Overclock.net banner
21 - 36 of 36 Posts

·
Registered
Joined
·
25 Posts
Discussion Starter · #21 ·
There are so many CPU related settings in the BIOS (like curve optimizer, that can be tested with corecycler), I wish I had a better understanding of such things to experiment a little bit with the settings. But I agree, with default settings, such errors should not appear.
 

·
Robotic Chemist
Joined
·
3,886 Posts
It is because the CPU spikes up to very high speeds for very short period of time with very light short loads. Under longer loads using multiple cores (like a game) it doesn't boost any single core as high as it does for short OS tasks.

It is odd that it is only file things though. The suggestion to try CoreCycler is a good one, you might even be able to add a positive offset to the PBO curve on one or two cores and get the system stable without losing much performance, if you wanted to try that.
 

·
Registered
Joined
·
230 Posts
Hi there!

Not really an overclocking issue I think (more an issue with boost mode), but I think it fits here.

There seems to be an issue with my new PC as soon as CPB (Core Performance Boost) is enabled (the option will let the CPU boost to 4.6GHz, a common procedure as it seems and also supported by my CPU - "AMD Ryzen 5 5600X, 6C/12T, 3.70-4.60GHz"). When I disable CPB, the CPU stays at 3.7GHz also in boost mode and all weird problems are suddenly gone.

The weird problems I discovered:
  • Unable to start Elite Dangerous
  • Unable to start Star Citizen (because of invalid game files) (game sometimes start when I set DRAM voltage to 1.45V - even with CPB enabled)
  • DISK WRITE ERROR when downloading/updating Steam games
  • INVALID DEPOT CONFIGURATION error in Steam when downloading/updating Steam games
  • Steam downloads are stopping, only to resume again in 5 minutes - if not many times in the middle of the download, then at the very end (e.g. 425MB/425MB) and I have to manually resume the download
  • Unable to pack/unpack 7z files when I use a password for the archive file (CRC/Data error) - this one I also tested on a Live Linux USB stick - it happens there as well, except I set DRAM Voltage to 1.45V - but in Windows it happens even with 1.45V

All of the problems are gone when I disable CPB in BIOS (of course no solution for me, I'm already in contact with the manufacturer of the PC, as the problem also occurs in Linux, so maybe it's a hardware issue / compatibility issue).

Today I get RAM modules that are on the compatibility list of the mainboard's website (the manufacturer of the PC used RAM modules that are not on the list, so I ordered some to do this test before I send the PC back).

I'm using an ASUS ROG STRIX B550-F Gaming Wi-Fi, an AMD Ryzen 5 5600X, an ASUS ROG STRIX RTX3080, a 500GB WD_BLACK SN750 M.2 PCIe 3.0 x4 SSD, a bequiet Dark Power Pro 11 1200W power supply and "CMK32GX4M2B3200C16" (2x16GB) - I know, they are not on the compatibility list. Also tested with "CMK16GX4M2B3200C16(Ver5.32)" from a friend already, but they are not on the list either (only "CMK16GX4M2B3200C16(Ver4.24)" is on the list). So I ordered G.SKILL "F4-3200C14D-16GTZN" for testing. They will arrive today.
*EDIT/UPDATE: Tried it with the G.SKILL RAM from the compatibility list - same error, unless I deactivate CPB :-/

Prime95/memtest86 are running for hours without any errors. It's also not about my disks, they are fine. I'm also able to play games (even with CPB on for many hours, even demanding ones like MS Flight Simulator. It's really "just" the issues I reported above.

I'm on BIOS 2006 (latest non-beta) - is it worth a try to update to 2401 beta bios? I can't find any details what has been changed with "AMD AM4 AGESA V2 PI 1.2.0.3 Patch A" or what the "Improve system stability" exactly means.

Thanks!
I'm wondering if you ran "Rizen Master" then when into your BIOS and otherwise tweaked PBO and/or CPB. You are not supposed to use Master then head for the BIOS or the other way, Worth checking out.
 

·
Registered
Joined
·
25 Posts
Discussion Starter · #24 ·
I'm wondering if you ran "Rizen Master" then when into your BIOS and otherwise tweaked PBO and/or CPB. You are not supposed to use Master then head for the BIOS or the other way, Worth checking out.
Sorry, I don't get what you're trying to tell me ^^
But I don't tweak anything and I've never heard of "Ryzen Master", so I guess I didn't use it ;-)
 

·
Registered
Joined
·
8 Posts
Hi.

Try this program: Releases · jedi95/BoostTester

run the following command in "command prompt" from the directory where you saved the file:
powershell ".\BoostTester.exe | tee Log.txt"

This will log to log.txt and you might see what core is causing problems.

For me this has been very useful in finding unstable cores, especially after I started logging to a file as the system seems more unstable when also doing file-access at the same time.
 

·
Registered
Joined
·
25 Posts
Discussion Starter · #26 ·
Hi mckajvah,

thanks for the tip - I let it run for 2 hours already, log file doesn't show any errors (and I can see how it cycles through the cores - in the log file and on Open Hardware Monitor).
How long do you suggest to let it run? Thanks!
 

·
Registered
Joined
·
8 Posts
Hi mckajvah,

thanks for the tip - I let it run for 2 hours already, log file doesn't show any errors (and I can see how it cycles through the cores - in the log file and on Open Hardware Monitor).
How long do you suggest to let it run? Thanks!
Sorry for not mentioning. Normally it crashes within a few minutes. If nothing has happened after 2h, then this test won't find anything. Sorry.
 

·
Registered
Joined
·
25 Posts
Discussion Starter · #28 ·
It is odd that it is only file things though. The suggestion to try CoreCycler is a good one, you might even be able to add a positive offset to the PBO curve on one or two cores and get the system stable without losing much performance, if you wanted to try that.
Despite sending the PC back to the manufacturer (which I will do), do you mind explaining the "add a positive offset to the PBO curve"?
CoreCycler ran without any issues for more than 8 hours over night (like 14 iterations or so).
By the way - the issue also appears with CPB enabled, but PBO disabled completely:
"Funktioniert" means none of the initially mentioned errors, "Fehler" means that all of the errors appear with this setting.
 

·
Robotic Chemist
Joined
·
3,886 Posts
Despite sending the PC back to the manufacturer (which I will do), do you mind explaining the "add a positive offset to the PBO curve"?
In the PBO curve optimizer you can tell the CPU to use more or less voltage per clock with an offset. A positive value means a lower clock for a given voltage (higher voltage per clock).

CoreCycler ran without any issues for more than 8 hours over night (like 14 iterations or so).
This is interesting. It also make me much less sure it is simply an unstable CPU core at stock boost, you might need to increase the SOC voltage (CPU can impact SOC and vice versa). Maybe try a +0.05V offset to the SOC voltage? As a diagnostic at least, though running a slightly higher than stock SOC voltage isn't too bad as long as you didn't want to push fclk or do any hardcore OC.

PBO is really just opening up the power limits. It doesn't really change single core short term boost behavior by itself.

Are all the power cables plugged into the motherboard properly? The extra 4 or 8-pin CPU power too?
 

·
Registered
Joined
·
25 Posts
Discussion Starter · #30 ·
Ok, thanks for the explanation.
I keep wondering why only file-involved operations (steam downloads, extracting 7z files with password, starting some games) cause this issue (and that it sometimes work with DRAM voltage set to 1.45V) - this is really strange. As said, I can play games for hours, even with CPB enabled (so I guess all needed cables are plugged in to the mainboard, but I can check again).

I did the "SOC voltage test" (raising it by 0.05V) already some weeks ago, didn't change anything as well (also in combination with DRAM voltage set to 1.45V instead of 1.35V).
As a reminder:
  • I can sometimes start Star Citizen in Windows with DRAM voltage set to 1.45V
  • When I'm using Linux, I can extract 7z files with password when DRAM voltage is set to 1.45V - but it never works in Windows with CPB enabled (CRC error).

I like your signature :-D
 

·
Robotic Chemist
Joined
·
3,886 Posts
Truly mysterious. DRAM set to 1.5V? It would be odd to have CPB impact the DRAM voltage, but if you notice improvements in that direction...

Some kits do default to voltages even higher than that.

Have you tried the other voltages?

VDDG IDO to 1.05V
VDDG CCD to 1.0V
VDDP to 0.9V

Maybe SOC to 1.15V

I would start to suspect other hardware, but it all goes away with CPB off so it is hard to look outside the CPU. Even the total power draw doesn't change that much with CPB on, not when doing something like updating Steam games.

As a long shot, have you tried a negative CPU voltage offset? That prevents the CPU from using the highest boost clocks. I found up to -0.625V to work well for keeping the max voltages down, but it does keep the CPU from boosting that last 25-75 MHz too.
 

·
Registered
Joined
·
25 Posts
Discussion Starter · #32 ·
Yes, tried DRAM voltage 1.5V as well, didn't change anything and I don't want to go higher honestly.

Thanks for your other tips - I will try them as well - but only out of interest and to provide some more information for the manufacturer (as I will send it back anyway). Because I think at stock/default settings, such errors should not occur ... and I don't want to restart my PC to disable CPB always when I want to update steam games and do another restart to enable CPB to get the full power of the CPU to play them :-D

I will keep you updated. Thanks again!
 

·
Registered
Joined
·
1 Posts
buy some good thermal paste then remove your CPU cooler clean thermal paste off then take CPU out see if all pin look ok then re paste it then get cooler back on don`t over tighten the CPU screws
retest system
make sure using 4 pin power plug on motherboard to not just 24pin from power supply
if that don`t help you then maybe just verry bad CPU or motherboard it self , you can test this if you have a mate that using am4 cpu you test parts on
 

·
Registered
Joined
·
25 Posts
Discussion Starter · #34 · (Edited)
Ok, this is going to be a long one :)

TLDR: Setting VDDCR CPU Voltage from 1.1V to 1.2V solves the issue (CPB enabled, everything works fine)

The longer version:

Side note: A suggestion from another forum (AMD) to disable HPET "High precision event timer" in Windows and to disable C-State of the CPU in BIOS didn't help either.

Then I started with the suggestion of @Asmodian

I started with default settings:
CPB: Auto
CLDO VDDG CCD: 0,9V
CLDO VDDG IOD: 0,9V
VDDCR-SOC: 1,0V
CLDO VDDP: 0,9V
DRAM Voltage: 1,35V
VDDCR CPU Voltage: 1,1V
-> Described errors appear

After setting the suggested values:
CPB: Auto
CLDO VDDG CCD: 1,0V
CLDO VDDG IOD: 1,05V
VDDCR-SOC: 1,15V
CLDO VDDP: 0,9V
DRAM Voltage: 1,50V
VDDCR CPU Voltage: 1,1V
-> Described errors appear

Then I decided to look into the "Asus AI Suite" for my mainboard (Windows application)
There is this option in "TurboV Evo" menu -> "Auto Tuning" that says "Auto System Level Up" (what a name), which I activated.
Basically it set CPU clock to 4200MHz (42x100), but left CPB to "Auto" as well - but base clock and boost clock were always 4.2GHz
And I noticed that VDDCR CPU Voltage was raised from 1,1V to 1,2V
All described errors disappeared and everything worked fine.
So the settings were:
CPB: Auto (but with a fixed clock speed at 4.2GHz)
CLDO VDDG CCD: 0,9V
CLDO VDDG IOD: 0,9V
VDDCR-SOC: 1,0V
CLDO VDDP: 0,9V
DRAM Voltage: 1,35V
VDDCR CPU Voltage: 1,2V
-> Everything works fine

After a BIOS reset to default settings, I then decided to look into the CPU offset voltage you've suggested
"I found up to -0.625V to work well for keeping the max voltages down, but it does keep the CPU from boosting that last 25-75 MHz too. "
I guess you meant -0.0625V (or even -0.00625V)? -0.625V seams a bit harsh :)
What I noticed after using -0.0625V is that the VDDCR CPU Voltage simply was 1,0375V instead of 1,1V and all the issues still appeared.

I did another BIOS reset to default settings and the only thing I changed afterwards was setting VDDCR CPU Voltage not to "Offset", but to "Manual" and setting it to 1,2V instead of 1,1V (maybe also something in-between works, but I haven't tested it so far).

Setting it back to 1,1V (even with Asus AI Suite, live during the Windows session, without a reboot and setting it in BIOS) instantly shows all the errors again (steam downloads failing, file extract failing, game start of Elite Dangerous failing). Setting it back to 1,2V again with Asus AI Suite -> everything works fine.

Question is: What does this mean? Is my CPU faulty in some way? Did it pass quality assurance just "barley" and the issues didn't appear and it only fully works when you provide more voltage to it?

Thanks for your assistance so far! :-D

*edit: What I also noticed after setting VDDCR CPU Voltage to 1,2V in BIOS: The display of "CPU Core Voltage" in Asus AI Suite stays fixed at 1,2V ... before it switched, don't remember exactly, between 0,3 (?) and even 1,3V. Guess it was because it was set to "Auto" in BIOS (and even BIOS displayed 1,328V as current CPU Voltage). Is the fixed voltage to 1,2V a problem? Will it somehow shorten the lifetime of the CPU?

*edit2: on a second thought: maybe the Mainboard doesn't switch voltages fast enough between 0,3V and 1,3V or doesn't switch at all (and then the CPU voltage is too low) and by setting it to a fixed value everything works fine?
 

·
Robotic Chemist
Joined
·
3,886 Posts
Very good work figuring it out! Yes, I meant -0.0625V. :eek:

So it was the CPU core voltage, but one of the middle steps that are so hard to test on Ryzen, or maybe bad transients. Have you tried a positive offset? 1.2V is very safe 24/7 voltage as long as you are not running at 90°C all the time. I would be fine with 1.25V fixed, or maybe a bit higher with good cooling.

I expect single core boosting to hit a lower peak clock speed with a fixed voltage, when I have used a fixed voltage on Ryzen I also used a fixed clock. I am not sure what it does when you set the voltage manually but don't set the frequency. You need that 1.3+V to get the fastest single core speeds out of your CPU. A nice all-core fixed OC is still a good prospect, but you cannot clock quite as high for the same light single threaded loads with a fixed voltage/frequency. My 5950X will use 1.5V for very short periods of time to briefly hit 5050MHz at stock settings, but trying to run it at anything close to that for a fixed OC would destroy it.

What is the max clock speed you see in HWiNFO when running at a fixed 1.2V?

Another idea is turning up LLC, it would reduce/eliminate voltage droop under load, but I am not sure it would help your CPU. It is in the DIGI+ menu in BIOS, do not go above LLC 3. Also setting the SOC to OC mode, so it never clocks down, might help. I have had stability improve a lot when overclocking fclk when setting the SOC to OC mode. DRAM too, powder down mode is harder than no power down. Maybe not too, it still seems like a CPU core clock-at-voltage issue to me, but it is possible the root cause of that is due to bad power delivery from the motherboard in some way. The ASUS ROG STRIX B550-F Gaming Wi-Fi should not be bad at power delivery for a 5600X, but it could be a bad board.
 

·
Registered
Joined
·
25 Posts
Discussion Starter · #36 ·
Hi there and sorry for late reply, weekend and stuff ;-)

Have you tried a positive offset?
No, not so far, only played around with fixed voltages and the -0.0625V offset, which didn't solve anything.

1.2V is very safe 24/7 voltage as long as you are not running at 90°C all the time. I would be fine with 1.25V fixed, or maybe a bit higher with good cooling.
Ok, thanks! I think the same after playing around with fixed voltages (see below). Temperature doesn't seem to be an issue here.

I expect single core boosting to hit a lower peak clock speed with a fixed voltage, when I have used a fixed voltage on Ryzen I also used a fixed clock. I am not sure what it does when you set the voltage manually but don't set the frequency. You need that 1.3+V to get the fastest single core speeds out of your CPU. A nice all-core fixed OC is still a good prospect, but you cannot clock quite as high for the same light single threaded loads with a fixed voltage/frequency. My 5950X will use 1.5V for very short periods of time to briefly hit 5050MHz at stock settings, but trying to run it at anything close to that for a fixed OC would destroy it.
That's very interesting. Just wondered earlier what would happen when I clock my "AMD Ryzen 5 5600X, 6C/12T, 3.70-4.60GHz" permanently at 4.60GHz, but didn't think about trying it ;-) If you're wondering what happens with a fixed voltage (e.g. 1.2V, but without a fixed frequency ... it reboots sometimes during CPU tests - see below).

What is the max clock speed you see in HWiNFO when running at a fixed 1.2V?
See tests below ;-)

Another idea is turning up LLC, it would reduce/eliminate voltage droop under load, but I am not sure it would help your CPU. It is in the DIGI+ menu in BIOS, do not go above LLC 3. Also setting the SOC to OC mode, so it never clocks down, might help. I have had stability improve a lot when overclocking fclk when setting the SOC to OC mode. DRAM too, powder down mode is harder than no power down. Maybe not too, it still seems like a CPU core clock-at-voltage issue to me, but it is possible the root cause of that is due to bad power delivery from the motherboard in some way. The ASUS ROG STRIX B550-F Gaming Wi-Fi should not be bad at power delivery for a 5600X, but it could be a bad board.
Thanks for the additional information. I will try SOC OC mode if I experience any stability issues in the future. I played around with VDDCR CPU Load Line Calibration in the Asus AI Suite Tool, but not in BIOS. Without raising CPU Voltage to 1.2V in addition, the mentioned errors still appear. But I'm currently at VDDCR CPU LLC Level 5 (somehow this has been auto-set? But I don't know when/where - but this is also the unstable "Another BIOS reset, BIOS default settings, but with D.O.C.P enabled AND CPU Voltage set to 1.2V" setup (see below).

The tests:

BIOS default settings, but with D.O.C.P:
Mentioned errors appear (so this isn't an option)
CPB: on
VDDCR CPU Voltage: 1.1V
CPU Frequency: 3660.1MHz to 4650.1MHz
CPU Voltage maximum (according to Asus AI Suite): 1.36V

3DMARK CPU Benchmark (CPU Profile):
Max. threads16 threads8 threads4 threads2 threads1 thread
5098 pts.5397 pts.4325 pts.3186 pts.1776 pts.907 pts.
68.6ms64.8ms80.9ms109.8ms197.1ms385.5ms

Using Asus AI Suite's option "Auto System Level Up":
Errors didn't appear anymore, system is stable during CPU Benchmark
CPB: on
VDDCR CPU Voltage: 1.2V (automatically set by "Auto System Level Up")
CPU Frequency: Fixed at 4200MHz
CPU Voltage maximum (according to Asus AI Suite): 1.2V

3DMARK CPU Benchmark (CPU Profile):
Max. threads16 threads8 threads4 threads2 threads1 thread
5563 pts.5541 pts.4409 pts.3193 pts.1650 pts.831 pts.
62.9ms63.2ms79.4ms109.6ms212.1ms421.0ms

Another BIOS reset, BIOS default settings, but with D.O.C.P enabled AND CPU Voltage set to 1.2V:
Errors didn't appear anymore, but CPU Benchmark wasn't stable (had 1 reboot during first test, but then 2 tests succeeded)
CPB: on
VDDCR CPU Voltage: 1.2V (manually set)
CPU Frequency: 2940.1MHz to 4650.1MHz
CPU Voltage maximum (according to Asus AI Suite): 1.2V
VDD CPU LLC: Level 5 - according to the slider in the Asus AI Suite Tool (didn't check this before during the other tests).

3DMARK CPU Benchmark (CPU Profile):
Max. threads16 threads8 threads4 threads2 threads1 thread
4787 pts.4776 pts.4132 pts.2978 pts.1663 pts.850 pts.
73.1ms73.3ms84.7ms117.5ms210.4ms411.5ms

...

So even the last test shows (regardless of the reboot that happened) that you don't operate a system like that (see benchmark results) :-D
Although it displayed 4650MHz as max. Frequency per core in Open Hardware Monitor, it seems that it didn't manage to get there during the CPU Benchmark.

Maybe I can try 1.25V or 1.3V with a automatic (non-fixed) frequency, but as far as I understood you, this is not how you overclock, right?

I'm totally happy with the "Auto System Level Up" settings (if I notice some instability in the future, I guess I would just raise CPU voltage from 1.2V to 1.25V - but until now every test ran stable), but what I need to know (because I need to decide if I should send the system back so that they can check it) ... could this be some latent issue with the CPU? Something that could get worse in the future? Can such problems be considered "normal" for AMD systems and I should be fine by raising CPU voltage a bit?

Thanks!
 
21 - 36 of 36 Posts
Top