Best stress test utility for overclocked 2500k in 2018y.

[BambOOs]

Hello!
I'm looking for best stress test utility to check my cpu's stability after OC.
Last CPU which I OCed was Q6600 - in those days I've been using LINX, Orthos and IBT.
Which utility is the best nowadays? What about Prime95 and OCCT with it's Linpack AVX stress test?
TIA.

 

[RejZoR]

RealBench from ASUS. Imo the only stability testing tool I can rely on.

 

[Regeneration]

The best stress testing utility for 2018 is Linpack Xtreme of course . Run it overnight with 80% of your RAM, if it passes, your system is 100% stable.

 

[John Naylor]

What is your goal ?

-If your goal is test stability in a multitasking environment using real world applications to get the best OC you can then RoG Real Bench is your best bet.
- If your goal is to be able to say you tested your OC under the most extreme and unrealistic conditions ... conditions which your CPU will never be exposed to in its liufetime, then test will one of, or even, ALL of the synthetics. Your stable OC will be significantly lower than it otherwise would be because no set of conditions will ever place such a single task loading on your PC again.

I am not saying there is no value to using synthetics... the question is, is their value to you ? As the saying goes, many folk "won't be able to sleep at night" knowing that there's some set of conditions which could be created, even artificially, that would make an OC unstable. OTOH, I have had 24 stable synthetic OCs fail under a multi tasking benchmark in < 20 minutes.

It could very will be that a OC limited by OC temps to 4.8 Ghz using synthetics, can reach 5.0 Ghz under RoG Real Bench. Yes, it might not get by synthetics because you hit temps that aren't comfy.... but do you care since PC will never see similar loadings again ?

To put it in simpler terms, if you live near the shore at 75 feet above sea level, when you add a trailer hitch to your compact SUV to haul your new 260 pound Skidoo to the beach 14 miles away, is there really any value to testing the hitch and engine / tranny / fuel mixture by hauling a 10,000 pound load up and over the Rocky Mountains ? Your hitch / SUV will never see that load nor the thin air of the Rockies.

Like most things PC related, there is no "best" for everyone... just what's best for you under the anticipated loads and conditions you expose your PC to.

 

[BambOOs]

Okey guys. Thank you for answers. I just overclocked my CPU to 4.5 GHz. Offset voltage mode, actually -0.030V. LLC on level 4. Additional Turbo Voltage on + 0.004 (to be sure MB won't add to much).
Rest of settings like C states etc. set correctly.

I'm wondering about Vcore voltage.

With current settings Vcore under full load (OCCT Linpack + AVX) gives 1.3V (CPU-Z) - I think it's stable at this level, but I will check it later with longer test.

But... sometimes I see ~1.34V jump in CPU-Z, when I for example launch some application. Is that jumps normal?

 

[cadaveca]

But... sometimes I see ~1.34V jump in CPU-Z, when I for example launch some application. Is that jumps normal?

yes, normal, and well within safety margins.

 

[RejZoR]

Problem with most stability testing tools is that they only hammer CPU. RealBench hammers all subsystems, even graphics one which is why I found it to be more reliable as it also heats up graphic card.

 

[BambOOs]

Okey, previous settings were not stable.
I changed LLC to level3. Current voltage under full load is 1.328V and it passed 7.5 h occt test so I guess it's stable this time.

I will try to lower the offset (or maybe increase it and back to LLC level4) to set Vcore under load to ~1.31-1.32. Maybe 1.33 is higher than needed.

 

[RejZoR]

Keep LLC at OFF or lowest possible level. LLC is really needed for extreme overclocking. We're talking DICE, phase or NL...

Also, make sure to load CPU, RAM and GPU at once. It's possible that system will pass CPU only test, but will be unstable when running games because you'll also put load on RAM and GPU.

 

[BambOOs]

Currently I set:

LLC level 4 (25%)
-0.005V offset

pretty stable.

Voltage under full load: 1.312-1.320V

But there are spikes to 1.344V while average load like opening apps or playing not demanding game. It just fluctuates: 1.336, 1.328, 1.344V.
Is there a way to get rid of these spikes?

 

[cadaveca]

Is there a way to get rid of these spikes?

No, and you shouldn't worry about it. It's actually what you WANT to see, as CPU load lessens. The current is converted to voltage, and helps to protect the CPU.

 

[BambOOs]

I just tried fixed voltage mode. And now it's seems that LLC is not working at all. There is no Vdroop at all. LLC is now set at 5 level (0%).

I set 1.31V and this voltage doesn't change even at full load. Is this normal? I thought that Vdroop is always presented.

 

[cadaveca]

That's what it is supposed to do, and NOT what you want for 24/7 use, as instead of the voltage fluctuating, the current does. That's for benchmarking purposes.

Today's boards offer "adaptive" voltage, which is actually what you really want, but is not possible on the P67 platform.

 

[BambOOs]

So LLC is not working when fixed voltage mode is enabled?

 

[cadaveca]

So LLC is not working when fixed voltage mode is enabled?

technically, it IS working, but at a set value. Proper loadline levels for any CPU sees voltage fluctuate according to workload, and the options you have for LLC vary how much it fluctuates (either above or below requested voltage). No fluctuations on voltage means that the current is what is fluctuating instead, and that can be dangerous for a CPU and lead to an earlier death than if proper loadline calibration values are used.

 

[BambOOs]

So with Level 5(0%) there should be high Vdroop (from for example 1.31 to 1.29) But there's not... Maybe my MB is broken or something?

 

[jboydgolfer]

id just apply your OC settings, then use the CPU as you would normally, doing whatever tasks you normally do, & if you encounter stability issues, adjust accordingly.. Using applications like P95, etc to "stress test" stability is silly, and a time waste, unless the OC is what your doing as a "project/hobby", in which case , im sure there are plenty of suggestions already posted. But if your not OC'ing as a hobby, using things that super stress your CPU in a way that will never be encountered outside of the stress application is like building a car that can operate under water, even though it will never go under water.

2500k is a great chip, i loved mine, i hope you have fun with it

 

[BambOOs]

Yes, you are right. I will just test in normal usage, gaming etc.

But I just wonder why this fixed voltage mode is not working - I mean there is no Vdroop at all no matter LLC settings I choose. At Level 5 (0%) Vdroop should be pretty high. It seems that Fixed Mode is broken. I even tried Beta Bios, but this not fixed this.

 

[Keullo-e]

Switched myself to gaming to test stability. Some heavier new games are one excellent way for testing, though for streaming I use the 100% stable clocks.

 

[BambOOs]

Ok guys, here is the little update to my OC results:

Actually I have one profile which I think is stable (99%). It passed 5.5 hours of OCCT Small Data Set test. In normal usage there is no problem at all until now.
Settings:
- Load Line Calibration at Level 3 (50%)
- Offset Voltage "-0.030"
- Additional Turbo Voltage set to minimum = +0.004V

Above settings gives some spikes to 1.336 V at average load. At full load - 1.320 - 1.328V.

Now I'm testing some other settings:
- Load Line Calibration at Level 4 (25%)
- Offset Voltage "-0.005"
- Additional Turbo Voltage set to minimum = +0.004V.

These settings gives some spikes to 1.344V at average load. At full load - 1.312-1.320V. (OCCT Linpack with AVX).

I wonder about load voltage (1.304V or even 1.296V) when OCCT Linpack without AVX is running. Why in this test load voltage is lower? Is Linpack without AVX less demanding than with AVX? Or maybe it's higher so Vdroop is higher? What do you think?

EDIT: Settings with Level 4 are stable - 7.5 H OCCT Linpack AVX passed.

Now I'm wondering, which profile is better to 24/7 use.
1st profile has lower spikes (1.336V) with average load.
2nd profile has higher spikes (1.336/1.344 sometimes) but lower full load Vcore - 1.312-1.320V.

 

[storm-chaser]

Im with John N on this one. You should benchmark and torture test specific to what your system will encounter in the real world. For some of us, that means as little as 20 minutes of AIDA64 torture test to see all is well and for others, that means endless rounds of synthetic torture tests and rigorous heat testing using Prime for days and weeks on end. Tech tip of the day: Feather touch

 

[BambOOs]

Update:

Both of my OC profiles are stable at load, but today, when I left my computer for awhile it crashed in idle state - black screen, no bsod, no restart, just black screen. At this profile, offset voltage was at -0.030 so idle voltage was lower than stock.

I increased offset to -0.025 for now and see will it help or not.

Is black screen crash normal when voltage is insufficient?
I though there will always be 0x124 BSOD.

EDIT: It wasn't idle voltage I think. With OC it is at 0.976V, without OC - 0.968V.

Which other settings may cause this black screen? Until now it only happened once, but..