A question for anyone with a 5600X.....

[NoddyPirate]

I had to go back and re-read my first post or two here to figure what has really changed - so a quick summary seems worth it!

I had used the stock single core boost voltage values as my upper limit for all scenarios under the Overclock on the basis of them demonstrably being safe continuous boost values for indivudal cores. If anyone thinks that this isn't sound logic then let me know!

STOCK:

Single-Core Boost: 4.649 GHz - Average Core VID 1.338 - Average VCore 1.345 - 54 Deg C

Multi-Core Boost: 4.262 GHz - Average Core VID 1.170 - Average VCore 1.194 - 61 Deg C


PBO Overclock Settings, including RAM OC:

PPT - 100, TDC - 65, EDC - 105, Curve Optimiser - Negative 25 All Cores, Boost Override - 175 MHz, CPU LLC - Extreme.

(vs Stock)

Single-Core Boost: 4.825 GHz (+176 MHz) - Average Core VID 1.317 (-21mV) - Average VCore 1.328 (-17mV) - 54 Deg C (+0 Deg)

Multi-Core Boost: 4.608 GHz (+346 MHz) - Average Core VID 1.294 (+124mV) - Average VCore 1.326 (+132mV) - 73 Deg C (+12 Deg)


I am basically seeing very smilar results to what I could achieve under a manual all core overclock - except that the average voltages are lower, the power draw is lower, the single core boost has been retained and increased, and it is also running cooler with my regular quiet fan curves vs when manually OC'd with all fans at full tilt.

(EDIT - And I still have a lot room to spare in all that I believe. If my Motherboard allowed higher Overrides - which its doesn't - I would think I could get 4.7 GHz all core and close to 5 GHz single core without too much trouble.)

It certainly seems to me that PBO is the way to push Ryzen rather than a Manual OC. You indirectly control the boosts by setting background limits and allowing the CPU to decide for itself what it can do. It doesn't prevent instability when you push it too far as we've already seen, but should make it less likely at least.

 

[NoddyPirate]

Also - I have a much better idea of the thermal capabilities of my cooler with all the data - with a very predictable Deg C/Watt ratio emerging. It's probably not a linear relationship of course, but projecting all my data out would suggest my cooler would allow the CPU to reach a max temp of about 95 Deg at a power draw of 140-145W.

All things being considered that really not bad at all. But, when you consider the actual power draw (not just TDP) across the Ryzen 3rd Gen suite - this is confirmation (as if it were needed) that the 5600X is the only 3rd Gen CPU that could appropriately be paired with a similar air cooler. Even in stock conditions, an AIO is absolutely necessary from the 5800X up.

If Overclocking is your thing, or you want to maximise the performance of any Ryzen CPU, then an AIO is also required - no matter what 3rd Gen chip you choose.

 

[Scott]

AMDs PBO has been held in very high regard in the latest versions. I think your hard work is showing just how good it actually is.

I've only ever properly played with overclocking on Intel.... and any automated overclocks are awful... all they do is add crazy voltage to get stability with an OC.

At least people reading through this thread can know that they should be able to overclock their chips using the simplified PBO options and get very decent results.

 

[NoddyPirate]

AMDs PBO has been held in very high regard in the latest versions. I think your hard work is showing just how good it actually is.

I've only ever properly played with overclocking on Intel.... and any automated overclocks are awful... all they do is add crazy voltage to get stability with an OC.

At least people reading through this thread can know that they should be able to overclock their chips using the simplified PBO options and get very decent results.

The AGESA updates in recent times certainly seems to have made PBO quite a bit more useful.

PBO seems to sit sort of halfway between Auto OC and Manual OC. You don't get to push things right to the cliff edge as you can manually, but can mitigate and perhaps even just plain avoid those crazy high voltages that come with Auto OC. Add in the retention of single core behaviour and it's just win-win really. It also makes you much more aware of the capabilities of, and the impact on, your Motherboard too.

This thread is messy enough with all my waffle as it is - the last couple of pages is where everything seems to have started coming together - but I'll leave it all be for now. It's been fun!

(I will inevitably play some more at a later date though I am sure! Given that in all cases I am still staying below the average Voltages seen under stock single core boosts, there is a good bit more performance hiding in there still me thinks.......)

 

[NoddyPirate]

I had to go back and re-read my first post or two here to figure what has really changed - so a quick summary seems worth it!

I had used the stock single core boost voltage values as my upper limit for all scenarios under the Overclock on the basis of them demonstrably being safe continuous boost values for indivudal cores. If anyone thinks that this isn't sound logic then let me know!

STOCK:

Single-Core Boost: 4.649 GHz - Average Core VID 1.338 - Average VCore 1.345 - 54 Deg C

Multi-Core Boost: 4.262 GHz - Average Core VID 1.170 - Average VCore 1.194 - 61 Deg C


PBO Overclock Settings, including RAM OC:

PPT - 100, TDC - 65, EDC - 105, Curve Optimiser - Negative 25 All Cores, Boost Override - 175 MHz, CPU LLC - Extreme.

(vs Stock)

Single-Core Boost: 4.825 GHz (+176 MHz) - Average Core VID 1.317 (-21mV) - Average VCore 1.328 (-17mV) - 54 Deg C (+0 Deg)

Multi-Core Boost: 4.608 GHz (+346 MHz) - Average Core VID 1.294 (+124mV) - Average VCore 1.326 (+132mV) - 73 Deg C (+12 Deg)


I am basically seeing very smilar results to what I could achieve under a manual all core overclock - except that the average voltages are lower, the power draw is lower, the single core boost has been retained and increased, and it is also running cooler with my regular quiet fan curves vs when manually OC'd with all fans at full tilt.

(EDIT - And I still have a lot room to spare in all that I believe. If my Motherboard allowed higher Overrides - which its doesn't - I would think I could get 4.7 GHz all core and close to 5 GHz single core without too much trouble.)

It certainly seems to me that PBO is the way to push Ryzen rather than a Manual OC. You indirectly control the boosts by setting background limits and allowing the CPU to decide for itself what it can do. It doesn't prevent instability when you push it too far as we've already seen, but should make it less likely at least.

So my Negative 25 all core Curve Optimiser offset turned out to be too much and was causing the occassional crash after 45 minutes of idle time it seems. I reduced it significantly in the process of troubleshooting the crashes, but I want to see where the limit actually is for my particular dose of silicon.

I have read that the 5600X can easily handle the full negative offset (limited to 30 units or about a 100-150mV reduction) where the more finely tuned 5950X can't (typically a negative offset of 5-10 units is all it can handle). But my specific CPU has two cores with basically bottom level CPPC tags - so it is most likely that it was just these two cores that couldn't handle the 25 unit offset.

CPPC tags basically tell the system which core is the best candidate for boosting and in which order additional core boosts should be applied. I know that cores 4 & 3 are the weakest on my chip, with 6 and 5 being the strongest performers. Ryzen Master indicates the strongest (and most overclockable) cores with a star or dot, but doesn't tell you anything about the weaker ones. ClockTuner for Ryzen did give me the full set of CPPC tags, but the simplest is just to go into HWInfo and find your perfromance order in there. Here is mine:

The Curve Optimiser in BIOS allows a 'per core' offset as well as the simpler 'all core' single offset entry. So, I shall first figure out the maximum 'all core' offset that remains stable (I'm down to -23 from -25 now and all seems well so far) - which should identify how far the weakest cores can be pushed, and then I can use the 'per core' function to increase the offset beyond that only for the cores that are more capable of handling a bigger undervolt.

So as an example, I might find that I can push cores 5 & 6 to the full available offset, with the other cores subjected to lesser offsets according to their ability.

So Ryzen allows you to indirectly control (and in my opinion this is the best way really) the behaviour of individual cores. A clever system!