Questions arising from overclocking 9600K related to the H100i PRO AIO

[andiey]

So I have my new 9600K rig to play around for the last couple of weeks. And I still love my old 3570K.

I overclocked the two systems and stressed test using CPU-Z and have found the followings:

9600K @ stock 4.3GHz - Temperature 48C (Max) for 30 minutes Cooler = CORSAIR H100i PRO
@ 4.8 GHz (ocd) - Temperature 64C (Max) for 30 minutes

3570K @ 4.2 GHz (ocd) - Temperature 65C(Max) for 60 minutes Cooler = CORSAIR H60 (1st generation, square pump no rgb)

My thoughts are GREAT especially the 3570K which I placed it inside a Bitfenix Prodigy case without any intake fans, just the exhaust fan attached to the radiator at the rear. Remark is that there is no graphics card in this rig, I took the graphics card and put it into the 9600K.

The questions which arise from overclocking and stress test the 9600K are

1) I found the default fan curve H100i PRO AIO was sensor'ing the H100i PRO. It was one of the selected items in a list where there are also 9600K Core#0, 9600K Core#1,...Asus Rog Strix Z30-F fan#1,..., CPU Package....
I like to know should I select CPU Package as the right sensor? It doesn't make sense that the dual fans' speed should be dependent on the pump temperature rather than on the CPU's, am I right or I am missing something here? Please advise.

2) After stress testing the 9600K, I stop the stress. The 9600K's temperature however, didn't fall back by very much. I was expecting it fell back to the range 31~35 like it was ilde. But what happened was that it fell back to the mid 40's, I am referring to the CPU Package, not individual core's. Individual cores did fall back to 30's with at most 2 out of 6 of them hovering around the mid 40's.

Can anyone advise please, I am worrying about the 9600K's behaviour.

 

[Deleted member 185158]

Max temp of only 64c?

From my seat, you have a pretty cool running processor. I see nothing to be alarmed about here.

 

[andiey]

Max temp of only 64c?

From my seat, you have a pretty cool running processor. I see nothing to be alarmed about here.

Yes, it is 64C on all cores, for 30 minutes. I don't see any points running the stress for a bit longer given it is new CPU.
Maybe, it is the 3x Intake fans and the 2x Exhaust fan/radiator and 1x exhaust fan does the job pretty well.

Do you know anything about the H100i PRO fan curve?

 

[oxrufiioxo]

You need to stress with something like aida 64 cpu/fpu... CPU Z doesn't even stress my 9900k much at all.

 

[Deleted member 185158]

Yes, it is 64C on all cores, for 30 minutes. I don't see any points running the stress for a bit longer given it is new CPU.
Maybe, it is the 3x Intake fans and the 2x Exhaust fan/radiator and 1x exhaust fan does the job pretty well.

Do you know anything about the H100i PRO fan curve?

I don't actually. I've never had an AIO unit before, I cannot help.
However I do manually set my fan curves in bios to my desires. If your specs are accurate, hit F6 once you are in the bios. (In other words I don't use the fan/oc software)

 

[sneekypeet]

Corsair based the fan curve on the coolant temperature. Software should allow you to set your own curve if you do not like what is happening now.

 

[andiey]

Corsair based the fan curve on the coolant temperature. Software should allow you to set your own curve if you do not like what is happening now.

So you're sure about that, the H100i PPRO means coolant temperature?

 

[oxrufiioxo]

So you're sure about that, the H100i PPRO means coolant temperature?

I actually have mine set to ramp up from 800-1200 depending on my GPU temps... My coolant never gets warm enough otherwise to make my Fans spin up. If you set it to package temp the fans will constantly ramp up and down depending on what RPM you set and what you're doing which is annoying. I would just leave it to coolant and use the balance preset or you can tinker with the fan curve based on coolant temperature.

 

[andiey]

This is from my office PC, I was trying to find the "CPU Package temp" for the RGB lighting.

I actually have mine set to ramp up from 800-1200 depending on my GPU temps... My coolant never gets warm enough otherwise to make my Fans spin up. If you set it to package temp the fans will constantly ramp up and down

Exactly, they are very sensitive.
That's why I need advices, I can't take my current setup for too long or the fan will be damaged more or less.

...depending on what RPM you set and what you're doing which is annoying. I would just leave it to coolant and use the balance preset or you can tinker with the fan curve based on coolant temperature.

Balanced preset is pretty stable that's true and relatively quiet. I'll try that for a few days and see how it goes then, thank you oxrufiioxo for your help.

 

[oxrufiioxo]

View attachment 138053

This is from my office PC, I was trying to find the "CPU Package temp" for the RGB lighting.


Exactly, they are very sensitive.
That's why I need advices, I can't take my current setup for too long or the fan will be damaged more or less.


Balanced preset is pretty stable that's true and relatively quiet. I'll try that for a few days and see how it goes then, thank you oxrufiioxo for your help.

Do you not have your h100i plugged in?

 

[biffzinker]

Try stress testing the overclock with ASUS's Realbench. It has caught an unstable overclock for me just running through the benchmark although you can run a stress test.

undefined

undefined

rog.asus.com

 

[Valantar]

I would avoid setting the fan speed relative to CPU temps at least without some sort of hysteresis (i.e. that the fans won't spin up until the temperature has reached the threshold for at least 5 seconds). As mentioned above any fan without hysteresis will constantly ramp up and down and be very, very annoying. Basing fan speeds on coolant temps (which take time to increase, and will never spike) is a lot more sensible, which is why that's Corsair's default approach. Unless your CPU is thermal throttling (which it clearly isn't - you're still a good 35C away from that) I would leave well enough alone. And if you're seriously worried about post-stress idle package temps in the 40s: stop. 40C has never ever hurt a CPU. This is 99% likely due to the AIO still retaining a bit of heat from being under load, which takes time to dissipate fully. Another ten minutes at idle and you'll be back to where you were. Stop worrying, leave it alone, enjoy your PC.

 

[andiey]

Do you not have your h100i plugged in?

H100i is my home PC. This is my office PC. But I guess they work the same aren't they?

 

[oxrufiioxo]

H100i is my home PC. This is my office PC. But I guess they work the same aren't they?

I mean setting up profiles will be the same and what not but your home PC will have a drop down setting for coolant temp your work PC will not.

 

[andiey]

I would avoid setting the fan speed relative to CPU temps at least without some sort of hysteresis (i.e. that the fans won't spin up until the temperature has reached the threshold for at least 5 seconds). As mentioned above any fan without hysteresis will constantly ramp up and down and be very, very annoying. Basing fan speeds on coolant temps (which take time to increase, and will never spike) is a lot more sensible, which is why that's Corsair's default approach. Unless your CPU is thermal throttling (which it clearly isn't - you're still a good 35C away from that) I would leave well enough alone. And if you're seriously worried about post-stress idle package temps in the 40s: stop. 40C has never ever hurt a CPU. This is 99% likely due to the AIO still retaining a bit of heat from being under load, which takes time to dissipate fully. Another ten minutes at idle and you'll be back to where you were. Stop worrying, leave it alone, enjoy your PC.

My mind is almost set at ease.
However, there was one minor thing I did notice when the stress was over.

When I started the system at 4.8GHz, HWiNFO reported 31~35C at idle. Then I started CPU-Z stress test and left it running for 30 minutes while I went to make tea. When I returned, I noted down the temperature of CPU package's of course not individual core's. That read 64C max. Then I hit stop. I was anticipating the CPU package temperature dropping back to the 31~35C range like the 3570K did. But no, it didn't drop back to that range, at that point, I had a feeling that for some unknown reasons to me, that the 9600K couldn't adjust itself, why it behaved that way? I am concerned about that.

And I forgot to mention that prior to the overclocking taking place, I turned on the Intel Rapid Storage Technology in the BIOS, would that be the cause of that issue?

 

[oxrufiioxo]

No idea use hwmonitor to verify temps it should drop back down relatively fast.

 

[andiey]

I mean setting up profiles will be the same and what not but your home PC will have a drop down setting for coolant temp your work PC will not.

Thanks bud, both you and Valantar have told me about the coolant temperature is in fact what "H100i Pro Temp" means sets my mind at ease. I think Corsair's bad nomenclature has led me wrongly to think that's the pump's temperature and one would naturally correlate that to the CPU's temperature by extension considering the pump is physically attached to the CPU. Now that both of you have told me that means the coolant's temperature has completely enlightened me upon the idea, it makes perfect sense to me: liquid's temperature delta (or change) happens gradually rather than electronic element's temperature which alters instantaneously. In order to avoid sparks of the CPU fans, it is definitely more desirable to alter the fan speed accordingly with the coolant.

Thanks buds

 

[Zach_01]

...it makes perfect sense to me: liquid's temperature delta (or change) happens gradually rather than electronic element's temperature which alters instantaneously. In order to avoid sparks of the CPU fans, it is definitely more desirable to alter the fan speed accordingly with the coolant.

Thanks buds

You got it right... thats how I also set it.
And dont forget that CPU package temp sensor is further away from the surface than the core temp sensors... Thats why is so slow to return to idle temp. The whole package, the socket itself, and the board has build heat that dissipates rather slow. The cores are just under the AIO water block and cool off faster...

Its all normal

 

[Valantar]

Good to hear you're feeling better about this. You really have no reason to worry.

As for returning to pre-stress idle temps, that takes time - idle temps (especially when the PC hasn't been running for that long) are normally with the AIO's liquid at some temperature between ambient and the CPU temperature, even if fan speeds are slow at idle. This is simply because the relatively tiny heat load of an idling CPU (likely anywhere from 2-3W to short-term 30W spikes for casual desktop use) isn't enough to meaningfully heat the liquid above ambient even when the fans are slowed down. After long-term stress on the other hand, the liquid has absorbed a lot of heat from the CPU, which is retained unless the fans are still spinning quickly. And even with water temperature controlled fan speeds, your fan curve isn't likely to run the fans very fast if the water in the loop is at 40-45C. Water has a high thermal capacity (which is part of why it (coupled with a good radiator) is a good cooling medium), so it's entirely possible for your water loop to actually keep the CPU slightly warm if the CPU goes from full load to idle while the loop is still dissipating the heat from the previous workload - simply because the warmer water in the loop is providing less cooling to the chip. And as the water temperature slowly drops, the fans will spin down, meaning the last part of the drop takes the longest time.

My custom loop (not controlled by water temp, but some fans are CPU temp controlled, some GPU temp controlled) normally idles the CPU at 35-40C (it does that odd Ryzen thing where the temperature spikes every ~10 seconds and then slowly drops down again) and the GPU at a steady 31-32. After prolonged stress it takes at least half an hour for temperatures to reach the exact same levels, but no more than 10 minutes to come within 5 degrees. The last little bit takes a long time. On the other hand, if I manually set my fans to full blast, it returns to normal relatively quickly and stays there after I switch them back to their normal slow idle speed.

 

[Zach_01]

After long-term stress on the other hand, the liquid has absorbed a lot of heat from the CPU, which is retained unless the fans are still spinning quickly. And even with water temperature controlled fan speeds, your fan curve isn't likely to run the fans very fast if the water in the loop is at 40-45C. Water has a high thermal capacity (which is part of why it (coupled with a good radiator) is a good cooling medium), so it's entirely possible for your water loop to actually keep the CPU slightly warm if the CPU goes from full load to idle while the loop is still dissipating the heat from the previous workload - simply because the warmer water in the loop is providing less cooling to the chip. And as the water temperature slowly drops, the fans will spin down, meaning the last part of the drop takes the longest time.

This is not the case... short of...
What you say could happen, but only if you set the wrong/undesired fan curve. The cool thing about fan curves is that you can set the entire fan rpm range to like 30C delta or to 5C delta for min-max rpm
If you try to set the fun curve with wide delta and bind the curve to water temp... then its wrong and what you are saying will be true.

You can cool the CPU with either wide delta (20~30 or even 40C) bound to CPU temp or narrow delta (5~10C) bound to water temp and do the appropriate cooling.
The difference is that at CPU temp bound curve the fans will spike rpm a little more as CPU temp bounce around. Then you must introduce rpm latency to smooth out the curve... Water can do that with out any rpm latency/hysterisis. Corsair iCue does not have any latency setting because binding the fan curve to water does the same thing.

There is no wrong or right here... only preference.

 

[andiey]

This is not the case... short of...
What you say could happen, but only if you set the wrong/undesired fan curve. The cool thing about fan curves is that you can set the entire fan rpm range to like 30C delta or to 5C delta for min-max rpm
If you try to set the fun curve with wide delta and bind the curve to water temp... then its wrong and what you are saying will be true.

You can cool the CPU with either wide delta (20~30 or even 40C) bound to CPU temp or narrow delta (5~10C) bound to water temp and do the appropriate cooling.
The difference is that at CPU temp bound curve the fans will spike rpm a little more as CPU temp bounce around. Then you must introduce rpm latency to smooth out the curve... Water can do that with out any rpm latency/hysterisis. Corsair iCue does not have any latency setting because binding the fan curve to water does the same thing.

There is no wrong or right here... only preference.

Thanks for the picture. I think I know how to do it now. I'll try to set it up and see if I can push the clock speed a little further.

Good to hear you're feeling better about this. You really have no reason to worry.

As for returning to pre-stress idle temps, that takes time - idle temps (especially when the PC hasn't been running for that long) are normally with the AIO's liquid at some temperature between ambient and the CPU temperature, even if fan speeds are slow at idle. This is simply because the relatively tiny heat load of an idling CPU (likely anywhere from 2-3W to short-term 30W spikes for casual desktop use) isn't enough to meaningfully heat the liquid above ambient even when the fans are slowed down. After long-term stress on the other hand, the liquid has absorbed a lot of heat from the CPU, which is retained unless the fans are still spinning quickly. And even with water temperature controlled fan speeds, your fan curve isn't likely to run the fans very fast if the water in the loop is at 40-45C. Water has a high thermal capacity (which is part of why it (coupled with a good radiator) is a good cooling medium), so it's entirely possible for your water loop to actually keep the CPU slightly warm if the CPU goes from full load to idle while the loop is still dissipating the heat from the previous workload - simply because the warmer water in the loop is providing less cooling to the chip. And as the water temperature slowly drops, the fans will spin down, meaning the last part of the drop takes the longest time.

My custom loop (not controlled by water temp, but some fans are CPU temp controlled, some GPU temp controlled) normally idles the CPU at 35-40C (it does that odd Ryzen thing where the temperature spikes every ~10 seconds and then slowly drops down again) and the GPU at a steady 31-32. After prolonged stress it takes at least half an hour for temperatures to reach the exact same levels, but no more than 10 minutes to come within 5 degrees. The last little bit takes a long time. On the other hand, if I manually set my fans to full blast, it returns to normal relatively quickly and stays there after I switch them back to their normal slow idle speed.

Great info, it's very useful reference.
Talking about water-cooled GPU, I guess I will get the NZXT Kraken Beacket and one of the AIO, I take it that helps reducing the operating temperature of the GPU by 50%.

I guess a video and a screenshot saves a thousand words:

I've finally managed to record the Stress CPU process along with the information of the H100i PRO and the RMx 850i. I set the 9600K to run at stock speed. Note the ambient room temperature as shown in the screenshot reads 24C coz the image quality is not superb. SMH,

Strange is that after I stopped the stress test, the clock goes to 4.4GHz instead of 4.3GHz during the test, any clue?

{What is the video format allowed on this website?

I guess a video and a screenshot saves a thousand words:

I've finally managed to record the Stress CPU process along with the information of the H100i PRO and the RMx 850i. I set the 9600K to run at stock speed. Note the ambient room temperature as shown in the screenshot reads 24C coz the image quality is not superb. SMH,

Strange is that after I stopped the stress test, the clock went back to 4.4GHz instead of 4.3GHz during the test, any clue?


https://aandiey.tumblr.com%2Fpost%2F189413237432

 

[JackCarver]

Strange is that after I stopped the stress test, the clock goes to 4.4GHz instead of 4.3GHz during the test, any clue?

Did you set an AVX Offset in BIOS?

 

[andiey]

Did you set an AVX Offset in BIOS?

What is that? What is AVX Offset?

Please tell me what should I set it to.

BTW, power consumption as you see is tremendous almost 300W. Fortunately,I have bought this 850W PSU.

This is the overclocked to 4,9GHz. I've found that to make it happens XMP_1 profile must be set. Is that something to do with the memory speed? Also the 4,900Mhz wasn't constantly holding, it dropped intermittently to 4,989Mhz 4,987Mhz is that what people called "CPU throttle"?

https://aandiey.tumblr.com%2Fpost%2F189425674667

 

[John Naylor]

What is your goal ?

a) You want to get your name listed on leaderboard at overlocking web sites ? In this case use synthetic testing for whatever qualifies to get you on the boards.

b) To have the best OC you can get to run applications and games. In this case use an applications based stress test like RoG Real Bench. Your final OC will be higher , your temps will be significantly lower. I have had 24 hour stable P95 OCs fail in 45 minutes under RoG Real Bench

For the best settings advice, Id recommend the CPU specific thread for your CPU on overclock.net

 

[EarthDog]

I've found that to make it happens XMP_1 profile must be set.

some boards have tweaked things when using xmp... like asus. When you enable xmp, it tells you specifically about core enhancement. You chose to enable it. You can disable it and overclock manually. This way you can control voltage and perhaps lower the heat as leaving core voltage on auto tends to use more than is needed.

Also the 4,900Mhz wasn't constantly holding, it dropped intermittently to 4,989Mhz 4,987Mhz is that what people called "CPU throttle"?

Nope. That's normal. It's called spread spectrum and a natural occurence in PCs. You likely have an option to disable it (but do not need to).