On overclocking an i9 13900K

[OxSteve]

Hello Techpowerup people,
Please, check my config first.

I tried to overclock my 13900K, and the result are not very good.
@stock, it runs P core(5.5Ghz) and E Core(4.3GHz) I figured how to have OCCT, Cinebench 2024 and Prime95 stress tests ok
I can have OCCT extrem test and Cinebench 2024 passed @ P(5.6Ghz) and E(4.4Ghz)
Only Cinebench is ok @ P(5.7Ghz) and E(4.5Ghz), in addition I run it as a default setting when I play games, tested for several weeks, and it 's ok
But it fails OCCT tests and Prime 95...
Should i even consider it a problem, Am i the only one in this situation? My initial though was, If it's stable for your uses cases, it s fine...well something might be wrong, may be I'll face with more demanding upcoming games issues in the future, and If a game or Windows crashes, Is it my OC setting or yet another bug???

I'd like to have the point of view of experienced overclocker : Which tests should i consider? Is Delid and custom Watercooling a solution ? Can it improve stability for the 13900K ?
Edit : May be my RAM @ 7200MT/S is too high for stability, : I have to reduced the Vram by 0.05V to pass OCCT test @stock ???
Thanks !

 

[dgianstefani]

Hello Techpowerup people,
Please, check my config first.

I tried to overclock my 13900K, and the result are not very good.
@stock, it runs P core(5.5Ghz) and E Core(4.3GHz) I figured how to have OCCT, Cinebench 2024 and Prime95 stress tests ok
I can have OCCT extrem test and Cinebench 2024 passed @ P(5.6Ghz) and E(4.4Ghz)
Only Cinebench is ok @ P(5.7Ghz) and E(4.5Ghz), in addition I run it as a default setting when I play games, tested for several weeks, and it 's ok
But it fails OCCT tests and Prime 95...
Should i even consider it a problem, Am i the only one in this situation? My initial though was, If it's stable for your uses cases, it s fine...well something might be wrong, may be I'll face with more demanding upcoming games issues in the future, and If a game or Windows crashes, Is it my OC setting or yet another bug???

I'd like to have the point of view of experienced overclocker : Which tests should i consider? Is Delid and custom Watercooling a solution ? Can it improve stability for the 13900K ?
Edit : May be my RAM @ 7200MT/S is too high for stability, : I have to reduced the Vram by 0.05V to pass OCCT test @stock ???
Thanks !

Run your E cores at 3.2 GHz, that's their point of peak efficiency. This will give you more power and thermal budget for the P cores without affecting real world performance too much as the E cores are mainly for background tasks. You should also consider doing a per core overclock - 1 core 6.2 GHz, 2 cores 6.0 GHz, 4 5.8 etc.

You should also tune your ring clock.

Delidding and custom watercooling will both greatly improve temperatures, but they're not required for a good overclock. Mainly you'll just see a few extra 100 MHz and a quieter system.

If you're unstable that's a overly ambitious frequency target, wrong voltages problem or clockstretching from not having enough voltage. At higher temperatures you also have more voltage leakage, so it's a bit of a runaway effect.

 

[OxSteve]

Thanks ! Not a second I thought about downclocking the E Cores , I'm gonna give it a try right now.

 

[BoggledBeagle]

Overclocking is a thing of the past, when there were huge performance reserves left by the manufacturers in the stock configurations of the CPUs.

Nowadays the CPUs are already overclocked out of the box FAR PAST any reasonable frequencies.

So these days you downclock the CPUs, not overclock them.

You may find that after overclocking and extensive stability testing, you may destroy or at least degrade your CPU.

If you do not believe me, the only cure is to try it yourself.

BTW you may find that when the CPU is power limited, it automatically chooses a certain RATIO of P and E core frequencies. That ratio is not random, but a result of some OPTIMIZATION. If I remember correctly it is something around 5:4.

Running E cores slowly may not bring anything positive to the performance of the CPU.

 

[Bobaganoosh]

I would definitely echo the per-core OC approach. Outside benchmarks, there aren't many situations that have a noticeable benefit from bumping up the all-P-core clocks. I've tested games at 5.5 to 5.7GHz and it isn't that noticeable, it definitely isn't worth it when you consider the heat increase. I've had much better luck with bumping 1-4 cores to 6GHz, 5-cores to 5.8, 6 to 5.7, 7 to 5.6, and 8 to 5.5. So all-core workloads are basically stock, which keeps power under control, but games that use less threads get much higher boosted speeds and those actually benefit from the 6GHz. You're mileage will vary there, so you may get a situation where you can do single-core to 6.2, 2 cores to 6, 3 cores to 5.9, etc. but you won't know until you test. It's also a good way to utilize the V/F point control, dropping voltage for all-core loads, and thus reducing power/heat. As for e-cores, benchmarks will have a good bump by getting them to 4.6GHz, but outside benchmarks, you won't notice much difference between that and 4.4GHz, so if dropping those a little gives you more stability on P-cores and gaming is the goal, go that route.

 

[OxSteve]

Overclocking is a thing of the past, when there were huge performance reserves left by the manufacturers in the stock configurations of the CPUs.

I must admit I read this several times, but I'm also from the past, I also made some undervolt testing, but no downclock. To be honest, I do not need to overclock, just I enjoy the chalenge, though this 13900K really starts to bother me. I already consider to run it at stock undervolt settings.

You may find that after overclocking and extensive stability testing, you may destroy or at least degrade your CPU.

Well, I hardly believe it, how could It be damaged with all Voltage, Intensity and Power protection ? May be if I set max temp @115C° (i don't) ? I do believe if we use values within allowed range, it should be fine, unless you are unlucky enough to have a bad component , which could happen. As I'm rusty in electronincs, I would be more happy If you have references, articles , feedbacks on that matter

Running E cores slowly may not bring anything positive to the performance of the CPU.

I will test this, but need more time . and it s not a purpose to me, It's more like enjoying the path of overcloking

Thanks for your answer !

 

[OxSteve]

I've tested games at 5.5 to 5.7GHz and it isn't that noticeable, it definitely isn't worth it when you consider the heat increase.

I did the same, 1-2% max fps increased from my stats. I understand It might seems strange, I'm not doing this to get better FPS, for I know my old eyes and brain cannot process them.. I enjoy learning how it works, thats all.

I'm trying the per-core OC approach, and i will also stop trying to overclock E core as well.
Thanks for your feedback !

 

[BoggledBeagle]

per-core OC approach

Too bad that there is no such thing as per-core voltage.