Killed my hardware?

[grippy54]

I’m setting up my first rig and am having some difficulty. I’d be very appreciative of any help. Yesterday I unpacked 3x R9 290’s and got them all to run 500khs each (2 on the board and 1 on a powered riser). I know that’s low, but was hoping to tweak the config the next day.

I let it run for about 4-5 hours and everything seemed fine. Temps were all under 90. I let it run overnight and when I got up, it had a black screen.

Now when I test them 1 at a time, I have 1 card that does 500khs, 1 that does 175khs and 1 that crashes the PC just after the Starting Windows screen. Could I have killed the cards somehow?

I’m running:

MSI 970A-G43
4GB RAM
Win 7 x64
3x Gigabyte Radeon R9 290 GV-R929OC-4GD
Thermaltake Toughpower Grand 1200 PSU
SGminer 4.0
14.1 beta drivers

I am executing setx GPU_MAX_ALLOC_PERCENT 100 and setx GPU_USE_SYNC_OBJECTS 1 before launching SGminer.

My config:

"intensity" : "13",
"vectors" : "1",
"worksize" : "256",
"lookup-gap" : "2",
"thread-concurrency" : "8192",

"gpu-engine" : "1000",
"gpu-fan" : "40-100",
"gpu-memclock" : "1350",
"gpu-powertune" : "20",

"temp-cutoff" : "99",
"temp-overheat" : "95",
"temp-target" : "90",

"api-port" : "4028",
"expiry" : "120",
"failover-only" : true,
"gpu-threads" : "1",

"log" : "5",
"queue" : "1",
"scan-time" : "60",
"temp-hysteresis" : "3",

"scrypt" : true,
"kernel-path" : "/usr/local/bin"

[Gamer67]

No need for "scrypt" : true, if using sgminer.

Also I think you might find that you get better performance using -I 20, -g 1 and TC: 24000

[Starscream]

r9 290 don't use -g 2, so you need -g 1 with tc of 20481 (can vary, but this 1 works very well for me) . Intensity will also go up to 18+.

BTW, your target temp is quite high, I'm running with:


"temp-cutoff" : "90",
"temp-overheat" : "85",
"temp-target" : "80"

There's really no need for you to get to 90 unless you overclock, which you really shouldn't.

Did you change their volt? Cause if you did, than each needs his own tuning and that could explain why your 3rd card crashes windows.

[grippy54]

OK… thanks. I tweaked my config and it helped, but I still have 2 problems hopefully someone can advise me on.

1. I now have 2 of the R9 290's doing 590khs. I'm pretty sure they should be doing better than that. If I raise my intensity over 13, my khs decrease.

2. My 3rd R9 290 keeps crashing windows just after the Starting Windows screen… I think when the video driver loads. When I plug a monitor into that card, I get a vertical red stripe down the screen about an inch wide.

So here is my config now:

"intensity" : "13",
"vectors" : "1",
"worksize" : "256",
"lookup-gap" : "1",
"thread-concurrency" : "24000",

"gpu-engine" : "1000",
"gpu-fan" : "40-100",
"gpu-memclock" : "1350",
"gpu-powertune" : "20",

"temp-cutoff" : "90",
"temp-overheat" : "85",
"temp-target" : "80",

"api-port" : "4028",
"expiry" : "120",
"failover-only" : true,
"gpu-threads" : "1",

"log" : "5",
"queue" : "1",
"scan-time" : "60",
"temp-hysteresis" : "3",

"kernel-path" : "/usr/local/bin"

[Starscream]

OK… thanks. I tweaked my config and it helped, but I still have 2 problems hopefully someone can advise me on.

1. I now have 2 of the R9 290's doing 590khs. I'm pretty sure they should be doing better than that. If I raise my intensity over 13, my khs decrease.

2. My 3rd R9 290 keeps crashing windows just after the Starting Windows screen… I think when the video driver loads. When I plug a monitor into that card, I get a vertical red stripe down the screen about an inch wide.

So here is my config now:

"intensity" : "13",
"vectors" : "1",
"worksize" : "256",
"lookup-gap" : "1",
"thread-concurrency" : "24000",

"gpu-engine" : "1000",
"gpu-fan" : "40-100",
"gpu-memclock" : "1350",
"gpu-powertune" : "20",

"temp-cutoff" : "90",
"temp-overheat" : "85",
"temp-target" : "80",

"api-port" : "4028",
"expiry" : "120",
"failover-only" : true,
"gpu-threads" : "1",

"log" : "5",
"queue" : "1",
"scan-time" : "60",
"temp-hysteresis" : "3",

"kernel-path" : "/usr/local/bin"

Change your lookup-gap setting to: "lookup-gap" : "2",

If that doesn't help increasing intensity (it should) than try using stock clocks:
"gpu-engine" : "947",
"gpu-memclock" : "1250",


When you reduce your TC, can you increase intensity?

As for your 3rd GPU, if you plug it alone to the system, in the main PCI-e slot, and it's not working than I'd put good money on that that it's a goner. Try plugging it in a different computer if you got 1 lying around (it should be the only card plugged) and see if the problem persist (don't attempt to mine, just see if it works).
It it does than... you know what I'm gonna say.

[MoneyMorpheus]

Your temp-target is too high!

Vrm temps at 90 ar probably sky high, check them with gpu-z if using windows. If vrm temps are over 110ºC I would seriously consider increasing the fan speed.

Also the cooler you run your gpus, the more they will run for you. They are supposed to stand up to 120ºC, but they are also supposed to be run for a few hours a day. Try to achieve a temperature of 70, 80ish at most. I usually run mine at 80-85% fixed fan speed.

And 4 gb of ram is probably too low for 3 290. Aren't you getting a lot of hardware errors?

[Tiscoin]

Grippy,
I have the same black screen problem with the same card GV-R929OC-4GD-GA. I tested a new batch file and it artifacted like f***. After reboot it black screens.

Only thing I that has helped me so far is the 2nd bios on the card. Just flick the switch and it works again. Keep me updated if you find some solutions or good mining setting.

[grippy54]

So I'm at 780khs right now. I was running the 14.1 beta drivers… well they suck. I went back to 13.12 and I could magically up my intensity to 20 without hw errors. I shot up over 100khs. Then I tweaked my thread concurrency to 30718, and put the gpu settings to stock for my card 1040/1250. I also upped RAM to 8GB.

Well that leads me to my next question. Should an R9 290 be hashing at 780khs? I see so many posts from people who claim 800-900khs from these cards. Am I missing something?

Here's my current config:

"intensity" : "20",
"vectors" : "1",
"worksize" : "256",
"lookup-gap" : "2",
"thread-concurrency" : "30718", 

"gpu-engine" : "1040",
"gpu-memclock" : "1250",
"gpu-powertune" : "20",

"temp-cutoff" : "90",
"temp-overheat" : "85",
"temp-target" : "80",
“auto-fan” : true,

"api-port" : "4028",
"expiry" : "120",
"failover-only" : true,
"gpu-threads" : "1",

"log" : "5",
"queue" : "1",
"scan-time" : "60",
"temp-hysteresis" : "3",

"kernel-path" : "/usr/local/bin"

[Gamer67]

Adjust the clock speed now.

You should under clock the core. Just start going backwards 5-10 mhz at a time until you find the sweet spot.


bump the memory up as high as it will go without crashing too.

[grippy54]

I have to say that under clocking the core is completely unintuitive… but it sure did work. Thank you Gamer.

I used MSI Afterburner to adjust the clock in real-time while sgminer was running. I just dropped it in the smallest increments it would allow me, and watched the result. I took it all the way down to 967 before it evened off and started to drop the khs.

I also upped the tc by 1, to 30719. Don't ask me why, but it helped. Going more than that dropped the khs. I'm now at 845khs.

I tried upping the mem clock up too, but I could not budge it at all from 1250 before the khs started dropping. That doesn't seem right to me. Most folks are over clocking the mem from what I can tell. Could I be doing something wrong there?

[Gamer67]

I have to say that under clocking the core is completely unintuitive… but it sure did work. Thank you Gamer.

I used MSI Afterburner to adjust the clock in real-time while sgminer was running. I just dropped it in the smallest increments it would allow me, and watched the result. I took it all the way down to 967 before it evened off and started to drop the khs.

I also upped the tc by 1, to 30719. Don't ask me why, but it helped. Going more than that dropped the khs. I'm now at 845khs.

I tried upping the mem clock up too, but I could not budge it at all from 1250 before the khs started dropping. That doesn't seem right to me. Most folks are over clocking the mem from what I can tell. Could I be doing something wrong there?

845 is not too shabby. The only way you will be able to go higher is if you can bump the mem up. Have you tried to run it at 1500 ?