Hello, I am new to overclocking and have spent the past week doing research and trying to learn about it. Got a couple of questions but what worries me the most at the moment are my temps. Hopefully this community can help me out and put my mind at ease. :)
System/Info:
H100i aio is mounted on the top rear-end of the case (directly above CPU). I have the aio pump set at extreme and iCue readouts are hovering at ~2580rpm. I used Arctic MX4 thermal paste with the "dot" method. The tube was bought around 3 years ago but the paste texture still seems normal. 9x ql120 fans in the case are setup as 3 bottom intake, 3 side intake, and 3 top exhaust (2 of which are mounted on the aio rad in a push configuration). I have all 9 fans set at a fixed 1000rpm (I'll set a custom curve once I find a stable OC).
OC Settings:
With everything set to default in BIOS (with the exception of RAM which I left on XMP profile 1), I tested with Aida64 (30 min) and Prime95 small FFT with AVX on and off (30 min) in order to establish a baseline for temps.
Aida64 Stock 30min (CPU, FPU, Cache, System Memory)
Prime95 Stock 30min (Small FFT AVX Off)
Prime95 Stock 30min (Small FFT AVX On)
9700K 5.0GHz 1.270V Test Results:
Running the 9700K at 5GHz and 1.270V, I stress tested with Cinebench R20, Aida64, and Prime95. Once I find a stable overclock I'll run some longer overnight tests but here are my results for now. Cinebench was easily passed with a custom 300 second loop. Ran Aida64 (CPU, FPU, cache, system memory) for 30 minutes. Prime95 with small FFT No AVX for 30 minutes was passed, however, Prime95 with small FFT AVX On BSOD'ed ~3 minutes into the test with thermal throttling at 100c on some of the cores. Would the BSOD be due to the thermal throttling or the general instability of the overclock? Due to the BSOD I'm unable to show the temps but as far as I remember, about half the cores thermal throttled at 100c prior to the BSOD.
Aida64 5.0GHz 1.270V 30min (CPU, FPU, Cache, System Memory)
Prime95 5.0GHz 1.270V 30min (Small FFT AVX Off)
Key concern here are the temps. Is this normal? Looking at other people's results, some of them are getting lower temps with lower end cooling solutions. I'm not expecting crazy custom water loop temps but I certainly also wasn't expecting temps this high neither. For example, saw another post boasting ~35c idle and ~85c Prime95 (same setup as screenshot directly above) on a 9700K with an air cooler. I mean, I have 9 fans. Surely that's gotta count for something right? I know ambient room temp also affects cpu temps but afaik 20c seems cool to me. I read another troubleshooting post where it was recommended to re-seat the aio and use the "X" shape method instead in order to evenly distribute paste throughout the cores. Could the problem possibly be with the 3 year old mx4 paste? If not, I was also thinking about increasing all fans to max rpm. But then again, I feel like that's just unrealistic when I'm trying to compare my temps to other users. I can try setting the avx offset to 1 or 2 to try and get the p95 test to stabilize but that's besides the point since I never run avx programs (mostly game on PC) and temps are more so the concern here, not stability. If people in the community think the "p95 small FFT AVX On" tests are unnecessary for my uses I'll negate them from my tests. I also ran Apex Legends and Rainbow Six Siege for around 2 hours each and they average around 75c and maxing at 80c. Is that normal? So the only solutions I've got right now are: buying new thermal paste, re-seating aio heat sink (possibly try different thermal paste shape), and setting a custom fan curve or running them at max. But before I go ahead with any of these, Ill wait to see what this community has to say first.
Another concern I have is the voltage. With CPU voltage set manually as 1.270V, loadline calibration at "turbo", and the rest of the OC settings mentioned above, I'm confused about the readings CPU-Z and HWinfo64 are giving me. Watching this video, I've come to understand that loadline calibration is used to compensate for vcore droop right? So if I have it set at turbo, shouldn't the voltage reading be somewhere close to 1.270V but still under it? If that's the case, then aren't the CPU-Z and HWinfo64 (VR VOUT is what I'm supposed to be reading right?) readouts wrong since one is reading 1.284V and the other is 1.196V? Although, on rare occasions when my PC is idle, CPU-Z will read 1.272V for a second. Also, I've read that on gigabyte z390 boards, "High" is actually the recommended setting for LLC, so would changing it to that solve the problem and possibly even solve the temp problems?
I've also stress tested the 9700K at 5.0GHz and 1.260V so if you care to read the temps, you can find them here. Only differences with 1.260V as opposed to 1.270V are the slightly lower temps and the same BSOD on Prime95 (small FFT AVX On) at ~5 minutes instead.
Reading the FAQ of this community, the x264 stress test is recommended instead of Prime95 due to it stressing and testing stability similarly but without the crazy high temps. Any thoughts on this?
Before writing this, I wanted to gather as much info as possible so the troubleshooting process would be easier on anyone who wants to help me but I realized I started to ramble haha and now I've got paragraphs of writing so I'm sorry about the wall post!
Any and all help will be tremendously appreciated!
Update 1 - Ran a few more tests after tweaking LLC settings and here are the results: https://imgur.com/a/7cOyAC4
submitted by Mornin___Wood to overclocking [link] [comments] System/Info:
- Intel i7-9700K
- Gigabyte Z390 Aorus Ultra
- 32GB (4x8) 3200MHz CL16 Corsair Vengeance RGB Pro
- Corsair H100i Platinum SE AIO
- Gigabyte RTX 2060 Super Gaming OC
- Corsair RM850X PSU
- Windows 10 Home
- Lian-Li PC O11 Dynamic
- 9x Corsair QL120 Case Fans
- Arctic MX4 Thermal Paste
- Ambient Room Temperature: 20c
- CPU Windows 10 Startup Temperature: 30-32c
- CPU Idle Temperature: 38-41c
H100i aio is mounted on the top rear-end of the case (directly above CPU). I have the aio pump set at extreme and iCue readouts are hovering at ~2580rpm. I used Arctic MX4 thermal paste with the "dot" method. The tube was bought around 3 years ago but the paste texture still seems normal. 9x ql120 fans in the case are setup as 3 bottom intake, 3 side intake, and 3 top exhaust (2 of which are mounted on the aio rad in a push configuration). I have all 9 fans set at a fixed 1000rpm (I'll set a custom curve once I find a stable OC).
OC Settings:
- XMP: Profile 1
- enhanced multicore performance: disabled
- cpu base clock: auto (100mhz)
- CPU clock ratio: 50 (all cores)
- AVX offset: 0
- uncore ratio: auto (for now, once i get a stable overclock i'll tweak this)
- intel turbo boost technology: enabled
- intel speed shift: disabled
- CPU Vcore loadline calibration: turbo
- CPU Vcore: 1.270V
- BCLK adaptive voltage: disabled
- SVID offset: disabled
- CPU C-states: enabled (for now, once i get a stable overclock i'll tweak this)
- Aida64 v6.25
- Prime95 v29.8
With everything set to default in BIOS (with the exception of RAM which I left on XMP profile 1), I tested with Aida64 (30 min) and Prime95 small FFT with AVX on and off (30 min) in order to establish a baseline for temps.
Aida64 Stock 30min (CPU, FPU, Cache, System Memory)
Prime95 Stock 30min (Small FFT AVX Off)
Prime95 Stock 30min (Small FFT AVX On)
9700K 5.0GHz 1.270V Test Results:
Running the 9700K at 5GHz and 1.270V, I stress tested with Cinebench R20, Aida64, and Prime95. Once I find a stable overclock I'll run some longer overnight tests but here are my results for now. Cinebench was easily passed with a custom 300 second loop. Ran Aida64 (CPU, FPU, cache, system memory) for 30 minutes. Prime95 with small FFT No AVX for 30 minutes was passed, however, Prime95 with small FFT AVX On BSOD'ed ~3 minutes into the test with thermal throttling at 100c on some of the cores. Would the BSOD be due to the thermal throttling or the general instability of the overclock? Due to the BSOD I'm unable to show the temps but as far as I remember, about half the cores thermal throttled at 100c prior to the BSOD.
Aida64 5.0GHz 1.270V 30min (CPU, FPU, Cache, System Memory)
Prime95 5.0GHz 1.270V 30min (Small FFT AVX Off)
Key concern here are the temps. Is this normal? Looking at other people's results, some of them are getting lower temps with lower end cooling solutions. I'm not expecting crazy custom water loop temps but I certainly also wasn't expecting temps this high neither. For example, saw another post boasting ~35c idle and ~85c Prime95 (same setup as screenshot directly above) on a 9700K with an air cooler. I mean, I have 9 fans. Surely that's gotta count for something right? I know ambient room temp also affects cpu temps but afaik 20c seems cool to me. I read another troubleshooting post where it was recommended to re-seat the aio and use the "X" shape method instead in order to evenly distribute paste throughout the cores. Could the problem possibly be with the 3 year old mx4 paste? If not, I was also thinking about increasing all fans to max rpm. But then again, I feel like that's just unrealistic when I'm trying to compare my temps to other users. I can try setting the avx offset to 1 or 2 to try and get the p95 test to stabilize but that's besides the point since I never run avx programs (mostly game on PC) and temps are more so the concern here, not stability. If people in the community think the "p95 small FFT AVX On" tests are unnecessary for my uses I'll negate them from my tests. I also ran Apex Legends and Rainbow Six Siege for around 2 hours each and they average around 75c and maxing at 80c. Is that normal? So the only solutions I've got right now are: buying new thermal paste, re-seating aio heat sink (possibly try different thermal paste shape), and setting a custom fan curve or running them at max. But before I go ahead with any of these, Ill wait to see what this community has to say first.
Another concern I have is the voltage. With CPU voltage set manually as 1.270V, loadline calibration at "turbo", and the rest of the OC settings mentioned above, I'm confused about the readings CPU-Z and HWinfo64 are giving me. Watching this video, I've come to understand that loadline calibration is used to compensate for vcore droop right? So if I have it set at turbo, shouldn't the voltage reading be somewhere close to 1.270V but still under it? If that's the case, then aren't the CPU-Z and HWinfo64 (VR VOUT is what I'm supposed to be reading right?) readouts wrong since one is reading 1.284V and the other is 1.196V? Although, on rare occasions when my PC is idle, CPU-Z will read 1.272V for a second. Also, I've read that on gigabyte z390 boards, "High" is actually the recommended setting for LLC, so would changing it to that solve the problem and possibly even solve the temp problems?
I've also stress tested the 9700K at 5.0GHz and 1.260V so if you care to read the temps, you can find them here. Only differences with 1.260V as opposed to 1.270V are the slightly lower temps and the same BSOD on Prime95 (small FFT AVX On) at ~5 minutes instead.
Reading the FAQ of this community, the x264 stress test is recommended instead of Prime95 due to it stressing and testing stability similarly but without the crazy high temps. Any thoughts on this?
Before writing this, I wanted to gather as much info as possible so the troubleshooting process would be easier on anyone who wants to help me but I realized I started to ramble haha and now I've got paragraphs of writing so I'm sorry about the wall post!
Any and all help will be tremendously appreciated!
Update 1 - Ran a few more tests after tweaking LLC settings and here are the results: https://imgur.com/a/7cOyAC4
