Bitcoin Forum

Is anyone else having trouble with their Diamond 5970 (from newegg)
Did you have to falsh the bios and replace thermal paste and add pads to VRM to make basic use of them ?

I can't raise voltage using atitweak nor radeonVolt,

On one of them, one of the core can't overclock past ~760mhz on 1.05v

First core get VRM throttled constantly to 550mhz at default voltage / default clock. (with temp at 76c)

How is this acceptable ?

Edit : They also look far from new : http://www.eggxpert.com/forums/ShowThread.aspx?PostID=777284#777284


TIM application ; I have no idea what species could have messed up the TIM this much.

There are holes with no paste on the GPUs with corresponding hole on the heatsink,  This is BAD.

http://imgur.com/a/0THwO#1
BIG:
https://i.imgur.com/HH3z6.jpg
https://i.imgur.com/00sbP.jpg

Note : I have not touched anything, only removed the heatsink. - These card are sold as new.

Watching, since I have the same card (but haven't been able to set it up yet).

Try AMD GPU Clock tool. Don't know if it works with 5970 but it's worth a try.

I think cgminer has built-in overclocking functionality as well.

I need to overvolt this on Linux.

Used cards for sure. Have you monitored each vrm temp? I'd wager a vrm isn't working (indicated by lower temp) if you can't get past 1.05v

Yup, sure looks like a used card.
That TIM on the GPUs isn't original; all ref 5970s I pulled the fansink off of had phase-change pads, not goop.
Also, good idea with the dead phase. Check VRM temps under load, if one of the 3 phases is a lot (20C+) cooler than the others you got a dead phase; the card will generally still run but unstable and OC like crap.
If that's not the case... idiot who replaced the GPU TIM didn't replace the thermal pads for the VRMs. Most important ones are the one for the core VRMs. the long one under the fan and the smaller of the strips above the vapor chamber. inductor, ram and pcie bridge pads aren't really that important. Used a bunch of a 1mm sheet of 3M 5591 I had lying around. worked like a charm.
In linux a slightly modified radeonvolt works for reading VRM temps and overvolting 5970s.
As for the holes in the paste... bet the monkey who applied the paste touched the die and didn't degrease it before applying new paste. that generally ends up looking *exactly* like this.

I have replaced the TIM on both GPU with MX-2 and added some to the pad I didn't replace.

I replaced only the pads on VRM with some ram pad I scavenged from a broken 4870.

The same GPU #1) that had lower clock is now defective at stock clock, ; the fan ramp up 100% after 2-3 share and a reboot is needed.

It look like the jelly pads from the other card are worst performing than what was already there.

I don't know what VRM is assosiated with what core, but the 3 VDDC in GPU-Z are arround 90c with one GPU(#2) running.

Fell so bad for you...

Looking to make some bucks and newegg burned you.

Time to replace the cards

Aye, that is rough. No reason the VRM should be at 90c with only 1 GPU going. Are all 3 VRM reporting about the same temp?

Have you contacted Newegg? There is 30day replacement there. And should be a warranty with Diamond if you can't accomplish anything with newegg.

I would return the damn cards, warranty from Diamond is poop

Wow, A company worse than XFX, never thought id see the day :)

Feel bad for you mate, Them duel cards are not cheap!

I have a few of these and mostly have found them to be okay so far. You DO need to replace the TIM and pads though. I have been working on this and will post when I find a great recipe. I also would appreciate any advice from people here. I found using 1mm thermal pads from phobya did not work if you use them on all ram, vrm, etc. The heatsink wouldn't make proper contact with the gpu chip. My next try will be 1mm pads on VRM and .5mm on everything else because I was not getting good contact with the GPU chip when I used 1mm pads all around.

For the record, mine looked new for each one I bought and I have 7. I currently have four running at 800/300 24/7 and they have been doing well with the stock heatsink, paste and pads(have not taken them apart yet).

I took apart the other four(1 is amd brand) in hopes of achieving better cooling and higher clocks. I'm shooting for a 24/7 stable 850/300 on air. I'll add that the rigs are in a cold basement that doesn't get above 15c or so even in the summer.

Once I start working on them again I will take some pics and post what I did, maybe we can create a work log thread for these cards.

Also, you better register them with diamond for the 5 year warranty ASAP just in case. Their customer service might suck, but you will get a new card or refund eventually.

How many time do i have to say this, Can you read?

1mm - ram
0.5mm VRM

Read what? Your mind?

I don't see you mentioning this anywhere else in this thread.  Perhaps you've mentioned it several times in other threads, but you shouldn't expect others to read your post history.

I expect anyone competent enough to actually redo their cards to search for the subject they dont know about.

Seriously, when you remove the OEM HSF, dont you see the different pads with different thickness?

Monkeys talk monkeys do

I mixed up ram and VRM in my post. Glad elastic pointed out even if his delivery was impolite. I wouldn't want anyone to mix that up and fry their card.

I'm known for being an ass. Nothing personal.

No worries, hell you got me straight with my cards and that's all I care about.

I've redone four of them now with shin-etsu X23-7783D on the GPU's, 1mm phobya thermal pads on the ram and .5mm fujipoly thermal pads on the vrm.

Cards have been clocked at 800/300 for over 24 hours with no shutdowns in an open rig. Bottom card max temps of 50c or so and top card at 55c. Once I get the extenders going and get them out of the open case I should be good to go.

Once I learned to heat up the shin-etsu in a glass of hot water I was able to spread it perfectly with a thin plastic card.

I'm not going to mess with my other 4 5970's just leave them stock and compare them to the ones I just redid see if all this was worth it.

The thermal pad I used are so elastic I don't think they'd have problem compressing between 1mm and 0.5mm

you might be right, the phobya pads have a rating of 7watt or something like that. They do not give very much so they didn't work on the VRM. Heatsink was not making contact with GPU.

Good luck.

Would using this on VRM a waste of time ?

http://ca.startech.com/Computer-Parts/Fans/Thermal-Pad-Heatsink-Paste-Alternative-Package-of-5~HSFPHASECM

look like shit, but that's all I can get around here,

0.7 W/m-k the Product Height is said to be 0.3mm at some places at others 0.5mm
Phase Change Temperature: 58°C

Have you tried resetting the fromulator? I've heard newegg throttled them to be able to sell them cheaper. You could always try replacing the MOSFET rectifier, they seem to go bad on most Diamond products pretty easily.

I'm sorry it wasn't exactly accurate, all I can say is that those cards were WAY to good to be true. When I saw them, I nearly fell over. At this point have you considered Watercooling? It seems that may be one of your only options in regards to cooling. As for the voltage locks and such, if fixing the temps don't work, do you think it could be a BIOS problem? A newegg RMA might be the best option here, and just hope they send you a good one.

Best of luck!

Bumping this because I just replaced the fan, thermal pads, and thermal paste on one of my 5970s.

The card in question looked pretty bad:

http://chrono.firex.org/images/g5970-1.jpg
http://chrono.firex.org/images/g5970-2.jpg
http://chrono.firex.org/images/g5970-3.jpg

Maybe there is variance in the design, but advice in this thread to use 1mm pads on the RAM and 0.5mm pads on the VRMs was wrong for this card.  I am also using phobya pads and with 1mm on the RAM everything was pushed up much too high and the 0.5mm on the VRMs barely made contact.  Booting the card confirmed this as GPU temps were very high.  

I pulled everything apart and opened up another 5970 for reference.  The memory padding did not look like 1mm, so I ended up using 0.5mm pad on all the ram and the VRMs.  The only place I used 1mm was in the two recessed areas that had thermal padding.  It looks like the controller chip between the GPUs had a 1mm pad as well but I'm using a 0.5mm there as well and I don't feel like pulling the card apart again because I'm out of AS5.

Temps are now fantastic.  Prior to this, the card would barely run stable mining at 725mhz with temps in the ~74C range, much higher than any of my other 5970s.  Now its running 800mhz at ~55C and the thermal paste hasn't even fully cured yet.  :)

so you used .5mm everywhere? I used 1mm just on the vrm chips(the little ones or whatever the fuck they are called) and .5mm everywhere else. of course i used the 1mm in those recessed areas you are talking about as well.

Then I used shin-etsu thermal grease on the GPU chips. They have been running at 820-410 stable for a while now with two on one msi 890fx board. One in first slot and one in last pci slot.

Seems to be okay, but I haven't gotten to messing with that radeonvolt thing you can supposedly do in linux to measure the gpu ram temps.

oh yeah I used .5 on the controller chip also.

on cgminer settings --gpu-fan 85 --gpu-engine 820 --gpu-memclock 410

gpu 0 66.5c
gpu 1 65.5c
gpu 2 72.5c
gpu 3 62.5c

now the one card that read 72.5 has just always had a higher reading, even when it was stock. I'm not sure if the heatsink just doesn't sit right there or it's the chip or whatever. I've redone four cards now in the same manner and they all are closer to 60c with pci-e extenders and 65c if they are on a board with a little bit of spacing, like one or two slots between the cards.

The two cards we did with pci extenders actually sit closer to 55c, but that's at 800/300.

anyone else notice the plastic sheeting all around the stock heatsinks that is slowly melting and looks  like maybe just MAYBE it was supposed to be taken off before they put the heatsink onto the gpu board?

These diamond cards we definitely just fucking whacked together in some factory and thrown out there for sale. I highly doubt there is any need for plastic sheeting to be adhering to the heatsink all around the thermal pads and copper heatsink plates. What purpose could it possibly serve? Except to protect the heatsink in transport before it was supposed to be secured to the GPU board.

I wish I knew how to post pictures on this board to show everyone.

Huh, I've never seen sheeting like that on any card I own. Goddamnit Diamond.

I had nothing but problems with a pair of 5970, it would randomly freeze my computer (dead keyboard, hard reboot) every 24 hours or less. I tried everything including running at stock speed(1.2gh) without success. Had a industrial blowfan at max setting blowing directly at the intake, temp never got higher than 80C.

So happy finally got rid of them, i have 2x6870 coming now, got them from amazon(for easy return) this time, If this also craps out i am done with bitcoin mining.

I did learn that the "plastic sheeting" is actually a mylar sheeting and is supposed to offer some type of thermal protection or something. i'm not sure what the hell it could be protecting, since the area with the sheeting doesn't come into contact with anything, but whatever.

I believe the idea of the plastic-looking material is to insulate against electricity, not heat.  The idea is to prevent components on the PCB from shorting against the heatsink.  I was looking at this when I had to sand down a defect on a 5870 (they were all like this).  Once I saw that most of the aftermarket waterblocks were just using exposed copper throughout the whole contact side of the block, I didn't worry about it so much and broke out the sandpaper.

Over-engineering at its finest.

Diamond said I voided the warranty when I open up the card. I did this because the fan sounded like a jet engine, one of them would not even spin it was so trashed.. That's when I discovered the card was full of dust, and the fan bearing was shot. Newegg claimed they would replace if as soon as the new ones came in(they were shocked of course). There not even on the page now. Diamond never sent the rma papers to me as promised. Good thing I bought 4 of these fixer uppers. The lady at newegg told me it was supposed to be used, "thats what the OEM means". I actually had to make her look up the definition of OEM before she believed me. I bet they sold over a thousand of these things. They definitely spent a lot of time cleaning the dust from the fans so they appear new from the outside. What a pain putting new fans, thermal pads and paste on these things. I get upset just thinking about this.

Those cards weren't used, they were remanufactured.  Whoever remanufactured them had some poor quality control though.  I ended up with 6 of those 5970s and all of mine were fortunately okay.  The 5870s (from the same remanufacture batch) have been a different story.  Wrong BIOS versions mismatched with the chip on many cards and 2 of them (out of 48) had heatsinks that weren't properly seated on the GPU (misaligned template).  Fortunately I was able to take care of them all without shipping them back.

^ This. Remanufacturing (ie, factory recertified, certified pre-owned, factory refurbished, etc) is a difficult process, but on solid state components, if done right, you get a basically new product for cheap.

Diamond is a brand I do _not_ buy, and this is just another reason I can add to my list on why I don't.

  Thought I'd chime in and say that thankfully my two have been alright.  One card at 825 and another at 815 with stock voltage.  The second card is a bit flaky but but unstable enough to mine with - it just likes to have a core disappear randomly on reboots and seems to throttle a little worse temp-wise than the other.  Then again maybe I just didn't have the fanspeed high enough (it was the lower of the two cards in the system).  Top one runs fine at 50s% but the bottom card requires nearly 85% fan to keep the VRMs cool enough.  Temps are upper 60s to mid 70s so no problems there.

  Will be moving them to better room shortly though (just in the bedroom for now which has a lot of temp fluctuations).

I user AMDOverdriveCtrl on Linux. Works like a charm for me.

Would you mind posting the patch required to achieve this? It seems the developer who originally wrote radeonvolt has disappeared, but I have a fork (https://github.com/runeksvendsen/radeonvolt) that I can apply the patch to, so fellow Linux 5970 owners can enjoy the fruits of your work (if you're interested in that).

God yes PLEASE get us a tool to read vrm temps for 5970's in linux, specifically linuxcoin. I'll donate some BTC for a working VRM reader and would also be willing to put up part of a bounty for un updated linux dustro for bitcoin mining that is for dedicated mining only, no internet browsers, no bitcoin client,nothing that is not needed for mining or tracking mining or monitoring rigs remotely, etc. just what's needed to mine and with the newest stable combo of drivers, etc. for cards.

Also, WTF is up with linuxcoin being so damn sensitive to moving a gpu or replacing one? It's like I have to reload a fresh copy every time I change anything on my rigs, totally sucks. The whole graphical interface error appears every time. I've used the whole initial adapters command and everything, never works.

 OK, rant over, VRM tool would be great though. I'm in for 5 BTC to anyone who gets it to work and provides me step by step to get it set up and working.

I am a super linux noob only started using it when I started building mining rigs.

Have you tried radeonvolt (https://bitcointalk.org/index.php?topic=10228.0;all)? It should work if your card has a Volterra VT1165 VRM chip (are there any non-reference 5970s?) and not work if your card doesn't have a Volterra VT1165 VRM chip.

I couldn't get it to run, I run linuxcoin because I'm a noob with linux and i can't do shit with it to be honest. i'm starting to hate it. i can't even find out what the damn root username  and password is, should be easy for an opensource system.

Every time one of my computers shutdowns unexpectedly, it's a 50/50 chance that i can get it going again without a fresh install. i always get this graphical interface error. xorg or whatever that crap is. I'm really getting annoyed with having to do fresh installs every few days. Linuxcoin also seems to be too damn unstable for mining, but maybe that's my Hardware skills, but my temps, etc, all seem great so it doesn't make sense.

I'm seriously starting to consider windows machines with ssds since the ssds are so cheap now and use so little power. plus they are lightning on boot.

Linux is just killing me at this point.

It's takes a while to learn. Nothing more to say about it really. As with any OS. My Windows knowledge didn't come by itself, nor did my Linux knowledge. It's mostly trial and error, and a lot of Googling.
If you want assistance, I'm sure there are plenty who are willing to help you, if you want to invest the time it takes in learning it. Just make a thread in the Mining software subforum. If you make a thread, include as much information is it as possible, the error messages you experience are especially important. If you experience Xorg errors, include the files /etc/X11/xorg.conf and /var/log/Xorg.0.log

I have a Linuxcoin rig running right now and I have never experienced stability problems. It's rock solid for me.