Skylake cache overclocking

TheF34RChannel

New member
I have questions about cache overclocking, which I neglected in my other thread:

Currently it sits at 4.1GHz 24/7.

1. Is cache overclocking useful for gaming?
2. It is okay to have it at 4.1GHz 24/7 with SpeedStep enabled so the core clock is obviously lower in regular use?
 
1/ Cache Overclocking has very little real world benefit. So personally I don't see any point. Processor speed increase of 100Mhz has more effect that's a 1000Mhz increase on Cache.

2/Speedstep is absolutely fine to keep on with an overclock.
 
the only thing i found with speed step "atleast on a 2nd gen i5" is you need a bit more voltage for the overclock than you do without it, but i like to have it enabled as it uses less power when not under load, and obviously you spend less time pumping voltages through the cpu.
 
1/ Cache Overclocking has very little real world benefit. So personally I don't see any point. Processor speed increase of 100Mhz has more effect that's a 1000Mhz increase on Cache.

2/Speedstep is absolutely fine to keep on with an overclock.


Cache clocking opens up the taps for fast memory.

Speedstep with a hardcore overclock should be disabled along with anything else that alters the clock speed for stability and manpoints.
 
Thanks guys!

Cache clocking opens up the taps for fast memory.

Speedstep with a hardcore overclock should be disabled along with anything else that alters the clock speed for stability and manpoints.

Excellent, although I doubt I need a cache overclock with 3GHz RAM then? I asked in the op because somewhere I've heard that the cache needs to be equal to or lower than the core clock, and I'm a novice to cache overclocking.

I considered permanently disabling those but since I often just use word to write research papers for weeks on end I don't need an overclock and extra voltage. That's why I left them enabled. I thought it might add to the longevity of the CPU.

Manpoints?
 
Thanks guys!



Excellent, although I doubt I need a cache overclock with 3GHz RAM then? I asked in the op because somewhere I've heard that the cache needs to be equal to or lower than the core clock, and I'm a novice to cache overclocking.

I considered permanently disabling those but since I often just use word to write research papers for weeks on end I don't need an overclock and extra voltage. That's why I left them enabled. I thought it might add to the longevity of the CPU.

Manpoints?
Manpoints=Benchmark Scores
 
Ah and so I've learned something new ha ha! I don't benchmark though.

So the cache needing to be equal to or lower than the core clock is a myth.
 
I don't recall which by head but I've set voltage manually. Or is there a specific cache voltage? Ring bus voltage I believe? I can check once home then.

Which brings me back to my needing a cache oc in the first place with 3GHz RAM?
 
Last edited:
I don't recall which by head but I've set voltage manually. Or is there a specific cache voltage? Ring bus voltage I believe? I can check once home then.

Which brings me back to my needing a cache oc in the first place with 3GHz RAM?

Cache is way, way, way, WAY faster than even the fastest RAM! DDR4 speeds are in the region of 30gb per second. That sounds fast but Skylake L3 cache even as standard is in the region of 300gb/sec. So about 10 times quicker :eek::rolleyes::o

Compare the latency numbers too...

http://cdn.wccftech.com/wp-content/uploads/2015/08/Intel-Core-i7-6700K_Stock_AIDA64.jpg

Cache is also functionally different, i.e., parts of the CPU 'look ahead' and try to predict what data will be needed by execution resources next, then load that data from memory while the business ends of the CPU are busy. If the guess is right, a 'page fault' is avoided and the CPU doesn't have to twiddle its thumbs waiting for data to be fetched from memory...

Good/bad cache design will make/break a CPU. cf., Bulldozer. A lot of die area on a CPU is left to the last level cache because of how crucial it is to good performance. And CPU designers put vast amounts of R&D into good cache design.

Having said all that - it might well be that OCing the cache doesn't make much difference to the performance - and that will be because it's already screaming fast (and very good at guessing too)
 
Last edited:
Yep. Other sites have done comparisons of Skylake cache speed and the largest performance difference was something like 1.2% So yeah, CPU core speed makes much more difference than cache speed.
 
Still, having your cache equal or close to your core clock should result in a more balanced OC, I understand. For example 4.5+ on the core and stock 3.9 on the cache would be a bit more unbalanced than 4.1+ on the cache, if that makes sense?
 
Back
Top