhz without it freezing?
By increasing the core voltage to 1.23V. That allows the card to run at higher speeds, but it carries far more risk than overclocking. As pointed out, simply increasing the clock speed will not damage the GPU in any way, except through excessive heat (1). Overvolting, on the other hand, can physically damage the GPU. I've heard the limit for these GPUs is 1.25V.

(1) The GPUs are actually rated for very high temps. Over 100C. And the card will shut them down if they get too hot anyway.
newbie
Activity: 42
Merit: 0
May 15, 2011, 07:12:31 PM
#6
I'm using linux.
What is the command to check the VRM temperature?

I tried phoenix and now im getting ~340MH/s at 800mhz.
I tried OC to 850 and 900 and the temp is still only 80-85, but its not stable, after a while X and phoenix (or poclbm if i use that) freezes and wont close even with kill -9, and the only way to get it going again is to reboot Sad
How did you get to 980Mhz without it freezing?
full member
Activity: 126
Merit: 100
May 15, 2011, 12:23:44 AM
#5
AMDOverdriveCtrl will set up a profile that you can configure and load; tailored to your particular situation.  by 'situation', i mean your case and the air-flow in it.

it's an invaluable tool:  http://sourceforge.net/projects/amdovdrvctrl/

(if you ain't using linux, you ain't using.)

if you use Hashkill as a miner, it will cut your card out for a rest period (of 20 minutes, as i recall) if it hits 90 degrees.

really - 90 is pushing it too hard.  take care of your tools...
newbie
Activity: 47
Merit: 0
May 14, 2011, 10:38:25 PM
#4
If I leave the card to manage the fan by itself it goes 45-50% and temp is 85-90 C.
When i increase fan to 100% temp goes down to 65C.
The fan noise is not a concern to me.

Am I assuming correctly that lower temp is better, making the card last longer?
Or is running the fan constantly at 100% going to damage the fan?

I already overclocked it to 800 (was at 750 i think).
Is the only risk in overclocking the increase in temp?
Its in an air conditioned room so the heat shouldn't be too much problem, but I don't want to OC it too much if its going to damage the card.

Im getting about 315Mhash per core.
Im using poclbm, but i just read about pheonix and modified poclbm.
Would those give me better hash rate?

Thanks and please excuse me if these are noob questions.
I have 2 HD5970 as well.

Set the fan between 80%-90% (I see no difference with 100%, expect more noise) constant.
If the core is over 100C, its bad for the card. But your main concern should be the VRM temperature, if its over 125C the card will start throttling because the card is crying for mercy. Faster fan = wears out quicker.

Again when you overclock look after the VRM temperature not going over 125C and Core under 100C. (Mine is at 980Mhz Core (@83C - 86C and VRM just cruising at 109C) and 500Mhz Mem at 1.23v for the past month because its winter here Tongue and getting around 468ish Mhash per core)

Don't know much about miners, I just use whatever. Using Poclbm + Win7.

Use furmark as your stability test, because in a way that's how bitcoin works. It also stresses the VRM like crazy compared to playing games. Use GPU-Z to monitor your temp.


newbie
Activity: 25
Merit: 0
May 14, 2011, 07:20:42 PM
#3
as long as the fan is running at a constant speed it shouldn't get damaged at all
donator
Activity: 1218
Merit: 1015
May 14, 2011, 07:19:06 PM
#2
[1]Am I assuming correctly that lower temp is better, making the card last longer?
Or is running the fan constantly at 100% going to damage the fan?

[2]I already overclocked it to 800 (was at 750 i think).
Is the only risk in overclocking the increase in temp?
Its in an air conditioned room so the heat shouldn't be too much problem, but I don't want to OC it too much if its going to damage the card.

[3]Im getting about 315Mhash per core.
Im using poclbm, but i just read about pheonix and modified poclbm.
Would those give me better hash rate?

Thanks and please excuse me if these are noob questions.

[1] A temp of 90+*C is  too high. 100+*C is when you'll probably start to see performance problems, instability, and possibly damage to the card. The fan is meant to run at 100% without problem, but the graphics card is not meant to operate at such high temperatures.

[2] Generally, yes, assuming you are not increasing the voltage. There is a risk of your computer becoming unstable and crashing if you overclock too high, but I'm not aware of any case (in cards of the past few years) where the driver doesn't just simply reset itself and go back to stock speeds. You'll be made aware of when you've overclocked too high when your screen blacks out for a second Wink

[3] That's a reasonable hash rate for that card. poclbm's hash rate is comparable to pheonix's. You should try it out, though, if you're interested. You could also try out some of the different settings others used with your card model posted on https://en.bitcoin.it/wiki/Mining_hardware_comparison
newbie
Activity: 42
Merit: 0
May 14, 2011, 06:37:37 PM
#1
If I leave the card to manage the fan by itself it goes 45-50% and temp is 85-90 C.
When i increase fan to 100% temp goes down to 65C.
The fan noise is not a concern to me.

Am I assuming correctly that lower temp is better, making the card last longer?
Or is running the fan constantly at 100% going to damage the fan?

I already overclocked it to 800 (was at 750 i think).
Is the only risk in overclocking the increase in temp?
Its in an air conditioned room so the heat shouldn't be too much problem, but I don't want to OC it too much if its going to damage the card.

Im getting about 315Mhash per core.
Im using poclbm, but i just read about pheonix and modified poclbm.
Would those give me better hash rate?

Thanks and please excuse me if these are noob questions.
Jump to: