[SOLVED] AMD 5800X slower and hotter for no apparent reason

[Switch87]

Hello everyone,

I should first say my mb is a MSI x570 meg unify. I have no OC done to the cpu, in fact the only changes in the bios are fan curve and XMP.

I've had the cpu since it launched. and when I got it I tested it a lot. I'll use the cinebench r23 multicore test as an example of what's happening.
Back then, (around november last year) my temperatures during the 10 min throttle test was around 81, maxing at 82º. The multi core score result would be around 15500.

Since then, I haven't messed with the pc. I haven't installed any new hardware or anything. A few days ago I updated my bios and I installed the new AMD chipset drivers. After doing this I decided to run R23 again just to check.
My temperatures now during the 10 minute test go to close to 90º in the first 30 seconds and stay there. My score is also lower (now it's around 15000). I also noticed during the test, on ryzen master, that the cpu clock speed maxes out at around 4500MHZ from a max of 4800. Maybe that part is normal, I don't remember how it was before.

This is extremely odd to me, I don't understand how could my performance, but especially the temps be so different in half a year without changing any settings. I don't use any cpu OC software or anything, only to monitor temps. Like I said, bios settings are at default, and I even rolled back to an earlier version of the bios to see if it would change anything. It didn't. So the only thing I can assume did it must be the AMD chipset drivers?

Any input would be appreciated.
Thanks

 

[Switch87]

Forgot to mention, my cooler is still the same as it was like everything else, a Noctua NH-D15

 

[SamirD]

Since a bios rollback didn't change it, yep I would suspect the chipset drivers.

This is why a real benchmark should be run outside of windows as windows can affect the outcome. Too bad there isn't a single linux benchmark live cd available that is free--I'd be all over something like that.

 

[Ripthruster]

Not sure where you are located but could summer time ambient temps be a factor? You mentioned doing the benchmark in November of last year vs late August. I only mention it because I was able to pick up a 3080 last November and all was good until this summer's last couple of heat waves and I noticed how hot my GPU and total system thermals were getting. Led me to change mid-tower cases and undervolt my GPU. The results were a vast improvement in performance and thermals all around.

 

[alceryes]

Since a bios rollback didn't change it, yep I would suspect the chipset drivers.

This is why a real benchmark should be run outside of windows as windows can affect the outcome. Too bad there isn't a single linux benchmark live cd available that is free--I'd be all over something like that.

Check out both Ultimate Boot CD and Hiren's Boot CD PE. Both have some benchmark tools.

For the OP.
~4500MHz is normal for an all-core boost.
Has anything else changed? Voltage a little different? LLC? Did you run it right from a reboot or let your OS sit for 3+ mins at desktop before running?

 

[SamirD]

Check out both Ultimate Boot CD and Hiren's Boot CD PE. Both have some benchmark tools.

I have but the tools are incomplete--just tests of certain things. Nothing on the level of furmark or cinbench that will truly benchmark more than a single device.