5GHz - well, now there's something you don't see every day...

[cekim]

So, I guess sometimes trying to keep the uncore frequency up with the core really holds you back... Was trying to tune up a 5930k's uncore to 4GHz and stepped back to a default, but hit the 4.6G turbo optimized default by accident.
Didn't break 45C package temp so, I thought... "hmmm, how far does this go?"

Answer:
more /proc/cpuinfo | grep -i mhz
cpu MHz : 4999.941
cpu MHz : 4999.941
cpu MHz : 4999.941
cpu MHz : 4999.941
cpu MHz : 4999.941
cpu MHz : 4999.941
cpu MHz : 4999.941
cpu MHz : 4999.941
cpu MHz : 4999.941
cpu MHz : 4999.941
cpu MHz : 4999.941
cpu MHz : 4999.941

5G/3.5G/2666 at ~53-56C package temp (cores range from 48-56) after 10m of stressapptest (which has now been run several times)

DIdin't really beleive it, so I wrote a quick for-loop to compare that does 5B pairs of a simple integer math in a single thread:
5960x @ 4.4 - 19.9s
5930k @ 5 - 17.8s
compute delta = 11.78% faster time
clock delta = 12% faster clock

Yep, it really went there 1.36V core (er, it's my throw-away non-asus x99 MB, but we won't talk about that... 😉 ) H110i with extra case fans blazing - it isn't quiet.

I set a max junction temp of 85C in the bios too... 5.2 and 5.1 booted, but eventually crashed in stressapp:
2016/02/20-00:16:09(EST) Stats: Found 0 hardware incidents
2016/02/20-00:16:09(EST) Stats: Completed: 19696728.00M in 600.05s 32825.32MB/s, with 0 hardware incidents, 0 errors
2016/02/20-00:16:09(EST) Stats: Memory Copy: 19696728.00M at 32827.12MB/s

[Chino]

That's a very nice CPU. 🙂

[cekim]

indeed!

[BarneyC]

Nice

[Arne_Saknussemm]

Nice CPU cekim! :cool:

https://rog.asus.com/forum/showthread.php?24372-5Ghz-Club/page7

[Nate152]

5.0GHz at 1.36v?

Yep we should ring the bell we have another cpu lottery winner. 😉

I don't see very many Haswell-E's hitting 5.0GHz, you and Arne are one of the few, congratulations you have her screaming along.

[Praz]

Hello

Google's stressapptest is for memory only. It does not test any type of CPU stability.

[Nate152]

That's true Praz I think cekim's cpu will need a little more voltage for a cpu stress test.

You can give Realbench a try it's a good cpu stress testing program plus it's ROG. 🙂

Run only the stress test first, the benchmark will require a little more voltage.

http://rog.asus.com/rog-pro/realbench-v2-leaderboard/

[cekim]

Nate152 wrote:
That's true Praz I think cekim's cpu will need a little more voltage for a cpu stress test.

You can give Realbench a try it's a good cpu stress testing program plus it's ROG. 🙂

Run only the stress test first, the benchmark will require a little more voltage.

http://rog.asus.com/rog-pro/realbench-v2-leaderboard/

Agreed, Nate and Praz, my daily driving adds another 5-10C of temp on my various Haswell's under full thread/core load beyond stresapp or even RealBench.

So, stressapp is just a sanity check for me. 1.35 got all the way through stressapp and crashed on shutdown, so its definitely right at the line there.

I don't have windows on this machine yet to do testing over there - or even my own stuff to do my linux stress. Will get to that.

This thing just refuses to budge on cache clock though - 3.5 or 3.5 are my options. Which is unfortunate because I went back and checked with my good 5960x:

a. 3.5 cache does indeed make core OC easier
b. but 4.75/3.5 performs verily similarly (slightly less) in my real-world loads to 4.4/4.2 on the same chip. (now 4.75 and 4.125 is just super 😉 but runs at 71C pacakge and fixed only, not adaptive)

[cekim]

Interesting... Moved the 5930k to an R5E and now my cache is back up to 4G without complaint so far. I ran some quick RealBench and surprisingly (because I've not seen stability in linux but crashing in RealBench before), I couldn't get past 4.8GHz stable (still the old MB, not the R5E), but admittedly, I tried half-heartedly between Walking Dead commercials.

I guess Rick and I will have to try again next Sunday.