Aida64 not such a great stress test.

manu2b

Member

​

Joined

Feb 6, 2011

• Nov 20, 2013

• #1

Well, I know Intel strongly advises using Aida64 for IB/Haswell CPUs, but I don't find this one reliable for stability testing.

Let me explain:

I can run it for 24hrs+ @4.8GHz/1.34v with no error whatsoever, but Cinebench crashes, and P95 needs 1.36v for a 12hrs+ blend run. Below that vCore, some worker will stop after a couple of hours.

Do you share the same experience?

 

EarthDog

Gulper Nozzle Co-Owner

​

Joined

Dec 15, 2008

Location

Buckeyes!

• Nov 20, 2013

• #2

Nope.

Run just the FPU test (select nothing else) and try it again see if it matches your other experience.

Also note, unless you specifically selected memory, you are not testing it on AIDA64 whereas blend in P95 of course tests memory.

 

OP

manu2b

Member

​

Joined

Feb 6, 2011

• Nov 20, 2013

• Thread Starter
• #3

Thanks E_D.

I did both FPU only and FPU/memory.

Same outcome: stable 24hrs+. Only a few degrees more when running FPU only (80ºC max instead of 77).

 

EarthDog

Gulper Nozzle Co-Owner

​

Joined

Dec 15, 2008

Location

Buckeyes!

• Nov 20, 2013

• #4

And if you run all stress tests there, what are your temps?

Are you on the latest version?

 

OP

manu2b

Member

​

Joined

Feb 6, 2011

• Nov 20, 2013

• Thread Starter
• #5

temps with all tests: 77°C
temps fpu only: 80°C
temps P95 blend: 80°C

yes, latest version.

 

EarthDog

Gulper Nozzle Co-Owner

​

Joined

Dec 15, 2008

Location

Buckeyes!

• Nov 20, 2013

• #6

Not sure what to say honestly.. As you know, you can try stress test A and it doesnt pass on Stress test B. You can say the same with applications as well. Stress testing isn't fool proof it is a general guide for stability. As always, testing the applications you actually use after a stress test should be the icing on the cake for stability.

 

W

Woomack

Benching Team Leader

​

Joined

Jan 2, 2005

• Nov 20, 2013

• #7

AIDA64 is working for me. About the same results in Prime95/XTU or AIDA64 but I never test for more than 2-3h so hard to say what will be after 24h+.

 

mimart7

Member

​

Joined

Jun 6, 2011

• Nov 20, 2013

• #8

You could try OCCT.

 

OP

manu2b

Member

​

Joined

Feb 6, 2011

• Nov 20, 2013

• Thread Starter
• #9

When Aida64 passes, all other apps I use (CS6, games...) are running without a crash.

But Cinebench and 3DM11/2013 physic tests need a bit more vCore (1.36v vs 1.34v) to pass in loop.

 

OP

manu2b

Member

​

Joined

Feb 6, 2011

• Nov 20, 2013

• Thread Starter
• #10

mimart7 said:

You could try OCCT.

Click to expand...

Same as P95: crash below 1.36v.

 

W

Woomack

Benching Team Leader

​

Joined

Jan 2, 2005

• Nov 21, 2013

• #11

I had something like that while benching 3DM Vantage and 3DM11 ... I thought that all is stable but in one test CPU was overheating and in the other one, CPU needed higher voltage

Stress tests are helping to check stability but it's always best to use couple of programs and set slightly higher voltage or slightly lower clock.
I have really often situation that everything is passing stress tests ( AIDA64, Prime95, IBT, etc. ) but in games I see various errors ( and games are not using hardware in 100% ).

In your case I would simply set 1.36V as it's the voltage that is passing everything.
You can also play with LLC options like lower level but higher voltage or higher level and lower voltage.

 

OP

manu2b

Member

​

Joined

Feb 6, 2011

• Nov 21, 2013

• Thread Starter
• #12

Thanks Woomack.

That's what I do. 1.36v with llc ultra high (same vcore in the bios and on load).

As temps never see 80ºC while stressing. I believe I am good to go.

Heck, I am even tempted to run it 24/[email protected]/1.42v (temps top at 88ºC). But it might be kind of a high vcore for Ivy Bridge...