Core 2500K - Are stability programs testing oc representative of actual use?

[zink77]

I've been playing around with my 2500K and I've done 12+hrs of maximum heat/power without a crash but whenever I run linx windows7 borks to blue screen. When I run games the utilization and temps are 40%-50% less then running burn in tests. Are burn in tests representative of 'stability' under actual use in the multi-core era? If no program you run ever pushes temps past 50-55C then aren't burn in tests more of a theoretical stability test? i.e. you can fail linx provided prime doesn't fail. All the programs I run never push core temp past ~55C, only the burn in tests on the Prime or linx push it higher.

I've safely o/c'd to 4.7ghz @ 1.4volts and run prime max power for 12+ hours straight, linx on the other hand just causes win7 to crash.

I don't think linx represents any actual software that could actually use the cores the that extent pushing temps to the point of instability.

 

[Belegost]

The standard OC stability tests represent programs designed to stress corner cases - high float stress, high fixed point stress, memory stress, etc. The goal in testing cases like these is that if you can run stable under the corner conditions, then it is reasonable to believe that all use cases intermediate should also be stable.

Certainly, it is possible that for the normal use scenarios you impose on the machine you will not run into the instability found at the edges. However, you take the risk that should your use scenario push you outside the stable region, you may suffer system instability, data corruption, and possibly even hardware damage.

Linx runs tests of linear algebra operations using linpack. These operations should heavily stress double and single precision float operation - cache, registers, and floating point units. Prime on the other hand stresses mostly integer operations, which are different registers, different instructions, and different processing units.

So using only Prime to judge system stability is like saying a car is safe to drive by testing the ability to go forward in a straight line and never turning the steering wheel. Maybe if all you do is drag race you're fine, but if your wife takes it for a trip to the supermarket it could crash... Much like a computer with an OC that isn't stable.

 

[Kenmitch]

LinX and/or Intel Burn Test(newest version) stress parts of the CPU that prime doens't touch like the AVX extensions. Altough it'll be awhile before we get apps that use the AVX extensions it's best to test for it to claim stability.

How are you testing with LinX? Intel Burn Test (latest version)? Default settings? Max memory?

What MB do you have? How is your overclock setup? Hard locked? Throttles?

Yes stress testing does subject your rig to a far greater load than everyday use.

 

[Concillian]

They are not representative of real world cases, but they're the best we have. They test specific functions. Other programs will generally use a wider array of functionality, but never will push the level of heat and stress that the stress testing programs will.

I have had cases where a computer was 100% prime95 stable for 24 hours, but a specific game would reliably crash the system in less than 10 minutes from launching. It was clearly doing something that Prime95 wasn't testing.

Since moving to Intel Burn Test and using more conservative distance to Tj specs I have had much better stability.

It's not worth pushing a machine to the point where it's on the edge, IMO. Sure it works fine now, but what about a few months from now when summer is starting, temperature warms up and you haven't cleaned the dust from your system in those 6 months? You're working on something important and get a blue screen and suddenly you have to stop what you're doing and start troubleshooting.

I much prefer the convenience of stability. Now I find linx stable, and then turn down the OC a little more than that at the same voltage just to be sure it's gonna be rock solid... even in the summer and when I skip spring cleaning on the case dust.

 

[VirtualLarry]

Prime on the other hand stresses mostly integer operations, which are different registers, different instructions, and different processing units.

I was pretty sure that Prime95 tests the FPU too. Possibly the SSE2 units as well.

 

[Puppies04]

There isn't one golden stress test you can run that tells you if your system is stable or not, usually running 2 or 3 differnt programs will give you a lot better idea as to the reliability of your OC.

If somebody posted here saying they were prime stable but still BSODing the first thing I would suggest is IBT.

If somebody posted saying they were IBT stable but still crashing the first thing I would suggest would be prime.

 

[fastamdman]

Just as a heads up puppies04, 99.99999% of the time if a system is stable on LinX/IBT it will be stable in p95 as prime95 does not push your system all that hard.

If you want a stable system, be LinX stable. Otherwise, turn your head to the bsod's and live with an unstable system.

In my personal opinion, instability is the plague and should be avoided at all costs. What if your in a match for a game and you bsod out of no where. What if you are typing a long email or document and then you bsod bam its gone. There are so many instances that bsod's can really ruin someones day that I don't think they should be allowed.

If a system is not LinX stable, it's not stable plain and simple. Just make adjustments until you can pass LinX, thats all ya gotta do.

Even if you don't bsod gaming, or web browsing it doesn't mean that at some point in time it won't happen randomly while loading up skype and a website, or alt tabbing from a game to check your instant messenger or whatever you do. If you are unstable, which you are, you will bsod, it's just a matter of when and how pissed you will be.

Also, a few mhz isn't worth the hassle of blue screens. Just get her stable.

Not a lot of 2500k/2600k's can do above 4.5ghz without a lot of voltage. Some can, some can't, either raise your voltage or lower your overclock. If you aren't stable for at least 2hrs in LinX, you aren't stable in my book. The difference between 4500 and 4700 is next to nothing and I guarantee you will be stable at 4500 at 1.4 🙂

 

[zink77]

If you want a stable system, be LinX stable. Otherwise, turn your head to the bsod's and live with an unstable system.

Temperatures are what are causing the instability if you run a thermal monitor + run the burn in test, it is the TEMPERATURE that is causing the crashes, temps are what causes the instability. Hence my main point. Burn in tests push CPU temps to places they will *never get to* using software. Any game I can run doesn't even come close to pushing temps to where prime and linx push temperatures. As far as I'm concerned if you can do prime 95 12+hrs your good to go. Linx pushes temps to where 99% of most real world programs can ever reach.

 

[zink77]

How are you testing with LinX? Intel Burn Test (latest version)? Default settings? Max memory?

What MB do you have? How is your overclock setup? Hard locked? Throttles?

Yes stress testing does subject your rig to a far greater load than everyday use.

Mobo: P8Z68-V PRO

Testing using "ALL" memory, 4 threads. My o/c setup is automatic (i.e. frequency turbos when using program, then idles back to 1600mhz and low voltage). Win7 when running LinX just crashes to blue screen with stop error 0x0000101 (not sure if thats the right number of zeros).

Temperatures are what is causing the instability/crashing, I run temp monitoring and watch the temperatures rise until crash.

 

[Concillian]

As far as I'm concerned if you can do prime 95 12+hrs your good to go. Linx pushes temps to where 99% of most real world programs can ever reach.

You posted here for a reason. Presumably that reason is to seek advice.

You got several people telling you the same thing (Prime95 alone is not enough to claim full stability). I have personally experienced this fact.

Then you post that you don't believe us and are going to do what you were going to do before you posted.

That's fine, you're free to do what you want, but it brings up an question in my mind... If you were going to ignore and seemingly argue against people (who have personal experiences) who know that Prime95 is not enough by itself, then why did you post in the first place seeking advice?

You will eventually get bitten by your unstable system. In my case I lost data too and needed to reinstall at a time where I really didn't want to spend the time reinstalling Windows. I hope that when your system has crashing issues due to it's instability, you are more fortunate and just need to change the clock down some.

 

[Kenmitch]

Mobo: P8Z68-V PRO

Testing using "ALL" memory, 4 threads. My o/c setup is automatic (i.e. frequency turbos when using program, then idles back to 1600mhz and low voltage). Win7 when running LinX just crashes to blue screen with stop error 0x0000101 (not sure if thats the right number of zeros).

Temperatures are what is causing the instability/crashing, I run temp monitoring and watch the temperatures rise until crash.

If you try it with just 1024mb of memory what happens? What kinda temps are you seeing?

The bsod is most likely due to a vcore that is just a little low.

I've been playing around with my 2700k this am at 4.8ghz. Using Intel Burn Test using " ALL " cores 100 runs at 1024mb, 100 runs at 2048mb is stable so far. Not sure if I'll do 4096mb or not as the system gets pretty unresponsive during the heavy load of IBT. I don't worry about temps too much as I'm water cooled but the hottest core so far today hit 73*C sometime during my testing.

 

[repoman0]

Just as a heads up puppies04, 99.99999% of the time if a system is stable on LinX/IBT it will be stable in p95 as prime95 does not push your system all that hard.

If you want a stable system, be LinX stable. Otherwise, turn your head to the bsod's and live with an unstable system.

In my personal opinion, instability is the plague and should be avoided at all costs. What if your in a match for a game and you bsod out of no where. What if you are typing a long email or document and then you bsod bam its gone. There are so many instances that bsod's can really ruin someones day that I don't think they should be allowed.

If a system is not LinX stable, it's not stable plain and simple. Just make adjustments until you can pass LinX, thats all ya gotta do.

Even if you don't bsod gaming, or web browsing it doesn't mean that at some point in time it won't happen randomly while loading up skype and a website, or alt tabbing from a game to check your instant messenger or whatever you do. If you are unstable, which you are, you will bsod, it's just a matter of when and how pissed you will be.

Also, a few mhz isn't worth the hassle of blue screens. Just get her stable.

Not a lot of 2500k/2600k's can do above 4.5ghz without a lot of voltage. Some can, some can't, either raise your voltage or lower your overclock. If you aren't stable for at least 2hrs in LinX, you aren't stable in my book. The difference between 4500 and 4700 is next to nothing and I guarantee you will be stable at 4500 at 1.4 🙂

Agreed with the main point, but both Prime95 and LinX are both still necessary. They do test different parts of the CPU - LinX touches many things that Prime doesn't and vice versa. Just a week or so ago I had an overclock that passed Linx for 100 minutes, with the regular high AVX temperatures. I turned on Prime95 just to be sure and one core failed in 8 minutes with temperatures about 10C lower.

But, either way, you should be able to run any stress test without failure or crashes in my opinion, otherwise there's always the chance that you're generating crap data somewhere and silently corrupting files, and of course there's always the chance you'll get a BSOD while you're doing something important.

Typically after my system is prime and linx stable I leave Prime95 on and start playing whatever game I'm into at the time - prime runs at low priority so the game will run smooth but this will ensure your CPU stays at 100% while also experiencing all the extra heat from the GPU. I might actually start doing this with Linx+AVX at low priority just to be extra sure. 100% stability is always more important than a few percent faster operating frequency, in my opinion

 

[Puppies04]

Temperatures are what are causing the instability if you run a thermal monitor + run the burn in test, it is the TEMPERATURE that is causing the crashes, temps are what causes the instability. Hence my main point. Burn in tests push CPU temps to places they will *never get to* using software. Any game I can run doesn't even come close to pushing temps to where prime and linx push temperatures. As far as I'm concerned if you can do prime 95 12+hrs your good to go. Linx pushes temps to where 99% of most real world programs can ever reach.

Temps are not causing the instability, in fact adding vcore and therefore increasing temps will stabilise the system (if the BSODs are being caused by insufficient voltage). You could quite happily hit tjmax and all that will happen is the CPU will underclock itself to artificially reduce the temp (although the life of the chip will be seriously compromised if you are regularily pushing the temp that high). I don't know where you got your information from but you are 100% wrong.