OverClocking CPU Non-K Skylake processor in 2020: Reported Stock Core Clock

[althepal01]

So, I just overclocked my cpu(non-k Skylake in 2020) to 4.25GHz and ram to 3067mHz and when I run 3DMark, it is reporting my stock core clock at 3.2 and maximum turbo core clock at 4.25. Why is it not reporting my stock core clock at 3.7? Cinebench R20 does something similar. Cinebench reports my CPU with this title, "2C/4T @ 3.2 GHz, Intel Core i3-6100 CPU" in the Rankings Table. Before the overclock my Cinebench was reporting 3.7, and surprisingly my score was higher before I overclocked my cpu. That was not the case for 3DMark, as my score greatly increased on 3DMark after the overclock. I would appreciate someone's expertise on this :))))))))))

Operating System: Windows 10 build 17763 (64-bit)
CPU Type: Intel Core i3-6100 @ 3.70GHz
Number of CPUs: 1
Cores per CPU: 2
Hyperthreading: Enabled
Motherboard: Z170 Pro4S
Memory: 8GB Corsair CMK8GX4M1A2666C16
Videocard: NVIDIA GeForce GTX 1060 3GB
Hard Drive: ADATA SP550 (120GB)
Hard Drive: ST1000DM003-1SB102 (1TB)

[porina]

Software can get confused by overclocks. Use either something like CPU-Z or hwinfo64 to look at the clocks while the bench is running. If it is running correctly, don't worry that some software doesn't understand it.

 

Edit: which Cinebench? R15 should show an increase. R20 might show a decrease, as the non-K OK breaks the AVX performance which R20 uses.

Edit 2: I just saw R20 mentioned. It's probably the AVX hit on non-K OC.

[althepal01]

24 minutes ago, porina said:

Software can get confused by overclocks. Use either something like CPU-Z or hwinfo64 to look at the clocks while the bench is running. If it is running correctly, don't worry that some software doesn't understand it.

 

Edit: which Cinebench? R15 should show an increase. R20 might show a decrease, as the non-K OK breaks the AVX performance which R20 uses.

Edit 2: I just saw R20 mentioned. It's probably the AVX hit on non-K OC.

I am using Cinebench R20. I don't know what AVX means but you're probably right about that. I will go ahead and test it with Cinebench R15. Here is a screenshot in the meantime of CPUz and Hwinfo64. 

 

Thanks for the quick response, I was expecting not to get an answer for weeks. First time using these types of forums.

 

[porina]

8 minutes ago, althepal01 said:

I am using Cinebench R20. I don't know what AVX means but you're probably right about that. I will go ahead and test it with Cinebench R15. Here is a screenshot in the meantime of CPUz and Hwinfo64. 

That seems to show you running at 4.25 GHz.

 

AVX is a set of instructions with dedicated hardware support. Unfortunately one of the side effects of the tricks used to do a non-K overclock is that the full AVX performance is not available. For software that makes heavy use of AVX, you might lose a lot of performance. Cinebench R15 didn't use it, but R20 does.

 

To put it in context, Intel since Haswell has good AVX performance, which AMD did not catch up with until Zen 2. Zen and Zen+ were about half the performance. I think the non-K OC penalty to performance would make it more comparable to Zen or Zen+ levels.

[althepal01]

1 hour ago, porina said:

That seems to show you running at 4.25 GHz.

 

AVX is a set of instructions with dedicated hardware support. Unfortunately one of the side effects of the tricks used to do a non-K overclock is that the full AVX performance is not available. For software that makes heavy use of AVX, you might lose a lot of performance. Cinebench R15 didn't use it, but R20 does.

 

To put it in context, Intel since Haswell has good AVX performance, which AMD did not catch up with until Zen 2. Zen and Zen+ were about half the performance. I think the non-K OC penalty to performance would make it more comparable to Zen or Zen+ levels.

I am getting the same 3.2 GHz recognition by Cinebench R15.

 

[porina]

14 minutes ago, althepal01 said:

I am getting the same 3.2 GHz recognition by Cinebench R15.

Don't worry about that. The important thing is are the scores higher with the OC?

[althepal01]

1 hour ago, porina said:

Don't worry about that. The important thing is are the scores higher with the OC?

I didn't test my CPU at 3.7GHz on CineBench R15 before I overclocked so I will need to restore CPU back to 3.7 and test it to tell if the scores are higher. My goal is to try to get my CPU to overclock to 4.4GHz if possible given that I have a basic cooler Hyper 212 EVO. I am trying to keep my voltage and temps as low as possible while keeping stable. I have tried to OC to 4.4GHz with voltage around 1.375-1.395 range and it booted into windows and I noticed my CPU temps was unstable fluctuating from 30 C to 61-63 C at idle and it eventually blue screened several times just sitting idle. Maybe 4.25 is the max I can achieve with this CPU, given my CPU cooling and how old my system is (around 3-4 years). My performance is being limited by having a single channel memory, which I should have thought about when I bought this PC but it was my first build.

[porina]

6 minutes ago, althepal01 said:

My goal is to try to get my CPU to overclock to 4.4GHz if possible given that I have a basic cooler Hyper 212 EVO.

It's been a while since I did a non-K overclock, also on a 6100 like yours. From memory I did get mine to around 4.7, which is typical for a Skylake CPU with air cooling. Looks like I was using a cooler even worse than a 212! https://hwbot.org/submission/3674669_mackerel_cpu_frequency_core_i3_6100_4737.92_mhz

From memory I had some trouble with the system after I got to that clock so I never really got around to benching it. I can't recall exactly, but I think the ram scales also with the bus speed, so remember to back off on the ram multipliers as you try to increase CPU speed.

 

I did find an old CB15 result of mine at 3.7-ish, scoring 413. https://hwbot.org/submission/4176103_mackerel_cinebench___r15_core_i3_6100_413_cb

 

Personally since I do care about AVX performance, I'd be inclined to run stock for general use. If the benefits outweigh the negatives, from memory 4.2 is easy at or close to stock voltages (up to 1.25v or so), but much above 4.2 you have to turn up the volts and the heat starts to build up faster also.

 

Dual channel ram is nice to have, but with only 2 cores to feed it probably makes less difference in this case than generally. If you can afford it, throwing in another 8GB stick of ram to allow dual channel might help a bit in some games.

[althepal01]

On 4/10/2020 at 12:01 PM, porina said:

From memory I did get mine to around 4.7, which is typical for a Skylake CPU with air cooling. Looks like I was using a cooler even worse than a 212! https://hwbot.org/submission/3674669_mackerel_cpu_frequency_core_i3_6100_4737.92_mhz

4.7 GHz on the i3-6100 at 1.2 V is insane! You must have won the silicon lottery. To even get your system to boot with those numbers is really unbelievable. The highest oc that I could get was 4.4 GHz and that was raising the core voltage just below 1.4 V. My chip literally hits it's limit at 4.4 GHz no matter how much I try to raise the voltage within the Intel specified safe range of 1.52 V, I could not get it to even boot at 4.5 GHz and above (I didn't go above 1.42 V for Core voltage). I settled for a 4.3 GHz final overclock, which allowed me to decrease my voltage to 1.36 V idle and 1.34 V at load. This is also with my FCLK(1.162 GHz), Cache(4.3 GHz), and RAM(2943 MHz as well as tightened timings 13-15-15-28-2T-270) oc'ed as well. I still can't believe that you got a 4.7 GHz oc at 1.2 V, unless CPU-z did not read your voltage correctly, that is quite out of the ordinary.

On 4/10/2020 at 12:01 PM, porina said:

From memory I had some trouble with the system after I got to that clock so I never really got around to benching it. I can't recall exactly, but I think the ram scales also with the bus speed, so remember to back off on the ram multipliers as you try to increase CPU speed.

Clearly, 4.7 GHz at 1.2V is not going to run stably let alone boot. And yes, BCLK overclock increases not only CPU speed but also FCLK, RAM, and Cache. I was not able to isolate FCLK frequency(only three settings in UEFI 400, 800, & 1000 MHz) multipliers and Cache multipliers. However, RAM I was able to decrease frequency as I was increasing BCLK. I later OC'ed RAM from its stock 2133 MHz and above its XMP Profile of 2666 MHz to 2964 MHz with shortened timings as stated above. I had to increase DRAM, VCCIO, & VCCSA ( 1.39 V, 1.24 V, & 1.25 V )  voltages to achieve overall stability as well.

On 4/10/2020 at 12:01 PM, porina said:

I did find an old CB15 result of mine at 3.7-ish, scoring 413. https://hwbot.org/submission/4176103_mackerel_cinebench___r15_core_i3_6100_413_cb

My stock 3.7 GHz CB15 score was 387 pts. A 26 point difference, let's say 6 points for the difference of 73 MHz. A 20 point difference, you definitely won the silicon lottery or maybe I just got the worst of the batch. I got a 460 score oc'ed and 182 single core. 

On 4/10/2020 at 12:01 PM, porina said:

Personally since I do care about AVX performance, I'd be inclined to run stock for general use. If the benefits outweigh the negatives, from memory 4.2 is easy at or close to stock voltages (up to 1.25v or so), but much above 4.2 you have to turn up the volts and the heat starts to build up faster also.

I'm quite the opposite, I don't really care about AVX performance. I only game, stream videos, code, and browse. At the same time, I admit that I don't know which applications actually utilize AVX and depend on that to the extent that if I disable AVX it won't run efficiently or very slow.

 

As for the temperature problem, i3-6100 definitely gets really hot once you increase the voltage which is inevitable for higher oc's. In fact, I could not sustain a 4.2 GHz oc without hitting max temperatures of 63-66 degrees at idle on HWiNFO and keep in mind that HWinfo doesn't report actual core temps and the reported package temperatures were probably giving me a margin of error of 5-10 degrees Celsius (according to a yt video made by Der8auer on OC'ing the i3-6100). So, I didn't dare try to even stress test 4.2 GHz knowing that the temperatures would sky rocket to above 80 degrees on HWiNFO.

 

Luckily there was a solution for that, DELIDDING!!! So I went out and spent 14 dollars on the good ole liquid metal thermal compound conductonaut. Along with a simple performance razor blade for less than a dollar. Now, the safest option would be to go purchase that 40 dollar delidding tool, however, it's not worth spending 40 dollars for a kit I'm only going to probably use once. Anyways the razor blade did the job, no scratches and the process went by very smooth, not too bad for the first time delidding. I actually enjoyed delidding a lot more than the actual OC, it was fun! Before and after photos: 

After delidding, my temps at idle under stock 3.7 GHz didn't go down much just a mere 1-2 degrees at idle however under load/stress tests max temps went down 10 degrees at 3.7 GHz. Now OC'ed to 4.3-4.4GHz at 1.36 - 1.40 V under stress tests like RealBench (I have found this test to the most stressful, heats up the system the most, and also throws blue screens much quicker if your oc is not stable), Prime95, & IntelBurnTest the system held up a whopping max temperature of 60 degrees. Keep in mind all these temps were recorded on HWinfo, however I do recommend using a digital multi-meter as HWiNFO is not reliable when reading voltages sometimes underestimating and sometimes overestimating voltages. While Temperature you can always assume a 10 degree margin of error for CPU core temps. That was impressive, I probably achieved a 20 or more degree temperature drop as a result of delidding, given that at idle 4.2 GHz I was experiencing max temps at 66 degrees.

 

[porina]

3 minutes ago, althepal01 said:

4.7 GHz on the i3-6100 at 1.2 V is insane! You must have won the silicon lottery. To even get your system to boot with those numbers is really unbelievable. 

Don't trust the CPU-Z report. I was using a lot more than 1.2V, but I don't know what that number would be. I tend to go up to 1.4V as my general benching limit, but may go higher for short amounts of time.