Ryzen 9 3950x not boosting first 4 cores

[Seneca]

Hi there.

 

I recently upgraded from an R7 2700X to an R9 3950X. Main purpose of my computer is CPU rendering with V-Ray, so I was really happy at first when I was getting 4.3GHz automatic boost clocks while doing light work and 4GHz all-core when rendering--at incredibly lower voltages too comparing to my 2700X. Temps won't go above 60 degrees C while all core rendering at 4GHz, which is really impressive.

 

I intended to keep clocks unlocked and let the computer control the optimum speed for the workload, but today I was noticing some sluggishness while doing basic administrative work like emailing and saving files on Photoshop. I fired up HWMonitor to see what was going on and noticed that all cores were boosting randomly, except for the first 4 (core 0 through 3) that were stuck and wouldn't go past 533 MHz (??). This is not the first time I noticed this behavior with this new CPU. Last time a simple restart took care of the problem. However, I can't afford to stop everything I'm doing to restart the computer every time this happens.

 

For now I quickly locked the clock speeds to 3.7Ghz in the BIOS just so I can keep working. The "snappiness" is back, but I would REALLY rather leave the clock speeds automatic to get the full benefits of having a 3950x.

 

So, is there a way to let the CPU boost automatically, but not let it go below a certain clock speed? Does anyone know what might be causing this issue of a few cores being "stuck" at low speeds?

 

I'm a bit of a noob when it comes to knowing what all the settings and knobs do in the BIOS.

 

My mobo is an Asus ROG Strix X470-f, BIOS is updated to the latest version, cooler is a Cryorig R1 Ultimate. Windows power mode is set to performance. "Asus Optimal" does exactly nothing in the BIOS for me (under EZ System Tuning).

 

Thanks in advance!

FEB 2021 UPDATE: There was a problem with my CPU. It stopped working suddenly after around 6 months of use. Computer would not turn on whenever RAM was in dual-channel mode. Tested the CPU in multiple motherboards and DIMMs to confirm that the problem was indeed with the CPU. I sent it back to the store where I bought it and they confirmed the issue and sent me a new R9 3950X. So if you found this on Google because you're running into the same issue, I recommend you RMA your CPU right away. Hope this helps.

Haven't run into the same issue with my new Ryzen 9 3950X yet!

Edited March 3, 2021 by Seneca
Providing update for people who might run into the same issue in the future

[boggy77]

use hwinfo, not hwmonitor. try to reset cmos. 

[.Apex.]

27 minutes ago, Seneca said:

I intended to keep clocks unlocked and let the computer control the optimum speed for the workload, but today I was noticing some sluggishness while doing basic administrative work like emailing and saving files on Photoshop. I fired up HWMonitor to see what was going on and noticed that all cores were boosting randomly, except for the first 4 (core 0 through 3) that were stuck and wouldn't go past 533 MHz (??). This is not the first time I noticed this behavior with this new CPU. Last time a simple restart took care of the problem. However, I can't afford to stop everything I'm doing to restart the computer every time this happens.

This is normal behavior as the CPU will underclock itself depending on load to keep the power consumption at a minimum, if you do a full core load you'll notice all of the cores will jump to the all core boost, which is lower than the single core boost, if you want to prevent the CPU from dropping the cores that low you could increase the Minimum Processor State in the Power Plan options

[.Apex.]

2 minutes ago, Syn. said:

This is normal behavior as the CPU will underclock itself depending on load to keep the power consumption at a minimum, if you do a full core load you'll notice all of the cores will jump to the all core boost, which is lower than the single core boost, if you want to prevent the CPU from dropping the cores that low you could increase the Minimum Processor State in the Power Plan options

The reason some cores were locked to 533mhz is because there was no load on them, it's a 16 Core / 32 Thread CPU, Photoshop and Emailing would barely take advantage of those cores..

[Seneca]

31 minutes ago, Syn. said:

This is normal behavior as the CPU will underclock itself depending on load to keep the power consumption at a minimum, if you do a full core load you'll notice all of the cores will jump to the all core boost, which is lower than the single core boost, if you want to prevent the CPU from dropping the cores that low you could increase the Minimum Processor State in the Power Plan options

Thanks Syn, minimum processor state was already set to 100%, so although this is closer to the issue, I don't think this setting was the cause.

31 minutes ago, Syn. said:

The reason some cores were locked to 533mhz is because there was no load on them, it's a 16 Core / 32 Thread CPU, Photoshop and Emailing would barely take advantage of those cores..

That would make sense if the computer wasn't noticeably sluggish while these cores were stuck at 533mhz.

 

I'm not the expert here, so it's possible that I'm very wrong, but as far as I know clock speed is independent from CPU load (as in, right now, my PC is idling at 4.3GHz with 0 to 1% utilization, as it should). Unless, of course, clock speeds are lowered to protect the CPU from overheating or over voltage, but I don't see why a CPU would underclock when in low load other than for power saving purposes. Given that the computer is set to high performance plan, it's a bit baffling that the underclocking is taking place.

[.Apex.]

3 hours ago, Seneca said:

That would make sense if the computer wasn't noticeably sluggish while these cores were stuck at 533mhz.

 

I'm not the expert here, so it's possible that I'm very wrong, but as far as I know clock speed is independent from CPU load (as in, right now, my PC is idling at 4.3GHz with 0 to 1% utilization, as it should). Unless, of course, clock speeds are lowered to protect the CPU from overheating or over voltage, but I don't see why a CPU would underclock when in low load other than for power saving purposes. Given that the computer is set to high performance plan, it's a bit baffling that the underclocking is taking place.

Can you use HWiNFO instead? as HWMonitor doesn't get much support and it's unreliable, and there is a specific Power Plan for Ryzen that is called Ryzen Balanced, it's intended to be used instead of the other plans because Ryzen controls its power states differently than Intel, also is there a setting in BIOS that lets you disable all those power saving features? for Intel the power states are called C-states which I think should be the same for Ryzen and there are other Intel specific technologies called Speedshift and Speedstep that also control how the CPU behaves on idle so maybe you could find similar features as they are usually on the same page.

 

Check your temperatures as well using HWiNFO if you think there is throttling involved, make sure to tick the "Sensors-only" box when you run HWiNFO so you could have a similar layout to HWMonitor

 

It seems like AMD's power state/saving technologies are called "CPPC2 (Collaborative Power and Performance Control) and Cool'n'Quiet" along with C-states

[.Apex.]

1 hour ago, Seneca said:

I'm not the expert here, so it's possible that I'm very wrong, but as far as I know clock speed is independent from CPU load

With the boosting technologies that Intel and AMD have the clock speed is very dependent on CPU load, unless you disable those features in BIOS then Windows can't take control of the CPU frequency

[Seneca]

Thanks @Syn. I'll take a look at all the things you mentioned. So far I haven't had the cores get stuck at 533Mhz again.

[Seneca]

Solved the issue!

 

Downloaded the chipset drivers directly from AMD's website as opposed to from Asus's page for my mobo. Asus's supposed AMD chipset driver download is a load of garbage as far as I'm concerned... You download it and unpack a bunch of unintelligible folders with random files and .exes. The chipset drivers I got from AMD actually has one .exe that installs everything you need.

 

That being said, the chipset update installed AMD's Ryzen Power Plan, which I think is what took care of the issue.

 

I believe that's it. I'll report back if it happens again.

 

Thanks @Syn. I also started using HWinfo instead of HWmonitor. It's reporting a higher (and more believable) voltage, so I see now why everyone says HWmonitor is unreliable.

[.Apex.]

3 hours ago, Seneca said:

That being said, the chipset update installed AMD's Ryzen Power Plan, which I think is what took care of the issue.

 

I believe that's it. I'll report back if it happens again.

 

Thanks @Syn. I also started using HWinfo instead of HWmonitor. It's reporting a higher (and more believable) voltage, so I see now why everyone says HWmonitor is unreliable.

Glad you got it figured out

[Seneca]

14 hours ago, Syn. said:

Glad you got it figured out

Spoke too soon  Happened twice in a row. I think I'm going to RMA my CPU before it's too late.

[Seneca]

UPDATE: There was a problem with my CPU. It stopped working suddenly after around 6 months of use. Computer would not turn on whenever RAM was in dual-channel mode. Tested the CPU in multiple motherboards and DIMMs to confirm that the problem was indeed with the CPU. I sent it back to the store where I bought it and they confirmed the issue and sent me a new R9 3950X. So if you found this on Google because you're running into the same issue, I recommend you RMA your CPU right away.

Haven't run into the same issue with my new Ryzen 9 3950X yet!