What stress test is better? CPU-Z or Aida64?

[Inversion]

Pretty self-explanatory question.

[Oshino Shinobu]

CPU-Z's stress test isn't that great. Use CPU-Z for stats and a more fleshed out test (like Aida64) for stress testing. 

[Bananasplit_00]

AIDA64 is better, much better

[WittsEnd]

Am I the only one that still uses Prime95? lol

[Inversion]

cool thanks

[faziten]

1 hour ago, WittsEnd said:

Am I the only one that still uses Prime95? lol

makes little sense to be honest.

The load it puts VRM into makes it closer to a VRM test rather than CPU test. I had many builds pass 12 hours of prime 95 to later fail in random tasks like converting H264 videos with handbrake. Plus, it's unsafe to run on new CPUs since we don't know when a Haswell AVX scenario will happen again.

 

 

[WittsEnd]

Quote

The load it puts VRM into makes it closer to a VRM test rather than CPU test. I had many builds pass 12 hours of prime 95 to later fail in random tasks like converting H264 videos with handbrake. Plus, it's unsafe to run on new CPUs since we don't know when a Haswell AVX scenario will happen again

Makes sense, i'm a little out of the loop with stress testers. I haven't done a stress test on a pc for a few years. Next time I need to do one though I will check out AIDA64.

[done12many2]

1 hour ago, WittsEnd said:

Makes sense, i'm a little out of the loop with stress testers. I haven't done a stress test on a pc for a few years. Next time I need to do one though I will check out AIDA64.

 

You will find that AIDA64 is a waste of time if you are used to Prime95 type loads.  Give something like RealBench stress test a go before wasting time with AIDA64.  AIDA64 requires just a little more voltage than what's required to boot Windows, which is why so many people like it.  Nobody want to put a real load on CPUs anymore.  They'd just rather talk about their high clock speeds.  

[Melodist]

25 minutes ago, done12many2 said:

 

You will find that AIDA64 is a waste of time if you are used to Prime95 type loads.  Give something like RealBench stress test a go before wasting time with AIDA64.  AIDA64 requires just a little more voltage than what's required to boot Windows, which is why so many people like it.  Nobody want to put a real load on CPUs anymore.  They'd just rather talk about their high clock speeds.  

Are you kidding me? Don't spread false information! My 6900k load with AIDA 64 is higher than with Prime 95, meaning I get higher temps using AIDA 64...

[done12many2]

1 hour ago, Melodist said:

Are you kidding me? Don't spread false information! My 6900k load with AIDA 64 is higher than with Prime 95, meaning I get higher temps using AIDA 64...

 

Bud I can tell you without a shadow of a doubt that you are lost.  Nobody in their right mind will believe that you get higher temps and a heavier load from AIDA64 than you would from Prime95.  It ain't happening.  

 

FYI, if your 6900k gets hotter with AIDA64 than it does with Prime95, you may want to contact Intel for a RMA.  Good luck.

[Melodist]

1 minute ago, done12many2 said:

 

Bud I can tell you without a shadow of a doubt that you are lost.  Nobody in their right mind will believe that you get higher temps and a heavier load from AIDA64 than you would from Prime95.  It ain't happening.  

 

FYI, if you're 6900k gets hotter with AIDA64 than it does with Prime95, you may want to contact Intel for a RMA.  Good luck.

Why? I'm getting the exact same Geekbench result as any other 6900k.

[done12many2]

1 minute ago, Melodist said:

Why? I'm getting the exact same Geekbench result as any other 6900k.

 

The fact that you just mentioned Geekbench speaks volumes.   Good luck man.  

[Melodist]

25 minutes ago, done12many2 said:

 

The fact that you just mentioned Geekbench speaks volumes.   Good luck man.  

LOL and what do you use your PC for? On Linustechtips? Neuroresearch? WTF DUDE? What is your fucking problem? Prime 95 is old and at the time I used it didn't use the CPU properly, that's a known issue, because if you were such an expert, you should've known. How do you check your CPU's performance other than Geekbenchm, Cinebench and Adobe Premiere? Let me guess, Crysis 3?

 

Read that thread before you begin to cock around:

 

http://www.overclock.net/t/1390486/aida64-vs-prime95

 

Their CPUs are probably all faulty.

 

And you are using Wolfram Mathematica to benchmark your CPU, Dr. done12many2.

[MageTank]

20 minutes ago, Melodist said:

LOL and what do you use your PC for? On Linustechtips? Neuroresearch? WTF DUDE? What is your fucking problem? Prime 95 is old and at the time I used it didn't use the CPU properly, that's a known issue, because if you were such an expert, you should've known. How do you check your CPU's performance other than Geekbenchm, Cinebench and Adobe Premiere? Let me guess, Crysis 3?

 

Read that thread before you begin to cock around:

 

http://www.overclock.net/t/1390486/aida64-vs-prime95

 

Their CPUs are probably all faulty.

 

And you are using Wolfram Mathematica to benchmark your CPU, Dr. done12many2.

Yes, let's use a thread that is 5 years old, on a CPU that doesn't support AVX, as concrete evidence for which is a more brutal stress test. 

 

The answer to OP's question is neither. Linpack MKL is more brutal than Prime95, which is worlds more brutal than Aida64. Anyone that says otherwise has clearly never ran MKL, and anyone without a soldered/delidded chip that supports AVX will likely be unable to do so. 

 

Since we seem to be extremely fond of using OCN as our source, enjoy: http://www.overclock.net/t/1411077/haswell-overclocking-guide-with-statistics

 

 

[iconoclastic]

Personally, I prefer Prime95 as a starting point.

 

AIDA64 can be stressful, but there's no point in running it in my select battery of tests. Nowhere near as stressful a properly configured Prime95. MKL Linpack (with current libraries from Intel's website) is BY FAR the most stressful thing you can do to a CPU. However, it's easy to pass, not sensitive to errors and in general, incredibly hot.

 

My test battery includes at minimum:

1 full pass of Prime95 blend at 15m/FFT, or roughly 26 hours;

large archive testing on binary data using WinRAR (it's sensitive to memory);

Linpack FLOPS to generate generic performance metrics;

and encoding a video or two.

 

My box runs 24/7, mining and usually some hyper-V instances crunching away at stuff. Stability never falters.

 

 

And about that "Prime95 is unsafe" crap- it is not unsafe, has never been unsafe, and will never be unsafe. I have a 4790K which I still use every day, and it has been active since I bought it near it's release. Overclocked, very tweaked RAM, and extremely stress tested. Trust me, if Prime95 could kill a CPU, it would have died a long time ago. There is no such thing as a piece of software that can physically damage a CPU. That is people being unable to accept responsibility for user error.

[MageTank]

Just now, iconoclastic said:

Personally, I prefer Prime95 as a starting point.

 

AIDA64 can be stressful, but there's no point in running it in my select battery of tests. Nowhere near as stressful a properly configured Prime95. MKL Linpack (with current libraries from Intel's website) is BY FAR the most stressful thing you can do to a CPU. However, it's easy to pass, not sensitive to errors and in general, incredibly hot.

 

My test battery includes at minimum:

1 full pass of Prime95 blend at 15m/FFT, or roughly 26 hours;

large archive testing on binary data using WinRAR (it's sensitive to memory);

Linpack FLOPS to generate generic performance metrics;

and encoding a video or two.

 

My box runs 24/7, mining and usually some hyper-V instances crunching away at stuff. Stability never falters.

 

 

And about that "Prime95 is unsafe" crap- it is not unsafe, has never been unsafe, and will never be unsafe. I have a 4790K which I still use every day, and it has been active since I bought it near it's release. Overclocked, very tweaked RAM, and extremely stress tested. Trust me, if Prime95 could kill a CPU, it would have died a long time ago. There is no such thing as a piece of software that can physically damage a CPU. That is people being unable to accept responsibility for user error.

I do not know who you are, but kudos for this post. It encompasses everything I try to instill in people when they ask about "stress tests". In my eyes, there is a difference between "stress testing" and "stability testing". Stress testing is exactly what the name implies, stress. I mostly use it for testing thermal solutions more than anything, while stability testing requires using several different applications (both synthetic and real world) to make sure your CPU is stable under every possible load. Little bit of Prime95 at various FFT lengths, little bit of Linpack for performance, and a ton of real world applications to make sure what I intend to use my hardware for, doesn't cause my system to crash.

 

I know plenty of people that pass Prime95 blends and Aida64 runs, only to crash mid-game under what is typically considered  "lighter stress". Prime95 is great because you can do an awful lot with various FFT lengths to simulate different kinds of load, but even that is not enough to completely test stability, and likely never will be. No one stress/stability test should be treated as the be all, end all. 

[Bananasplit_00]

Just gona say, P95 crashes shit that dosent crash in real world cebarios from my experience. Crashes some of my OC's so I guess "I didn't overclock well enough" and then they never crash under any load I happen to put them through, no games, no video rendering(ok that's tops once a year but that's never crashed), no browsing, no benchmarks, no streaming... I see no point in pushing way too much voltage because "it crashes this one stresstest that tickles the fairies the right way so it's NOT RIGHT!!!"

[MageTank]

11 minutes ago, Bananasplit_00 said:

Just gona say, P95 crashes shit that dosent crash in real world cebarios from my experience. Crashes some of my OC's so I guess "I didn't overclock well enough" and then they never crash under any load I happen to put them through, no games, no video rendering(ok that's tops once a year but that's never crashed), no browsing, no benchmarks, no streaming... I see no point in pushing way too much voltage because "it crashes this one stresstest that tickles the fairies the right way so it's NOT RIGHT!!!"

I've had some of that in reverse, where my GPU's pass 3DMark, Superposition and Unigine, but crash immediately on Overwatch, lol. Nothing wrong with people having their own preferred stress tests, as long as they make sure their system is stable for their own intended usage. 

[aisle9]

3 minutes ago, MageTank said:

I've had some of that in reverse, where my GPU's pass 3DMark, Superposition and Unigine, but crash immediately on Overwatch, lol.

I've had that exact thing happen more than once. Playing an entire match with a black screen and occasional red or blue outlines of other players was fun for me. Not so much for my team.

[Bananasplit_00]

10 hours ago, MageTank said:

I've had some of that in reverse, where my GPU's pass 3DMark, Superposition and Unigine, but crash immediately on Overwatch, lol. Nothing wrong with people having their own preferred stress tests, as long as they make sure their system is stable for their own intended usage. 

GPU benches i have the same experience, but not CPU ones stress tests