Bitcoin Forum

Hi,

3x Saphire R9 290 , H61 PRO BTC. After power failure all 3 cards detected but
only 2 working. After coldreboot I cant see 3th card anymore. If I put this
card on other MB i can't get video signal - no output to monitor.
Before I run cards at 1000/1500 powertune 10 at 880Kh/s
I use BAMT linux.

Please advise. Thank you.

They are broken.

Very helpful.

What more do you want me to say? It's pretty obvious whats happened, you've had a power failure/surge while maxing your cards out and fried them - it's not an uncommon thing to happen.

You've tried them in another mobo and they still didn't fire up, they are broken.

I like your 2nd post more ;) You said is a common problem?
Well, never happen to me till now. Usually the problem after power failure  when power come back on
and then off and on again. Never happen to broke something when power off.

If they are still under warranty see if you can RMA them, you might have trouble trying to RMA two cards, but you should be able to recover 1 at least, just tell them you were rending some video and the card stopped working - do not tell them you were mining with them.

Then maybe in a month or so try to RMA the other broken card.

Just one resist to work other 2 are ok.

How can we prevent in future this to happen? UPS?

Thank you.

I have 3x60GH bfl singles an one day after 15 minutes power failure allover fucking neighborhood one of them dropped to 33gh/s. not worth of shipping back to RMA since shipping from eastern europe to usa is aroun $150 and it will take way to long.
I'm using UPS since then!
it doesn't need to be anything special, just something that will give you enough time to stop the mining and power off your rig.

True, you can get it for 50$. Compared to price of GPU is nothing.

It's not just the pci-e extension cable by chance?

alot of UPS's have a usb cable and an app that can be installed on the host computer to automatically shut it down if power is lost before the battery dies.

Nope.

I remember back in the 90s we used to hook our computers up to surge protectors and sometimes UPSs.. do people still do that?

I use surge protectors just so we can plug more PSUs into one outlet! :P

But really, it all depends on the quality of your PSU. A nice, high quality PSU will make a difference. Our Rosewill Capstone 750W have 'Active PFC, Over Temperature / Current / Voltage / Power Protection, Under Voltage Protection, and Short Circuit Protection". Never had an issue with them, including 2 losses of power.

To heck with a PSU tho, set up a bank of AGM marine batteries..  Maybe use your btc to buy a few solar panels on overstock to keep them charged, and suddenly you can cut power consumption out of your profit/loss

time to RMA the card.

Avoid pushing the card that much.
a power outage most likely won't kill a GPU if you are not pushing them to the limit.

It is one of the reasons I don't OC my GPUs, it all depends on your income, if you can afford loosing one GPU, and instead push it to the limit to gain more, it is up to you.

in my case, GPU are way too expensive here and almost impossible to get, to risk them, I rather don't OC them at all, also use the best PSU quality you can afford, single rail, dual conversion (DC-DC for 5V and 3.3) are better. (you will see the DC DC converters close to the low voltage heatsink, it is easy to see them through the fan hole).

BTW: In one of the farms I built for a client, one card died, same as you, in my case, there was a BUG (literally, a biological one <GRIN>, some sort of BIG ANT) that crawled into the VRM area and fried (obviously VRM are close to 90 C easy!), I removed the bug (remainings), and apply some contact cleaner (spray) wait it for dry, apply pressure air to clean what it is left of the bug, and that was it, card back to life, maybe it is not something very common but check it out just in case.

I forgot to mention: don't handle the card when it is HOT, wait until it cools off, apply mechanical pressure to a gpu when at high temperature is looking for trouble.

Thank you induktor, very helpful.

 First thing I did after failure was putting clocks back
to stock 947/1250. Does it count pushing card to rise memory clock or just core clock?

I think quality of PS is mandatory. Two of my cards are connected to profi PS from COMPAQ proliant server;
The card which die was connected to standard pc PS. So Your theory and my praxis handshake here ;)

And the name BUG in software actually come from real biological BUG they found in first mechanical computers where bug block some relay contact.

I had a similar problem with my R9 290. Now I bought two 280x... They are also not so loud... That is important for me becouse I am mining in my room :) but I turn off R9 290 when I go to sleep ;)

I don't have a 290x but I've read that they are loud, you should consider replacing fans with japanese, dual ball bearing, server class fans, they are expensive, but incredibly powerful and very thick so even at low RPM they push a enormous amount of air.

I usually have 7970 working at 85% fan speed, after the mod, the fan rarely needs to go above 50%, at that speed the noise is not noticeable, but at 100% ( 5000 RPM) sounds like a jet engine taking off hehe :)

Nice! a proliant server PSU!, I will love to have one of these!, there is nothing better!.

did you try to boot with another card as primary and see if the driver sees the other card?, or if the other card is plugged in the OS don't start?
if it does, you can run some GPU testing on the secondary (probably failed) GPU to be sure.
did you change the bios to backup? (2)
just checking, probably you gonna have to RMA it, but it would be better be sure that there is nothing you can do to get it back to life.
indtk

It would be nice to post a link to those japanese server class fans please.

Tried any combination with cards - no success.

I have RMA card today to my dealer.

Finally received new card from dealer!!  \o/