The Pentium G4400 was simple to overclock, and i managed to reach a stable 4.7ghz on the stock intel cooler under prime95 and linpack load. Everything was amazing, even overclocked my ram to 3200mhz CL15-15-15-30-CR1. Everything ran fine. After that test, i decided to test a Core i5 6600T, since they were cheap, and I wanted to see what the voltage tolerances of the T series processors would be like. Needless to say, i was very impressed. The CPU made it to 4.5ghz without breaking a sweat, at 166.7bclk. Temps were fine, and managed to pass my previous stability tests. However, I noticed one very significant difference between the Pentium G4400 and Core i5 6600T. 

When overclocking the Core i5 6600T, it's L1 cache speed drops by over 75%. Thinking this was a random fluke, i retested my G4400 at stock and overclocked speeds. My pentium saw a performance boost across the board when overclocking, as its cache speed actually increased at 4.6ghz compared to its 3.3ghz stock, but the Core i5 6600T went from 820GB/s, to less than 240GB/s. I have tried absolutely everything I could think of to remedy this, and I am completely out of idea's. Is it the board forcing the CPU Cache to some sort of default low cache multiplier that I cannot touch? Or is it the nature of the 6600T to lose cache speed upon overclocking the core? If anyone else has had this problem, I would appreciate any insight you could offer.

Screenshots of my G4400 and 6600T below.

G4400 Before:

G4400 After:

Core i5 6600T Before:

Core i5 6600T After:

Is it a problem with the board itself? Or is it a bios issue that ASRock can solve with an update? I would appreciate any comment that can help figure this out, as I am completely out of ideas.

Sure thing. 

Stock: (Turbo Boost 3.5ghz, not overclocked)

Overclocked: (50%, 150 BCLK)

Thank you for taking the time to help me out with this. 

EDIT: More validation of cache speed, from XTU:

Ignore the 100C, after overclocking these locked Skylake CPU's, temperature reporting is not available. Package temps are still at 35C idle, under 70C at load.

The options to touch cache multi in XTU are grayed out. I can only touch the voltage, and even 1.4v has not helped. 

edit: In fact now that I think about it I remember reading that benchmarks were not reflecting the benefit of the overclock on certain CPUs with BCLK overclocking and the posts were not exclusive to ASRock boards either. It may be an inherent flaw with the overclocking method. There still may be hope for a fix though. I won't be able to confirm until I have a SKY OC enabled board to test with my 6600k.

Let us know what happens with the incremental increase test I suggested above.

http://postimg.org/image/5tbh4uyth/" rel="nofollow">

for comparison - skyoc bclock 177.9  - 

http://postimg.org/image/fqcstb6rh/" rel="nofollow">

Option 1: It's Aida64 being unable to properly read the cache after BCLK is adjusted.

Option 2: Something funky is going on in the bios, and touching BCLK too much changes cache speeds.

Option 3: Built in safety feature on the CPU's themselves, that is not present on Pentiums. 

I have someone with an MSI board that will be giving me results tonight with his Core i5 6400. If MSI boards have the exact same issue, then option 3 is probably the result. 

Mine was reading it at exactly the same speed as my core clock. HWinfo64 also showed the cache multiplier at x27, same as my core multiplier. So the programs see the cache speed as what it is supposed to be, but the aida64 bandwidth test claims it is going much slower than what it is supposed to be. Almost 75% slower in my case. 

There is still a possibility that the problem is with Aida. I will reach out with them and see if they can answer any questions. 

My 6600T goes from 1600 single thread, to 2000 single thread. It then goes from 6500 multi thread, to 8000 multi thread. This is from a 50% BCLK adjustment. Though to be fair, CPU was boosting to 3.5 on single thread test, and 3.3 on multi thread, so the difference between the two were only 500mhz and 700mhz respectively. 

It will take me some time to swap back to my pentium and test, as this is an ITX case and a little difficult to work in. I'll try to get the results for you though.

Thanks for all the time you have put into this, I will PM tech support and point them to this thread 

http://postimg.org/image/vnybd74gr/" rel="nofollow">

Enabling Sky-OC lowers the cache multiplier to allow more overclocking headroom. This is by design. If this were not the case your cache multiplier would be locked at the maximum turbo multiplier of the CPU, not an issue on pentium and celeron class CPUs where this is 1:1 but on all CPUs with turbo boost it can result in obscenely high cache frequencies.

For example:

i5 6400

Stock core multiplier: 27

Turbo core multiplier: 33

Cache multiplier: 33

Without Sky-OC enabled (feature no longer supported) the cache multiplier remains at 33 while the core multi is locked at 27 so the result:

BCLK: 150 

Core: 150 x 27 =  4.05ghz

Cache: 150 x 33 = 4.95ghz 

And the difference between core and cache frequencies grows even further the higher you set your BCLK. This means that your cache will hold back your OC headroom long before you reach the maximum OC your core can handle. With Sky-OC enabled however the cache multiplier is lowered to a point where it will not hold back your overclocking potential. The problem is that it does lower your cache performance quite significantly until reach or surpass the stock Cache frequency by raising your BCLK.

Sky-OC has been discontinued by ASRock and is no longer supported so use at your own risk if your board still has the feature.

Hi. Just wanna chime in after I update my Z170 Gaming K6 (paired with i5 6500 cpu) to the latest UEFI P7.40 (for the sake of meltdown/spectre patch). The SkyOC feature is still there but the cache multiplier preset itself to the highest value. This is crazy.

Because of this, i'm now unable to overclock. Sigh. RIP SkyOC.

Before (P7.20):

Now (P7.40) - notice the high cache clock: