[SOLVED] CPU not scaling Ryzen 3800x on ASUS ROG STRIX X470-F Gaming

273 · 02-12-2022, 07:41 PM

I'm sorry, I don't think the explanation that it might hit 4.5 when I'm not looking, so to speak, holds any water. If a CPU is supposed to hit 4.5 then, to my mind, it's supposed to do so for at least a few seconds especially when it's powered and cooled properly - if I run a single-thread mprime I expect to see 4.5 for at least a few seconds. My graphics card, for example, can be seen to hit 2020GHz or higher for a minute or so when conditions allow.
The CPU does manage about 4.21 on all cores sustained, which is cool, but there again it seems to throttle slowly down despite the temperature not getting above 70C.
As it stands I have only ever owned desktops with AMD processors, this one was to be the exception but then Ryzen happened and I am, generally, happy but for goodness sakes, why, why can't AMD just actually let me see what I thought I was paying for?

computersavvy · 02-12-2022, 09:01 PM

So according to your ideas this means my 3600 is not performing up to standards, even though it is holding these figures 24/7 @~60C and ~60% load on all threads?

Code:

$ inxi -Cxxx
CPU:
  Info: 6-core model: AMD Ryzen 5 3600 bits: 64 type: MT MCP smt: enabled
    arch: Zen 2 rev: 0 cache: L1: 384 KiB L2: 3 MiB L3: 32 MiB
  Speed (MHz): avg: 4046 high: 4067 min/max: 2200/4208 boost: enabled
    cores: 1: 4026 2: 4060 3: 4042 4: 4064 5: 4047 6: 4067 7: 4045 8: 4040
    9: 4055 10: 4053 11: 4041 12: 4020 bogomips: 86234
  Flags: avx avx2 ht lm nx pae sse sse2 sse3 sse4_1 sse4_2 sse4a ssse3 svm

Or that my intel laptop (i7-9750H) is not performing up to standards (again running 24/7 @ ~75C & 60% load on all threads). Maybe I should be ranting at intel because there is a bigger difference between the max rated speed of their cpu than I see in performance when compared to the AMD cpu with almost identical loads. Note that the 3600 is posting 86234 bogomips vs the i7 @ 62399 bogomips.

At one point it was suggested that you post the output of "inxi -Cxxx" as I just did, but you have failed to even show what the system is actually doing. Instead you just complain about it. Are you possibly exaggerating and don't want us to see that documented?

Code:

$ inxi -Cxxx
CPU:
  Info: 6-core model: Intel Core i7-9750H bits: 64 type: MT MCP smt: enabled arch: Coffee Lake
    rev: A cache: L1: 384 KiB L2: 1.5 MiB L3: 12 MiB
  Speed (MHz): avg: 4143 high: 4202 min/max: 800/4500 cores: 1: 4202 2: 4157 3: 4176 4: 4157
    5: 4148 6: 4144 7: 4154 8: 4035 9: 4081 10: 4100 11: 4198 12: 4173 bogomips: 62399
  Flags: avx avx2 ht lm nx pae sse sse2 sse3 sse4_1 sse4_2 ssse3 vmx

You are expecting perfection and not realism. Get your head out of your A~~ and look at things realistically instead of idealistically.

273 · 02-12-2022, 10:17 PM

Quote:

Originally Posted by computersavvy

So according to your ideas this means my 3600 is not performing up to standards, even though it is holding these figures 24/7 @~60C and ~60% load on all threads?

Code:

$ inxi -Cxxx
CPU:
  Info: 6-core model: AMD Ryzen 5 3600 bits: 64 type: MT MCP smt: enabled
    arch: Zen 2 rev: 0 cache: L1: 384 KiB L2: 3 MiB L3: 32 MiB
  Speed (MHz): avg: 4046 high: 4067 min/max: 2200/4208 boost: enabled
    cores: 1: 4026 2: 4060 3: 4042 4: 4064 5: 4047 6: 4067 7: 4045 8: 4040
    9: 4055 10: 4053 11: 4041 12: 4020 bogomips: 86234
  Flags: avx avx2 ht lm nx pae sse sse2 sse3 sse4_1 sse4_2 sse4a ssse3 svm

Or that my intel laptop (i7-9750H) is not performing up to standards (again running 24/7 @ ~75C & 60% load on all threads). Maybe I should be ranting at intel because there is a bigger difference between the max rated speed of their cpu than I see in performance when compared to the AMD cpu with almost identical loads. Note that the 3600 is posting 86234 bogomips vs the i7 @ 62399 bogomips.

At one point it was suggested that you post the output of "inxi -Cxxx" as I just did, but you have failed to even show what the system is actually doing. Instead you just complain about it. Are you possibly exaggerating and don't want us to see that documented?

Code:

$ inxi -Cxxx
CPU:
  Info: 6-core model: Intel Core i7-9750H bits: 64 type: MT MCP smt: enabled arch: Coffee Lake
    rev: A cache: L1: 384 KiB L2: 1.5 MiB L3: 12 MiB
  Speed (MHz): avg: 4143 high: 4202 min/max: 800/4500 cores: 1: 4202 2: 4157 3: 4176 4: 4157
    5: 4148 6: 4144 7: 4154 8: 4035 9: 4081 10: 4100 11: 4198 12: 4173 bogomips: 62399
  Flags: avx avx2 ht lm nx pae sse sse2 sse3 sse4_1 sse4_2 ssse3 vmx

You are expecting perfection and not realism. Get your head out of your A~~ and look at things realistically instead of idealistically.

I am expecting that when I see a figure of 4.5GHz in the marketing I should see that from the actual CPU. The CPU is marketed as being 4.5 as opposed to 4.4 for the 3700x - AMD don't even get into cooling or power delivery and simply sell them as such. I do not count the fact that a CPU may hit a speed on one clock cycle as it hitting that speed.
Oh, I installed it just for you:

Code:

inxi -Cxxx
CPU:
  Info: 8-core model: AMD Ryzen 7 3800X bits: 64 type: MT MCP smt: enabled
    arch: Zen 2 rev: 0 cache: L1: 512 KiB L2: 4 MiB L3: 32 MiB
  Speed (MHz): avg: 3288 high: 4257 min/max: 2200/4559 boost: enabled
    cores: 1: 4039 2: 3516 3: 4257 4: 4020 5: 2155 6: 2943 7: 1991 8: 3674
    9: 4163 10: 3594 11: 3584 12: 3847 13: 2129 14: 2898 15: 2025 16: 3775
    bogomips: 124566
  Flags: avx avx2 ht lm nx pae sse sse2 sse3 sse4_1 sse4_2 sse4a ssse3 svm

273 · 03-18-2022, 09:25 PM

I have since done some googling and it seems that I did get a bad CPU or the motherboard just does not work with it properly. I may have to try the Windows software to see whether I have a hardware issue.

273 · 09-04-2022, 06:25 PM

I revisited this and discovered that the Ryzen CPU seems to thermal throttle at 70C (I really need to clean my dust filter properly) and now I'm managing to see 4.24GHz, for a few seconds, on all cores wit ha faster fan profile. So, seems I may need to reapply my thermal paste and set my case fan speeds faster since I don't see 4.4GHz any more running mprime.
That typed, other than Carmageddon being a bit choppy, and I think that's running on DOSBox or something?) everything seems to run OK.
Edit: With a new, more aggressive, CPU fan profile just using the machine I have seen 4.47GHz. So I may reapply my thermal paste, clean my case fans and get them running better and see how it goes.
I also forgot to mention that I revisited this and worked it out because in Kernel 5.19 I can see what seems to be a more standard CPU temp and, for the first time, the CPU fan RPM. Still no case fan RPM but I can play that by ear.

273 · 09-21-2022, 06:25 PM

With a more aggressive fan profile I now see CPU speeds of 5.4GHz shown while the usage fro that core is showing single digits. So, I suppose, AMD aren't lying they're just being disingenuous - how does a higher clock speed while idle help gaming etc.?

273 · 10-14-2022, 10:03 PM

I will just rant further here rather than wasting space anywhere else that I now see "5.54GHz" regularly as the speed of the highest core but it is always when the CPU isn't being taxed at all - so AMD are selling CPUs at the speed then can travel downhill, basically.

boughtonp · 10-16-2022, 07:11 AM

Not sure whether it's relevant here, but there was a recently fixed kernel bug that was artificially limiting Ryzen CPU performance due to an out-of-date workaround still being in place.

The patch was included in v6.0 released two weeks ago, but I don't know if it has or will be backported to any LTS kernels.

273 · 10-16-2022, 07:14 AM

Quote:

Originally Posted by boughtonp

Not sure whether it's relevant here, but there was a recently fixed kernel bug that was artificially limiting Ryzen CPU performance due to an out-of-date workaround still being in place.

The patch was included in v6.0 released two weeks ago, but I don't know if it has or will be backported to any LTS kernels.

I forgot about that, thanks, I will read up on it and check kernel versions.

boughtonp · 10-16-2022, 07:33 AM

Quote:

Originally Posted by 273

I revisited this and discovered that the Ryzen CPU seems to thermal throttle at 70C (I really need to clean my dust filter properly) and now I'm managing to see 4.24GHz, for a few seconds, on all cores wit ha faster fan profile. So, seems I may need to reapply my thermal paste and set my case fan speeds faster since I don't see 4.4GHz any more running mprime.

It seems mprime is for calculating prime numbers - is that actually a good tool for testing?
(Seems too niche/specialized - benchmarking should be as real-world as possible to be useful?)

Also, you're using "inxi -Cxxx" to determine speeds, but that only provides a snapshot - I'd expect continuous measurements to provide more information on any trends/behaviour? Especially if correlating with particular activity/throttling/etc. I don't see mention of what you're using to obtain temperature info?

I need to do similar testing, and currently planning on using collectd, but I've not yet looked into things properly, so don't know if there's better options.

273 · 10-16-2022, 09:47 AM

Quote:

Originally Posted by boughtonp

It seems mprime is for calculating prime numbers - is that actually a good tool for testing?
(Seems too niche/specialized - benchmarking should be as real-world as possible to be useful?)

Also, you're using "inxi -Cxxx" to determine speeds, but that only provides a snapshot - I'd expect continuous measurements to provide more information on any trends/behaviour? Especially if correlating with particular activity/throttling/etc. I don't see mention of what you're using to obtain temperature info?

I need to do similar testing, and currently planning on using collectd, but I've not yet looked into things properly, so don't know if there's better options.

I'm monitoring speeds and temperatures with conky, calling sensors periodically the CPU data it can't get as well as an XFCE sensors plugin set to show the highest CPU core speed at that moment on the panel.
I use mprime because it allows choosing how many threads to use but the results (only seeing higher CPU speeds on cores showing single digit usage) seems to be the case when running VirtualBox or having Google Earth and a 4K video open, for example.