Linpack vs. Prime95

[BigDMadinko]

I understand IntelBurnTest is a lot more stressful than Prime95. I have a Q9550 @ 3945 Mhz on a P5Q Pro, Zigmatek Rifle, and using Real temp 3.00. The problem is I'm not sure what the difference in temperature should be while running a stress test with each of the two programs. Prime can get to a maximum of 64 50 59 59. Running BurnTest I get as high as 73 70 71 72.

[GAR]

Burntest causes more heat, not sure how, but it does. Anyway, if your system is Burntest stable, then you are good to go!

[OverDozeD]

I dont trust real temp, using core temp instead
There's ~10 degrees difference between Prime95 and linpack for me
Actually, I dont like linpack, nothing can load CPU (for home use) as much as it does

[BigDMadinko]

I dont trust real temp, using core temp instead
There's ~10 degrees difference between Prime95 and linpack for me
Actually, I dont like linpack, nothing can load CPU (for home use) as much as it does

I'm starting to feel the same way myself, it doesn't make much sense to unnecessarily be stressing the processor to the level that linpack does, as you say, for home use.

[ReverendMaynard]

I'm starting to feel the same way myself, it doesn't make much sense to unnecessarily be stressing the processor to the level that linpack does, as you say, for home use.

The point is, that it does a great job of taking that cpu to the outer limits of operation so you get almost instant stability feedback to help you dial in your overclock.

[Tekdemon]

I dont trust real temp, using core temp instead
There's ~10 degrees difference between Prime95 and linpack for me
Actually, I dont like linpack, nothing can load CPU (for home use) as much as it does

Other programs don't load the CPU as much but the point is to load it up and see if it's stable, so you know that in months and months of regular loads it wouldn't error out.

Of course part of the problem is that the extra heat it's generating causes some instability, but it's probably best to run it at linpack stable speed anyway.

[roller11]

The point is, that it does a great job of taking that cpu to the outer limits of operation so you get almost instant stability feedback to help you dial in your overclock.

Your stability and OC are *strictly* relative to the progs you run on a daily basis. So unless you need to solve a system of linear equations, linpack won't help you find your stable OC.

[ReverendMaynard]

Your stability and OC are *strictly* relative to the progs you run on a daily basis. So unless you need to solve a system of linear equations, linpack won't help you find your stable OC.


You're alone, or with very few people on that train of thought roller, so think what you like and deny the fact that linpack is a legitimate stability tool.

[OverDozeD]

For me stability=Prime95 Blend test (5hr) + memtest(DOS 5 loop) + Crysis benchmark tool 10 loop maxed out

[mattkosem]

I've been well served by:

1) 1 hour prime95 8k FFTs not in-place with 1600mb allocated (open the app, select custom, and adjust the max box from 4096k to 8k to get this configuration)
2) 10 passes of linpak at maximum memory
3) 400% coverage with windows memtest with one instance per core, each instance with an equal share of your total available system ram.

I've never experienced a crash on a system that can pass those tests, Intel or AMD, and it doesn't take ages to finish them all.

--Matt

[Mescalamba]

For me, if it survives 35 mins OCCT, its ok.. cause more simply cause only one thing, more heat.. stability check is really fast thing. When you run it longer, it only proves if your cooling is sufficient or not. :)

[NaMcO]

If you want to see if your system is stable after a few weeks of use and run Vista;

1) Open Control Panel
2) Go to Administrative Tools
3) Double-click on Reliability and Performance Monitor
4) Click on the "Reliability Monitor" tab.
5) Enjoy

I never used linpack nor Prime95 to test my system's stability (although i DID use them already), i rather look at that graphic and check for a perfect 10.0 score :up:

[BigDMadinko]

The point is, that it does a great job of taking that cpu to the outer limits of operation so you get almost instant stability feedback to help you dial in your overclock.

I understand what the point of running a stress test is, or I wouldn't be doing it, but I wonder if it makes much sense to be stressing to these levels if Prime95 can be doing the same thing without heating my chip to near meltdown levels. Its faster with IntelBurntest, and I like it for this reason, but other then the convenience factor, isn't Prime95 also stressing the cpu to its limits, at least the limits that a home pc could possibly reach?

I was wondering primarily about the temperatures, its easy to find what other people are getting while running Prime95, but can only find vague mentions on the internet for what my temp limit should be with Linpack.

[Cyber-Mav]

stress testing varies from person to person i guess. me personally i want a overclocked system for more cpu speed but i also was 100% stability. so i ran intel burntest for 22 hours and after that passed i ran prime95 for 76 hours in blend mode. and before i even put the os on the system i ran memtest for 9 hours. hehe was funny though expecially running the intel burntest stresser, the q6600 @ 3.2ghz i have is cooled by an arctic cooling freezer 7 pro on pwm fan control. fan was spinning up and down all night when i went to sleep.

IBT gets the temps 10c ish higher than prime did. although need prime to catch a few memory controller instabilities. IBT is dam good test for the cpu. after all the whole point of a stress tester is to hammer the cpu good n proper till it errors out. thats when you have to increase voltage or downclock or get the cooling sorted (all all 3).

oh freezer 7 gets me q6600@3.2ghz@stock vcore of 1.2875v up to 72c in hot weather when i run IBT.

[bluehaze]

For me Linpack/Burntest is nice for initial stability testing just because sometimes testing with Prime takes hours to find an unstable OC but I find by testing with Burntest/Linpack first if I pass a 20 pass test with max workload then I am close on being 24hr Prime stable so I don't have to run the test and then start over when it fails 8 hours in. I usually end up bumping vcore up 1 or 2 notches to be prime stable for 24 hours after passing 20 pass Linpack and it saves alot of time.

Others have different results though, it seems for some that are Prime stable will still fail Linpack, I have never had this happen so who knows but I've come to the conclusion that for me personally I will use both just to ease my mind so that when a game or app crashes I know it's not due to my OC.

[OC Nub]

I noticed on my e8400 I could pass Prime Blend countless hours then fail IBT or LinX. Now with a core2quad I can pass ITB and LinX only to fail Prime Blend. Haven't had any problems with crashing or errors after passing Linx. So if that passes I call it stable enough for me, at least for a few days then it bugs the hell out of me until I pass both Prime and LinX

[Kougar]

That's just it, I use both. Prime Blend is good for checking the memory subsystem, not just the CPU. IntelBurn is only good for the CPU and won't detect almost fully stable RAM... I can set RAM settings that fail a memory test program but don't show errors in IntelBurn. There isn't any single one program that definitively prooves a system is as stable overclocked as it would be at stock.


Your stability and OC are *strictly* relative to the progs you run on a daily basis. So unless you need to solve a system of linear equations, linpack won't help you find your stable OC.

But what if you run programs you never have before on a routine basis? How would you know with any reasonable level of confidence that it wasn't an overclock issue causing the problem but just problems with the program itself... revert to stock every single time some program fails to work properly?

Game develoeprs like Valve can't even design their games to be halfway stable, even with stock system L4D likes to lock up. And TF2 would always crash upon exiting the program for Vista 64bit (& Windows 7 64bit) users until Valve finally patched it March 18th... more than a year after the game was on the market. This made tools like Vista's "Reliability Monitor" completely useless because their games were constantly crashing... and that was just Valve. There are certainly worse offenders out there.

[BigDMadinko]

I personally like IntelBurnTest, its so fast and you know you're getting maximum stress. Its hard to go back to prime after using it, it takes too long. The only thing that bothers me is that there is a point where I feel like I'm putting excessive stress on my chip, and I feel like I'm at least getting close to this point when I'm running burn test.

[mickrussom]

Your stability and OC are *strictly* relative to the progs you run on a daily basis. So unless you need to solve a system of linear equations, linpack won't help you find your stable OC.

Wrong.

Most programs used on a daily basis do not contain internal self checking, nor do they stress the CPU much. I find confidence testing systems with prime95/intelburntest to be a great way to make sure the system is not going to have issues.

The main problem with systems is memory, and without ECC is much harder to tell if the memory is stable. With these math programs, there is self checking, so if the memory is bad something like this will catch errors.

What a terrible comment.

[zalbard]

Way to necropost... :shakes: