• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

A bit confused about curve optimiser on my 5950x

Soldato
Joined
12 Feb 2014
Posts
2,932
Location
Somewhere Only We Know
As the title suggests, im a bit confused about curve optimiser on my 5950x, I instantly set -15 per core across the board and instantly failed prime95 small fft's, even with curve optimiser disabled I was failing, so i ditched prime 95 and started using various other tests like intel burn in test which i know tests avx, and also 3 different versions of cinebench, i started the tests at -10 and passed everything, so now apart from 2 cores, everything is passing at -15 including prime95, ive still yet to adjust the other 2 cores......go figure ?????

my PPt is at 250 and using 232w according to hwinfo
my edc is at 170 and using 170
my tdc is at 165 and using 155 according to hwinfo

earlier on I was using a guide on youtube were he pretty much ramped all of these upto 500, could that of caused instability ?
 
After a lot of time trying to get the best out of my 5900x i learnt a few things...

First, are you using Ryzen Master or the Bios to change the settings? generally RM works well but there can be a few options that need to be set in the bios as well.


in the Bios, enable XMP, PBO in Advanced mode and Motherboard limits

In RM choose the curve optimiser option on the left, select the PBO and the ppt tdc edc limits should change to suit.

without anything else running, run the curve optimiser. this will take around an hour,

once it completes click to accept the results and then click apply and reboot


Hopefully this will complete without crashing!

(be careful of using any of the other presets in RM as it will always disable the CO... there is no option to keep it enabled except in the CO preset)


next try running some benchmarks like R23 or Performance Test, or Prime95, OCCT etc

while running the tests keep HWinfo open and monitor the core speeds and at the bottom for WHEALogger errors.

If the system crashes or errors are present check the event viewer - windows logs - system for which core (APIC ID) if crashing

(half the APIC ID to get the core no.) and increase the CO offset by 5 of that core and test again

Now for PPT and TDC, the standard PBO setting is fine. I have however learnt that the EDC is better kept under 200A, for my 5900x 180A seems the best, you will find the core clock speeds will run higher if you get the EDC limit and the CO offset right

Also, just turning the limits up to 500 will probably blow your mainboard VRMs and the CPU to the moon...


A Final note if the WHEALogger errors are BUS errors then your FCLK (fabric clock) could be too high.. i had mine set to 2000 with DDR4000 ram but it wasnt happy no matter what voltages i tried, changed it to 1900 and ddr3800 (always stay coupled for FCLK and DDR)
 
Last edited:
Thanks, im not using ryzen master, I didnt know you could use it to set CO, I thought it was just and overclocking utility, and ive always been a bios person for overclocking, so thats where ive also been adjusting CO.

everything passed at -15 so im now working on -20 core by core, I know its a slow process, but getting through all the cores in an evening, im just running a couple of passes of CB15 and 23, 15 mins of prime and 15 mins of IBT, i'll do a more through test when everything is where I want it, to be honest, id be happy if almost every core made it to -25.
 
I did per core curve optimiser using CoreCycler to test stability and adjusted to which core failed till it didnt fail it took sometime going through the cores

before that I used normal tests to verify and even if it looked stable but crash few times per week mainly when idle since running CoreCycler I havent had any crashes for weeks 7 cores I have at -30 others range from -16 to -29
 
Last edited:
I did per core curve optimiser using CoreCycler to test stability and adjusted to which core failed till it didnt fail it took sometime going through the cores

before that I used normal tests to verify and even if it looked stable but crash few times per week mainly when idle since running CoreCycler I havent had any crashes for weeks 7 cores I have at -30 others range from -16 to -29

Thanks, ive seen CoreCycler, how long did you let it run before adjusting the CO for another core ?

Did you use the default Prime95 6 mins per core ? and did you use AVX, or SSE ?
 
Last edited:
Thanks, im not using ryzen master, I didnt know you could use it to set CO, I thought it was just and overclocking utility, and ive always been a bios person for overclocking, so thats where ive also been adjusting CO.

everything passed at -15 so im now working on -20 core by core, I know its a slow process, but getting through all the cores in an evening, im just running a couple of passes of CB15 and 23, 15 mins of prime and 15 mins of IBT, i'll do a more through test when everything is where I want it, to be honest, id be happy if almost every core made it to -25.

Ryzen Master allows you to adjust almost every setting you get in the bios for CPU and memory... I know its hard to trust a windows program over directly in the bios but it does speed up making ajustments. Plus the Curve optimiser option will go through all the cores one by one and test what they are stable at all in one pass.

It is still possible to get instabilities but it gives a solid ball park to work from..

and technically, ryzen CPUS use undervolting to allow it to overclock itsself, this is why you find higher core clocks from lower offsets and limiting the EDC
 
Thanks, ive seen CoreCycler, how long did you let it run before adjusting the CO for another core ?

Did you use the default Prime95 6 mins per core ? and did you use AVX, or SSE ?

cant remember where I saw the guide I set CoreCycler to run 3 minutes per core prime95 and let it run 5 Interations everytime core failed I dailed it back by 1 till no error I did the adjusting all in the bios much prefer that as I like windows to run as lean as possible :)

its been fine for weeks now , before I seemed to settle for -24 all cores passed tests etc but would randomly restart on idle few times a week so decided to try core cycler and do per core instead
 
Ryzen Master allows you to adjust almost every setting you get in the bios for CPU and memory... I know its hard to trust a windows program over directly in the bios but it does speed up making ajustments. Plus the Curve optimiser option will go through all the cores one by one and test what they are stable at all in one pass.

It is still possible to get instabilities but it gives a solid ball park to work from..

and technically, ryzen CPUS use undervolting to allow it to overclock itsself, this is why you find higher core clocks from lower offsets and limiting the EDC

Just gave RM a try, went to the bios and set PBO to active and motherboard limits etc as you explained, I then downloaded RM selected CO per core, inputted my PPT, EDC, TDC values, hit optimise, got an instant black screen, had to switch the PSu off, upon restart the bios had reset.
 
cant remember where I saw the guide I set CoreCycler to run 3 minutes per core prime95 and let it run 5 Interations everytime core failed I dailed it back by 1 till no error I did the adjusting all in the bios much prefer that as I like windows to run as lean as possible :)

its been fine for weeks now , before I seemed to settle for -24 all cores passed tests etc but would randomly restart on idle few times a week so decided to try core cycler and do per core instead

Thanks, I kind of read a bit of a guide which said to run each core for 3 to 5 mins, the 2 main tests that find errors the fastest FFT 128 and FFT 720, also to run them twice before making adjustments, once in SSE mode and once in AVX2 mode.

I prefer to run settings in the bios too, windows boot is a good benchmark in itself.

This was the first run, I cant work out if this is core 3 or core 6, so ive dialed them both back a little.

QbvHxZo.jpg
 
Last edited:
Thanks, I kind of read a bit of a guide which said to run each core for 3 to 5 mins, the 2 main tests that find errors the fastest FFT 128 and FFT 720, also to run them twice before making adjustments, once in SSE mode and once in AVX2 mode.

I prefer to run settings in the bios too, windows boot is a good benchmark in itself.

This was the first run, I cant work out if this is core 3 or core 6, so ive dialed them both back a little.

QbvHxZo.jpg

that would be core 3
 
Just gave RM a try, went to the bios and set PBO to active and motherboard limits etc as you explained, I then downloaded RM selected CO per core, inputted my PPT, EDC, TDC values, hit optimise, got an instant black screen, had to switch the PSu off, upon restart the bios had reset.
thats not ideal is it... selecting PBO in ryzen master should automatically apply the appropriate limits. i could only assume there was a bios setting you had made that may be conflicting... i can only recommend to load defaults in the bios (i assume it has already...) and try again with out changing any settings apart from in RM..

but it seems your trying CoreCycler, which is a good alternative. Id still try the RM CO test, running the test doesnt automatically apply any changes unless you accept them at the end.. it would be good to make sure it doesnt crash again tho.

I have seeming got my PBO boosting sorted but am getting idle crashes instead now! .. seems windows / the PSU get confused when all the cores go to sleep and it thinks its shutdown
(also like to point out i have a 12700K system that is also randomly crashing on a seemingly stable setup) shame we cant leave these things alone eh! - or the CPU makers shouldnt nobble the core clock speeds out of the box so much!
 
thats not ideal is it... selecting PBO in ryzen master should automatically apply the appropriate limits. i could only assume there was a bios setting you had made that may be conflicting... i can only recommend to load defaults in the bios (i assume it has already...) and try again with out changing any settings apart from in RM..

but it seems your trying CoreCycler, which is a good alternative. Id still try the RM CO test, running the test doesnt automatically apply any changes unless you accept them at the end.. it would be good to make sure it doesnt crash again tho.

I have seeming got my PBO boosting sorted but am getting idle crashes instead now! .. seems windows / the PSU get confused when all the cores go to sleep and it thinks its shutdown
(also like to point out i have a 12700K system that is also randomly crashing on a seemingly stable setup) shame we cant leave these things alone eh! - or the CPU makers shouldnt nobble the core clock speeds out of the box so much!

This is sort of normal behaviour, low load crashes, most people test CO using things like cinebench, prime, OCCT etc which are all high load applications, and then get no load or zero load crashes, thats why with CoreCycler they recommend testing in SSE mode, which is not completely zero load, but also not really what I would consider as high load, my fans for instance are not raging away, my CPU temp right now on the core currently being tested is 65oC, the others are sitting at idle at 35oC, and then test again in AVX2 mode which is high load, will hit about 85oC then.

 
Last edited:
This is sort of normal behaviour, low load crashes, most people test CO using things like cinebench, prime, OCCT etc which are all high load applications, and then get no load or zero load crashes, thats why with CoreCycler they recommend testing in SSE mode, which is not completely zero load, but also not really what I would consider as high load, my fans for instance are not raging away, my CPU temp right now on the core currently being tested is 65oC, the others are sitting at idle at 35oC, and then test again in AVX2 mode which is high load, will hit about 85oC then.

Yeah, the confusing new world of undervolting overclocks and idle power c states.. I have switched to typical current idle which should stop the shutoffs... Time will tell
 
that would be core 3
When running 768K in SSE mode im passing on every core.

When running 768K in AVX2 mode its failing on every core saying there was not enough power to sustain the core, is there a way around this ? maybe setting a +0.0625v offset on the CPU voltage, or do I need to back off on the CO for every core, currently they are at -20 for every core excpet core 3 which is at -15.

9NfOi9L.jpg
 
Last edited:
When running 768K in SSE mode im passing on every core.

When running 768K in AVX2 mode its failing on every core saying there was not enough power to sustain the core, is there a way around this ? maybe setting a +0.0625v offset on the CPU voltage, or do I need to back off on the CO for every core, currently they are at -20 for every core excpet core 3 which is at -15.

9NfOi9L.jpg

Pretty sure I didn't run that test. I also have motherboard limits set to disabled just wanted to improve without going above what power is used at stock and slightly be above stock performance

It's been probably 4 weeks now and used everyday and left at idle for hours so far no issues
 
Pretty sure I didn't run that test. I also have motherboard limits set to disabled just wanted to improve without going above what power is used at stock and slightly be above stock performance

It's been probably 4 weeks now and used everyday and left at idle for hours so far no issues
Thanks, I tried running it again with AVX and SSE and it passed, it just failed on AVX2, so I disabled CO in the bios and ran it again, and it still failed lol, I think its a bug personally, its passing everything im throwing at it.
 
Posted the same thing over on another site on the dedicated Core Cycler thread, quite a simple fix really, there is no 720 AVX2 test, it was just simply confusing it, I was advised to set it to 768K and it ran through fine, nearly finished with all the testing now, i'll be at -30 for every core except 1 which will be at -18.
 
Back
Top Bottom