6900K overclock? 4.5Ghz.

[Nicklas0912]

Hello.

I would like to ask what other with 6900K is getting.

mine will 4.5Ghz @1.41 volt.
125x36
Cache is left at Auto, no idea how high it go, but the volt for it is --> 1.18.
It get around the same CPU Score as the 5960x@ at 5.1Ghz!
is the chance good to get a better chip if I get a other one? or should I just be happy 😉
is with Custom Water cooling setup.

[cekim]

I am seeing 4.5 stable at 1.35 on a 6950x. Other than heat, it is reasonable to expect that the 6900 should be similar, if not slightly better. I can boot/run at 4.6/4.7 @ 1.37, but it will break eventually. I have not tried higher voltages yet.

Your result looks typical to good so far with very limited data. It seems some reviewers were not even able to hit 4.5 at all. So, your mileage may vary on looking the horse in the mouth.

[Nicklas0912]

#2 Yea, I see too that the 6950x is overclokcing similar of even better than every other 6900K, that just how it is I thnk !:)

of the 4.5Ghz at 6900K I get the power of a 5960x at 5/5.1Ghz wich is not that bad, I mean they really impoved performance per Ghz.
and I think I wont find a 5.1Ghz 5960X chip if I get that instead.

[cekim]

Nicklas0912 wrote:
#2 Yea, I see too that the 6950x is overclokcing similar of even better than every other 6900K, that just how it is I thnk !:)

of the 4.5Ghz at 6900K I get the power of a 5960x at 5/5.1Ghz wich is not that bad, I mean they really impoved performance per Ghz.
and I think I wont find a 5.1Ghz 5960X chip if I get that instead.

It depends pretty drastically on what you do AND on exactly how the two systems are configured.

I can produce a 15-18% difference on my 5960x by doing nothing other than changing memory timing. Similar story with cache.

You can see this same story playing out in the RealBench numbers. There are 5930ks that outscore 5960xs because of living on opposite ends of the "ideally configured" spectrum.

The 6900x is a much easier sell. You get better speedstep, better idle a bump in IPC. Once/if they (Asus) sort out the IMC a little better, I think it would be a good choice if you don't already have a 5960x. The memory controller is touchy right now. A step back in ease of setup with OC'd memory.

[Nicklas0912]

cekim wrote:
It depends pretty drastically on what you do AND on exactly how the two systems are configured.

I can produce a 15-18% difference on my 5960x by doing nothing other than changing memory timing. Similar story with cache.

You can see this same story playing out in the RealBench numbers. There are 5930ks that outscore 5960xs because of living on opposite ends of the "ideally configured" spectrum.

The 6900x is a much easier sell. You get better speedstep, better idle a bump in IPC. Once/if they (Asus) sort out the IMC a little better, I think it would be a good choice if you don't already have a 5960x. The memory controller is touchy right now. A step back in ease of setup with OC'd memory.

Yea, I see that 🙂 just every one I find get the same score.

But you think is a bad overclock/scaling? I see that CACHE was on max 24.
or I will be more lucky with a new 6900K :)? can easy get a new one.

[Nicklas0912]

do you gain much more performance with more Cache? like x26 instaed of 24.

[cekim]

Nicklas0912 wrote:
do you gain much more performance with more Cache? like x26 instaed of 24.

Yes, but again it depends on application. So far I've not gotten the cache beyond 3.8 and I've left it at 3.5 to keep voltage reasonable.

The difference I measured in various applications between 3.0GHz and 3.5GHz was about 6-8% (changing nothing else). Not earth shattering, but my first objective was to get the 6950x to provide per core performance equivalent to the 5960x, which took some tuning...

I assume you mean x26 with 125BCLK? or 3250MHz? If your CPU is anything like mine, you should be able to do 125x28 or 3.5GHz, but mine required either fixed 1.250 or offset + 0.250 so far.

I need to experiment more to see if I can bring that down (though 1.3v is pretty typical for the 14nm process it seems).

[Nicklas0912]

cekim wrote:
Yes, but again it depends on application. So far I've not gotten the cache beyond 3.8 and I've left it at 3.5 to keep voltage reasonable.

The difference I measured in various applications between 3.0GHz and 3.5GHz was about 6-8% (changing nothing else). Not earth shattering, but my first objective was to get the 6950x to provide per core performance equivalent to the 5960x, which took some tuning...

I assume you mean x26 with 125BCLK? or 3250MHz? If your CPU is anything like mine, you should be able to do 125x28 or 3.5GHz, but mine required either fixed 1.250 or offset + 0.250 so far.

I need to experiment more to see if I can bring that down (though 1.3v is pretty typical for the 14nm process it seems).

125x24 Max was it last time I checked, but can try rasied it and do some 3dmark to see 🙂

[aktivexkontrol]

Hi Nick,

Have you tried running it with AVX and AVX2 intructions?

Please try Prime95 27.9 or 28.7 versions

[kcmcc]

I'm stable at 4.4
keeping voltage at 1.25
bclk keeping at 100

Trident Z 3200 mhz with working xmp

corsair AIO 115i

when i jump to anything above 1.25 to reach 4.5 its very unstable.
cache is at 32 min 33 max

Am I forced to up the bclk? I was hoping not to since bclk increases everything not just the core clock, correct?