1usmus Power Plan for AMD Ryzen - New Developments

[Weshya]

Ryzen Master did not change the algorithm for determining core ranks.
Also, I do not see any changes in the power profiles. "Duty cycle" is also disabled by default for Ryzen Balanced. That is, the problem was presented to the public not as a bug but as a features.
Perhaps they did not have time.

Thank you for your reply. Is "Duty Cycle" available when choosing Ryzen Balanced + Best performance power mode, or at least in "AMD High Performance" power plan ? In any case, this power mode slider definitely changes the behavior of the power plan in use.

Edit : It would be great to know how this power mode slider affects your universal power plan too. You haven't observed any change at all when using the "Best Performance" ? Am i the only one to notice even more aggressive core boosts among other things ?

 

[Artex]

Gonna call this a win!

 

[X570-3900X-DE]

"Best Performance" ? Am i the only one to notice even more aggressive core boosts among other things ?

I did say this also that it was better...to go to best performance. So its worth to check

Gonna call this a win!

View attachment 137729

should be a little higher... check the setting in the power plan go to best performance, i did gain there also i am max 7381 with 1867mhz as PBO off no OC.
Please kill all tasks that are not needed to make sure you get the real thing...

i would recommend also that you use 1usmus memory calculator for ryzen, that also again me more there and on the rest tests... you can see my test in some post above, I hope you get to it also, as i was very happy to see 4650 first times...

as you have the same board and same CPU should be close then, you did take F10 not other F10C, F10A or so BIOS

 

[Artex]

I did say this also that it was better...to go to best performance. So its worth to check


should be a little higher... check the setting in the power plan go to best performance, i did gain there also i am max 7381 with 1867mhz as PBO off no OC.
Please kill all tasks that are not needed to make sure you get the real thing...

i would recommend also that you use 1usmus memory calculator for ryzen, that also again me more there and on the rest tests... you can see my test in some post above, I hope you get to it also, as i was very happy to see 4650 first times...

as you have the same board and same CPU should be close then, you did take F10 not other F10C, F10A or so BIOS

Thanks! I will definitely try the calc and power settings. I'm just happy to see these cores boosting to 4.6. I did update to the F10 bios from F10C.

 

[X570-3900X-DE]

yes i was also very happy the first time.. as before all i never got more then 4400 or so if i was lucky. if you need to see my BIOS check here... maybe some stuff you missed

now i get 4650 on a few cores... from time to time... the other setting in the plan kicked it up

let me know if you need more help then

 

[panni]

Hey @1usmus thank you for all you're doing for the community.

I've just read about the 1.1 version of your power plan (got a 3700X on a Gigabyte X570 AORUS Pro, still 1.0.0.3 ABBA, 1903).
After doing a binary comparison of your initial release of the power plan ("1usmus Ryzen Power Plan.pow") with the new one ("1usmus Ryzen Power Plan pre-1909.pow"), it's exactly the same (SHA1: 5730d3b21e88c9c0d4c4b8d4cb385a12c3e092a3).

Did I interpret the article wrongly or has the non-universal power plan not changed *at all*?

If this is intended, that might not be how the users see it, as I've seen people looking for ways to delete duplicated power plans - which happens after using your first release, then install.bat of the 1.1 release. Perhaps that should be clarified?


Thanks

 

[willgart]

Hey @1usmus thank you for all you're doing for the community.

I've just read about the 1.1 version of your power plan (got a 3700X on a Gigabyte X570 AORUS Pro, still 1.0.0.3 ABBA, 1903).
After doing a binary comparison of your initial release of the power plan ("1usmus Ryzen Power Plan.pow") with the new one ("1usmus Ryzen Power Plan pre-1909.pow"), it's exactly the same (SHA1: 5730d3b21e88c9c0d4c4b8d4cb385a12c3e092a3).

Did I interpret the article wrongly or has the non-universal power plan not changed *at all*?

If this is intended, that might not be how the users see it, as I've seen people looking for ways to delete duplicated power plans - which happens after using your first release, then install.bat of the 1.1 release. Perhaps that should be clarified?


Thanks

hey, quick question for you...
I also have a Ryzen 3700x and what I dont understand is the lack of power during benchmarks.
for example, with R20, my 2 cores used by windows are reaching 4250 mhz only, (score 470)
and in multithread, I can get between 3800 and 3900 mhz. (score of 4500)
and so my results out of the box are lower than the other benchmarks
based on the different articles on the subject, I expected to have 4000 to 4100mhz on all cores during benchmark

can you share the frequency you get out of the box?
and for me the PBO maxed out did not change anything at all.
is it the same for you?

after, when I manually overclock, I can push all the cores to 4250mhz and hwinfo , after few hours of activity, show that 2 cores reached 4400 mhz. so its possible to be high, but the benches cant do it!

I'm not sure if the problem is the cpu or the MB (asus tuf gaming x570 wifi for me)

 

[nko]

Guys i've got a question about temperature.
My set is: MSI B450 Tomahawk Max (last stable bios and last AMD Chipset update) and AMD Ryzen 7 3700x

And i have a big difference (something like +/-10°c) between Ryzen Master CPU temperature report and MSI Bios (and other tools like HWinfo, AIDA64,... )
And more annoying, temperature report by those tools and bios is pretty unstable constantly moving between a large range of -5°c +5°c around the middle value.

That's a quite annoying problem at least for bios, more than the apparently "wrong value", those temperature are used to adjust fan curves settings and with those fast fluctations fan speed in quite unstable even with a 0.3 latency (maximum possible setting in my bios)

Some of you also encounter this kind of problem ? Do you have a solution ?

Thank you for reading and thank you for your answer

.

 

[X570-3900X-DE]

same here temps reading is not easy not sure tool are right... i have 2 die and get different all over, i gave up

ok i got mine on water, maybe you set it on value % min and max if that goes and say after this 50C go on this % value

if you could set the value as you like you just say do this % after this temp, i had my last system pump on this setting with the tems on CPU

 

[willgart]

Guys i've got a question about temperature.
My set is: MSI B450 Tomahawk Max (last stable bios and last AMD Chipset update) and AMD Ryzen 7 3700x

And i have a big difference (something like +/-10°c) between Ryzen Master CPU temperature report and MSI Bios (and other tools like HWinfo, AIDA64,... )
And more annoying, temperature report by those tools and bios is pretty unstable constantly moving between a large range of -5°c +5°c around the middle value.

That's a quite annoying problem at least for bios, more than the apparently "wrong value", those temperature are used to adjust fan curves settings and with those fast fluctations fan speed in quite unstable even with a 0.3 latency (maximum possible setting in my bios)

Some of you also encounter this kind of problem ? Do you have a solution ?

Thank you for reading and thank you for your answer

.View attachment 137924

I'm seeing the same issue
the temperature is jumping 10deg, slow down 10deg in 5 sec and jump again etc... a constant cycle.
and the ryzen master tool dont display these changes.
I'm using the default ryzen cooler, and I can see that the fan speed is following with small variations, so both are linked (the fan make it cooler).
this occurs as soon as there is a little activity, and because there is no perfect idle case in windows, this jumps always occurred ;-)

 

[Artex]

I'm seeing the same issue
the temperature is jumping 10deg, slow down 10deg in 5 sec and jump again etc... a constant cycle.

Seeing this issue verbatim - jumps 10 degrees, drops slowly in a few-degree increments and then jumps 10 degrees over again.

 

[willgart]

Seeing this issue verbatim - jumps 10 degrees, drops slowly in a few-degree increments and then jumps 10 degrees over again.

ok, finaly when windows is really in IDLE mode, no more background process running after the startup (ie waiting 15/30min) its now stable.
no more up/down.
so it took some times to windows to terminate all its startup activities.

 

[nko]

For me this temperature instability is constant even after windows startup, computer in idle (0% - 1% CPU usage) or in charge.
And this is the same with all performance plan (windows, ryzen, or 1usmus universal), no difference on this.
This had certainly something to do with the way bios read the temperature sensor with that CPU... Quite strange... and quite annoying

 

[willgart]

For me this temperature instability is constant even after windows startup, computer in idle (0% - 1% CPU usage) or in charge.
And this is the same with all performance plan (windows, ryzen, or 1usmus universal), no difference on this.
This had certainly something to do with the way bios read the temperature sensor with that CPU... Quite strange... and quite annoying

maybe its related to the idle temp and fan curve setup. now I'm below 50deg with normal usage of my PC (word, mails, browser...) and no more spikes.

another element I found, the PCH temperature and fan (fan on the motherboard for the x570 chipset) was low after the reboot and it appear to be higher and constant,
I was around 800rpm and now 1700rpm. its like the x570 chipset is now stable.
so maybe its related to the temp of other elements on the motherboard, when too cold, there are some variations causing variation to the CPU, once the warmup phase done, all the components are stable and so the CPU too.

 

[529th]

Thread planning policy used for short-term threads in heterogeneous systems has been changed. Also we force scheduler to use the best cores (if possible) .
By default, the auto option is enabled there. That is, windows may use the best cores, or it may not.

How can I force the scheduler to use the best core?

 

[willgart]

How can I force the scheduler to use the best core?

its already the case.
but, in fact the system search for the best 2 cores, (the best side by side sharing the same cache)
so even a single thread is spread to 2 cores by default.
in my case its the best core 2 and 4 (as per the AMD ranking) which are mainly used
during tests, the core 2 deliver the same performance has my core #1. so even if its ranked 2 its the same result.

 

[X570-3900X-DE]

i am still not sure if the tools show right... i got 20c jumps, well I will not look to much for that better. As I did new tests and the system gained more... just did change my monitor but don't think that is why... ok i did text with XMP and XMP off but for some reason it like the XMP setting with custom value more...

so crazy how much more have...

when I started i here had CPU-Z 543 single 8392 Multi, now i have 555 single and 8516 so i think the temps are not the issue..

 

[Wickedt]

I would love for somebody to write a program to get the temp from HWInfo and the top 3 processes from windows at that temp.
These CPU'S are weird as hell, but they kick major butt in anything they touch.

hmmmm maybe grab the top 2 cores frequency too.

 

[korzychxp]

I have Ryzen 3600X. It can't do 4400Mhz on cinebench single core.

usmus power plan didn't help.

Windows 1909 update ruined RAM latency (AIDA test) for about 1-1.5ns.

1909 Update ruined game performance - in Far Cry Primal i have 2 fps less in benchmark.

I have newest bios for B450 A Pro.

Even +200 boost override doesn't work. Anyone can help? Please PM me if we can't discuss it here.

 

[willgart]

I have Ryzen 3600X. It can't do 4400Mhz on cinebench single core.

usmus power plan didn't help.

Windows 1909 update ruined RAM latency (AIDA test) for about 1-1.5ns.

I have newest bios for B450 A Pro.

Even +200 boost override don't work.

same issue with my 3700x
I'm not able to push a single core during benchmarks using PBO and other auto tuning options in place. (no matter the overclocking option I try, the PBO config, VRM setup etc etc etc... I pushed everything to the max but no change!!! and most of the time its worst)
I dont know if the problem is the CPU or the MB.
its like the system didnt try at all to push the core used.
but if I let my PC active for few hours and I do normal activities, HWinfo64 records a spike at 4350mhz and sometimes 4375mhz. but never during the benchmarks.
and based on the 3700x reviews, I'm far below what they reached in R20 and other benchs.

currently my best setup is:
slightly overclocked RAM
CPPC and other 1usmus recommandations activated
Ryzen Balanced plan selected, the 1usmus one gives me lower performance.

 

[Calmmo]

I finally gave this a try, custom plan vs ryzen balanced is giving me a small boost in pretty much every benchmark. I've seen 4650 once in regular use and once also reported in 3dmark




Now if only PBO actually worked and didn't give reduced scores in every single bench..!..

 

[X570-3900X-DE]

nice, close to mine...

 

[HoloTheWise]

Does anyone happen to know if there are still benefits to be gained using 1usmus' updated power plan on the latest chipset/bios/OS updates? Or have all the changes been corrected and merged upstream?

Windows 10 v2004 - OS Build 19037.1 (12/06/2019)
Asus X570 C8H - Bios v1201 (12/03/2019)
AMD X570 Chipset Drivers - v1.11.22.0454 (11/25/2019)

The updates are newer than the publication of the v1.1 power plan so I'm a bit lost as to which to use (amd balanced vs 1usmus balanced).

 

[willgart]

Does anyone happen to know if there are still benefits to be gained using 1usmus' updated power plan on the latest chipset/bios/OS updates? Or have all the changes been corrected and merged upstream?

Windows 10 v2004 - OS Build 19037.1 (12/06/2019)
Asus X570 C8H - Bios v1201 (12/03/2019)
AMD X570 Chipset Drivers - v1.11.22.0454 (11/25/2019)

The updates are newer than the publication of the v1.1 power plan so I'm a bit lost as to which to use (amd balanced vs 1usmus balanced).

with all the updates from Asus and AMD I got.
now I'm using the AMD standard or high performance plans, they performs better than the 1usmus one

in CPU-Z I get 504 to 510 points in single thread, while I got 470 with the 1usmus plan. (3700X used)
also I can see a big difference in the core used during the tests

but at the end, no matter the setup, a manual overclocking performs better ;-)
I change the thermal paste and use the AMD stock cooler, I can reached 4350mhz to all cores which works most of the time (R20 failed above 4300mhz)

 

[X570-3900X-DE]

my 3900X is still better with the 1.1 universal plan