Thoughts on disabling cores/hyperthreading/turbo boost on 5930K for better overclock

Toggle sidebar Toggle sidebar
C

Cidona

Distinguished
Apr 17, 2015
40
2
18,535
Dear forum.

I'm narrowing in on making a purchase for a system with 5930K processor. I'm mainly figuring on the 5930K rather than a 4970K as I want access to lots of memory (work software).

While the programs I'm using love lots of memory, for the most part they are single threaded operations that I'm doing. I'm wondering if disabling 2 of the 6 cores on a 5930K would allow for a better/higher/more stable overclock for the remaining 4 cores? Also, would the full 15MB L3 cache be available to the remaining active 4 cores?

Any other thoughts/concerns/ideas?

Similar thinking with the hyperthreading.

With the turbo boost - would this be irrelevant/switched on or off when overclocking?

Thank you for any thoughts/experience in this regards!!
 
Sort by date Sort by votes
MarkW

MarkW

Judicious
Dec 7, 2009
6,308
0
32,960
When you buy a car, do you take out 2 of the spark plugs?

Any core that is idle is not adding to the heat generation.

Any thread that is not a thread, again, idle, is not adding to the heat generation.

I never, ever recommend disabling cores.
 
Upvote 0 Downvote
C

Cidona

Distinguished
Apr 17, 2015
40
2
18,535
lol. Well, while I wouldn't take out 2 of the spark plugs, there would be cases where I might make a purchase and not use all the parts/features in the purchases but where it was still worth getting the package for the features that I was using.

It was with the thought process that you mentioned (i.e. if a core is switched off/idle it is not adding the the heat generation) that would seem to allow for a higher overclock on the remaining 4 cores.

I am anticipating having the build performed by a professional builder (but want to be as informed as possible for discussions, etc. with them). My thought/understanding (lacking) of when overclocks are done is that they are typically tested, etc. for stability with various demanding benchmark programs, etc. So it would seem that they were to set it based on the heat from the 4 cores rather than the 6 (or 12 if including hyperthreadeing) it would allow higher clock speeds.

Do you think it would create problems for the memory controller or such? What issues would you think might come into play?

Thanks again for the reply.
 
Upvote 0 Downvote
MarkW

MarkW

Judicious
Dec 7, 2009
6,308
0
32,960
No. The memory controller is separate from the CPU's themselves.

What I was getting to in my earlier post is that an idle core is now put into an almost completely off status. This is what is called being put into a C6 state. Voltage actually drops to 0 volts. Deep sleep. Off for all intents and purposes.

Now the question I have is this. Are you certain that your business software is single threaded and not capable of running multiple threads?
 
Upvote 0 Downvote
C

Cidona

Distinguished
Apr 17, 2015
40
2
18,535
Thanks for the link Mark. I'm in the middle of something at work at the moment but will check it out asap.

In regards to your query as to whether I'm certain that the software is single threaded... There are a few different programs that I regularly use. Nearly all of them have a few aspects that I use that do make some use of multiple cores, but those features are few.

When I'm working I generally have a desktop gadget 'Top Process Monitor' which shows the top 5 CPU processes and the % of CPU use by each. There are times where I'm waiting up to 20 minutes for an import or such and where that Process Monitor shows the application at a constant 12.4% to 12.6%.

I rarely see the CPU go above 25% ( I have a Quad Core with Hyperthreading so shows as 8 cores); and even then the 25% is usually total of a couple of applications running at the same time.

I would want the ability to utilize the 4 cores to the max (when software is able to); however the way I'm looking at it is if 90% of my work is single threaded and if I could get and extra 15% (for example) single threaded performance (using 4 cores rather than 6); that it would be a 'saving' in time of 13.5%. That would be a bigger percentage time saving that even if the multi core was twice (or any speed faster). With the multiple threaded features only accounting for 10% of my work, no matter how many cores I threw at it, it couldn't save me 13.5% :)

Again, your thoughts are greatly appreciated!

 
Upvote 0 Downvote
C

Cidona

Distinguished
Apr 17, 2015
40
2
18,535
Upvote 0 Downvote
MarkW

MarkW

Judicious
Dec 7, 2009
6,308
0
32,960
You want those cores to wait up when there is work to do. That is the entire point of all of this. When there is something that needs to be done, it needs a place to do it or everything will be forced to work by essentially time sharing the core(s) that are on. You do not need to manage any of this. The I7-5930K knows how to do C6/C7 by default. If one setting in the BIOS is enabled to allow that to happen, it will do what you want on power savings, but still have all cores available if needed. Within thousandths of a second after a core is done working, it will be at 0 volts used. And it will stay there until more work needs to be done.

I guess the other point that needs to be made is that if you overclock your CPU by 20%, you are not likely to see a 20% increase in performance. The I7-5930K stock speed is 3.5Ghz. I found a page where during a review of the CPU, they overclocked it. They did not disable cores or anything like that, and they still got the CPU to 4.55Ghz. That is probably beyond the limits of most overclocks that can be done. Be aware, this is like the final page of a long review. Just trying to skip to what is relevant.

http://www.bit-tech.net/hardware/2014/09/03/intel-core-i7-5930k-and-core-i7-5820k-revie/9
 
Upvote 0 Downvote
C

Cidona

Distinguished
Apr 17, 2015
40
2
18,535
Thanks Mark, but the reason I don't see the C6/C7 as being a solution for what I'm looking for is, if one was to enable C6/C7 in the BIOS and then proceed to overclock the CPU, as part of the testing, etc. to find a stable clock there would be stress tests on all cores; at which point those cores in low power state are going to 'wake up' and if there are 12 thread (6 physical cores + hyperthreaded), then the overclock is going to be limited by the heat dissipation from the 12 threads. Since this is hardly ever a real life situation that I can use much it doesn't seem to make sense to have that drive the overclock.

It would seem that one could get a higher clock just operating 4 cores (a) since you would be creating less heat than 6 cores and (b) since you could 'cherry pick' the best 4 cores of the 6 (see that article in my previous post).
Understood that I will not get the same performance from the 4 cores as I would with the 6 when in a multi-threaded application.

In looking at that article you linked it seems that there is a fairly linear pattern in quite a few of the benchmarks between the 3.5GHz/4.5GHz 5930K. Not perfect but certainly a substantial increase in most between the stock and overclock.
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom