It seems that you're using an outdated browser. Some things may not work as they should (or don't work at all).
We suggest you upgrade newer and better browser like: Chrome, Firefox, Internet Explorer or Opera

×
coughs

okay, a little curiosity i could not manage to find an answer too. Copilot managed to provide options but i'm actually without any clue what might be best in my gamer live..

With intel cpu's, and especially their latest generations being very well known as hot heads, i was wondering how i could avoid or at least find a mid-way to too much sound annoyance. I'm used to a 10900k, which i cool with the well-known Arctic liquid 280 and this setup almost never presents any problems. I probably buy a MSI 360 liquid cooler for the new system i'm planning because i think this will respond better with the MSI board. An Air cooler might do too but my experience, especially with the 3090Ti which also comes with its own radiator, is enough to say that in that respect I already made up my mind.

The options Copilot provided are actually quite sane in their own right, not anywhere near calling an answer for 14 year olds hallucinant when the AI search engine in question cookes up a story about when he went to school Dutch will know what i'm talking 'bout .....

The options

1. Downclock your cpu
2. Lower the TDP
3. Disable the boost clock (this option i'm very curious about, does anyone do this?)
4. Change the cooler setting in the bios from watercooled to regular (dubious as what?)
5. Enable Power saving Features
6. An option of your own making

What would you do?
This question / problem has been solved by Dark_art_image
avatar
P. Zimerickus: The options

1. Downclock your cpu
2. Lower the TDP
3. Disable the boost clock (this option i'm very curious about, does anyone do this?)
4. Change the cooler setting in the bios from watercooled to regular (dubious as what?)
5. Enable Power saving Features
6. An option of your own making

What would you do?
2 is not an option unless you are shopping for new, all else you can potentially do with existing setups.
Anyway, I'd go for 2 (if possible) and 5.
Option 4: this distinction likely allows for more aggressige "boost clocking" since the system can assume the liquid cooling will better handle the added heat. So it might boost higher or add another core to the boosted ones.

I'm not convinced that downclocking actually helps, because if the CPU is idle, it will clock down anyway, and if it's not, then it'll just have to churn for a longer time at reduced speed, so it may actually consume more power for the same task, or ramp up another core because it cannot handle the demand without. It might help if it is waiting for I/O or memory, because it may not be able to fine-tune the clock this much between memory cycles, etc., IOW: YMMV depending on type of load.

I'd definitely look at the OS power options and set the CPU governor to second-lowest. That introduces a slight delay before the full clock runs, but ramps down also with only a slight delay. The lowest just takes too long to get up to speed, while the higher ones just keep burning for too long and ramping up gets triggered too easily (one little load spike ramps up all cores, and they keep running full tilt for 1 minute or so, and then comes the next spike so it'll only slow down when actually idle). The higher settings may be what you want if you actually do high-load tasks, but for an office machine that mostly runs text editing and web browsing, that is not noticable.

The boost clock only comes into play if there isn't enough load to distribute evenly, so there may be 1 or 2 cores fully loaded while the others idle. Thus boosting them may be sensible if it is expected that the load can be finished faster. If it just runs a busy loop, then rats. If more cores can be saturated, then it'll not boost, anyway.

This means: parallelize if possible. When compiling, use the option to create more worker threads (between number of cores and twice that to account for I/O times, more then twice doesn't help), and I'm sure image / video editing software can also parallelize, even though these might end up choking on memory access. Games have no parallelization options so they probably can't do that (yet). Things like CIV could definitely benefit from it in terms of speed alone.
Post edited February 03, 2025 by Dawnsinger
I turned off boost as a rule, knowing that I was able to do so being one of the main things I wanted to make sure of even back when I started considering a new computer, not to mention before choosing the CPU and motherboard. But since I run BOINC projects and the CPU is always at 100% on all cores, boost obviously only meant it running past its normal maximum and possibly alternating between slightly higher speed and normal or even lower when it got too hot or consumed too much, so higher consumption and temps for minimal benefit.
And since I came from an old and weak CPU and the difference is immense, turning down TDP was also a clear choice, especially after seeing a thorough consumption vs. performance comparison for the particular CPU (Ryzen 8700G) which stated that at 45W TDP it has an overall performance score only 7% lower than at the full 65W. But, speaking of those BOINC projects, it seems like Asteroids@home makes it exceed those limits somehow, which allowed me to compare some things, so running MilkyWay@home makes it typically stay below 50C, clock speeds at the maximum of 4200 and reported CPU consumption right where it should be, just under 45W, but running Asterioids@home, despite the cTDP staying at 45W, made it report consuming about 60W and temperatures being around 60C, despite the reported clock speed being 3800 on all cores. And splitting, making it run some Asteroids and some MW tasks in parallel, pushed it even more, speed at 4200, consumption approaching the normal maximum of 65W and temps exceeding 65C, approaching 70C.
So, anyway, if you can lower TDP and can take the performance hit, it seems to have a massive effect, as it well should.
Of course, downclocking is also a basic solution, but lower TDP is likely more efficient, letting the system maximize performance while staying within that envelope.
I disable boost on my Ryzen 4500u laptop. The 4.0GHz boost is way too high for a simple laptop cooler and it start spinning the fan even on simple tasks like open several bloated websites, since the "base clock" is 2.4GHz, it's around the perfect compromized frequency. 2.8 to 3.0 would be better but I can live waiting a few more miliseconds for a website to load.

On the i5 8250u laptop disable the boost meaning having a max of 1.6GHz, because Intel, so I limit the frequency to 2.4GHz (wich I did a bunch of tests to determine the sweet spot) on software (Linux TLP).
Both of the above have a large inpact on battery run times. Another reason to limit the max frequency is that some workloads use all the available clocks for no reason (no performance gains), and even only 1 core boosting heats a lot.

On the desktop I limit the max TDP on the BIOS/UEFI options because the boost frenquency isn't very high to begin with. Depending on the load the CPU single thread clock up to 3.7GHz and all core to around 3.0 - 3.2GHz.

These are my cases and of course, have nothing to do with yours. If you are running a 10900k I assume you want every last drop of performance possible otherwise you would be on a sane CPU. With that said, any tweak you can do to reduce power and temps will impact the performance.
If I was in your shoes (besides never get a 10900k in the first place) I would limit the max frequency to around 4.0 to 4.5GHz, this value would require some testing to find the sweet spot between power and performance.
My reasoning is the follow: since the 10900k is running well above it's efficiency peak, any decrease in clocks will bring a greater decrease in power used, say less 10% performance will yeld 35% less power used.
Simply limit the max TDP will also work in bringing the temperature down but still allow the CPU to clock a few threads very high, that's probably the best compromise between a snappy system that can get high FPS's and not too much temp (but high single/dual thread still heats up a lot). I would guess a limit of 120-150W would be around the default on most non-k CPU's.
Another option would be to skip PL2 (fast power limit) entirely.

As you noted by the above, I don't like high end "gaming" CPU's (nor GPU's for that matter) and have no use for them. I also don't play the lastest and greatest, so my opinion may not making much sense to you.
One of the best things I did to reduce the computer power is to stop playing games at high rates just because. When it's needed I still do otherwise 60fps/Hz is fine, not gonna lie, was a bit hard at first but since I also did stop playing competitive games and doing big sessions, after a while felt fine.

The best tip I have for you is: hook the computer to a power meter/kill-a-watt and play with it. It doesn't need to be a fancy device of very accurate for that matter but faster response is helpfull in computers due the fast changes in load. Keep in mind that the power you measure will be smoothed by several internal power supplies but you may find a lot of interesting things. Despite what my GF says, I believe bigger is not always better.
Post edited February 04, 2025 by Dark_art_
avatar
Cavalary: -snip-
Perhaps "Asterioids@home" is using some instructions running on the GPU.
Some chips BIOS/UEFI TDP setting doesn't include the integrated graphics in it's envelope (I assume most desktop chips).
I don't know if I'm missing something, but, why don't you downvolt it?
And, don't know it that's the case but, don't think that water cooling will do magic: it is more efficient than air cooling, yes, but either can't bring temperature below ambient temp.
Post edited February 04, 2025 by Judicat0r
Im using a 7950X and just setting a power limit in Windows is pretty effective. Think I've set it to 95 Percent max for everyday use (or something like that, not a whole lot lower) and need about a third less power this way. I just set it to high performance when I need it then.

That said, im using the same Processor at work and it runs at 100 percent for days or weeks sometimes, all good. They know when to throttle down I guess...
Great responses guys! thanks !

I'm actually treading back on lowering the tdp i think in my case it might be more beneficial to avoid a situation where CPU runs at 100% load whereas i have no trouble with limiting my GPU through tdp means
avatar
Dark_art_: I disable boost on my Ryzen 4500u laptop. The 4.0GHz boost is way too high for a simple laptop cooler and it start spinning the fan even on simple tasks like open several bloated websites, since the "base clock" is 2.4GHz, it's around the perfect compromized frequency. 2.8 to 3.0 would be better but I can live waiting a few more miliseconds for a website to load.

On the i5 8250u laptop disable the boost meaning having a max of 1.6GHz, because Intel, so I limit the frequency to 2.4GHz (wich I did a bunch of tests to determine the sweet spot) on software (Linux TLP).
Both of the above have a large inpact on battery run times. Another reason to limit the max frequency is that some workloads use all the available clocks for no reason (no performance gains), and even only 1 core boosting heats a lot.

On the desktop I limit the max TDP on the BIOS/UEFI options because the boost frenquency isn't very high to begin with. Depending on the load the CPU single thread clock up to 3.7GHz and all core to around 3.0 - 3.2GHz.

These are my cases and of course, have nothing to do with yours. If you are running a 10900k I assume you want every last drop of performance possible otherwise you would be on a sane CPU. With that said, any tweak you can do to reduce power and temps will impact the performance.
If I was in your shoes (besides never get a 10900k in the first place) I would limit the max frequency to around 4.0 to 4.5GHz, this value would require some testing to find the sweet spot between power and performance.
My reasoning is the follow: since the 10900k is running well above it's efficiency peak, any decrease in clocks will bring a greater decrease in power used, say less 10% performance will yeld 35% less power used.
Simply limit the max TDP will also work in bringing the temperature down but still allow the CPU to clock a few threads very high, that's probably the best compromise between a snappy system that can get high FPS's and not too much temp (but high single/dual thread still heats up a lot). I would guess a limit of 120-150W would be around the default on most non-k CPU's.
Another option would be to skip PL2 (fast power limit) entirely.

As you noted by the above, I don't like high end "gaming" CPU's (nor GPU's for that matter) and have no use for them. I also don't play the lastest and greatest, so my opinion may not making much sense to you.
One of the best things I did to reduce the computer power is to stop playing games at high rates just because. When it's needed I still do otherwise 60fps/Hz is fine, not gonna lie, was a bit hard at first but since I also did stop playing competitive games and doing big sessions, after a while felt fine.

The best tip I have for you is: hook the computer to a power meter/kill-a-watt and play with it. It doesn't need to be a fancy device of very accurate for that matter but faster response is helpfull in computers due the fast changes in load. Keep in mind that the power you measure will be smoothed by several internal power supplies but you may find a lot of interesting things. Despite what my GF says, I believe bigger is not always better.
Well after rereading your post it seems i have done the same with my new cpu, i'm actually limiting it to 5 ghz, See my upcoming new build post