Hashrate drops, only reboot helps

[Meeho]

I am getting good hashrates with no HW errors and low rejects for hours with my 6850 with cgminer-kalroth 3.7.3, but then my hashrate almost halves.Restart of cgminer doesn't help, nor does connecting to another coin/pool. GPU utilization stays at 99-100, but temps are low @~74 instead the usual ~85C. Only a reboot fixes it, but the problem returns eventually.

Here is my config:

{
"pools" : [
	{
		"url" : "stratum+tcp://eu.wafflepool.com:3333",
		"user" : "abcd",
		"pass" : "x"
	},
        {
                "url" : "stratum+tcp://us.clevermining.com:3333",
		"user" : "abcd",
		"pass" : "x"
        }
]
,
"xintensity" : "133",
"worksize" : "64",
"lookup-gap" : "2",
"thread-concurrency" : "5761",
"gpu-threads" : "1",
"temp-cutoff" : "92",
"failover-only" : true
}

 

[Bonksnp]

I have the exact same problem with my 270s on BAMT 1.5. Couple people recommended lowering memory settings but that didn't help me. I just have to baby sit it.

 

[_r2h]

Did you

"screen -ls"
"screen -r xxxx.cgminer"

to check to see if a card was dead or sick?

 

[Meeho]

Are those Linux commands? I'm on Windows 7. How can I check?

 

[_r2h]

Yes, I assumed you were using BAMT.


Sounds like a driver crash.

 

[Meeho]

Is there a way to check? Or some tweaks to prevent it if that is indeed the case? I'm on 13.12 and SDK 2.9. The GPU is slightly OCed, but handles gaming/benchmarks w/o problems and cgminer shows no HW errors.

 

[Pitbully]

Your driver is crashing, probably due to too high of an overclock or possibly the wrong settings in your miner.

 

[Meeho]

Do you see anything wrong with my settings above? I'm not sure it is the OC that is the problem, since I've run the card for a month before and it worked fine. It was reporting a small number of HW errors with those settings, so I've tweaked them a bit. I got higher WUs and no HW errors, but then this started happening. And updated the miner from an old version, huh, many changes, I know.

 

[_r2h]

You want zero hw errors. IMHO it's probably the symptoms of barely stable OC. I'd back'm down and see what happens.

 

[Meeho]

OK, I'll try to lower it.

 

[Carbon_Rod]

Whenever I got HW errors on my 7950's I found out it was due to the VRMs overheating. I had to back down on OC and I tried backing off the voltages a bit to get rid of them completely... I thought nothing of them at first considering I used to get only one HW error every few days of running.

 

[Meeho]

Before the last couple of posts I've started a test run with a mix of original and new cgminer settings, with the same OC. It has been 19h so far and all is well. Got 2 HW errors, which was expected since I used to get a couple with the original settings, but highest yet WUs and WUE of 97%.

The problematic drops used to occur within the first 20 hours. If these settings hold, I'll keep them. It is interesting, though, that these seemingly less stable settings (due to non zero HW errors) end up being faster and actually more stable long term.

VRM temps are <92C currently. At least I think they are VRM, since GPU-Z only labels them GPU temp 2 & 3.


At 14h:

 

[edge929]

I've seen cgminer do all sorts of weird things seemingly at random. More likely than not, I always had an unstable overclock be the culprit but the one time I noticed my hash rate drop, over time, was due to the intensity. Lowering it from 19 to 18 on a 270 of mine fixed the hash rate issue.

I've done something as simple as removing the USB risers on my cards, zip tied some cables together and reinstalled the risers EXACTLY how they were before, same position, same cables used, same PCIe slots and had all sorts of problems getting cgminer stable. As far as it was concerned, only a reboot happened. I went from having 1-week stable settings to redoing them again just because I decided to clean up some cable clutter. I'm afraid to touch it now

 

[Justintoxicated]

What usually works for me when this happens. Re-apply voltage settings in MSI Afterburner, then restart cgminer. I use kalroths 3.7.3 build and finds setting xintensity works better than intensity and appears more stable. I can actually brows the internet on my PC while minign with xintensity 400, while getting better rates than using intensity 20.

 

[Pitbully]

I've seen cgminer do all sorts of weird things seemingly at random. More likely than not, I always had an unstable overclock be the culprit but the one time I noticed my hash rate drop, over time, was due to the intensity. Lowering it from 19 to 18 on a 270 of mine fixed the hash rate issue.

I've done something as simple as removing the USB risers on my cards, zip tied some cables together and reinstalled the risers EXACTLY how they were before, same position, same cables used, same PCIe slots and had all sorts of problems getting cgminer stable. As far as it was concerned, only a reboot happened. I went from having 1-week stable settings to redoing them again just because I decided to clean up some cable clutter. I'm afraid to touch it now

I've noticed the same thing at times...I've even moved one card that has been mining for months, to another rig and have to retune the entire thing. Drives me crazy.

 

[Meeho]

One of those lovely PC quirks

What usually works for me when this happens. Re-apply voltage settings in MSI Afterburner, then restart cgminer. I use kalroths 3.7.3 build and finds setting xintensity works better than intensity and appears more stable. I can actually brows the internet on my PC while minign with xintensity 400, while getting better rates than using intensity 20.

Yes, I also get better performance with xintensity.