Question Question concerning CPU temps and PBO for Ryzen 9 5950x

[boagz]

I recently purchased a ryzen 9 5950x for my mini-itx build. Just for some background info my build uses the lian li micro-atx case and the x570-I itx motherboard with the scythe fuma CPU cooler. I also live in south carolina and ambient temp tends to hover anywhere from 76 to 80 degrees in my apartment. When not performing any tasks on my Win 10 OS the CPU temp seems to hover around 44-49 degrees (with the occasional spikes to upper 50's 60's that quickly come down. This just seems to happen randomly with windows?). When I actually go to do something like startup a program like say unreal editor or davinci resolve my temps can suddenly jump to as high as 85 degrees before settling back down to a more reasonable temp after the initial spike. My question is are these significant jumps in temp (from mid 40's to mid 80's in some cases like described) just part of the precision boost technology and normal temp spikes with stock PBO settings? Should I try and adjust the PBO settings to lower these initial spikes? Or are these spikes a little too high and maybe more the result of improper cooler setup? The reason I was thinking more PBO was because most of the time my computer does idle around 44 (sometimes lower, sometimes slightly higher) which doesn't seem too high to me given the build and my environment. I also ran prime95 for about 20 minutes on blend and avg temp was around 85 with max 91

 

[tennis2]

Do you have the stock fan config (1 fan in front intake and 1 fan rear exhaust) or have you added a 2nd front intake fan?

I'd recommend doing a PBO Curve Optimizer undervolt. It won't reduce power consumption but it does optimize power consumption so the CPU will boost to higher frequencies. After that, you could adjust max frequency or power consumption in PBO settings to keep temps under control.

View: https://www.youtube.com/watch?v=jTkDqARDkGg&t=688s

 

[boagz]

Do you have the stock fan config (1 fan in front intake and 1 fan rear exhaust) or have you added a 2nd front intake fan?

I'd recommend doing a PBO Curve Optimizer undervolt. It won't reduce power consumption but it does optimize power consumption so the CPU will boost to higher frequencies. After that, you could adjust max frequency or power consumption in PBO settings to keep temps under control.

View: https://www.youtube.com/watch?v=jTkDqARDkGg&t=688s

I have 1 front intake fan (though its a larger 140mm fan), 1 120mm intake fan on the bottom of the case under the GPU and then the scythe cooler is pretty big so it's basically functioning as my exhaust since I can't fit another back fan in there.

Thanks for your thoughts on the matter. I haven't looked into the PBO stuff much so maybe I'll have to. So overall you think it sounds more like a PBO issue from what I described?

 

[tennis2]

The Lancool 205M doesn't have bottom fan mounts....just the PSU intake.

Also, what would you guestimate is the time duration of the 85C "initial spike"?

 

[boagz]

The Lancool 205M doesn't have bottom fan mounts....just the PSU intake.

Also, what would you guestimate is the time duration of the 85C "initial spike"?

Ah crap, wrong case. Here is the correct one. And I would say a couple seconds off the top of my head. I'll test it a little more when I get home.

 

[Viking2121]

Idle temps really don't mean much as the system will lower the fan speeds, But thats about right for a 5950x in a hot room like that, The way PBO works, it will try to keep boosting if temps allow, You can play with the voltage curve and it will allow for a slightly lower temp, but it wont make a huge difference, I would look into playing with the EDC, TDC and PPT settings, I would only mess with that in the AMD Agesa section in the bios rather than the other spots or Ryzen Master though I don't recommend it, Most boards do have the same settings for these in 2 or even 3 spots and they will conflict sometimes or default to whatever the Agesa sections sets it to.

EDC, TDC and PPT are all power tables that can be adjusted, I would still play with the curve optimizer, any little bit will help, can save power with minimal if any loss of performance but run cooler in return, but you will have to play with it, someones setting on the internet might not work ideally for you. But its free tuning, why not give it a go, I find it fun. Still tweaking with the Curve optimizer on my 5900x and 5700G system and I've had them for over a year now.

Or you can leave it, the 85C is quite normal for these chips, they are designed to boost aggressively up to 90C, Once that 90C is reached it will slowly downclock to base speeds if temps don't improve, but if it remains at 90C at base speeds, it will throttle until it cools down or worse shut down the system.

 

[boagz]

Thanks for the input.

Or you can leave it, the 85C is quite normal for these chips, they are designed to boost aggressively up to 90C, Once that 90C is reached it will slowly downclock to base speeds if temps don't improve, but if it remains at 90C at base speeds, it will throttle until it cools down or worse shut down the system.

Ya, this is what I was trying to figure out I guess. Just how aggressive these CPU's will boost their clocks and if these sort of 30-40 degree spikes were typical. I guess part of the really annoying thing is with Windows since my CPU seems to randomly spike to 70 even 80 degrees when not doing anything but only every once and a while. I think it has to do with some background programs windows is running for a short period of time that for some reason spike my CPU temps (even though CPU utilization is only %10 or lower usually when I check it in task manager).

 

[tennis2]

I would look into playing with the EDC, TDC and PPT settings,

PPT is the one that will reduce your temps since that's the PBO power limit.

On that note, I do wonder what your "CPU package power" is maxing out at currently under load.

Also, what GPU?

 

[KyaraM]

Small cases and powerful internals very often lead to high temps. However, in your case there is nothing out of the ordinary, temps are where you would expect them, so no reason to worry.

Btw, I'm pretty sure you mixed Celsius (PC temps) and Fahrenheit (room) there. Please stick to one as far as possible. But at the very least, mark them appropriately (40°C instead of just 40) so people can more easily differentiate what you mean.

 

[boagz]

PPT is the one that will reduce your temps since that's the PBO power limit.

On that note, I do wonder what your "CPU package power" is maxing out at currently under load.

Also, what GPU?

I have a GTX 1660 Super

Small cases and powerful internals very often lead to high temps. However, in your case there is nothing out of the ordinary, temps are where you would expect them, so no reason to worry.

Btw, I'm pretty sure you mixed Celsius (PC temps) and Fahrenheit (room) there. Please stick to one as far as possible. But at the very least, mark them appropriately (40°C instead of just 40) so people can more easily differentiate what you mean.

Cool. And sorry, my bad. I'll try and remember that for the future

 

[drea.drechsler]

Thanks for the input.



Ya, this is what I was trying to figure out I guess. Just how aggressive these CPU's will boost their clocks and if these sort of 30-40 degree spikes were typical. I guess part of the really annoying thing is with Windows since my CPU seems to randomly spike to 70 even 80 degrees when not doing anything but only every once and a while. I think it has to do with some background programs windows is running for a short period of time that for some reason spike my CPU temps (even though CPU utilization is only %10 or lower usually when I check it in task manager).

Do be aware it's considered normal for Ryzen 5000 CPU's to run upwards of 90C in heavy work...AMD's said this is to be expected and it's by design. When you watch and read what Robert Hallock is saying it sounds like it's pretty much temperature seeking: it keeps boosting till it see's 85-90C as long as it's also within the other boost parameters (PPT, TDC, EDC).

That's why you can use PPT, TDC and/or EDC settings to tweak temperature, although possibly limiting performance at the same. The point is to keep in mind is the processor is designed to run as high as 90C so don't be fearful of seeing it do so in extreme workloads.

That also seems to affect how Curve Optimizer works: because the boost algorithm is temp seeking I don't think it exactly results in lower temperature operation even though it does lower voltage at the high end of the V-F curve. It should, however, be able to perform better because it will stay boosted to a higher clock longer before reaching that temperature because the voltage is lower. Assuming, of course, you've set up your PPT, TDC and EDC to allow it to.

And yes, Ryzen is a very 'boost happy' processor since it eagerly boosts single cores at a time to max clocks to dispatch transitory workloads and return the core to sleep. When it does you might see a sudden rise or even spike in temperature that just as quickly drops away. Windows is very active even when the system seems idle so it happens pretty frequently. Each temp spike is pretty minor in terms of heat load so you have to watch average core temperatures to get a better idea of the thermal state of the CPU.

 

[tennis2]

And that affects how Curve Optimizer is working: because it is temp seeking I don't think it exactly results in lower temperature operation even though it does lower voltage at the high end of the V-F curve. It should, however, be able to perform better because it stay boosted to a higher clock longer, before reaching that temperature.

Even assuming "steady state" performance. A lower vCore = less power draw = lower temp at the same frequency. You're right though, the boost function operates on temp, power, current, and frequency limits (HWinfo64 shows these). Because curve optimizer lowers power and temp at a given frequency, the boost algorithm may likely see that as additional headroom to push frequency higher until those limits are met again (likely the temp limit in this scenario). That's why PPT is the easiest(?) way to actually lower temps.

My use of curve optimizer ended up lowering the temp, power, and current limits such that the frequency limit is controlling, so I actually did reduce power/temp without having to touch PPT.

@boagz Did you try removing your cooler (inspect thermal paste spread), clean off thermal paste, reinstall thermal paste, and remount the cooler?

 

[drea.drechsler]

... That's why PPT is the easiest(?) way to actually lower temps.

Just keep in mind it's not free but also reducing performance potential. I prefer tweaking down EDC with my 5800X, it seems to result in improved sustained performance through a video rendering while also keeping temperatures in check. There does seem to be some minor loss in light, bursty processing but that's really very hard to bench.

But the point is to remember that hitting (even) 90C isn't bad as AMD's designed it into the CPU...especially if not seen except in extreme and rare work loads. And also to look at average core temp to see the true thermal load of the processor, not the temp surges and spikes as transitory processes come and go but alone are really very light thermal loads. So don't go killing the performance you paid for with the processor by shooting for arbitrarily low temps.

 

[tennis2]

^ absolutely.

 

[boagz]

@boagz Did you try removing your cooler (inspect thermal paste spread), clean off thermal paste, reinstall thermal paste, and remount the cooler?

I have not. Partly because it sounds like that really isn't the issue the more you guys tell me about this processor and how it works. I guess overall the temps aren't really that bad considering my setup/ambient room temp and how this CPU boost technology works? It's also not the easiest to remount the cooler in this case/build and hard to see how I'm lining things up to screw in the cooler.

 

[tennis2]

Never know.

Another good test would be to remove the side panel and run Prime95 small FFT. That would mostly eliminate the case airflow from the equation