Silicon Lottery: Roughly 20% of Intel Core i7-10700K achieve a stable 5.1GHz 24/7

Wonder how many achieve 5.0GHz and 4.9GHz respectively. Most people don't know and the rest don't care about 100MHz but it's nice that enthusiasts have a choice.

 

68% and 100%

 

20% hitting 5.1 is pretty good considering how high that is, I was hitting 4.5 on my sandy and ivy bridge CPU's when they where relevant and that was amazing then.

 

I wonder how these numbers compare to the 9900K and 9900KF. Would be interesting to see such metrics, to check how much comet has/hasn't improved compared to last gen, since those overclocked pretty well already.

Maybe rocket lake will make use of this headroom and will implement something equivalent to AMD's PBO? That would be a nice idea, though I guess PL2 already kinda does that.

 

i would love to find a detailed changeset from gen 9 to 10, but i can't find it.
Once you would read that l1 cache doubled, moved from 8 way associative to 16 way associative and things like that. I have no idea what is changed now.

 

Thankfully my 10700K hits 5.1GHz at 1.295v. Good temps with 74 degrees C on the hottest core during Cinebench..though using a 360 AIO in the Arctic Freezer II...

 

Just tried to push it to 5.2GHz at 1.3v...booted into Windows but froze during the Cinebench R20 run...It also pulling more than 1.3v in HWinfo at 1.327v. I had LLC at Turbo....Do not want to push more vcore as I am more than happy at 5.1GHz all core stable...

 

R20 uses AVX which raises voltage on intel processors. Unless you use adaptive voltage(and offset the addition of core voltage) or lower the avx multiplier in the UEFI, you will have a harder time keeping stable versus non-avx workloads.

Also, you want to use the LLC setting that causes the least amount of core voltage fluctuation from idle to full load. Highest setting is not always the best and can sometimes produce more heat/use more energy than necessary.

 

Llc back in haswell days didnt affect cpuv but cpu input voltage, I think it hasn't changed since.

 

Consistent with other CPUs since Coffee Lake with minor variation.
https://siliconlottery.com/pages/statistics

BTW, CB is pretty bad as a stability test, in fact it's as bad as it gets. I'd say the bare minimum is SmallFFTs without AVX in Prime95. Obviously, AVX stability testing must be impossible for Intel's 10 cores.

Pseudo-stability in CB can be achieved with low voltages. It won't be actually stable. You'll still get crashes and maybe BSODs. I worked with more than a dozen Coffee and Cannot Lake CPUs and none was fully stable until it went through 10-12 hrs of AVX testing in prime95, and yes, it will affect normal usage.

 

Do always find it funny, that they keep trying to push the GHZ when they reached this cap back in sandy bridge days, my 2700k got to 5ghz, Knew some people who could push a little more than that but it was rare. But they just pushing the thermal limits of the design, either need a new one... or as I've been saying for years just more cores and get devlopers to use them effectively

 

4.5 was bare minimum on those CPU's. I have 2 3770K's, one hits 4.7 and one 4.9

 

Maybe i lost the lottery i even delidded the 3770k but temps easily hit between 70 and 80c at 4.5ghz with a 240mm aio that's as far as i was comfortable with.

 

Yeah something off there. I ran a delidded one in a custom loop at 4.9/1.416v for a few years and temps stayed in the 60's in IBT. The 4.7 was cooled with an NH-D14 running at 1.44v and even it barely got into the 70's.

 

Been a while since i had that system now but i remember being around the same voltage somewhere between 1.4 and 1.45 oh well i always find i get shafted compared to hilbert here and others in forums even my 2 ryzen+ CPU's and flare-x kits refuse to go any higher than 3266mhz ram speed.

 

haswell/broad uses fivr (fully integrated voltage regulation), with skylake desktop they went back to the old voltage regulation, presumably for thermal reasons. intel server chips still used fivr as far as i know

 

My 9900K won't even hit 5Ghz stable with an H100i cooler 24/7. Kinda sad.

 

I must have missed where SL provides what test(s) is/are considered 'a rigorous stress test routine'.

Seriously, provide verified empirical data.