Question Ryzen 2700x High Spiking Idle Temperatures.

[heslipsean]

Hi, I've had the 2700x for about 3 months now and I'm little confused with the idle temperatures. The temperature seems to start at around 45-50C, which is already pretty high for an idle temp, but then spikes continuously up to 60C. The temperature will then drop back down just below 50 then spike up to 60 again immediately, this happens on about a 15 second cycle. I'll link my MSI Afterburner file below so you can see what I'm talking about. Also, the temperatures do not even increase that much under load, the CPU stays at the 55-65 mark when playing games, never goes any higher.

https://drive.google.com/open?id=19IDrEF-pyZSU_wh9NOREVth1tA3_b05E

My BIOS is at stock settings other than my DOCP for RAM. However, when I use the "EZ Tuning" feature to automatically overclock the CPU, this problem goes away. I believe this has something to do with the built in precision clock boost feature. Using the EZ tuning gives me much nicer idle temperatures, but limits my clock speed and voltage significantly in comparison to the stock settings.
The motherboard is the ASUS X470 Prime Pro

Its probably worth mentioning that my games aren't running great either, CS:GO, Rocket League and the Division 2 all run at a very high FPS but stutter very frequently, which is quite off putting for me.

Any help would be greatly appreciated. Thanks.

Specs:
ASUS Prime X470 Pro mobo
ASUS Strix RTX 2060 GPU
Ryzen 7 2700x with stock cooler.
Corsair TX650m PSU
2xSSD 1xHDD
16GB DDR4 3000mhz RAM
Corsair Carbide 400C

 

[mtracy1991]

Did you install the CPU correctly with the latch placed down?
I have the blackhawk ultra case with about 12 fans and it doesn't go any higher than 60C when gaming whereas it sits around 30-40 idle. Is your room hot? What is your intake & exhaust setup within the case?

List your specs as well for your system.

 

[heslipsean]

Did you install the CPU correctly with the latch placed down?
I have the blackhawk ultra case with about 12 fans and it doesn't go any higher than 60C when gaming whereas it sits around 30-40 idle. Is your room hot? What is your intake & exhaust setup within the case?

List your specs as well for your system.

Yes the CPU is seated correctly and the case is the Corsair 400C with an intake and an outtake fan, the temperatures are fine when they are not spiking, the temperature begins to fall down towards the 35 mark and then spikes up into the 50s every few seconds, I can also audibly hear the cooler ramp up when this happens, as if I had just started a game or something.

https://drive.google.com/open?id=1q0oQ75j4IC4J6oQOMS_Nuv5RGQyjVKuy This is the temperatures with the EZ Tuning settings applied, the temperatures are much more normal although there is still the occasional small spike

EDIIT- Games still do not run great with these settings applied though. I cant win really

 

[heslipsean]

I managed to band-aid fix this by using the "power saver" power plan, and then switching back to balanced or high performance whenever I need to do anything strenuous. Power saver seems to be the only thing that will stop these spikes, I have found nothing else so far.

 

[mdd1963]

any chance these spikes are merely occurring with assorted routine WIndows (or an installed application) housekeeping duties, such as downloading/installing updates, background file indexing, AV canning , etc? (YOu'd need to monitor HWMonitor and/or task manager to know when the system is truly idle....; lots of little tasks happen for several minutes upon turning on a system, especially if anywhere near the 2nd Tuesday of a month, or a Creators/semi-annual update, etc... (both of which recently apply)

 

[NightHawkRMX]

Those spikes are not going to cause any harm.
Just make sure they are under 80c under load
Likely things in the background are causing this, whats idle cpu usage?

 

[junglist724]

First, all Ryzen CPUs have 2 temperature readings, Tctl(T Control) and Tdie. Tctl is actually meant to be the primary temperature reading, but Tdie is your actual cpu temperature. On many Ryzen CPUs Tctl is actually Tdie plus an offset, for the 2700X Tctl = Tdie + 10C and I think the cpu temperature read by Afterburner would actually be Tctl so you should actually subtract 10 degrees from whatever that reads to get your actual temperature. I think AMD did this because they don't want servers or prebuilt workstations to throttle or overheat because of the really simple fan controls in a generic BIOS that never expected an 8-16 core desktop cpu or a 32 core server cpu to exist. My 2700x system is currently running FreeNAS so I can't check afterburner, but I would definitely install HWinfo64 and check the Tdie reading to be sure. HWinfo64 is hands down the best hardware monitor, unlike hwmonitor which doesn't even monitor BCLK.

Precision boost changes clocks and voltages up to every 1ms, which is 1000x faster than msi afterburner even checks sensors by default, and even at the lowest 100ms sensor check interval you still don't really see the whole picture temperature/voltage/frequency wise. GloFo's 12nm process kind of sucks and requires up to 1.55v to hit the max 4.35GHz lightly threaded boost clocks so naturally temperature will spike when it does that. Plus the temperature reading is showing you how hot the hottest core is running at. Also don't be alarmed by the massive amount of voltage required for the 1-2 core boost clocks because it's safe when only used on a couple cores at lighter loads and precision boost knows that.

It's basically impossible for precision boost to reach unsafe temperatures because it will throttle back clocks/voltages within 1ms latency, whereas with a fixed OC like with EZ Tuner it's totally possible to slowly kill your chip.

You always want to be running the windows balanced power plan with Ryzen 2nd gen, since precision boost can vary voltage(to a degree) and frequency on a core by core basis. If you want the higher boost clocks when you're using fewer cores, the rest of the chip needs to be able to downclock to give those cores headroom to boost higher. With the high performance power plan all your cores will likely never boost past 4.2GHz even with a single core load since the other 7 cores are trying to idle at over 4GHz, when they could be dropping to 2.2GHz.

 

[rigg42]

First, all Ryzen CPUs have 2 temperature readings, Tctl(T Control) and Tdie.

I believe this is inaccurate. This appears to be on a model by model basis. My 2600x and 1600 have the "CPU (Tctl/Tdie)" value instead of a separate value for each. According to the HWiNFO Author this means the CPU has no offset.

https://www.hwinfo.com/forum/Thread-Is-CPU-Tctl-value-is-still-a-Tdie-value-on-Ryzen

I believe only certain X chips have this offset to keep fan speeds up and temps safe for PB/XFR. It seems strange that the 2600x doesn't appear to have an offset though. You'd think it would being an x chip and all. It will ramp up to the mid 90's c under stress load with stock cooler/settings. The cooler really should be inadequate at full stress load with the stock boost settings though. Every piece of monitoring software I've tried shows the same temp as the "CPU (Tctl/Tdie)" value shown in HWINFO64. I have a lot of experience overclocking different Ryzen CPU's and the 1600 and 2600x that I have show nearly identical temps at 1.2 vcore at similar clocks using the exact same case and surrounding components. 3.9 for the 2600x and 3.8 for the 1600. Both read high 70's peaks in real bench with those settings.

 

[junglist724]

I believe this is inaccurate. This appears to be on a model by model basis. My 2600x and 1600 have the "CPU (Tctl/Tdie)" value instead of a separate value for each. According to the HWiNFO Author this means the CPU has no offset.

That's why the next, next sentence says:

On many Ryzen CPUs Tctl is actually Tdie plus an offset, for the 2700X Tctl = Tdie + 10C

And not ALL Ryzen CPUs. If you don't have a cpu that has an offset then you don't have a way to check if other monitoring software is using Tctl or Tdie for their general "cpu temperature" reading because for you they are one and the same.

 

[Texas_NightOwl]

My question is, why with very little software installed beyond MS Office 2010 and Gigabyte's motherboard software (and all processes killed) is Tctl/Tdie, which matches CCD 1 & 2 Tdie except during a spike, jumping by 10 to 15 degrees Celcius for a second every few seconds?

Even if there is something in the chip that COULD actually jump in temperature that fast, there is no point in having the fan speed up over it because it would take minutes before the heat could be carried away. So why not use the max of CCD 1 & 2? Or use a moving average of Tctl so that we don't have to be ANNOYED with a loud fan every 5 seconds?

Is it possible that the spike in Tctl is a defective CPU?

 

[Flayed]

Just add another intake fan, you may as well if you are concerned about temps

 

[drea.drechsler]

I believe this is inaccurate. This appears to be on a model by model basis. My 2600x and 1600 have the "CPU (Tctl/Tdie)" ....

This is true... I think the offset changed starting for some of the Ryzen 2000 chips. It's really a bit confusing, but HWInfo64 can be relied on to accurately reflect temperature reports from the CPU (since it reports both Tctl and Tdie) so I only use it.

Also, Zen processors are known to spike temperatures frequently. This is especially true of Zen 2 because of the very small thermal mass of the separate 7nm CPU dies. The spiking is a result of the Zen boost algorithm, something AMD has called a 'rush to idle'. When a processing load comes along it will boost to max clock (and high voltage for stability) for a few mS then pull clock (and volts) back quickly to finish it and go back to idle. Since there are a lot of threads constantly running in modern OS's somethings always waking up and in need of CPU time so it accomodates if it can.

 

[InvalidError]

I think AMD did this because they don't want servers or prebuilt workstations to throttle or overheat because of the really simple fan controls in a generic BIOS that never expected an 8-16 core desktop cpu or a 32 core server cpu to exist.

Which was a pretty dumb idea on AMD's part. You do realize that before a PC or server will boot on a chip newer than what was planned back when that BIOS was current, the BIOS needs to be flashed to support the new chips first, so whatever fan speed controller offsets may have been required could simply have been lumped in at that time instead of fudging reported numbers. IIRC, AMD got rid of offsets on most chips with an AGESA update to reduce the amount of unnecessary confusion it created.

 

[Texas_NightOwl]

Just add another intake fan, you may as well if you are concerned about temps

This thread is about temperature spikes. The CCD chip temperatures don't change. It is the Tctl temperature that spikes 10 - 15 degrees (18 to 27 F) for a fraction of a second. NO COOLING system will handle that!!! It drops back down to the CCD chip temperature within a second and water nor air will do anything about a 1 sample spike. However, it causes fans to loudly jump in speed for a couple of seconds. You can set stupid fan curves to compensate, but that means that your fan is running loud all the time because you set to 60% for temps from 25 to 80 degrees, or you set for a lower % over those temps and your CPU runs up to 80 pretty fast and you have the fan spiking again unless at 80 degrees you set fan to very loud 100% speed.

Most people don't adjust their fan curves and don't seem to have a problem with fan spiking every few seconds. AMD is likely to have Tctl use a moving average because 1) increasing fan/pump speed will do nothing for a 1 ms spike in temperature and 2) it would cause the problems we are having for everyone.

So I am guessing either defective AMD CPUs have a needlessly spiking Tctl or defective motherboards/BIOS are somehow corrupting the Tctl value they read from the CPU.

 

[NightHawkRMX]

I haven't had any of the issues with random fan ramps above with my 1200, 1600, or 2600.

Keep windows in Ryzen balanced mode and latest bios revision.

 

[Texas_NightOwl]

I haven't had any of the issues with random fan ramps above with my 1200, 1600, or 2600.

Keep windows in Ryzen balanced mode and latest bios revision.

Ryzen balanced mode or even power saver mode doesn't prevent it. I think it is either defective CPU or defective motherboard.

 

[drea.drechsler]

Ryzen balanced mode or even power saver mode doesn't prevent it. I think it is either defective CPU or defective motherboard.

It's not defective, it's by design.

Watch this for some excellent tech background on what's going on and why it's perfectly normal.

View: https://www.youtube.com/watch?v=iZI9ZgwrDYg

It can be annoying when it makes fans pulse as he complains about in this video, also providing more tech background on the subject.

View: https://www.youtube.com/watch?v=ORHYffg5ipM


The way to 'fix' it with fans is to set a curve that ignores the temperature pulses at idle.

 

[Schlachtwolf]

I have a 2700x and a case modded with great airflow and it idles at 32 and spikes to 42, sometimes higher or lower and my observations are as follows

1. Windows, Steam, CC cleaner etc etc are always doing background stuff so it may look like the PC is doing nothing but it is. Turn everything off and after a few minutes it idles at 32c.

2. Cores have to be woken up to get a temp reading.... spike.

3. Gaming is no more than 65c...thats what really counts, what happens under load.

4. Don't worry about it, under 85 nought is going to happen.

5. If it really bothers you get a better cooler, the Wraith is good but my temps went down by 10c with better airflow and a 240mm AIO.

Cheers

 

[Texas_NightOwl]

It's not defective, it's design.

The way to 'fix' it with fans is to set a curve that ignores the temperature pulses at idle.

If it is not a defect, then why don't most people have the problem?

You are seriously wrong if you think most people go into their BIOS and set a fan curve that says from 50 to 75 degrees have a fixed fan speed so their fan doesn't speed up every 10 seconds when the Tctl jumps from 50 to 65 for a fraction of a second. In fact, I bet few people even touch their fan curves. Yet, you don't hear them complain about the loud fan noise of the AMD Wraith Prism fan. (I will watch your videos though since I like buildzoid.

I have a 2700x and a case modded with great airflow and it idles at 32 and spikes to 42, sometimes higher or lower and my observations are as follows

1. Windows, Steam, CC cleaner etc etc are always doing background stuff so it may look like the PC is doing nothing but it is. Turn everything off and after a few minutes it idles at 32c.

2. Cores have to be woken up to get a temp reading.... spike.
....
5. If it really bothers you get a better cooler, the Wraith is good but my temps went down by 10c with better airflow and a 240mm AIO.

Since it was a new install it was easy to just zap the SSD and completely reinstall Win 10 Pro. Without any software installed the spikes were there. So can't blame Steam, etc. And just about everyone runs Windows yet don't have the problem. Not sure cores have to be woken up to read temp, but even if they do it is done by BIOS in order to apply the fan curves and happens for all CPUs and motherboards. Besides AMD would have to know that was going to always happen and build their chip so it doesn't cause a 10 degree temperature jump or use a moving average to average out those spikes.

I tried all power plans after the OS reinstall. Only changing Maximum Processor State from 100% to 99% gets rid of the problem. But that also keeps any of the cores from going over the CPU base frequency.

 

[drea.drechsler]

If it is not a defect, then why don't most people have the problem?

You are seriously wrong if you think most people go into their BIOS and set a fan....

I can't say anything about what 'most people' say think or do and don't pretend to speak for them, all I can say with any certainty is what's worked well for me and the systems I've worked with. I've seen the temp pulsing in every one and those where it pulsed fans I've set curves that ignored it, simple as that.

If you're so sure you've a defective processor then the thing to do is call AMD and get it an RMA. I'm not sure what you're trying to accomplish here.

 

[Texas_NightOwl]

If you're so sure you've a defective processor then the thing to do is call AMD and get it an RMA. I'm not sure what you're trying to accomplish here.

I have messaging with Gigabyte tech support (because at first I thought they were corrupting the Tctl values that they were reading) and am still waiting for response from AMD since I contacted them yesterday when I found that people have been having this problem with other brand motherboard boards and other AMD CPUs over the years.

 

[jon96789]

I have a AMD 3900X. When running in Balanced or Power modes, the CPU will idle at ~50 degrees and spike to 65 degrees. I noticed that the voltage is pretty high, ~1.487 volts.

When the system is in Power Saver mode, the idle temps drop to 38-40 degrees and remain there until a load appears. In Power Saver mode, the voltage drops to 0.9-1.2 volts.

So i think that what you are experiencing is normal for the new Ryzens.

 

[Texas_NightOwl]

I have a AMD 3900X. When running in Balanced or Power modes, the CPU will idle at ~50 degrees and spike to 65 degrees. I noticed that the voltage is pretty high, ~1.487 volts.
...
So i think that what you are experiencing is normal for the new Ryzens.

Maybe so, but why aren't more people complaining about the fan RPM constantly changing? AMD must know it too, so why aren't they using a moving average on Tctl so that it doesn't spike. After all, there is NO way ANY cooling system is going to lower a 15 degree temperature spike in under 1 second (the amount of time it is generally elevated by 15 degrees). So why not gradually increase Tctl over a 5 or 10 second period if the chip really is heating. That would eliminate the spiking entirely.

Also, the motherboard makers would know it. So why are they providing only 3 to 5-degree tolerances away from the fan curves before adjusting fan speed? Why aren't they doing what AMD should be doing and either moving average the temperature or requiring the temperature spike last a few seconds before adjusting the fans?

 

[Texas_NightOwl]

I'm not sure what you're trying to accomplish here.

One thing I did find out about half an hour ago is that the AMD Wraith Prism fan that my 3900x uses has a "fan overclocking control" and what it does is changes the max speed from a fast 2,650 RPM, to a finger chopping 3,600 RPM. Mine came with that switch in that position. It was especially loud because it was increasing its speed about 1,500 RPM with each spike. I put it in the low-speed position and I can still hear it surge every 10 seconds or so, but it is not nearly as loud. Almost tolerable.

 

[Flayed]

Having the fan speed up and slow down all the time sounds really annoying. Setting a fixed fan rpm sounds like the way to go, or possibly getting a better heatsink/fan or both.