Did AMD's recent chipset driver improve temps and performance?

[A Computer Guy]

I'm doing my monthly updates and noticed after AMD's chipset update (5.02.19.2221) on my x570 my 5950x is running a few degrees cooler and scoring better in CPUz. +300 points in multi-thread.

For some reason my 5950x had always been running 300 points behind in the CPUz multi-thread benchmark. Now it's fixed. Anyone else see similar improvements?

 

[freeagent]

I never use their chipset drivers, just what winupdate gives. With my 5900X I was at 704 SC and 103xx MC.

 

[oxrufiioxo]

300 points in that benchmark is super margin of error the wind could have literally just been blowing the right way....

 

[tabascosauz]

I'm doing my monthly updates and noticed after AMD's chipset update (5.02.19.2221) on my x570 my 5950x is running a few degrees cooler and scoring better in CPUz. +300 points in multi-thread.

For some reason my 5950x had always been running 300 points behind in the CPUz multi-thread benchmark. Now it's fixed. Anyone else see similar improvements?

View attachment 291459

Maybe if you ran it 100 times on the old one and 100 times on the new one and took the average of 300 points. CPU-Z is not a consistent benchmark - it's good for a quick look and that's it.

2-3C temp difference is also well within run-to-run variance for CPU-Z, especially since you are only Benching and not using the Stress button. CPU-Z scores also vary depending on how long after boot you ran it and CPU temp at start etc. The bench lasts 10 effing seconds, idk what sort of conclusions anyone could ever draw from that.

Been over this before, free performance doesn't just randomly show up for old Ryzen generations. Rule of thumb, if you think your Ryzen got suddenly faster due to x, it hasn't - unless you did extensive and repeated testing to eliminate all other variables.

 

[A Computer Guy]

Maybe if you ran it 100 times on the old one and 100 times on the new one and took the average of 300 points. CPU-Z is not a consistent benchmark - it's good for a quick look and that's it.

I didn't realize CPUz was so inconsistent. I usually use it's benchmark to take a quick look from time to time to see if updates negatively impacted my systems and since I put my 5950x in my x570 it was pretty consistent about being 300pts under the baseline where as on my B550 it was usually in the 12000+.

2-3C temp difference is also well within run-to-run variance for CPU-Z, especially since you are only Benching and not using the Stress button. CPU-Z scores also vary depending on how long after boot you ran it and CPU temp at start etc. The bench lasts 10 effing seconds, idk what sort of conclusions anyone could ever draw from that.

The temp observation was just letting my system idle. This time of year it hovers 36c to 38c and was dipping down to 32c. Of course there can be some other explanations for this as well as I have been tweaking fan profiles and pump speeds.

Been over this before, free performance doesn't just randomly show up for old Ryzen generations. Rule of thumb, if you think your Ryzen got suddenly faster due to x, it hasn't - unless you did extensive and repeated testing to eliminate all other variables.

They could always be tweaking and fixing things in driver updates not that one would necessarily think it would unlock some earth shattering performance. I didn't do much extensive testing but casually wondering if anyone had similar observations of seemingly minor improvements. I had some R15 scores saved and also noticed a bit of improvement below prior PBO experiments I had conducted last year.

 

[tabascosauz]

I didn't realize CPUz was so inconsistent. I usually use it's benchmark to take a quick look from time to time to see if updates negatively impacted my systems and since I put my 5950x in my x570 it was pretty consistent about being 300pts under the baseline where as on my B550 it was usually in the 12000+.

The temp observation was just letting my system idle. This time of year it hovers 36c to 38c and was dipping down to 32c. Of course there can be some other explanations for this as well as I have been tweaking fan profiles and pump speeds.

They could always be tweaking and fixing things in driver updates not that one would necessarily think it would unlock some earth shattering performance. I didn't do much extensive testing but casually wondering if anyone had similar observations of seemingly minor improvements. I had some R15 scores saved and also noticed a bit of improvement below prior PBO experiments I had conducted last year.

I only started paying more attention to CPU-Z for Ryzen, and it's generally just not a good way to gauge performance. In terms of clocks it can be quite fickle being especially variable in the ST test for no reason. You can really see the extent of CPU-Z being a "quality" benchmark when core usage dips below 100% during MT, or score suddenly plummeting to 0 and slowly climbing back up a few minutes into the Stress test. Then there's the inexplicable dip in ST score on Windows 11 compared to Windows 10, while Cinebench ST is still relatively consistent.

In terms of temps, CPU-Z Bench is way too short to come to any reasonable conclusions. Cinebench by comparison may not be a good stability test, but it's decent if you just want to look at some numbers, and the temp climb should be slowly tapering off by the end of a single R23 MT loop (or well into equilibrium by the end of the ST test).

Chipset drivers are more focused on the software side of the firmware handshake (ie. AGESA's interaction with Windows). So the fixes that have historically come out via chipset driver are usually along the lines of fixing Windows scheduler bugs, the CPPC idle activity filter, etc. - not fundamental changes to performance that are contained in AGESA.