cancel
Showing results for 
Search instead for 
Did you mean: 

Processors

Re: Ryzen 10°C temperature fluctuations / spikes when running idle

I have a 1700x with a Noctua NH-D14 (A nice cooler, if you're looking for a replacement to the AMD stock cooler ... mind you, get the D15), and I do not get this behaviour you speak of.  Maybe you didn't get good thermal contact with your cooler?  The thermal capacity of the cooler should soften that temperature spike.  Are you using the AMD Ryzen Balanced plan?  Not even sure if that's recommended anymore, or if it will make a difference, but I've been using it since I got my 1700x.  Also, BIOS updates might help, namely the 1.0.0.4 update coming soon, especially since it comes with some microcode updates.

I agree the default fan profiles are asinine (at least on my Gigabyte B350 they are), but CPU temps can spike fast, and the motherboard can't predict if that temp-spike is fleeting. This is not an AMD only thing, try using any Intel CPU that comes with a boxed cooler, and you'll see some nutty behaviour with that fan.

0 Kudos
hardcoregames_
Big Boss

Re: Ryzen 10°C temperature fluctuations / spikes when running idle

My CPU came with the wraith cooler, I used MX-4 with it and my CPU is very thermally stable with it.

Even when a CPU bound game is played my cooler never gets very warm at all.

texasnightowl
Adept I

Re: Ryzen 10°C temperature fluctuations / spikes when running idle

That is why I suspect it is either a defect in the CPU (of people that have the problem) or in the motherboard that is reading the Tctl wrong.  Good thermal contact might cause high CPU temperatures, but not spikes of 15 degrees C (27 degrees F) that last less than a second. (BTW, the cooler that comes with the CPU has the thermal coating pre-applied.) There is no cooler that can soften spikes because the temperature drops before any heat can be transferred from the chip internal to the chip external. 

I, and many others, have tried all the plans.  With the 3rd Gen Ryzen they recommend using the Ryzen plans because it samples every 1 ms instead of 15 ms.  BTW, if you set Maximum Processor State from 100% to 99% in any of your power plans it will cause the CPU to run about 10 degrees cooler and get rid of the spikes.  But your processor will never run faster than its base frequency.  No automatic speed boosting of any cores.  A lot to give up just to keep fan from loudly varying speed.

Yep, I just updated BIOS the day it came out (day b4 yesterday I believe).  People would be screaming if everyone's fans were to be going up and down.  Also, there are people that didn't have the problem that suddenly get it.  Again, possibly pointing to a defect.

The default fan curves aren't nutty if the processor were showing say the max of CCD 1 & 2 temperature (which are stable with workload) or were to be averaged so that the Tctl temperature doesn't spike (because, again, there no reason to try to increase cooling for a temperature change that is only there for 1 second or less...nothing is going to lower the temperature in under 1 second and the temperature will be back to its lower value within 1 second anyhow).

Re: Ryzen 10°C temperature fluctuations / spikes when running idle

There ARE coolers that can absolutely deal with those temp spikes, this is not debatable; my aging Noctua easily does, as does hardcoregames' Wraith using good thermal compound. In those temp-spike situations, the CPU is probably only drawing (for example) 25w on each of two cores for a short period ... coolers can pretty easily deal with 140+w of heat for extended periods of time, meaning those 25w spikes are peanuts, and that the thermal-barrier is obviously capable of conducting 140w+ of heat ... all of that depends on heat transfer efficiency (CPU to cooler via thermal paste) and thermal capacity (of the cooler itself).  While it won't change a 15°C spike into a 1°C spike ... it might turn it from a 15°C spike into an 8°C spike.  The thermal capacity of a relatively-cool CPU cooler is considerably more effective for dealing with temp spikes than the fan speeding up ever will be, hence why "big" coolers are good (and why there is a market for them, as well as high-quality thermal paste).

Now, regarding thermal paste: just because it is pre-applied, does not mean you will get always get good contact.  I'd say, in general, the pre-applied stuff is a "safe" option, leaving the most room for installation error, but it can still work poorly.  For the $10 (max) it costs for high-quality thermal paste, it can't hurt to re-do it to see if you get better results.  At best, the pre-applied stuff isn't fantastic from a thermal-conductivity perspective.  At worst, during installation there was some excessive rocking, and the paste didn't apply well.  Just saying, and if you're not getting near your 4.6GHz boost, I'd say this is plausible (much more plausible than a robot screwing up soldering on the TIM).

FWIW: there is a new  power plan put out by 1usmus; not that it will resolve the issue you're experiencing, but it might be another piece of the puzzle (and "his" claims, as well as testimonials, are that things are just generally better thermally).  The plan "requires" you to set a few things in your BIOS (the one you just installed, assuming the manufacturer implemented the features).  Also, Windows 1909 should be installed to help with core hopping (if you're not afraid of a 0-day Windows patch  ).

While I don't know your specific board's implementation, I beg to differ on the fan profiles generally being acceptable; I was "lucky" not to have gotten a boxed cooler with my 1700x, but I've built a lot of Ryzen PCs (50-ish), and with the stock Wraith coolers, the fan profiles are almost always unacceptable.  Also, you have the "best" Wraith cooler (most of my builds are lower-end Ryzen CPUs); lower-end CPUs have less capable coolers.  Also note, your manufacturer's default fan profiles might actually be good, my Gigabyte ones for my board are not.  The Wraith coolers themselves are actually surprisingly high quality, but the profiles don't match the lower-end coolers well (and, again IMO, the stock thermal paste is lacking).

Edit: As for "why spin up the fan for a 1 second long spike?"  Because the motherboard doesn't know if that 1s spike is going to last 1s, or 20 minutes, and has to get on top of it ASAP, especially since the motherboard doesn't know if you threw a blanket over your computer, or if your dog laid on top of your intake, or if your cooler is a liquid-system connected to your 4 GPUs that you just started mining for Dogecoins, etc etc.  You're unhappy with a 1s spike ... but what about a 2s spike? 5s?  You can adjust how much of a spike you are willing to accept with fan profiles, but by default, it makes sense that they respond to imminent "danger" ASAP.

0 Kudos
texasnightowl
Adept I

Re: Ryzen 10°C temperature fluctuations / spikes when running idle

The fan profile I set up, for now, is 20% for temps 0 to 40 degrees, linear increase to 45% at 42 degrees (Gigabyte doesn't support higher slope than that), then flat to 72 degrees, then linear increase to 100% at 78 degrees.

A better thermal contact may facilitate faster heat transfer from the chip even in 1 second.  I won't argue that.  But a water cooler is not going to get colder water to the CPU and a CPU fan is not going to remove extra heat from the heat sink in 1 second.    So waiting 2 or 3 seconds to react, or using amoving average of that duration, won't substantially affect continued load handling, but it would eliminate all the spikes I have seen. 

Using fan profiles to smooth out the spikes requires that you run your fan faster (louder) all the time or your CPU hotter all the time because over expected idle (except chrome open with 20+ tabs) operation ranges (48 to 58 degrees, I live in Florida where we don't keep our house 21 degrees) you need to allow for 15 degree spikes.  Hence 58+15 =  73 degrees.

0 Kudos

Re: Ryzen 10°C temperature fluctuations / spikes when running idle

I think you're greatly underestimating how fast a heatpipe transfers heat; it's shocking. As soon as there is a temp differential between the TIM and the bottom of the cooler, thermal conduction starts happening.  You're saying ~1s spikes result in ~15°C increases ... that means that at 0.1s it's up by ~1°C, and at that point, the cooler starts "pulling" heat from the TIM. 0.5s later, the TIM would have been at ~8° hotter, but the cooler has been pulling heat away already, so maybe it has only gotten to 6°C, and now the cooler is transferring heat even faster (larger temperature delta). Maybe by now the fan is ramping up to deal with the heat that is getting to those fins, or maybe not.  I don't think you're understanding how fast heat transfer is in a heatpipe.


AMDs decision to base CPU cooler fan speeds on the temperature of the surface the cooler is attached to is a logical one.

-= Start Quote =-
Using fan profiles to smooth out the spikes requires that you run your fan faster (louder) all the time or your CPU hotter all the time because over expected idle ...

-= End Quote =-
... or get a more effective cooler and/or better thermal paste, and run your fans slower and your CPU will run cooler.  I honestly believe re-doing your thermal compound will greatly reduce the issues you're experiencing.

I know it's not a good comparison, but my chip should be notably worse than yours: With my 1700x (which runs hotter idle than your 3900x ever will), using Noctua's thermal compound, the CPU fan is at 521RPM (near stall, and inaudible) basically all the time (currently, with AutoCAD running, 20 tabs in Opera, Thunderbird, MS Outlook, Excel, and random other crap idling ... the CPU is at 37°C). This room is 22.5°C (smart flood sensor, old timey thermometer, and laser temp gauge all agree), the case temp (TMPIN0) is currently 34°C, and my 1070 GPU is at 50°C.  When the CPU gets to 45°C or something like this, the CPU fan linear ramps up to max RPM (IIRC it is 1,200RPM, which my computer never gets to).  Spent 20 minutes running OCCT, so 100% loaded CPU for ~20 minutes (CPU stock), the CPU fan is at 983RPM, case is 36°C, and the CPU peaked at 65°C.  I would expect if I installed your CPU in my case, with my cooler, it would result in similar performance (Source: https://old.reddit.com/r/Amd/comments/ceuay1/shoutout_to_noctua/eu5crih/ ).

Also, an "idle" temp of 48°C is not fantastic for that CPU, even with the stock cooler, depending on case cooling (TMPIN0 is a good indicator of general case cooling).

Not trying to be adversarial, just letting you know the issues you're having might be easily resolvable, with cheap and easy solutions (or slightly less cheap fancy coolers).

0 Kudos
texasnightowl
Adept I

Re: Ryzen 10°C temperature fluctuations / spikes when running idle

I live in Florida and sit under a ceiling fan.  I feel perfectly comfortable at 29 degrees (house thermostat in another room so probably 30 or so in my office). That would account for about 10 degrees in the CPU temperature. I didn't see any plastic over the stock cooler's thermal pad.  I might pull it and just replace the pad with compound since I have since heard that it can turn hard and actually destroy the chip if you try to take it off without warming up the chip first (old wives' story?). 

0 Kudos

Re: Ryzen 10°C temperature fluctuations / spikes when running idle

29°C? OMG wow, I would be dying.  I do appreciate you using Celsius though; this Canadian appreciates it .

Anyhow, you won't wreck the CPU if it's not warm, but it can't hurt either.  That might be the case with a thermal adhesive, or even thermal tape, but not thermal paste.  And you most definitely didn't leave the plastic cover on ... your CPU would basically instantly overheat.

As I'm lifting, I kindof turn the cooler a bit; probably not necessary, but again, doesn't hurt anything (unless your cooler's bottom surface isn't flat).  Will be nice to see what result you get refreshing the thermal paste (please come back and let us know).

0 Kudos
texasnightowl
Adept I

Re: Ryzen 10°C temperature fluctuations / spikes when running idle

This jpg shows what I was trying to indicate would be desirable.  Apparently, I will have to return my Gigabyte X570 Master and get the MSI X570 Ace to get it.FanSpeed.jpg

0 Kudos
o_mally
Journeyman III

Re: Ryzen 10°C temperature fluctuations / spikes when running idle

Just disables turbo boost
0 Kudos