New PC Specs - your comments welcome

[martin-w 3,236]

Throttle was wrong word , most of them get unstable with high OC at 80-85C and the risk of a dead CPU.

A lot of dead good CPUs during Intel XTU with AIO and custom loops at ambient 20-22C with +1.45v 80-85C core temp.

i dont recomend to run stresstests at over 75C, optimal max 70C

 

Appologies Martin.....

 

 

:smile:  Fair enough Hasse.

[martin-w 3,236]

I have double checked now back at my PC - I have set:

- 46x all-core multiplier

- 100MHz BCLK

- VCore fixed at 1.25v (!!!!!)

- 2666 MHz XMP profile

 

LLC is set at Level 1 - I think this means the actual voltage varies.

 

My results running AIDA64 stress test for 1 hour are:

- max temps 78/80/77/76

- average temps are high 60s low 70s (the max temps are very sporadically reached, and only momentarily)

- "core voltage" is flipping between 1.248v/1.264v under the AIDA64 load, and 1.232v at idle (where CPU fq drops to 800MHz)

- using P3D, CPU temps are 60s, low 70s

- ambient temp will be low 20s (MB sensors read 28'C)

 

I went straight in at a voltage of 1.3v. tested as stable, and made one adjustment to 1.25v. Absolutely no stability issues under a 1 hour AIDA64 test. But this voltage seems too low! I'm worried I'm missing something, but above is what I set in the BIOS and monitor using CPUID. Any thoughts?

 

 

Hi Rich.

 

Personally, I say no need to set FPU, almost none of us encounter AVX instruction sets in daily use. Not unless you are running something like Adobe Premier, or some advanced features of Photo Shop.

 

Having said that, I find that even running FPU in Aida, doesn't generate the extreme temps something like Prime 95 Small FFT's do, so feel free if you wish to. The temp increase in regard to my rig isn't great with FPU on in Aida, compared with off. But as I say, not necessary and wont guarantee stability with the software we ordinarily run, "that doesn't utilise AVX". Asus will give you the same advice, as do Intel.

 

Stress testing should be as analogous as possible to the software we ordinary run day to day. That way you are stable in the software you built your PC to run. For example, you can be 100% stable in Aida and Prime95 for many hours, and then fail Handbrake in seconds. For me personally. in the past I have been 100% stable in Aida64/Prime95 and then crashed in BF4 within a few minutes, necessitating higher voltage. 

 

With the above in mind, I would highly recommend RealBench. RealBench runs tests that are as close as possible to the way we use our PC's in day to day use. And yes, one hour of RealBench stress testing, as instructed by Asus in their overclocking tutorial, is "usually" all you need. I tested my overclock in RealBench after each incremental voltage change, and then one final hour of testing when I had achieved my suspected stable overclock. Utterly stable ever since.

 

https://rog.asus.com/tag/realbench/

 

 

It's worth remembering that the true test of stability is that your system runs your day to day stuff in a stable fashion. With that in mind, I suggest regarding the next few weeks as a trial period, an assessment period, within which you run your favoured applications.

 

It's perfectly okay to use RealBench even if you don't have an Asus MB.

 

I use CPUz for monitoring voltage frequency etc. And RealTemp for temp.

 

Regarding you overclocking results and temps, it seems my NH-D15S is matching your results, which pleases me greatly. If you fancy doing a direct comparison running RealBench, let me know. Would be interesting and useful info.

 

Something like...

 

CPU fans at max

RealBench for 10 minutes or Aida if you prefer. Both of us with the same settings.

Measure temp in RealTemp

Take a note of ambient temp

 

Would be interesting to see the D15S and H110 compared. we can take into consideration that your voltage is lower than mine.

[james_2k 36]

I'm stressing FPU, CPU, Cache and System Memory. I know I need to do longer than 1 hour. These are initial tests. Temps on AIDA are prob about 10*C above P3D. My explanation may have been a little off initially on that one!

 

try it with just fpu selected. Will give the highest heat load. It's a worst case scenario, and good for checking that despite what you may run in the future, you won't hit the temperature limits

[Jetset408 108]

Again, thanks all - this turned into a very useful thread! Good ole Avsim!

 

 

 

With the above in mind, I would highly recommend RealBench.

 

Thanks - I will use RealBench for sure. I have it downloaded but not got round to running it yet!

 

 

 

It's worth remembering that the true test of stability is that your system runs your day to day stuff in a stable fashion. With that in mind, I suggest regarding the next few weeks as a trial period, an assessment period, within which you run your favoured applications.

 

Agreed. I only use my PC for Microsoft Office, internet browsing, emails, and Prepar3D! So its very possible I don't find its limits. I guess stress testing for me is to cover off the "what if" scenarios in the future.

 

 

 

use CPUz for monitoring voltage frequency etc. And RealTemp for temp.

 

CoreTemp, Corsair Link, AIDA64 and CPUID for me. Probably enough!

 

 

 

Regarding you overclocking results and temps, it seems my NH-D15S is matching your results, which pleases me greatly. If you fancy doing a direct comparison running RealBench, let me know. Would be interesting and useful info.

 

Absolutely - very happy to participate. Perhaps lets take it off line through PM's to run this then we can report back? I'm a bit pushed for time due to other commitments at the moment, but within the next few weeks, lets aim to do this. It will be a piece of mind for me and a useful exercise hopefully for others!

 

 

 

try it with just fpu selected. Will give the highest heat load.

 

Good tip - thanks!

[martin-w 3,236]

Absolutely - very happy to participate. Perhaps lets take it off line through PM's to run this then we can report back? I'm a bit pushed for time due to other commitments at the moment, but within the next few weeks, lets aim to do this. It will be a piece of mind for me and a useful exercise hopefully for others!

 

 

 

Okay Rich, drop me a PM when you're ready to have a bash.  :smile: This will be the first time I've had the opportunity to compare directly with someone with a H110. Nothing too complex, should just take a few minutes. I've always said, that if I ever opt for AIO, H110 it would be.

[Jetset408 108]

I actually have an H115i but I assume that's still ok for you? Will do on the PM!

[martin-w 3,236]

Yep, even better Rich. I believe it's the improved version of the H110.