Bitcoin Forum

So I just installed the card and am only getting around 200 khash/s with cgi miner versions both 3.2.2 and 3.7.2

The card freezes my computer at intensity 13 which was recommended by many sites so I've been running it at intensity 11.

Here is the .bat file.

cgminer --scrypt -o stratum+tcp://usa-1.liteguardian.com:3335 -u Sancho.1 -p x --intensity 11

--thread-concurrency 8192 --worksize 256 -d 0,1 -g 2


Can someone please help me out here? This card was a lot of money and I have no idea how to get it mining at capacity, which should be around 1100-1500 khash/s. :(

Thank you so much ahead of time!

I guess you are using Windows, to get more than 200KH/s you have to use intensity 20. But you will need to solve your freezing problem first. Sorry can't help. I use Linux.

Yeah - using windows... hmm, I forgot about the OS - I will try it, thank you.

Edit: Was actually getting almost 1000 khash/s a minute ago before comp froze - with windows 7.


hmmm

GPU mining is not good ATM :(

Not for BTC.

Yeah, I'm mining scrypt.

One important question before troubleshooting, is your card new or did you buy it used?

I definitely got it used, ymer. Supposedly lightly.

I wonder if that's part of the problem? Maybe some settings are out of whack? Or maybe it's just fried? Hopefully the former.

I can get 450 khash/s now at intensity 11 and haven't frozen yet - but everything above that crashes me.

Do you have big enought PSU ? Also this freezing happens to me when one of the power connectors going to GPU is burned.

My PSU is 850 watts - should be more than enough, right?

The power connectors should be okay, they are new. But I suppose it's a possibility.

Could the fact that I used two 6+2 pins from the PSU to the card instead of the two 8 pins that came with the card have anything to do with it? No, right?

My hunch is that some asshole sold you a bad card, probably flashed with wrong settings, probably it overheats, probably has some power issue that makes it freeze...

This would sadden me greatly. :(

Could the fact that I used two 6+2 pins from the PSU to the card instead of the two 8 pins that came with the card have anything to do with it? No, right?

change the content of your batfile to include the following
"
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECT 1
cgminer --scrypt -o stratum+tcp://usa-1.liteguardian.com:3335 -u Sancho.1 -p x --intensity 13

--thread-concurrency 8192 --worksize 256 -d 0,1 -g 2

"

I had it holding steady at 1000 khash/s at intensity 12.

F9, I tried your suggestion and it still crashed. :( Looked good for a second though at 1400. I was in the process of replying here to tell you it was working when it crashed. :(

Using the PSU dedicated 6+2 dedicated pins is better than using the card adapter, as long as you have all the power pins populated and your PSU is really 850W it should work just fine, what model PSU is it?

Yes, for 1 GPU 850 watts should be enough.

Yeah, it's just the one GPU. I wanted to run another with it but it was nvidia and I thought it would be too risky.
\
My PSU is a Cooler Master Silent Pro M 850W Power Supply with Modular Cabling (RS850-AMBAJ3-US).

I got good results with following parameters


start.bat

maybe this example could help you....

the red parameters reduced my amount of rejected files a lot....

What temp does the card reach?

I'm pretty intimidated about using intensity 17 when 13 isn't working. >_<
I'll try it though.

Are you running Linux or Windows?



F9, I've been trying to keep the temp around 85 if possible. I can't say how high it reaches at intensity 13 because it freezes soon after. EDIT: Although, it just crashed at intensity 12 around 88 degrees. Dammit!

Might as well try at a lower memclock and work us upward, try changing it to:
"
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECT 1
cgminer --scrypt -o stratum+tcp://usa-1.liteguardian.com:3335 -u Sancho.1 -p x --intensity 13

--thread-concurrency 8192 --worksize 256 -d 0,1 -g 2 --gpu-memclock 1210

"

Thanks, but I actually already had it at memclock 1210. I should've mentioned that already.

I can't believe it crashed at intensity 12. :(


DjNorad, I tried your settings as well (also with memclock at 1210) and it held up for a couple of minutes before it froze. :( Thanks though. Maybe I'll try it at intensity 13 next.

88 degree is quite hot, and quite understandable that it crashes.
How is the airflow? is it in a closed cabin? if so try open the side window/panel and check the temp, if not try to lower the memclock down another 100

Ok man you need to stop checking cgminer settings, it's not a problem with the intensity or the cgminer variables, something is wrong in your system hardware wise if you are on a fresh installation with fresh drivers.

Yeah, I'd been keeping it at 85 or lower before.

I already have the side panel open. I'll try lowering the memclock more. I also have the fan operating at 100% for now.

Thanks.

Hmmm...

Well, I did just uninstall a bunch of nvidia drivers (which was sketchy because windows kept trying to auto reinstall them no matter what I did) but I'm fairly certain I got rid of them (all nvidia files)  before taking out the card and replacing it with the radeon and installing the newest drivers. I don't think that has anything to do with it but thought I would mention it.

As far as hardware, I don't know what it could be besides the card itself.

Well before you wipe out the drive try playing a game on max graphic settings and check if the system freezes.

Will do.

Was able to run it at intensity 13 for much longer before stopping the program at 86 degrees at memclock 1010 but the khash/s rate didn't improve much.

If you are using Windows, first of all download and install "MSI Afterburner Latest Beta" here: http://event.msi.com/vga/afterburner/download.htm (http://event.msi.com/vga/afterburner/download.htm). 

Disable the Overdrive in the AMD control panel, then in afterburner open the settings.

In the General tab make sure that the following are selected:

Synchronise settings for similar graphics processors,
Start with Windows,
Enable hardware control and monitoring,
Enable low-level hardware access interface (user mode),
Unlock voltage control,
Unlock voltage monitoring,
Force constant voltage,
Restore settings after suspended mode,
Disable ULPS,
Unofficial overclocking mode disabled.

Then on the fan tab set a custom fan curve to make sure that the fan is upto 100% around 75C to keep things nice and cool.

Once all that is done set the core voltage to 1050, power limit to +0, Core Clock to 980, Memory Clock to 1350 and fan speed to auto (if you dont hear the fans kick in when the card starts working just set manually to 100%). 

Once all the settings are entered click apply, then make sure the padlock icon in the bottom right is set to unlocked, click save and select slot 1.

With regards the mining, I am using Multiminer and have the following settings in advanced miner settings Scrypt:
It used BFGMiner underneath so if you prefer to run command line the same settings should work for that (I believe CGMiner uses the same command strings also, perhaps someone else can clarify).  The advantage of Multiminer is it easily allows you to switch coins on the fly and allocate different pools to each GPU if you wish.

I am getting a steady 1330KHs or so with these settings on my MSI 7990 which is a malta reference design.  Undervolting is the key to dealing with the heat, even when it seemed to be running cool on higher voltage the VRM temperature was peaking over 100C, not good in my opinion but at 1.050V everything is running well.

It is worth downloading GPUz also as it lets you look at the other temperature sensors on the card, i.e. the VRM sensors.  If these are overheating they will fry and cause you serious problems so keep an eye on them. 

Hope this helps you get up and running, feel free to come back to me if you are still struggling and I'll try and help some more!  FYI, do not go above I13, I just found it errors to hell.

Aleks

Sounds like the issue is caused by overheating then. Would follow Aleks aadvice and try undervolting and find a stable setting where the temp is manageable and then eventually try increase the memory speed to obtain better hash rate.

I have been fiddling with the settings to reduce heat and yes, I think you're both right - that seems to be the issue. Thanks a lot.

I did not have the beta version so I couldn't mess with voltage - but downloading it now and I suspect this will do the trick.

Thanks to ALL and I'll report back results!

7990 is voltage locked. At least mine is. I can change it to 1.17 at lower clocks but that's it.

The sweet spot for mine is 1100/1500 for ~760kh/s per gpu. -I 13, -g 2, thrconcurr. 8192. Powertune is not necessary.

Thank you so much - the voltage really was the key. Getting 1350 now with similar settings. :)

Going to download GPUz now and see what the VRM sensors look like. Thanks!

Glad it worked out :)

In regards to the VRM sensors, is 87 degrees too much? What temp should I be shooting for?

Thanks to all again!

<90* on pre-200 cards is fine. <95* for the newest generation. 100* is generally the breaking point. Heat + time will reduce MTBF, but it's not too big a deal... MTBF will be waaaay beyond the lifespan of the card from technological progression.

No problem dude.  Your VRM temp is probably fine around 87C.  What kind of case do you have also?  I really had to get some serious airflow through the case to keep things in order.  I've got 6 120mm fans feeding air through, though I've resorted to taking the side of the case off and putting a room fan blowing into the case, the way the cooling works on that card it superheats the interior of the case so if you don't vent it you will have some serious problems! 

Good luck!

can you please tell me what does MTBF stand for?

Mean (average) Time Before Failure. They're generally given on components with relatively short lifespans... rechargeable batteries, lightbulbs, solar panels, some computer components. Most computer components won't have MTBF listed because they generally outlast their usefulness and are subject to a huge variance depending on how they're used (a GPU in a mining rig will obviously have a much, much lower MTBF than a GPU in a "light-use" PC grandma uses to check her emails.... unless she smokes in a dusty house with cats).

MTBF is useful in guesstimating how long a given component will function at manufacturer specs, but it's always worth noting it's very luck-based, and if you have 100 GPUs hashing, the "MTBF" for one out of the hundred cards failing is 100x shorter than MTBF - so if a card has a "true" MTBF of 25 years, and you have 100 of them, you'll probably be replacing ~4 per year. This isn't quite accurate, though, since failures/time tends to make a bell curve. In this case, maybe failures are dramatically more likely to occur at the 10+ year mark, so before that ten-year mark, maybe you'll be replacing 1-2 cards of the 100 each year, and once you hit the ten year mark, maybe you're replacing 5 cards per year.

I should note that I had to watercool mine to get 1.5Mh/s @ 1100/1500. I can't seem to adjust the voltage in afterburner though, would be nice if I could get it down a bit. My vrms stay around 65-75 even under water. One bank is 8-10C cooler than the other; I noticed this on air too.

Man these run hot! Unfortunate that the factory heatsinks are entirely too small with way too much thermal grease smeared on them.

With what I have it at now, the VRM temp is sitting around 89/91.5.

My tower = Thermaltake Chaser MK-1 VN300M1W2N
I have had both sides off and probably will keep them off now. That's funny you mention the room fan, I was just talking about that possibility with my girlfriend. Does it really work well?

Thank you so much

Box fans are better when you want to move "fresh" air in and hot air out of a room full of rigs. They move a lot of air and aren't very "directional." It's kind of like using a full-size oven for one piece of toast. A PCI-slot fan would probably drop temps ~3-5*C and be much quieter and energy-efficient. Cheaper, too.

Thanks, didn't even know they made pci slot fans - I'll check it out!

Any recommendations?

You'd probably want an intake PCI fan, which is pretty uncommon. This assumes your gfx card is "normal" and takes air in from the face of the fan, blowing it out the PCI vents. This is a double-slot one... there should be a single-slot version somewhere, but this is what you'd want: http://www.frys.com/product/4764059?source=googleps&gclid=CIuc4vjY9rsCFWRk7AodxVgAaQ

You usually can't effectively reverse the orientation of these kinds of blower fans, so you'd probably want one designed as an intake fan (else you'll need a creative DIY solution). You might not be able to get around wanting to put an "air pipe" in for the PCI vent, though.... this solution non-ideally takes air which will partially be what the GPU fans just exhausted, though it's way better to have more 22*C air over a 90*C component than less 18*C air over a 90*C component. Ideally, you'd want to create a plastic or metal air-tight pipe which has a mouth over the PCI vent(s) being used, then jutting out the side of the case so you're sure you're getting ~100% fresh air. If you get DIY like that, you'd probably want to mount another normal case fan over the end of the "air pipe."