Advanced search

Message boards : Number crunching : Energy loss

Author Message
GPUGRID
Send message
Joined: 12 Dec 11
Posts: 91
Credit: 2,730,095,033
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwat
Message 30863 - Posted: 16 Jun 2013 | 17:11:28 UTC

Well, I don´t have many where I live, but when it happens, it ruins all the units beeing processed. That would be 10 long run units, such a waste. There is any way to prevent this? The units just fail, driver reset, machine hangs up and so on. Very frustrating.

flashawk
Send message
Joined: 18 Jun 12
Posts: 297
Credit: 3,572,627,986
RAC: 0
Level
Arg
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 30864 - Posted: 16 Jun 2013 | 17:58:14 UTC - in response to Message 30863.

Are you saying you're power went out or you're computer crashed? When I first built the 4 computers I have now, there were all kinds of problems. I used the best components I could buy but it took a lot of fine tuning to get them stable and cool enough to run 24/7 365.

Have you looked at the logs in "Event Viewer", the most helpful are the Windows logs and the Application logs, check their dependencies too. One big issue I had that was hardware related was the voltage being supplied to my RAM, it was set to Auto in the BIOS (Asus M5A99FX PRO R2.0), the RAM required 1.5 volts and the motherboard was supplying 1.45 volts, I manually upped it to 1.575 volts (4 sticks, 32GB) and my stability increased 10 fold. You can safely take most memory to 1.65 volts with passive cooling as I have done it many times.

With the types of loads were putting on our computers, memtest won't catch these kinds of problems, I don't think there are many stress tests that even compare. I run 2 GPU-GRID models and 6 CPDN models 24/7 365 per computer on all 4 machines and seem to have been very lucky with crashes and spontaneous reboots, it took a few months but the logs in Event Viewer helped.

GPUGRID
Send message
Joined: 12 Dec 11
Posts: 91
Credit: 2,730,095,033
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwat
Message 30866 - Posted: 16 Jun 2013 | 18:31:51 UTC

No I mean, power cuts. Energy goes down on all the house, the block, etc.... when it comes back, all crashes on both machines. It cant recover from a previous saved point, it seems.

TJ
Send message
Joined: 26 Jun 09
Posts: 815
Credit: 1,470,385,294
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30867 - Posted: 16 Jun 2013 | 18:37:23 UTC - in response to Message 30863.
Last modified: 16 Jun 2013 | 18:39:13 UTC

Well, I don´t have many where I live, but when it happens, it ruins all the units beeing processed. That would be 10 long run units, such a waste. There is any way to prevent this? The units just fail, driver reset, machine hangs up and so on. Very frustrating.

I had power outages a few years ago, but things that run on the CPU and (I let write down to disk every 5 minutes), could be restarted and resumed where they where. I didn't now what happened with GPU jobs as I didn't had them them.

What you can do is to use a UPS for the crunching rigs and in case of a power outage and you are at home you can safely power-down the rigs.

I almost bought two UPS myself but there haven't been outages for the last 5 years. If it doesn't happen often than I should leave it as it is and take this as a loss. However it is pitiful.
____________
Greetings from TJ

flashawk
Send message
Joined: 18 Jun 12
Posts: 297
Credit: 3,572,627,986
RAC: 0
Level
Arg
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 30868 - Posted: 16 Jun 2013 | 18:55:38 UTC

http://www.newegg.com/Product/Product.aspx?Item=N82E16842102134

I'm sure you're computers would require a more powerful UPS, this will give you an idea. They come with their own software or you can configure what's built in to Windows or Linux, set BOINC to not run on batteries. I have 4 of these

http://www.newegg.com/Product/Product.aspx?Item=N82E16842102048

If you do decide to get some sort of UPS, make sure it's compatible with you're PSU, all newer PSU's are active PFC (APFC).

GPUGRID
Send message
Joined: 12 Dec 11
Posts: 91
Credit: 2,730,095,033
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwat
Message 30869 - Posted: 16 Jun 2013 | 20:19:03 UTC

Thanks for the tips guys :)

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30919 - Posted: 22 Jun 2013 | 12:32:17 UTC

As far as I understand a UPS runs down the battery even when there is no power loss, i.e. it decouples both grids completely. That means the rather expensive batteries will have to be replaced within a few years. This alone might make it more expensive than simply loosing a few WUs every now and then.

However, I don't know if there are "light UPSs" which would bypass the battery as long as power's still there.

MrS
____________
Scanning for our furry friends since Jan 2002

Profile dskagcommunity
Avatar
Send message
Joined: 28 Apr 11
Posts: 456
Credit: 817,865,789
RAC: 0
Level
Glu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30925 - Posted: 22 Jun 2013 | 16:29:07 UTC
Last modified: 22 Jun 2013 | 16:34:45 UTC

There is always a very little loading on the battery because the type of the batteries needs that. Depending on the sort of battery the lifecyle of them varies between 1 and 5 years. Sadly i had one with 1-1,5 :( the UPS maked more troubles then running the server on normal powerline -_- But the newer APC are rather good. The batteries have a good lifetime. Normal the UPS Builder say you have to change the batteries every 3 years, neither they are shown as still good. It can happend it shown as good but fail after very short time when the power is down (single cell defect). Costs of batteries are varies too very wide, so you have to look before buy the UPS ;) Its been a long time ago but i think mine cost about 180Euros or more but that was a fortress UPS not an APC. We have some APC in my work, but i dont buy the batteries for them so i cant say what we pay for the 750, 2200, 10000 Models.
____________
DSKAG Austria Research Team: http://www.research.dskag.at



Crunching for my deceased Dog who had "good" Braincancer..

nanoprobe
Send message
Joined: 26 Feb 12
Posts: 184
Credit: 222,376,233
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 30930 - Posted: 23 Jun 2013 | 4:48:31 UTC - in response to Message 30868.

http://www.newegg.com/Product/Product.aspx?Item=N82E16842102134

I'm sure you're computers would require a more powerful UPS, this will give you an idea. They come with their own software or you can configure what's built in to Windows or Linux, set BOINC to not run on batteries. I have 4 of these

http://www.newegg.com/Product/Product.aspx?Item=N82E16842102048

If you do decide to get some sort of UPS, make sure it's compatible with you're PSU, all newer PSU's are active PFC (APFC).

I have 3 of their 800 watt models. Love 'em. With all the momentary power interruptions and lightening strikes we have here having a reliable UPS is a must.

terencewee*
Send message
Joined: 29 May 12
Posts: 8
Credit: 21,605,500
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwat
Message 30933 - Posted: 23 Jun 2013 | 7:51:46 UTC

UPS-Wikipedia

I'm using line-interactive UPSes, even for my Internet connectivity. Quite cheap and can last ~3+years before needing a change (best just to recycle). Place UPS @ well ventilated location.

Do try to get those that offer USB connectivity - so that you can set your computer:

To automatically shutdown during power-outages
or
Config BOINC to suspend work when using battery
or
Run a script/batch for everything else



____________
terencewee*
Sicituradastra.

Post to thread

Message boards : Number crunching : Energy loss

//