[OC] Per Core Multipliers

[Mushious]

Hallo,

 

Is it alright to OC using different multipliers per core or should they really all be the same? What's the advantage of keeping them all the same if so?

[Yeater]

Hallo,

 

Is it alright to OC using different multipliers per core or should they really all be the same? What's the advantage of keeping them all the same if so?

it comes down on what application you use

[Suffokation]

They should all be same.

It'd be tricky getting the voltages correct.

Why would you need seperate core clocks anyway?

[Mushious]

it comes down on what application you use

 

If I'm stable with decent temps with 47x on core 1 and 2 and 46x on 2 and 3, does it really matter?

 

 

They should all be same.

It'd be tricky getting the voltages correct.

Why would you need seperate core clocks anyway?

 

Was playing around with the auto OC thing Asus does, notice it prefers split multipliers and seems to run happy.

[QueenDemetria]

Hallo,

 

Is it alright to OC using different multipliers per core or should they really all be the same? What's the advantage of keeping them all the same if so?

I experimented with this a while ago. I didn't have any problems with setting different multipliers on cores, but I found it to be pointless since the faster cores were being held back by the slower cores. I imagine if you could force programs to use certain cores(like setting the Affinity in Windows) you could make it work well, but it would be a lot of work to have only slight gains. If you just want to mess with it I'd say go for it since tinkering is always fun, but I wouldn't do it if you are expecting huge performance gains.

[Mushious]

I experimented with this a while ago. I didn't have any problems with setting different multipliers on cores, but I found it to be pointless since the faster cores were being held back by the slower cores. I imagine if you could force programs to use certain cores(like setting the Affinity in Windows) you could make it work well, but it would be a lot of work to have only slight gains. If you just want to mess with it I'd say go for it since tinkering is always fun, but I wouldn't do it if you are expecting huge performance gains.

 

Was wondering about the per core performance, if the faster ones get held up by the slower cores, just seems like a waste, voltage increase for nothing. But as you say, tinkering is fun :3

[Django]

For SNB I used stepped multi's of 46,47,48,50 because of temperature limits with 4 cores at 4.6GHz meaning if running all 4 cores loaded it would run at 4.6GHz at 1.36V IIRC, any higher would result in higher temps and possible thermal throttling.When using less cores it could run faster with a single threaded app running 5.0GHz with the voltage increasing as necessary. Cores that were not being used would sit in C6 and be power gated off.

 

For the IVB system it hit a frequency wall at 4.7GHz but temps were okay so ran it with flat 47 multi's, 47,47,47,47.

 

In a stepped multi system with extra turbo voltage or adaptive mode the extra voltage only appears when running less than 4 or 3 cores for quads, when running all 4 cores the voltage is as would be for running flat multi's.

 

Whether the 10% or less increase in performance using app's which use less than 4 cores is important to you, only you can answer that.

[DildorTheDecent]

By default Intel has 'split multipliers' so that advertised turbo speed is for two cores only. My theory behind it is as more cores become active for a workload you can drop 200MHz and it will perform similarly.

 

I prefer to have my cores synced as this gives better scores for benchmarking. Highest i've ever ran is 49/49/49/49. The stock config is 38/38/37/36. I keep it at 46/46/46/46 for daily usage.

 

Wouldn't recommend auto OC. It is fine for doing some short term testing. Auto OC is generally not recommended since it applies far too much voltage for a final OC to be used daily. 

[Mushious]

By default Intel has 'split multipliers' so that advertised turbo speed is for two cores only. My theory behind it is as more cores become active for a workload you can drop 200MHz and it will perform similarly.

 

I prefer to have my cores synced as this gives better scores for benchmarking. Highest i've ever ran is 49/49/49/49. The stock config is 38/38/37/36. I keep it at 46/46/46/46 for daily usage.

 

Wouldn't recommend auto OC. It is fine for doing some short term testing. Auto OC is generally not recommended since it applies far too much voltage for a final OC to be used daily. 

 

Auto OC was pushing 1.275 for 46x across all cores, got it down to 1.245, still feels really high.

[DildorTheDecent]

Auto OC was pushing 1.275 for 46x across all cores, got it down to 1.245, still feels really high.

That's not bad. 1.245 for 4.6GHz is pretty good. If you want get it lower or go higher. http://linustechtips.com/main/topic/41234-intel-haswell-4670k-4770k-overclocking-guide/

 

That guide will help you. It applies to your CPU. Don't worry