Topic: Yet another noob with mining troubles using 7950s (Read 1469 times)

Congrats, at least it's working now .

tubbyjr: Thanks a lot

Well, um... excuse my colorful use of language but fuck me!

Over night - without me changing ANYTHING at all, or rebooting, or installing anything - well, I encrypted a new volume with TrueCrypt but I somehow doubt that has much to do with GPU mining - suddenly my cards run at any intensity without issue.

What? Yes. I have no idea why.

I read a while ago in some thread that another guy who had similar problems found his cards working perfectly after he just kept using them, aka playing games. I dismissed that as the deluded ramblings of an idiot. Now I think he might not have been bullshitting after all...

"Proof" pics (of course I don't have any screens of the old runs with tons of HW errors but I didn't join and ask for help here just to amuse myself):

@I15
https://i.imgur.com/z6jLqAK.png

@I18
https://i.imgur.com/MGEkFjH.png

@I20
https://i.imgur.com/Ze5dbrG.png

@I12 - the only intensity I could run until I tried again just now
https://i.imgur.com/Ux8Cj9Z.png

Admittedly I20 is a bit slower than I18 but notice that I get 0 HW errors in all intensities now, while I got hundreds within seconds @ I13 before and still around 10/hour on I12.

As for the cgminer .bin filename, that changed from "scrypt130511Tahitiglg2tc4032w256l4.bin" to "scrypt130511Tahitiglg2tc22400w256l4.bin" at any intensity above 12, so the thread concurrency was adjusted properly it seems. I can still fine tune later, right now I'm just stunned and starting to think I found a glitch in the Matrix

And for the record, still the same CoolerMaster PSU, still the same wiring, still the same other components, still the same drivers, still the same SDK, no changes in clocks or voltages, nothing.

Me no getty but me rikey.

Any day not mining is a great sunken cost in the mining world, so get it running as soon as you have time :p. Hopefully it all works.

Just a quick update, I didn't get to try around much more yet but I did end up buying some new hardware to accomodate all of the cards.

So I got 1 x16 riser (I heard one card has to be connected via x16 to work correctly? No matter, only cost 1€ extra anyway) and 4 x1 risers, all of them powered; as for the other hardware I opted for the Sempron, 4GB Crucial Value and the Asrock 970 Extreme 4, all powered by an Enermax Platimax 1350 because the 1500 is not available anywhere - again it comes with a split 12V rail but I'm still optimistic it'll work. So yeah, hopefully my troubles will soon be a thing of the past.

I'll keep this updated. Wish me luck

And again thanks for your input guys, I'll try your suggestions as soon as I can this weekend.

Seems to match up http://www.techpowerup.com/gpudb/307/radeon-hd-7950.html

Although I can't find your BIOS, must be a new one. http://www.techpowerup.com/vgabios/127391/sapphire-hd7950-3072-120907-1.html Closest one to it.

Wow thanks a lot guys I thought I was all on my own here. I'll try your suggestions asap, just got a few things to take care of first.

Does anything seem off in those GPU-Z screens? Both cards were mining when they were taken...

Card #1
https://i.imgur.com/EtziL4F.png https://i.imgur.com/sBKFJgM.png

Card #2
https://i.imgur.com/LgD9vma.pnghttps://i.imgur.com/Uv0mG9T.png

I noticed the GPU clock having gone to default but nothing else.

setting lookup-gap 2 and TC 20480 in conf file saved me when I had similar problem with my 7950

Just read you have a coolermaster. I'm usually not a PSU-Nazi(have a cheap rocketfish in my gaming computer), everything in my mining rigs is crap EXCEPT PSU. Coolermaster PSUs are very bad. They can say it's 850W, and those and those Amps, but you're better off getting at least a bronze-rated PSU.

Safe mode, uninstall the gfx drivers you have and sdk, use driver cleaner or sweeper from guru3d.

Download and install http://support.amd.com/en-us/kb-articles/Pages/latest-catalyst-windows-beta.aspx

These are working perfectly for me, and they include the SDK right in the newest versions. Only thing I can think of is some driver conflict, I would be surprised if it's a hardware issue, as you seem to have top notch hardware, while I'm running my mining rig on 6 year old CPUs and Motherboards.

Well first of all thanks for helping out

Unfortunately it didn't work without further adjustment and I have to go to work now but I'll play around more with it later.

https://i.imgur.com/K8lIUXI.png

GPU-Z reports everything as fine btw (925/1250 MHz).

And I forgot to mention I tried powertune before too, set that slider to +20% but no difference again... sigh.

Also, check the cards on GPU-Z. More than on one occassion have I been sold duds. Cards branded as a certain card, but only share the GPU chip of that cards, and have some shitty memory clocks.

Try following settings when running cgminer on these cards:

-I 13 -g 1 -w 256 --thread-concurrency 22400 --gpu-powertune -10

play only with intensity or gpupowertune settings.

...and one more bump, I just have trouble believing nobody ever had an issue like this. I'll stop bumping after this one though. Thanks again for any advice at all.

Bumpity bump, hope this is allowed... nobody has any ideas?

Well, I just bought 2 more of those cards since they're really low on stock. And I guess I'll buy the new mobo (the very likely bad guy here) tomorrow. So yeah, still, if anyone has suggestions: please let me know, otherwise it's more trial and error and that sucks.

Hey guys, just joined today because I can't seem to work out the problems I'm having with my non-dedicated scrypt mining rig. I hope someone here can point out what I'm missing.

So, the situation: I only got into mining crypto currencies very recently, all I've mined so far are COYE, LTC and 42 on a GF 560 Ti 448 @ around 225-240kh/s.

Of course I wanted more so I bought 3 new Sapphire Boost 7950s from two different retailers and one x16 powered riser, and I was going to buy new hardware for a dedicated rig but right now I don't dare to until I can say for sure what the problem is.

I'm running an i7 970 Hexa on a MSI X58 PRO-E SLI (also supports Crossfire/X, newest BIOS.. from some time in 2011 iirc) board with 3 x16 slots and 24GB of RAM, using 7 x64, with plenty of ventilation for my usual needs and a CoolerMaster Real Power M850 power supply.

The PSU supplies 12V on 6 rails: rail 1,2,5 and 6 supply 18A each, rail 3 and 4 supply 28A. Rail 3 and 4 are only connected to the two PCI-E 8 pin connectors, 5 and 6 are for the two PCI-E 6 pin connectors and 1 and 2 are for CPU and mobo, at least according to the manual.

I'm running cgminer-3.7.2-windows using "cgminer.exe --scrypt -o stratum+tcp://...:... -u ... -p ... -I 12"

That's the only intensity I can run, regardless of which of the 3 cards I use, regardless of whether I use two 6 Pin Connectors or 2 8 Pin->6 Pin Connectors (rail 3&4) or a mix of both. Everyone else seems to be running -I 20 on these GPUs and getting 550-750 kh/s, I'm stuck at 245... if I try to raise the intensity even just 1 step things go crazy with hardware errors, on all 3 cards, no matter which one I try, and the WU/m go down dramatically.

I have tried the following drivers/SDKs:
- Driver 13.12, SDK 2.9 => initially installed; can run -I 12 at most @ 235kh/s/GPU.
- Driver 11.6 => can't detect GPU/install driver in Windows
- Driver 11.12 => can't detect GPU/install driver in Windows
- Driver 12.4, SDK 2.7 => same as 13.12
- Driver 12.8, SDK 2.7 or 2.8, not sure anymore => same as 13.12
- Driver 12.10, SDK 2.8 => 5% faster (235 kh/s => 245 kh/s)
(I've used Guru3D's Display Driver Uninstaller to remove the garbage between installs and from the old Geforce that was in there once)

I've also installed a 2nd 7950 at the same time (w/o Crossfire bridge) and curiously enough they both behave exactly like a single one does making me think it's probably not a power supply issue, they both work only at -I 12, both produce 245kh/s, both have a few HW errors, 10 per hour maybe, didn't really count. WU around 200-250/m each depending on mood it seems.
They're both connected using one 8pin->6pin cable from the 12V3/12V4 lines (28A each) and one 6 pin cable (18A each).

So I'm leaning towards the motherboard being the culprit. FYI it only supports PCI-E 2.0 but that shouldn't be a problem right? I know it's a really old motherboard but my PC is running quite fine on it, except for this mining business.

Like I said, I want to buy new hardware for a dedicated mining rig anyway but until I can be sure what's causing this I can't afford to potentially throw away even more money.

Oh yeah, and I haven't tried the riser yet because I can't imagine it helping much when the board is able to power 2 cards apparently without problems.

And I forgot to mention I also tried slight (+/- 100MHz) over- and underclocking, not touching any voltages though, with no noticeable changes.

-------------------------

As for the new hardware, I'm on the fence between taking the cheapest, most low power solution possible or paying a few more bucks and getting something that can still be used as a PC on the side, or if I turn out to be unlucky or too stupid for mining. So I'm thinking either Sempron 145 and ASROCK FM2A88X Extreme6+, or AMD A6-6400K Black Edition and ASROCK 970 Extreme4 (AM3+). I haven't thought too much about the motherboard though, I just think I'd prefer one with 3 x16 slots so I can install the highest and lowest card normally and just take the middle one out using a riser to keep everything coolish.
Then RAM, I always hear 4GB are enough but the cgminer readme states you should have enough RAM to match your VRAM, so I'd need at least 8 (or actually 9) GB right? So which is true?
I'm entirely open to any suggestions here, no components are nailed down yet, and as for the PSU I'm leaning towards a 1000W Corsair for ~150€ or a 1200W Be Quiet or Superflower Platinum for ~225€.


Sorry for the long wall of text, I just wanted to explain what I tried as best I could.

Thx in advance for any suggestions and better luck mining than I had so far