Bitcoin Forum
December 05, 2016, 08:40:47 AM *
News: Latest stable version of Bitcoin Core: 0.13.1  [Torrent].
 
   Home   Help Search Donate Login Register  
Pages: [1] 2 »  All
  Print  
Author Topic: Sign of 5970 failure?  (Read 2249 times)
Cobra
Sr. Member
****
Offline Offline

Activity: 289



View Profile
May 04, 2012, 05:08:59 PM
 #1

I have a Win7 rig with a couple Diamond 5970's that have been cranking away for months, no problems at all. Both cards run about 380 Mh/sec per core clocked at 830 core and 300 memory at stock voltage, stable for months.

Now without any changes 1 of the gpu cores looks to be having an issue: With the above clocks it was jumping all over between 550-830Mhz, gpu load cutting in and out. I back the clock down to 820/300 and it is cranking away but load is fluctuating between 90-96% (previously pegged at 99%), 64c temp, fan is fine but it is only generating 304Mh just on this core. At present after a cold boot the speed quit fluctuating and it is stable at 820/300 as per GPUZ but lost its gusto for some reason. Other core on this card is completely normal.

I tried changing clocks back to stock at 725/1000 and when starting a new mining session it will still not hit full load.
Running at 725/1000 a core is making 304 Mh/sec is that typical for stock? When overclocked to 820/300 GPUZ would report that speed and I would still only get 304 Mh/sec. Has anyone seen something like this?

Power supply is Rosewill 1300 Lightning, powering only these 2 cards.

I bought the card on ebay and have sales receipt from previous owner and registered the card. It is refurb but should still be under warranty. In this case if I try to send it back would they even have a card to send me to replace this? I can imagine if they did find one laying around it would take months to get a working card back.

Based on the above info would you consider the card bad and would Diamond agree?

Thanks in advance for any feedback

Edit: Now it looks like another GPU is doing the same thing, so narrowing down to either whole card going, or power supply. In cgminer I have gpu 0-3. And 0 and 2 are now problematic, are these the same card? And the clocks on this new problematic core are jumping all over the place and the core as listed in the OP is stable at 725/1000. So the card could have a power issue or something with VRM's?
1480927247
Hero Member
*
Offline Offline

Posts: 1480927247

View Profile Personal Message (Offline)

Ignore
1480927247
Reply with quote  #2

1480927247
Report to moderator
1480927247
Hero Member
*
Offline Offline

Posts: 1480927247

View Profile Personal Message (Offline)

Ignore
1480927247
Reply with quote  #2

1480927247
Report to moderator
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction. Advertise here.
1480927247
Hero Member
*
Offline Offline

Posts: 1480927247

View Profile Personal Message (Offline)

Ignore
1480927247
Reply with quote  #2

1480927247
Report to moderator
P4man
Hero Member
*****
Offline Offline

Activity: 504



View Profile
May 04, 2012, 05:20:57 PM
 #2

Test the card with Furmark or something. I do suspect its the gpu or vram going bad, but I wouldnt use bitcoin benchmarks to get an RMA.

Cobra
Sr. Member
****
Offline Offline

Activity: 289



View Profile
May 04, 2012, 07:03:09 PM
 #3

Test the card with Furmark or something. I do suspect its the gpu or vram going bad, but I wouldnt use bitcoin benchmarks to get an RMA.
Something is off, was only getting 13fps with the default 1080 preset in that app. I didn't see a way to see more specifics with that app.

So after going from driver 12.3 (originally working fine) went to 12.4 (no good at all, opencl didn't detect cards), now went to 11.10. Now with same config hit with the cpu bug (cgminer 2.4) and 3 out of 4 gpu's are fine. No issue with cpu on 12.3 driver so I am back on that.

I'm convinced it is the card for now. I will look for another app that may be able to detect what specifically might be wrong, that app did show poor performance but that's it.
I'm going to run it for rest of month as is in the crippled capacity until I shutdown operations for summer.  With power price jumping to .19/kwh too much even for downclocking in my opinion.
oldDirty
Full Member
***
Offline Offline

Activity: 182


roundhouseminer


View Profile
May 05, 2012, 12:24:33 AM
 #4

You can have a look with gpucapsviewer, maybe it can give some info's.

CrunchingCulture

/// FREE ASSANGE \\\
Cobra
Sr. Member
****
Offline Offline

Activity: 289



View Profile
May 05, 2012, 05:12:38 AM
 #5

You can have a look with gpucapsviewer, maybe it can give some info's.
Nice benchmark tool. I ran the OpenCL benchmarks on all 4 GPU's at 1920X1080 8XAA and it ran fine on all of them at about 60fps when clocked at 830/300.

It looks like I will have a hard time proving this card is defective, so far all I have is the degraded mining performance on one core when it was fine for months.
Maybe lack of 65Mh/sec is not worth making a fuss over if I can't find any additional solid proof of failure.
Garr255
Legendary
*
Offline Offline

Activity: 952


What's a GPU?


View Profile
May 05, 2012, 05:22:34 AM
 #6

For benchmarking, run this 4k: http://ftp://ftp.untergrund.net/users/thygrion/yup-ethos.zip

xD

“First they ignore you, then they laugh at you, then they fight you, then you win.”  -- Mahatma Gandhi

Average time between signing on to bitcointalk: Two weeks. Please don't expect responses any faster than that!
MrTeal
Legendary
*
Offline Offline

Activity: 1246


View Profile
May 05, 2012, 05:41:58 AM
 #7

I assume you checked the VRM temps in GPU-Z? The 5970s start trottling at 125C on the VRMs, which can happen when the GPU itself is much cooler.
Cobra
Sr. Member
****
Offline Offline

Activity: 289



View Profile
May 05, 2012, 11:33:06 PM
 #8

I assume you checked the VRM temps in GPU-Z? The 5970s start trottling at 125C on the VRMs, which can happen when the GPU itself is much cooler.

You win! I have to admit I overlooked this. I was too focused on looking at GPU temp.  It is hitting 125-130C, all of the others are between 70-80c.
That may be fixable by replacing thermal pads right? and may not actually be a bad GPU or when this happens it is a sure sign of failure? If so maybe that will be good enough proof to get an rma.

This also explains why a new mining session will be fine for a minute or so at expected hashrate and drop off a minute or so in.

Thank You for pointing this out to me, I will send a 1BTC token of appreciation your way for solving this mystery for me (or asking the right question). Address?

MrTeal
Legendary
*
Offline Offline

Activity: 1246


View Profile
May 06, 2012, 12:10:04 AM
 #9

I assume you checked the VRM temps in GPU-Z? The 5970s start trottling at 125C on the VRMs, which can happen when the GPU itself is much cooler.

You win! I have to admit I overlooked this. I was too focused on looking at GPU temp.  It is hitting 125-130C, all of the others are between 70-80c.
That may be fixable by replacing thermal pads right? and may not actually be a bad GPU or when this happens it is a sure sign of failure? If so maybe that will be good enough proof to get an rma.

This also explains why a new mining session will be fine for a minute or so at expected hashrate and drop off a minute or so in.

Thank You for pointing this out to me, I will send a 1BTC token of appreciation your way for solving this mystery for me (or asking the right question). Address?



1GBfFqd3t1Gpm2TcNUfy6T4vKZDUPnM5os

You should be able to replace the thermal pads, and that might help. The two main row of VRMs require a 1mm pad, while the one isolated all by itself needs a 0.5mm one, although squishing a 1mm one in there probably wouldn't hurt anything. I force the fan speed on mine, just because the card controls the fan speed only on the GPU temps, and doesn't increase the speed even if the VRMs are overheating.
Cobra
Sr. Member
****
Offline Offline

Activity: 289



View Profile
May 06, 2012, 12:36:40 AM
 #10


1GBfFqd3t1Gpm2TcNUfy6T4vKZDUPnM5os

You should be able to replace the thermal pads, and that might help. The two main row of VRMs require a 1mm pad, while the one isolated all by itself needs a 0.5mm one, although squishing a 1mm one in there probably wouldn't hurt anything. I force the fan speed on mine, just because the card controls the fan speed only on the GPU temps, and doesn't increase the speed even if the VRMs are overheating.

Tip Sent.....http://blockchain.info/tx-index/4904256/d84a4e859c746a9e8c14eba6bca7f7bf2d1e2a5f8f130d48ef038829787c6ac2

I will probably see if it is possible to get an RMA first since it is under warranty, registered, and have original receipt. I imagine it may be a long while before anything happens, they will probably send me a power hungry 6990 or something if they can't find a 5970, if so will be putting that up for sale. If they deny the rma for whatever reason I will try those steps, thanks for the specifics on that process.
Vwlopez3
Member
**
Offline Offline

Activity: 78


View Profile WWW
May 06, 2012, 04:31:57 AM
 #11

I had a similar problem with the VRM's of one of the GPU's on the 5970 shooting through the roof. I tried watercooling but eventually it creeped up. Decided best approach was just not to mess with the voltage of that card. RMA is the best choice though.

19G8b28qHHM7yACCwwWfikGv7QimEsHyga
jjshabadoo
Hero Member
*****
Offline Offline

Activity: 519


View Profile
May 08, 2012, 01:23:49 AM
 #12

You can try, but you're unlikely to get an RMA from diamond. For sure I can tell you that if you registered it with them and put ebay as the vendor they will not honor the warranty. Ebay people who say warranty should get the scam label.

Also, diamond doesn't have any 5970's I heard they are offering a 6970, which is bullshit or making you pay for a 7970, which is also bullshit.

XFX gave me a 6990 which looked brand new for a 5970 I RMA'd with them.

I have a diamond RMA in  process right now and will post in this section when i hear back.
Cobra
Sr. Member
****
Offline Offline

Activity: 289



View Profile
May 08, 2012, 05:40:37 AM
 #13

You can try, but you're unlikely to get an RMA from diamond. For sure I can tell you that if you registered it with them and put ebay as the vendor they will not honor the warranty. Ebay people who say warranty should get the scam label.
Thanks for the info. I did get the card from Ebay but the original receipt has Newegg on it so that is how I registered the card. And yes please do let me know how the RMA with Diamond goes and what they offer you.  I had hoped they would offer a 6990, I will report back what I am told as well. If anything will keep on running it in a slightly crippled capacity, 684 Mh/sec isn't too bad.

Aseras
Hero Member
*****
Offline Offline

Activity: 658


View Profile
May 08, 2012, 03:28:08 PM
 #14

fat chance of getting a 6990. they'll probably just ship you a 6950 or 6970 and claim it's comparable and tell you to F off.
bulanula
Hero Member
*****
Offline Offline

Activity: 518



View Profile
May 08, 2012, 08:22:18 PM
 #15

You can try, but you're unlikely to get an RMA from diamond. For sure I can tell you that if you registered it with them and put ebay as the vendor they will not honor the warranty. Ebay people who say warranty should get the scam label.

Also, diamond doesn't have any 5970's I heard they are offering a 6970, which is bullshit or making you pay for a 7970, which is also bullshit.

XFX gave me a 6990 which looked brand new for a 5970 I RMA'd with them.

I have a diamond RMA in  process right now and will post in this section when i hear back.

Warranty on fleabay ? LOL.

And rightly so. Guy probably blew a VRM from too much overclocking.

If I was manufacturer I would refuse warranty and give 0 compensation. I doubt it was manufacturer that did not apply cooling to VRM.

This "break it" and "RMA it" behaviour is causing higher prices for all of us, including people like me who OC sensibly and don't run to the manufacturer when I screw a device ...
Cobra
Sr. Member
****
Offline Offline

Activity: 289



View Profile
May 09, 2012, 12:55:50 AM
 #16

Warranty on fleabay ? LOL.

And rightly so. Guy probably blew a VRM from too much overclocking.

If I was manufacturer I would refuse warranty and give 0 compensation. I doubt it was manufacturer that did not apply cooling to VRM.

This "break it" and "RMA it" behaviour is causing higher prices for all of us, including people like me who OC sensibly and don't run to the manufacturer when I screw a device ...

I did not say anything about a warranty from Ebay. If you look at my post a few up from here you will see that I said I registered it with the Newegg receipt that I received with the card when I got it from Ebay.

I feel my overclocking was safe and sensible. 830core and 300memory at STOCK VOLTAGE is pretty safe, ran great for months. The other Diamond I have is still going strong, no problems. It could be an issue that these were refurbs to begin with. GPU temps never above 72c and ambient room temp at 70F, open frame rig, box fan on high blowing directly on the cards.

It doesn't look like I'm the only one that will send the card back if it is under warranty, read the previous posts. I would have agreed with you if I would have jacked the voltage or did any "extreme" overclocking. In that case I would not attempt an RMA.

It was so nice of you to stop by and make this contribution. I should have locked the thread after Mr. Teal provided the answer. Someone with a scammer tag comes and talks about ethics, classic. I don't care how that happened or why and will not investigate it to find out more. I simply don't care.
Aseras
Hero Member
*****
Offline Offline

Activity: 658


View Profile
May 09, 2012, 03:17:02 PM
 #17

the VRM on the 5970 is a weak link. they tried to save a buck a card by going to a 3 phase power rather than the 4 phase power that is on the 5870 and many cards VRM run way too hot and die a early death because of it.

I consider it a design defect, the manufacturers should eat the cost of the rma's luckily in the 5970 days many offered lifetime or 5 year warranties. many companies dont do that anymore, especially after the nvidia chip fiasco and various problems with ATI.
bulanula
Hero Member
*****
Offline Offline

Activity: 518



View Profile
May 09, 2012, 03:38:58 PM
 #18

the VRM on the 5970 is a weak link. they tried to save a buck a card by going to a 3 phase power rather than the 4 phase power that is on the 5870 and many cards VRM run way too hot and die a early death because of it.

I consider it a design defect, the manufacturers should eat the cost of the rma's luckily in the 5970 days many offered lifetime or 5 year warranties. many companies dont do that anymore, especially after the nvidia chip fiasco and various problems with ATI.

Indeed. This is why I avoid 5970s and only have 5870s.

Wondering if 7990 will be a compromise to save $$$ again when GTX 690 totally owned all ATI has and almost 96% peformance of SLI GTX 680.

They can't afford to skimp on the 7990, I think ...
oldDirty
Full Member
***
Offline Offline

Activity: 182


roundhouseminer


View Profile
May 09, 2012, 05:43:52 PM
 #19

But the 5970 gives you a big bonus, 1600shaders x2 on one PCIe Slot. A single 5870 can never compare to this, no matter how high it is clocked.
I prever to run the 5970 in stock gpu clocks, reduce the memoryspeed and be fine.

CrunchingCulture

/// FREE ASSANGE \\\
The-Real-Link
Hero Member
*****
Offline Offline

Activity: 530



View Profile
May 14, 2012, 11:08:31 PM
 #20

You can try, but you're unlikely to get an RMA from diamond. For sure I can tell you that if you registered it with them and put ebay as the vendor they will not honor the warranty. Ebay people who say warranty should get the scam label.

Also, diamond doesn't have any 5970's I heard they are offering a 6970, which is bullshit or making you pay for a 7970, which is also bullshit.

XFX gave me a 6990 which looked brand new for a 5970 I RMA'd with them.

I have a diamond RMA in  process right now and will post in this section when i hear back.

Just had two of my cards finally be replaced after a month+ of downtime / failure.  Diamond said they were both outright overheating and crashed their systems (after they slowly wouldn't mine for me anymore).  Got two 6970s in return which, while may not be perfect, is better than 0 MH/s.  Indeed I'd have to pay to step up to 7970s (which I wonder how much it would've been)...

Yeah you really have to watch the VRMs on the 5970s.  Even in a cool room they will still run ~20C or so hotter than whatever the hottest temp of the cores are.  You might be able to push them to 130C but they'll quickly throttle back. 

While I've not personally replaced my thermal pads, I do have the cards running in a cool room finally.  With the VRMs, the lower temp the better Wink

Oh Loaded, who art up in Mt. Gox, hallowed be thy name!  Thy dollars rain, thy will be done, on BTCUSD.  Give us this day our daily 10% 30%, and forgive the bears, as we have bought their bitcoins.  And lead us into quadruple digits
Pages: [1] 2 »  All
  Print  
 
Jump to:  

Sponsored by , a Bitcoin-accepting VPN.
Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!