Question Zen 3 and CTR

Makaveli

Diamond Member
Feb 8, 2002
4,550
873
126
I've just installed this and been playing with it a little bit.

What are you thoughts on CTR and have you used it?

5800X CTR Golden Sample.png

The Diagnostic game me this so far.

5800X CTR Golden Sample2.png

I've not done any of the tuning yet and curve optimizer is also turned off. The only thing on is PBO2 + 200Mhz.


 

uzzi38

Platinum Member
Oct 16, 2019
2,360
4,962
116
I have no trust in CTR at all and I don't see a purpose for it over just PBO2+CO.

Worse, I know people who's chips have degraded and are no longer stable at stock voltages as a result of this tool's testing. Not even the clocks and voltages suggested by it.
 
  • Like
Reactions: Tarkin77

Makaveli

Diamond Member
Feb 8, 2002
4,550
873
126
I have no trust in CTR at all and I don't see a purpose for it over just PBO2+CO.

Worse, I know people who's chips have degraded and are no longer stable at stock voltages as a result of this tool's testing. Not even the clocks and voltages suggested by it.
I did see a complaint about that in another forum. But that user also allowed it to run 1.55 vcore into chip and was not monitoring it. So even with its recommendations one still has to watch the settings applied.
 
  • Like
Reactions: lightmanek

Det0x

Senior member
Sep 11, 2014
905
2,378
136
This is a post i have written on a other forum, but i can paste it here also since its regarding static OC vs CTR vs PBO CO :)

I agree with you its better to run a static OC in programs that use all threads all the time, both in regards to clocks and temps/watt.
But i dont think a static CCX overclock is better in medium-high to low threadcount programs and/or games.. Nevermind the singlethread.

With my PBO CO settings on normal ambient temps i'm getting ~12100 points in Cinebench r20, which means my allcore effective clock is around 4650mhz (~1.6% lower then your average static OC @ 4725mhz.

But in apps/games that use something like 28,24,20,16,10 or even 8 threads etc i'm pretty sure my effective clock with PBO CO is much higher then with a static OC of lets say 4750mhz..

One other problem i find with a static OC is that if you maximize your clocks for a "light AVX workload" (~4700-4800 mhz) like Cinebench is, it will fail in a "heavy AVX workload" (~4400-4600 mhz) like prime95/y-cruncher etc. And if you maximize for "light" games/low threadcount (4800mhz+) it will fail in everything else.

Clocktuner 2.0's P0,P1 and P2 states tries to fix this but in my findings its too slow/buggy to switch between the different modes effectively..
Just try to run CPU-Z bench and see the singlethread suffer because it dont switch fast enough. And if you lower the "hangtime" it will start switching profiles all the time and lower your overall performance.


With my lucky/better than average 5950x, i think PBO CO is the way to go.. but that's just my 2cents on this matter :)
 

Makaveli

Diamond Member
Feb 8, 2002
4,550
873
126
This is a post i have written on a other forum, but i can paste it here also since its regarding static OC vs CTR vs PBO CO :)

I agree with you its better to run a static OC in programs that use all threads all the time, both in regards to clocks and temps/watt.
But i dont think a static CCX overclock is better in medium-high to low threadcount programs and/or games.. Nevermind the singlethread.

With my PBO CO settings on normal ambient temps i'm getting ~12100 points in Cinebench r20, which means my allcore effective clock is around 4650mhz (~1.6% lower then your average static OC @ 4725mhz.

But in apps/games that use something like 28,24,20,16,10 or even 8 threads etc i'm pretty sure my effective clock with PBO CO is much higher then with a static OC of lets say 4750mhz..

One other problem i find with a static OC is that if you maximize your clocks for a "light AVX workload" (~4700-4800 mhz) like Cinebench is, it will fail in a "heavy AVX workload" (~4400-4600 mhz) like prime95/y-cruncher etc. And if you maximize for "light" games/low threadcount (4800mhz+) it will fail in everything else.

Clocktuner 2.0's P0,P1 and P2 states tries to fix this but in my findings its too slow/buggy to switch between the different modes effectively..
Just try to run CPU-Z bench and see the singlethread suffer because it dont switch fast enough. And if you lower the "hangtime" it will start switching profiles all the time and lower your overall performance.


With my lucky/better than average 5950x, i think PBO CO is the way to go.. but that's just my 2cents on this matter :)
After I load the new bios on my board I will enable CO then start that testing per core which will take a couple hours to find the limits of each one.
 

Makaveli

Diamond Member
Feb 8, 2002
4,550
873
126
Curve optimizer -15 + PBO + limit PPT

I get same MT performance as stock with better ST/lightly-threaded while using less power.

Win, win, win.
I'm testing out CO now.

Are you doing -15 on your best cores?

what about the rest?
 

Noid

Platinum Member
Sep 20, 2000
2,308
118
106
{ copied from the 5000 builders thread }

Just ran the diagnostics ---- with a hard reset after 5 mins
( never completed )
Passed it's own 10 min stability test at it's selected 4350Ghz
( boost mode )

And he says it's " safe " ? lol

Oddly ---- it's says I have a " bronze" sample.
( maybe a indicator of it's underclocking ability ? if so, Ok. I believe that )

The analysis is questionable.
Afterall, I have not see anyone yet to surpass my results.

It just crashed again using the " tuning " button.

Apparently, I can't run this at my current settings lol.
( Or, it's undervolting so badly --- it's unstable )

Uninstalled
(I'm re-running 95 again -- now @ 4.5 GHz with boost - 69C)
 
Last edited:

Det0x

Senior member
Sep 11, 2014
905
2,378
136
This is a post i have written on a other forum, but i can paste it here also since its regarding static OC vs CTR vs PBO CO :)

I agree with you its better to run a static OC in programs that use all threads all the time, both in regards to clocks and temps/watt.
But i dont think a static CCX overclock is better in medium-high to low threadcount programs and/or games.. Nevermind the singlethread.

With my PBO CO settings on normal ambient temps i'm getting ~12100 points in Cinebench r20, which means my allcore effective clock is around 4650mhz (~1.6% lower then your average static OC @ 4725mhz.

But in apps/games that use something like 28,24,20,16,10 or even 8 threads etc i'm pretty sure my effective clock with PBO CO is much higher then with a static OC of lets say 4750mhz..

One other problem i find with a static OC is that if you maximize your clocks for a "light AVX workload" (~4700-4800 mhz) like Cinebench is, it will fail in a "heavy AVX workload" (~4400-4600 mhz) like prime95/y-cruncher etc. And if you maximize for "light" games/low threadcount (4800mhz+) it will fail in everything else.

Clocktuner 2.0's P0,P1 and P2 states tries to fix this but in my findings its too slow/buggy to switch between the different modes effectively..
Just try to run CPU-Z bench and see the singlethread suffer because it dont switch fast enough. And if you lower the "hangtime" it will start switching profiles all the time and lower your overall performance.


With my lucky/better than average 5950x, i think PBO CO is the way to go.. but that's just my 2cents on this matter :)
Wanted to explore this alittle further, so i started to test the clockspeed+corescaling in cinebench23 for my PBO CO 5950x. (started 30min multicore benchmark and forced the threads around with affinity in taskmanager)

Do note that this is a "light AVX" workload, which is a heavier workload than 99% of the games out there, so my PBO CO clockspeed would actually be higher in real games.

Recorded sustained effective clocks with 1 thread per core (this is how windows assign threads in light loads)
1 core load = 5080 mhz
2 core load = 5060mhz
3 core load = 4950mhz
4 core load = 4832mhz
5 core load = 4825mhz
6 core load = 4800mhz
7 core load = 4800mhz
8 core load = 4772mhz
9 core load = 4748mhz
10 core load = 4744mhz
-
-
-
-
-
16 core load = 4650mhz
Recorded sustained effective clocks with 2 thread per core
1 core load = 5050mhz
2 core load = 4930mhz
3 core load = 4882mhz
4 core load = 4798mhz
5 core load = 4771mhz
6 core load = 4750mhz
7 core load = 4748mhz
8 core load = 4707mhz
9 core load = 4685mhz
10 core load = 4669mhz
11 core load =
12 core load =
13 core load =
14 core load =
15 core load =
16 core load = 4650mhz

threadscaling-2.png
 

Attachments

Last edited:

ASK THE COMMUNITY