5970 mining thread

[-ck]

Whatever, a GPU pushed too hard will produce errors, not rejects.

[1715028127]

1715028127

[1715028127]

1715028127

[1715028127]

1715028127

[Proofer]

Forking thread from Mining Support to here...

also,  who said that running the fan at 100% is good...  good for the guy that is going to sell you another card maybe.

Generally good idea to keep card as cool as possible for longest lifespan. Fans are cheap to replace and if fan fails, the card will soon shut off, likely without damage. I've had 7 5850s running 100% w/ fan @ 100% since last April, no problem. Heat will also reduce PSU efficiency in some cases, and max wattage output.

100%?  Comments?  (Replacement fans are US$ 15 each on eBay.)

[P4man]

Fan bearings will wear out sooner or later. I strongly suspect the relationship with fan speed is at least linear, possibly super linear as vibration is likely a very important factor (avoid speeds which produce more mechanical noise, humming etc).

Of course you can get lucky. Or maybe that poster has some advantages most other miners do not, like a dust free environment or he lubricates them regularly. Either way there is little doubt in my mind a fan at 100% wont last as long as one on 50% and its not like fan failure is a rare thing. Its equally true low temps will help the card live longer, but it all depends how high your temps are. I get <60C with fans at 50% (non ref cards), its pointless to increase fan speed in those circumstances.

[minero1]

2 x 5970s + 3 x 5850s
Motherboard MSI 790FX-GD70
CPU PHenomII x4 965 (i know, it was a combo deal)
BAMT 0.5
cgminer 2.3.1-2
#01
core: 750MHz
mem: 250MHz
Vddc: 1.0 Volts
Temps: 66ºC
hashrate: 690MH/s

#02
core: 725MHz
mem: 250MHz
Vddc: 1.0 Volts
Temps: 66ºC
hashrate: 666MH/s

They were working on windows until i installed the third 5850, then the whole rig kept BSODing so i installed BAMT instead

[trouserless]

3x5970

all configured the same, MH/s is +/- 2% between all three

bamt 0.5
cgminer 2.3.1f
clocks/mem 835/300
vddc 1.050 (not 1.163 as I previously mistakenly posted)
temps: 68-72C
MH/s 758

[minero1]

3x5970

all configured the same, MH/s is +/- 2% between all three

bamt 0.5
cgminer 2.3.1f
clocks/mem 835/300
vddc 1.163
temps: 68-72C
MH/s 758

wonder whats you MHash/W with that amount of voltage

[trouserless]

3x5970

all configured the same, MH/s is +/- 2% between all three

bamt 0.5
cgminer 2.3.1f
clocks/mem 835/300
vddc 1.163
temps: 68-72C
MH/s 758

wonder whats you MHash/W with that amount of voltage

Two of them are in the same rig with two 5870s running at 950/300/1.05V for ~437MH each  - total with a 1250W PS pull 1KW +/- 10W.  The whole rig crunches ~2367MH/s which makes for nice round math @ 2.36MH/W

[portron]

2 x 5970s + 3 x 5850s
Motherboard MSI 790FX-GD70
CPU PHenomII x4 965 (i know, it was a combo deal)
BAMT 0.5
cgminer 2.3.1-2
#01
core: 750MHz
mem: 250MHz
Vddc: 1.0 Volts
Temps: 66ºC
hashrate: 690MH/s

#02
core: 725MHz
mem: 250MHz
Vddc: 1.0 Volts
Temps: 66ºC
hashrate: 666MH/s

They were working on windows until i installed the third 5850, then the whole rig kept BSODing so i installed BAMT instead

Anything over 11.12 with BSOD with more then 4 GPU's on windows.

[The-Real-Link]

  Hey guys, sorry to bump this general 5970 thread but had an interesting issue come up.

  Been mining with a few of them fine for months now.  Two were starting to show massive crashes and instability (810 core, 500 memory, stock volts) despite temps being in the low to mid 70s.  Bumped everything back down to stock but the issues remained.

  Thankfully they were both Diamond models from Newegg so I could show legit purchase receipts.  So anyway, pulled the cards, sent 'em back...and waiting.  This was maybe a month and a half ago.

  Once stuff arrived at their RMA office, I figure testing does take time so I wanted to be patient.

  Now that it's been a month, I finally was back in touch with Diamond seeing as how I've heard nothing via email back.  Their tech, Chris, was quite helpful and apologized.  Now here's the concern for others of us out there with these cards.

  He said normally if a card is bad (and both of mine were confirmed as being such), they'd replace it immediately (makes sense).  But Diamond themselves are still waiting from AMD for more cards and they are either not coming or taking forever.  As such, we can only wait.

  An upgrade was offered but since they also have no stock of 6990s, I can't go that route.  He said he'd look into offering a 6970 or a 7970 which, given the price, warranty and newness of a 7970, would seem stupid for me not to accept should they validate it.  

  I know I'd probably get maybe 75% of the hashing power of a 5970 out of a 7970 but at this rate, I can't see any other choices open to me.

  Just a heads up for anyone else running on these old cards.  Should they bite the dust, you might be in for quite a wait in terms of a replacement, if any, or maybe an upgrade.  

  While I figured this was true to an extent when buying the cards, I had presumed manufacturers still had some stock left.  Apparently not.  

[DeathAndTaxes]

Thanks for the info.   Yeah I figured the 5970 would eventually be out of stock but I always assumed they would give you a 6990.  A 6970 is a dog of a deal.  At least with 7970 you get some improves power efficiency and easier cooling (due to single GPU design).

[MrTeal]

Does anyone know of a source for the VT1157SF VRMs used in this cards? I had one go out, and need a replacement. Emailing Volterra didn't help, since they do not have distributors for their chips.

We do not have any distributors.   Volterra only sells to select approved customers under a NDA due to limited resources.  We intentionally limit the # of customers we work with in order to provide superior customer support. At this time, we are unable to engage with your company. We apologize for the inconvenience.

[Mousepotato]

3x5970

all configured the same, MH/s is +/- 2% between all three

bamt 0.5
cgminer 2.3.1f
clocks/mem 835/300
vddc 1.163
temps: 68-72C
MH/s 758

Why do you run them so high?

[Gomeler]

Does anyone know of a source for the VT1157SF VRMs used in this cards? I had one go out, and need a replacement. Emailing Volterra didn't help, since they do not have distributors for their chips.

We do not have any distributors.   Volterra only sells to select approved customers under a NDA due to limited resources.  We intentionally limit the # of customers we work with in order to provide superior customer support. At this time, we are unable to engage with your company. We apologize for the inconvenience.

Good luck. Volterra makes great VRMs but they are a pain in the ass to deal with. EVERYTHING is under NDA from them. People have spent countless hours searching for the pinouts on their controllers to remove OCP/OVP. If the documentation is that difficult to acquire I imagine the chips are impossible to get short of yanking them off another card.

[DeathAndTaxes]

If the documentation is that difficult to acquire I imagine the chips are impossible to get short of yanking them off another card.

That is likely the only place you are going to find them.  Ebay routinely has dead (for parts only) 5870s and 5970s.  Tend to go for $150 or so each.  Annoys me when I am searching because I think I found a deal.

[zvs]

3x5970

all configured the same, MH/s is +/- 2% between all three

bamt 0.5
cgminer 2.3.1f
clocks/mem 835/300
vddc 1.163
temps: 68-72C
MH/s 758

Why do you run them so high?

1.163?  wtf..

i have 5 5970's, and 7 of the cores can do 830 @ around 1.02V

[MrTeal]

Interestingly, I removed the offending VRMs, and I went from measuring a short across the GPU to a more expected ~4 ohms. I just couldn't wait to win an auction on a parts card so I buttoned it up and fired it up. It won't display an image and I can't control the fan, but the other GPU is detected and I can mine on the other GPU. I was expecting the BIOS to detect the missing VRMs and simply not post.

Now I'll just wait to try and pick up a completely dead card for $20 if I can. Trying to repair the extra GPU isn't worth dropping $100 on a card for 3 VRMs.

[this time]

Interestingly, I removed the offending VRMs, and I went from measuring a short across the GPU to a more expected ~4 ohms. I just couldn't wait to win an auction on a parts card so I buttoned it up and fired it up. It won't display an image and I can't control the fan, but the other GPU is detected and I can mine on the other GPU. I was expecting the BIOS to detect the missing VRMs and simply not post.

Now I'll just wait to try and pick up a completely dead card for $20 if I can. Trying to repair the extra GPU isn't worth dropping $100 on a card for 3 VRMs.

How do you measure a short across a 5970 gpu? I have a 5970 with a bad core.

[MrTeal]

Interestingly, I removed the offending VRMs, and I went from measuring a short across the GPU to a more expected ~4 ohms. I just couldn't wait to win an auction on a parts card so I buttoned it up and fired it up. It won't display an image and I can't control the fan, but the other GPU is detected and I can mine on the other GPU. I was expecting the BIOS to detect the missing VRMs and simply not post.

Now I'll just wait to try and pick up a completely dead card for $20 if I can. Trying to repair the extra GPU isn't worth dropping $100 on a card for 3 VRMs.

How do you measure a short across a 5970 gpu? I have a 5970 with a bad core.

I just measured using a DMM across one of the ceramic caps on the back of the GPU. With the bad VRMs the card wouldn't boot though,so your issue is likely not the same. Doesn't hurt to test though, you don't even have to remove the backplate.

[johnyj]

Did anyone tested 5970 voltage stages and what is the maximum stable clock for each stage?

This one is a big heat and noise generator, I guess reduce the voltage can help a lot ...

[trouserless]

3x5970

all configured the same, MH/s is +/- 2% between all three

bamt 0.5
cgminer 2.3.1f
clocks/mem 835/300
vddc 1.163
temps: 68-72C
MH/s 758

Why do you run them so high?

1.163?  wtf..

i have 5 5970's, and 7 of the cores can do 830 @ around 1.02V

I messed that up, I have 5870's in that rig and read those voltages instead (which are 1.163).  The 5970's are all set to 1.050.  Good eye