I just bought a 7900x. Games worked well for a few days with normal core clocks but I did notice things running slow at times I just didn't think much of it, my temps were fine. Suddenly FPS in games tanked and Ryzen Master showed my clocks at 546mhz, despite running demanding games. I looked into it and found similar threads on this forum where people experienced the issue on the 3900 series and some 5900 however they weren't able to find a conclusive solution.
Since the low core clock stays the same even after a reboot I figured it might be a problem with BIOs. I flashed the newest version for my motherboard and saw some improvement where the cores will now reach to around 2700-2800mhz. Sometimes 3.1ghz but never higher. That's still ridiculously low for a new CPU.
Things I've tried to resolve the issue:
In BIOs all readings say the cores are operating at 4700mhz but that clearly isn't the case.
Here's the frequencies in hwinfo, the current core clock isn't accurate at all:
And here are my temperatures and voltages:
My system specs:
AMD Ryzen 9 7900x
ASUS TUF GAMING X670E-PLUS
PNY RTX 3080 12GB
ADATA XPG 16GB (2x16GB) DDR5 5200MHz CL38 Lancer
Corsair HX1000i 1000W
EDIT: I think I'm zoning in on the problem. The cpu IS throttling but for seemingly no reason still, it could be that my motherboard has a faulty sensor. Turning off PROCHOT allows me to run the processor at 4.7ghz.
I'd love more suggestions and potential solutions/causes.
Solved! Go to Solution.
The retailer asked for the CPU and motherboard and tested it. They couldn't find any errors and everything ran as expected. Being stubborn I wanted to try the products again and attempt to rule everything out (again). They sent along some complementary thermal paste and I got to work. First attempt and everything was fine, the next day it was back to throttling. The first attempt I'd added 4 spacer washers to hold the radiator in place more evenly (2 on each screw), just because I thought it looked better. Turns out that's why it worked. When it was throttling again I had removed one spacer on each screw because the two spacers didn't allow for the top filter to fit on the case and I wanted that filter on. My conclusion is that either the radiator itself or the fans came in contact with either the RAM or the MB and caused it to malfunction sending a throttling signal to the CPU.
TLDR; 420 radiator was the culprit, too tight a fit. Two spacers on each screw that holds the radiator case plate in place created enough distance to the MB/RAM to solve the problem. Accepting this as the solution until it stands the test of time or the problem reappears.
You might want to try asking on an overclockers forum (overclock.net, overclockers.uk are somewhere to start).
This isn't about overclocking, I just want the standard performance. If you think they might be able to help though, I will.
Won't hurt to try, they usually figure out quirks fairly quickly.
Hmm... Why is your VDDCR_VDD so high for that load?
Or is mobo counting that you run 5500 while CPU for some reason thinks that voltage that it gets not enough to sustain these clock and stretches clock WAAY down? But core VID's are also looking more or less normal. Strange
Something is definitely wrong there. CPU doesn't feel limited by neither EDC, TDC or PPT... Mobo reports that it delivers correct voltage. Cpu reports that it requests normal voltage, but for some reason voltage requested dropped from 1.47 to 1.37?
Can it be that something is lying in report?
I don't know. I'm afraid I'm completely lost in this. At one point before I flashed the BIOs the issue went away and a ryzen master test would bring the frequencies to where they should be. After another reboot it was back to the same problem. This was without changing any settings. It's like something is telling the CPU to throttle because temps are too high without them actually being high.
Can you try on different mobo somehow?
Maybe it is your motherboard issue?
I can't, only got the one and no one around that's got an am5 socket available. If it is the mb my only option is sending it back, it's brand new though but it could be faulty.
There could be something wrong with the power delivery system. You can try swapping the EPS12V to one of the other connectors on the PSU and see if that does anything. It looks like you are limited to around 80W on the processor.
To rule out some sort of glitch in the boosting algorithm or interaction with Windows, you can try doing a quick manual setting. In BIOS set the CPU voltage to 1.35V. With a manual voltage setting, the CPU won't boost, it'll just run at the base clock of 4.7 GHz regardless of load. 1.35V is low enough that you shouldn't burn up under a 12-core load.
If you can get decent performance with a manual setting, then the PSU and motherboard are able to supply the amperage needed to the CPU, and the issue is related to the boosting. If it cannot, then there is either something wrong with the power coming from the PSU and the EPS12V cable, or the 8-pin slots on the motherboard itself.
I tried setting the bios setting to 1.35 but it still throttles to 2800ish. Turning off PROCHOT allows me to run at 4.7ghz without interuptions but it's not very safe.
A friend had me do something similar in ryzen master and then it ran as normal for a while until the next reboot, then it was back to throttling at 545mhz.
I'll try switching the cables tomorrow.
Did you also happen to make sure that you have updated the chipset driver? Sometimes that can lead to wonky behavior as well.
If it is still throttling on manual however, then it is likely just not physically getting the power it needs. Could be a bad port of the PSU, bad cable, or the motherboard.
Yes I did, chipset is up to date. It would suck to have to rma parts on a brand new pc, but such is life. I'll keep looking into the power, strange that when not thermal throttled by PROCHOT it runs well though.
That is strange. When you set a manual overclock, it should ignore temperature restrictions as well. So that effectively rules out the PSU or the EPS12V ports on the motherboard, if those were bad, it would never boost. The most likely culprit now is a bad motherboard temperature sensor as you had initially stated.
The only other thing I can think of is: Do you happen to have the dynamic OC switcher engaged? Where the computer is switching from a boost setup to a manual overclock at a certain amperage setting? That setting is usually under Extreme Tweaker/CPU Core Ratio (Per CCX).
I haven't touched that setting, if it was engaged by default then I had it, otherwise I didn't. I've sent in my mb and cpu to my retailer (the retailer requested the cpu as well) and they're going to test it themselves. Counting from thursday it'll take me around 5 work days before I have an answer and another 20 if they decide to send it to the manufacturer.
1- Run HWINFO64 as adminstrator, sensors only
2- Run cinebench R20 or R23 multicore benchmark
3- See what HWINFO64 say as Maximum clockspeed
If way below expected:
1- Remove Ryzen master and do not install it anymore
2- Set BIOS to default, reboot
3- Clear CMOS by removing battery, shorting pins or pushing clrcmos button depending on your motherboard (computer has to be unplugged from wall socket)
4- Enable EXPO/XMP/DOCP, leave other settings default (you can enable virtualization and such, but do not change anything else like PBO)
5- Check you have connected power supply to motherboards CPU PWR connector
6- Run HWINFO64 as adminstrator, sensors only
7- Run cinebench R20 or R23 multicore benchmark
8- See what HWINFO64 say as Maximum clockspeed now
Done all of that already, as listed above. There is no change
What were the clockspeeds, volts and temps while running R20 or R23? Screenshots?
He sent screenshots in topic start.
But i wonder if behaviour differs between different stress/bench applications
CB 15/20/23, Prime 95, Blender, Y-Cruncher etc. . .
Yes I know, but I would have wanted to see clockspeeds while running R20 or R23. It is a different thing if they are 'similar', 'same' or something else.
Tried it myself. Games didn't get CPU clockspeed very high, but 3D Mark and (what is free for everyone) Cinebench pushed CPU clockspeed straight to my CPU's manual OC all-core max.
So what could cause low clockspeeds?
- Not enought power (watts)
- Too much temp (something wrong with cooling?)
- software not requiring high clockspeed (what I was asking to exclude)
- Some setting in BIOS
- Some setting in Ryzen Master/other software/registry
- Eco mode?
- Something else?
How to go through these, step-by-step?
Hmm. From all of this.
- Not common weak software as it still gets max nominal frequency on every core (but not effective clock)
- Not temperature for sure, as Ryzen 7000 have new boost behaviour that will target 95 degrees or power limit, depending on what will be first.
- As Ryzen Master reports OC Mode - Default, CO - OFF, there is nothing abnormal with these settings.
But it technically can be ECO mode... But there are 2 modes available, 65 and 105W. And his PPT on load is higher than 65 and lower than 105. Problem is that by default BIOS setup there should not be ECO mode enabled.
I attempted different settings in Ryzen Master to see if it would make a difference. Eco mode on and off, PBO, CO. Nothing mattered. I've since sent it to my retailer since they wanted to test it. Hopefully they find the problem. If so I'll post what they've said.
As you say it's not temperature and to add to that argument I've got a 420 radiator that can blast the heat away.
To settle the Ryzen Master argument, this problem occured before I installed that application as well. That is to say, it occurred before I EVER installed it. It's weird to say the least
You have good points, but I try to eliminate everything:
BIOS controls clockspeeds and voltages, but also Ryzyn Master does those things. I remember issues where those two clashed and caused issues.
So to eliminate one, I would get rid of the one you can uninstall, Ryzen Master, and see what BIOS alone does when allowed to work alone.
What I find strange about this is that CPU does not seem like it is stressed at all. Temps are low, clockspeeds are low... that is why I asked to stress with Cinebench.
Could be a bug somewhere.. Something keeping the CPU from 'waking up'. I would try different tasks at this time, before moving forward.
The HWINFO64 screenshot up above is from when I ran Cinebench R23. No matter the benchmark it throttled to the same value. Rougly 2.7ghz or 545mhz.
I'll add to that I could trick the CPU into working normally by flipping between profiles in Ryzen Master. It would always reset after a reboot however. With or without Ryzen Master installed the problem persisted.
The retailer asked for the CPU and motherboard and tested it. They couldn't find any errors and everything ran as expected. Being stubborn I wanted to try the products again and attempt to rule everything out (again). They sent along some complementary thermal paste and I got to work. First attempt and everything was fine, the next day it was back to throttling. The first attempt I'd added 4 spacer washers to hold the radiator in place more evenly (2 on each screw), just because I thought it looked better. Turns out that's why it worked. When it was throttling again I had removed one spacer on each screw because the two spacers didn't allow for the top filter to fit on the case and I wanted that filter on. My conclusion is that either the radiator itself or the fans came in contact with either the RAM or the MB and caused it to malfunction sending a throttling signal to the CPU.
TLDR; 420 radiator was the culprit, too tight a fit. Two spacers on each screw that holds the radiator case plate in place created enough distance to the MB/RAM to solve the problem. Accepting this as the solution until it stands the test of time or the problem reappears.
Interesting. Thanks for posting this. Great troubleshooting work.