Turbo OC: Single Core Stability Testing?

[eddiew]

This might be a totally rookie question, in which case I apologise...

My i5 has established a pretty nice overclock of 4.5 GHz on ~1.28 vcore, which I think is pretty good going. However, going beyond that I end up hitting a thermal ceiling long before what I suspect is the upper speed of the processor.

Been investigating the option of 1/2/3/4 core turbo modes as available in my UEFI bios, and set that to try for 4.7GHz with a single core load, or 4.5 with 2+. Just to test the theory. I might gun for 5.0 single core if it works

However, I can't for the life of me find a way of stability testing a single heavy thread. IBT and Prime, when running single threaded, have an awful tendency to core-hop. I get momentary jumps up to 4.7, but for the most part the multiplier stays at 4.5, probably because the load is getting distributed and the CPU thinks all four cores are in use

Is there any known application that will forcibly sit itself on a single core, max it, and NOT get shunted around by OS management?

Thanks!

 

[MichaelM]

I was thinking about this as well and would like to know how people do it but I think most just go for an OC on all cores.

Maybe if you disable 3 cores in the bios then boot up and see what you can do with a single core active on your chosen voltage, then do it with 2 and 3. But do cores differ in their ability to overclock so the turbo could pick a different core/s to the bios and you end up unstable anyway?

 

[Orcvader]

You can set the affinty in task manager to allow only 1 core to run the stressing software, just go to task manager, go to the process tab, find the one which is your stressing software (needs to be running), right click and set affinity.

 

[eddiew]

You can set the affinty in task manager to allow only 1 core to run the stressing software, just go to task manager, go to the process tab, find the one which is your stressing software (needs to be running), right click and set affinity.

/headdesk

Knew there would be something obvious, thank you!

Of course, that does mean running stability tests 4x... Not sure if it'll even be worth it for the ~8% boost...

Other obvious question is: if my single threaded stressing app tends to core hop, is there any reason a single threaded load such as a game wouldn't do so too? Since if it does, I'll STILL only see the 4x turbo speed

 

[Orcvader]

Games doesn't stress the CPU as much as IBT/Prime, so I'm guessing after that core reaches a certain temp, the CPU moves the workload to a cooler core. No idea what the real reason though...

 

[eddiew]

Quite possibly a futile exercise then Shame, it feels like something to blame on the Windows thread scheduler more than anything else!

 

[eddiew]

Lol, ok, well it turns out that mucking around with different speeds for 1/2/3/4 core turbo is a fantastic way to end up with a highly unstable system. And if you leave the voltage on auto, it spikes to some scary levels (booted me into windows at 1.42 which I feel much too high for comfort on this chip!)

Would have been nice to have a 4.5x4 or 5.0x2, but I think that achieving it is likely to shorten both the CPU's and my lifespans and take more time than will ever be recovered by that slight boost to single thread performance

Imma go enjoy my new chip now