Prime 95 alternatives for testing OC stability

Soldato
Joined
12 Apr 2007
Posts
12,606
Hi all is this still the go to stress tester?
Built up a new pc, 6600k, z170, 16gb team group ram.

My outgoing pc would always crash in prime, but was perfectly stable in games and 3dmark, so I just used 3d mark to fine tune the OC.. just wondering what people are using these days.


Thanks
 
realbench, intel burn test, aida64

Test in a range of tools, I have had systems 24hr stabs in prime, then crash instantly in realbench.
 
Hi all is this still the go to stress tester?
Built up a new pc, 6600k, z170, 16gb team group ram.

My outgoing pc would always crash in prime, but was perfectly stable in games and 3dmark, so I just used 3d mark to fine tune the OC.. just wondering what people are using these days.


Thanks

ROG realbench is good. It's a real world test as opposed to a completely synthetic test like P95.
As a result, realbench will give you a better idea of whether you are stable enough.
Someone once said, "If I make a program that crashes CPUs at stock settings would you feel the need to downclock them?" I think that really sums up stability testing, if it doesn't crash for what you do then it's all good!
 
I wouldn't use XTU or AIDA64 as they aren't demanding enough to test for game stability. Realbench and x264/x265 encoding are what I use for game stable tests. Though sometimes realbench cries about not having enough ram for some reason. But I have a 4.5ghz overclock that will happily do 8 hours of realbench and hasn't ever crashed in a game. However prime95 and occt will crash instantly.

Prime95 and occt are the most stressful and hot. Use them only if you want mega stability. I have a 4.2Ghz ovetclock that is prime stable in case anything demands it.
 
Asus RealBench here as well.

Originally when I built my rig a few weeks and and started with overclocking at 4.5Ghz I thought it was a bit odd or lucky that I was only getting max temps of 55c and all my fans not kicking in to reflect on the stress testing, I decided to take the former approach and then do a bit of more reading into it and then decided to give Realbench a go - now I'm glad I did, not only did this stress test everything the way it was meant to be done but also let me experience the integrity of my 6600K.

Liam.
 
I wouldn't use XTU or AIDA64 as they aren't demanding enough to test for game stability. Realbench and x264/x265 encoding are what I use for game stable tests. Though sometimes realbench cries about not having enough ram for some reason. But I have a 4.5ghz overclock that will happily do 8 hours of realbench and hasn't ever crashed in a game. However prime95 and occt will crash instantly.

Prime95 and occt are the most stressful and hot. Use them only if you want mega stability. I have a 4.2Ghz ovetclock that is prime stable in case anything demands it.

On the flip side XTU can test for things that encoding type tests won't necessarily - I usually use a mixture of XTU and some kind of x264/x265 encoding type test - I've never really found anything that won't pass both XTU and Cinebench that then fails something like OCCT - though for me a pass is when I've upped the voltage slightly from what initially appears stable.
 
The key is to use a range of tools that stress in different ways.

Don't just think it's stable enough for games. You don't know what you will use your system for in the future. You might get into transcoding your media with handbrake, or editing your game play footage for YouTube. Instability will strike at the worst time and range from mild inconvenience to catastrophic, so it's best to over do it.

It should only take a few overnight runs to really prove it's stable.
 
Its basically a gaming machine/media center.. so game stable is the acid test.
Unfortunately I can't afford a graphics card till next pay day, so ive run some cinebench tests, scoring about 720 with it overclocked to 4.4ghz, ram at 3000.

Almost seemed too easy, just enabled xmp, and ramped up the cpu speed, no manual voltage tweeks. It crashes at 4.5ghz thought, maybe needs a bit more vcore, temps dont go above 60c, idle is about 27c.

Can't really test games untill I get a graphics card though.
 
Ok strangely im now at 4.5ghz stable, at least in cinebench which is odd as windowsr was crashing during booting when I first tried, no vcore mods, although I think vcore might be on auto, everything is stock other than xmp on, multi at 45.

Has it 'got used to it' somehow? Temps still low with max of about 60c..
 
Before I purchase a XEON X5670 (in two months time) I updated my Gigabyte X58-UD5 from Bios version F7 to F13 in readiness. As we all know updating bios wipes out all the previous overclock settings.

Since the update and testing overclocking, I have been running several stress tests on my i7 920. Previously I used Prime from 3.5 GhZ up to 3.8 GHz overclock with no problems. Now under the new BIOS when running 8 threads I seem to get an occasional problem with a worker thread.

The error is

FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.

The first time this happened on Worker thread 7, the next time it happened on Worker thread 6. It happens after a couple of hours.

After that when I leave Prime running and it does not happen again with 7 cores running continuously for 12 hours overnight. If I switch off Hyper-Threading and run on 4 cores I do not receive any errors. Not ever.

I have tried another test called HeavyLoad and do not have any problems no matter how many threads I have running.

My RAM is G.Skill 6GB DDR3 NQ PC3-10666C9 1333MHz (3X2GB) Triple Channel DDR3 (F3-10666CL9T-6GBNQ).

https://www.overclockers.co.uk/g.sk...ple-channel-f3-10666cl9t-6gbnq-my-030-gs.html

I never get any blue screens, so I am wondering if it is getting flaky. I run it at 1.64v. Perhaps I should go to 1.5v?

Thoughts anyone?
 
Last edited:
I use RealBench and x264 stability test. I do an hour on RealBench during the day to confirm that my OC is not too unstable and to confirm max vcore and temp. Then I run x264 v2 overnight to test stability.
 
Back
Top Bottom