Author

Topic: 5970 - VRM Temps @ 105C? (Read 1866 times)

hero member
Activity: 1078
Merit: 502
June 24, 2012, 12:23:46 PM
#14
Things seem to work well with the 1/3rd Mem/engine clock speeds Smiley
legendary
Activity: 952
Merit: 1000
June 24, 2012, 12:19:53 PM
#13
the mem clock should be 1/3rd of your Engine clock Smiley

should be? I've seen no difference on my 5870s and 5970s running at 150, 160, 300, or 1/3 engine. I just leave them all at 300 (seems the most stable).
hero member
Activity: 1078
Merit: 502
June 24, 2012, 09:16:06 AM
#12
the mem clock should be 1/3rd of your Engine clock Smiley
hero member
Activity: 518
Merit: 500
June 24, 2012, 08:24:51 AM
#11
how can i read the VRM temps under linux (BAMT)?  Huh

You can't. Use Windows.
yxt
legendary
Activity: 3528
Merit: 1116
June 22, 2012, 05:49:27 AM
#10
how can i read the VRM temps under linux (BAMT)?  Huh
legendary
Activity: 952
Merit: 1000
June 12, 2012, 09:36:54 PM
#9
VRM temps are going to be higher than core temps. 105 is actually pretty normal.

Having said that, I don't like them that high, esp for 24/7 mining. @860,300, there's no reason you can't undervolt. Try 0.95-1.00 range. Puts way less stress on the VRMs.

*scratch* I have to drop card to 740-150 to be stable at 0.95v

Each card is different. Even going to 1.05 for 840,300 will still lower VRM temps.

oh, for curiosity sake... why does everyone use 300mem?  for me, on SDK 2.1, 840/300 w/ 256 worksize is about .5% improvement over 840/150 w/ 128 worksize..

300 is the mem speed when the card is at idle (lowest voltage). Any speed below 300 doesn't reduce power consumption in any noticeable way. For you, it's whatever's fastest. Some run 150, 160, 300, or some say 1/3 of your engine clock. I think it depends on your core speed, SDK, kernel, vectors, and worksize.

I run -w 256 -v 2 @ 925,300 @ 0.953V on SDK 2.5 with 11.4 drivers. I've tried 150, 160, and 924,308, and I can't see any difference. 300 is the most stable for me, so I stick with that. That answer your question? Tongue
zvs
legendary
Activity: 1680
Merit: 1000
https://web.archive.org/web/*/nogleg.com
June 12, 2012, 07:09:17 PM
#8
VRM temps are going to be higher than core temps. 105 is actually pretty normal.

Having said that, I don't like them that high, esp for 24/7 mining. @860,300, there's no reason you can't undervolt. Try 0.95-1.00 range. Puts way less stress on the VRMs.

*scratch* I have to drop card to 740-150 to be stable at 0.95v

Each card is different. Even going to 1.05 for 840,300 will still lower VRM temps.

oh, for curiosity sake... why does everyone use 300mem?  for me, on SDK 2.1, 840/300 w/ 256 worksize is about .5% improvement over 840/150 w/ 128 worksize..
zvs
legendary
Activity: 1680
Merit: 1000
https://web.archive.org/web/*/nogleg.com
June 12, 2012, 07:06:59 PM
#7
as for me...having 2*5970
1st - @840 1.049v - no more overclocking
2nd - @840 1.062 - no more (850 but at 1.075v)

each cooling system was tunned: old termalpads were replaced with new ones, MX-4 spot at termalpads/VRM, removed still backplate (now PCB is openned - no more heat air beatwean still backplate and PCB near VRM`s zone), and 3400 rpm 92mm fan blowing to the PCB VRM`s zone.
after that - 86 C.
It`s real to beat the 90C or lower.  Wink

PS 105C very bad - guaranted crash of DrMos Volterra in mounth or more.  Cry


well, at least someone has cards similar to mine then

after reading some of these posts, i thought i must have the worst 5970s in existance
legendary
Activity: 2026
Merit: 1005
June 12, 2012, 04:43:33 PM
#6
as for me...having 2*5970
1st - @840 1.049v - no more overclocking
2nd - @840 1.062 - no more (850 but at 1.075v)

each cooling system was tunned: old termalpads were replaced with new ones, MX-4 spot at termalpads/VRM, removed still backplate (now PCB is openned - no more heat air beatwean still backplate and PCB near VRM`s zone), and 3400 rpm 92mm fan blowing to the PCB VRM`s zone.
after that - 86 C.
It`s real to beat the 90C or lower.  Wink

PS 105C very bad - guaranted crash of DrMos Volterra in mounth or more.  Cry
sr. member
Activity: 373
Merit: 250
June 12, 2012, 02:43:46 PM
#5
Just undervolt as much as possible while keeping at the clocks you want to keep them at.  Don't want your VRMs blowing out and ending up with a dead card.
legendary
Activity: 952
Merit: 1000
June 12, 2012, 09:26:11 AM
#4
VRM temps are going to be higher than core temps. 105 is actually pretty normal.

Having said that, I don't like them that high, esp for 24/7 mining. @860,300, there's no reason you can't undervolt. Try 0.95-1.00 range. Puts way less stress on the VRMs.

*scratch* I have to drop card to 740-150 to be stable at 0.95v

Each card is different. Even going to 1.05 for 840,300 will still lower VRM temps.
zvs
legendary
Activity: 1680
Merit: 1000
https://web.archive.org/web/*/nogleg.com
June 12, 2012, 09:24:39 AM
#3
VRM temps are going to be higher than core temps. 105 is actually pretty normal.

Having said that, I don't like them that high, esp for 24/7 mining. @860,300, there's no reason you can't undervolt. Try 0.95-1.00 range. Puts way less stress on the VRMs.

*scratch* I have to drop card to 740-150 to be stable at 0.95v
legendary
Activity: 952
Merit: 1000
June 12, 2012, 07:48:31 AM
#2
VRM temps are going to be higher than core temps. 105 is actually pretty normal.

Having said that, I don't like them that high, esp for 24/7 mining. @860,300, there's no reason you can't undervolt. Try 0.95-1.00 range. Puts way less stress on the VRMs.
member
Activity: 81
Merit: 10
June 12, 2012, 02:34:38 AM
#1
I'm just curious at whether this is bad for my card.
I currently have my card (5970) clocked at 860 Core, 300 Memory and all temps seem to be fine.
Mining at around 790 - 800 mhash/s.
Core is @ 50C.
VRMs are at around 100 to 107C.

Is this a bad range for my VRMs? I am mining 24/7.
Jump to: