With multiple GPUs 1 card is always slow - Solution found / please lock
-
[quote name=“d2” post=“17238” timestamp=“1372014842”]
[quote author=Nulunas link=topic=113.msg17095#msg17095 date=1371955994]
[quote]
[quote author=Nulunas link=topic=113.msg16921#msg16921 date=1371889051]
That and I’m ahead of FTC mining schedule, even with a card that I need to RMA. Weird, but I’ll take it.
[/quote]May I ask what problems you are encountering with that video card?
[/quote]It is hashing at about 10% to the others. They’re all spaced up and externally air cooled, and what bugs me is that it is at the cool end of the line-up.
It’s a 7970, it has the same setting as the other cards, no hardware errors, approximately the same temperature as the other 3 on the same board. I’m running it 20/35mhz under instability with the others as well.
Aside of this card, this rig has given me no problems.
[/quote]Yea, that does sound like a bad card. Have you tried flashing the bios? May be worth attempting before sending it back.
[/quote]So, after about 6 hours worth of testing, this is what I got.
I found my peak numbers as well for hashing, and it’s running rather nicely now for the first 3 cards. Each card has dual-BIOS; they’re on the back-up BIOS in the screenshot (just switched over BIOS in the last 2 hours). This is identical hashing ratios to the main BIOS.
In my testing, I disconnected cards one at a time, but each card would hash like the 4th one if left alone. When I did 2 or 3, the last one would hash like that 4th one.
This is Windows 7 Ultimate, installed in the last 2 weeks, and used AMD’s auto-detect for drivers.
TL;DR: CGminer is fine, hardware is fine, settings are fine.
I have yet to stumble on a similar case, but I’ll see what I can do tomorrow after work. If anyone else has an idea, I’m all ears.
EDIT:
CGminer is fine, seemingly.
Cards are fine, each go to 700kh on standard BIOS.
CMD instructions are fine, but I’ve been catching minor tweaks bumping my hashes.
PCIe configuration is fine, tested nearly every card to each port, no difference.
Windows Hardware Acceleration is greyed out, controlled entirely with AMD Vision.
Remote Desktop Connection is off, but no changes in hashes.Issue: seating
[b]Solution:[/b] Each time the rig is turned off for whatever reason, I may have to change the seating of the cold card to the other free PCIe on the motherboard. I only have to do this whenever a card goes cold. Occasionally, a restart is all that is needed.
[attachment deleted by admin]
-
in the cgminer folder delete all the bin files.
run cgminer without thread-concurrency and shaders .and check the folder again.
you should now have four bin files. one for each card.what are they named? i suspect cgminer to put low thread concurrency on end card or current setting is not memory optimised
have you runned gpu_alloc and the otherone i cant remember before running cgminer?what memory do you have installed? and what are your cgminer run command?
edit: forget last part. saw picture after (on phone…)
-
[quote name=“svennand” post=“17688” timestamp=“1372202290”]
in the cgminer folder delete all the bin files.
run cgminer without thread-concurrency and shaders .and check the folder again.
you should now have four bin files. one for each card.
[/quote]I don’t get one for each card, I just get the one. It goes for a TC of 22400, and I’m still getting a slowdown on minimal instructions.
I noticed that through CGminer, that the last listed card is going at something around 350 / 500. I cannot get that to change, through CGminer or the .bat file.
EDIT: The TC of 22400 has jumped me another 10kh on the 3 good cards, so some nice tweaking bonus. Thanks svennand. :D
I’ve been looking rather hard for the same thing on other forums, and I stumbled on a possible Windows Hardware Acceleration issue. I have to stick a monitor to the rig, and that’s a little out of the question in the time area of this post. Remote Connection gets greyed out of a lot of display things.
-
I would try to run cgminer with just --scrypt --auto-gpu --auto-fan -o pool -u -p
Do you still get that low hash on one of the cards?Seems your config script is pretty cluttered with alot of in my opinion unessecary commands.
-
Somehow missed this. :-[
Dont’t use RDP for remote management - it causes all sorts of issues. Not least the lack of fan speeds/temps. You may find that it’s RDP that’s the cause of the problem.
I use TightVNC which is free and does the job.
-
[quote name=“svennand” post=“19084” timestamp=“1372677617”]
I would try to run cgminer with just --scrypt --auto-gpu --auto-fan -o pool -u -p
Do you still get that low hash on one of the cards?Seems your config script is pretty cluttered with alot of in my opinion unessecary commands.
[/quote]The only one in there I haven’t used is auto-gpu, but otherwise with bare minimum instructions I tend to get 550,550,550,80 kh respectively on the cards. Which means that it is more or less the same as the OP configuration, just that it isn’t tweaked out.
[quote author=Nutnut link=topic=2189.msg19089#msg19089 date=1372678658]
Somehow missed this. :-[Dont’t use RDP for remote management - it causes all sorts of issues. Not least the lack of fan speeds/temps. You may find that it’s RDP that’s the cause of the problem.
I use TightVNC which is free and does the job.
[/quote]
Hmm, I know it isn’t doing anything to the hash speeds, since I still run about 2100-2200 kh reported on the pool.
I was thinking about putting PocketCloud on there, it’s like TightVNC, except through for smartphones. I like it a lot, and on top of that, I’ve tried to set-up under TightVNC once and a couple of others with no success. I could give TightVNC another shot.
-
Similar issue, let me try to revive this thread
Powercolor 280x, 13.11, win7 ultimate 64bit, cgminer 3.72
In all my multi gpu rigs, one card is always couple of percent slower, i.e. all cards are 720khash/s, except of one which is 670-90khash/s
one cgminer instance per gpu. Exactly the same cgminer.conf with all gpus, and all gpu’s are identical.
edit: will start another thread… no need to reply here…