Processors

[kirts]

Hi,

As per the subject line, manual overclock of 4.475GHz is significantly outperforming PBO of 4.65-4.70GHz all core. How is this possible?

Clocks are verified across multiple programs such as hwinfo and core temp and the dreadful Asus AI suite 3.

Benchmark program is cinebench R23.

The system is 100% stable and also curiously the PBO which goes higher than the manual overclock gets 5-6c lower.

Many thanks in advance.

[goodplay]

Bit of a trick question, as in mobo model. whether bios/software OC ?

Some monitoring apps are known to reduce perf. on some benchmarks (e.g. cinebench (it is documented))?

You may do better asking on some Overclockers forums (they'll want specifics).

[kirts]

All changes are done directly to the BIOS.

The motherboard is Asus ROG Crosshair VIII Hero Wi-Fi.

The environment is exactly the same in both test cases and the results are consistent. There is a huge gap in performance of a range of 10-12%.

Good point about overclockers forums, I guess that's my next stop but I thought to ask here in case it's a known issue or other people came across it.

Thank you again.

[goodplay]

Could be just down to 'silicon lottery', but someone in an OC forum can probably help get the best out of it ?

[Thanny]

PBO is not an overclock.  It just lets your CPU consume more power, which allows it to maintain higher clock speeds under full load.  And unlike a manual overclock, it does not lower the max boost clock when only one or two cores are active.

The 5950X with PBO does not maintain the clock speeds you claim under an all-core load.  It will be at around 4.4 or 4.45GHz on all cores.

Either the monitoring programs you're using are wrong, or you're not reading the information correctly.

[kirts]

Thank you Thanny, I can reassure you that I can read correctly simple values I find it highly unlikely that all of the 3 tools (that I mentioned in my first post) are wrong unless there is some kind of a bug at reading PBO values?

Whether PBO falls under the category of overclock or not, I guess is not directly relevant. The fact is, all core jumps between 4.650-4.7GHz in PBO while my manual overclock was steady at 4.475GHz yet it performed better.

I have done some tuning (negative voltage and tuned the curve) among other things which may explain the higher than usual clocks under PBO. I also have open loop with dual pump and 2 460 radiators among other things which can keep the temperatures relatively low (100% usage brings me to 79c). My understanding is that PBO will keep the clocks if there is temperature room and that is definitely what all of my tools report.

Which brings the outstanding question of what is going on here.

[Robinhood]

Download Ryzen Master. Go to advanced view. Close out of all other monitoring applications. Do a Cinebench R23 run with your manual overclock, and do one only with PBO enabled. Send a screenshot of the Ryzen Master window while the test is running, where the frequencies of all cores are visible. Then please post the results of the two tests.

PBO boosts cores dynamically. Some cores will get a higher frequency during certain workloads; although during a Cinebench test all cores should be around the same frequency. Generally, doing a manual overclock will yield better performance in all-core workloads, but PBO will perform much better in lightly threaded tasks where it can boost one or two cores to 5 GHz temporarily, which your manual static overclock can't achieve.

[Thanny]

Changing the voltage might be negatively affecting the stability, which will hurt performance before it causes any kind of crashing.

You should check PBO performance at stock voltage first, then start reducing it again, stopping when you lose performance.