Hardware Canucks

Hardware Canucks (http://www.hardwarecanucks.com/forum/)
-   Troubleshooting (http://www.hardwarecanucks.com/forum/troubleshooting/)
-   -   prime95 rounding errors on a stock system (http://www.hardwarecanucks.com/forum/troubleshooting/54894-prime95-rounding-errors-stock-system.html)

kpamir June 19, 2012 02:22 PM

prime95 rounding errors on a stock system
 
I do freelance 3d work for clients and I have recently hit a pretty big snag with my PC.

I have a intel x58 system and it has worked very well for me for the past year or so but I have been having issues lately with renders erroring out and not finishing.

The rig consists of the following parts.

intel 970
sabertooth x58
24 gb corsair kit 6x4gb dimms that have just been replaced with 3x 8gb dimms.

the rest of the rig is fairly standard, SSD, 2tb hdd for storage, evga 560ti video card.

I ran prime 95 over the weekend and yesterday while at work and I get rounding errors on 3 of the cores it runs on.

Keep in mind that the system is running at stock.

Temps are fine, and at load they get to around 75 celsius now that it is summer.

I would like to narrow it down to what would be the possible issue. Is my motherboard the culprit?

Since I got the same error with my old set of ram. The only difference was that with the old set of ram the Prime95 calculations completely stopped when I got the error. And the error was on every core running.

Where as with the new ram I only got the error on 3-4 of the 12 "cores" running, and the calculations continued after the error.

Since the rounding error occurred regardless of which ram I used I assume the issue might be something else.

I also did a render test with Maya, and the scene that failed with the old ram also fails with the new ram.

Any suggestions would be great.

Squeetard June 19, 2012 02:36 PM

You are running the torture tests? Try the ones that test lots of ram and the ones that stress the cpu without using ram, see if that can tell you whether the culprit is ram or cpu.

I'd also try tweaking the bios settings and see what gives.

First, up the cpu voltage a bit and see if that helps, or underclock it. Then go conservative on the memory timings, maybe up the ram voltage a touch too.

Hopefully you can isolate the problem, or even make your system stable.

Adzsask June 19, 2012 02:38 PM

Manually set RAM voltage/timings? then try re seating next, few times I have had crashes due to a stick of RAM getting loose.(LAN lots)

kpamir June 19, 2012 03:00 PM

Quote:

Originally Posted by Squeetard (Post 635926)
You are running the torture tests? Try the ones that test lots of ram and the ones that stress the cpu without using ram, see if that can tell you whether the culprit is ram or cpu.

I'd also try tweaking the bios settings and see what gives.

First, up the cpu voltage a bit and see if that helps, or underclock it. Then go conservative on the memory timings, maybe up the ram voltage a touch too.

Hopefully you can isolate the problem, or even make your system stable.

Thanks for the input, I will try the individual tests tonight and tomorrow morning, hopefully that narrows it down for me.

I set my timings to the default that the sticks are rated for which is 10/10/10/27 at ddr3-1600 at 1.5v and fiddled around with the settings a couple of times already.

I will play around with the cpu settings as well.

Thank you sirs.

Krisso June 19, 2012 08:12 PM

Quote:

Originally Posted by kpamir (Post 635929)
Thanks for the input, I will try the individual tests tonight and tomorrow morning, hopefully that narrows it down for me.

I set my timings to the default that the sticks are rated for which is 10/10/10/27 at ddr3-1600 at 1.5v and fiddled around with the settings a couple of times already.

I will play around with the cpu settings as well.

Thank you sirs.

May also want to check the command rate as well. 1 board I've had only ran at the rated timings if the command rate was 2T.

enaberif June 19, 2012 09:05 PM

Small FFT is better for CPU
LinX is great for ram.

sswilson June 19, 2012 09:30 PM

Quote:

Originally Posted by Krisso (Post 636019)
May also want to check the command rate as well. 1 board I've had only ran at the rated timings if the command rate was 2T.

Yeah... seems to be an intel motherboard thing... the great majority of memory seems to be binned @ 2t, while it seems that motherboards want to select 1t if left at auto... definitely something to check.

kpamir June 20, 2012 12:25 PM

Yeah I double checked the settings last night, and they were set to what the ram is rated for.

I bumped up voltages on all the necessary things, but I have yet to run any tests to figure out the problem, hopefully this weekend I can get around to that.

kpamir June 25, 2012 07:25 AM

I did some testing over the weekend and the CPU stress test went fine, while the ram heavy tests seemed to have restarted my system.....but I have a strong sense that the power just went out and then came back on and thats what caused the restarts.

Regardless I bumped up the voltages a little more and tested it last night while I slept and it seems to be fine now.

I am going to test it again tonight with an actual scene to render, hopefully my problems are solved.

Appreciate the input fellas.

Squeetard June 25, 2012 09:33 AM

Weeee. Watch your temps now though.


All times are GMT -7. The time now is 05:10 PM.