Dynamic PX. Higher frequency. More stable operation. Lower temperature.
Added 2021-04-09 11:18:19 +0000 UTCThis video compares Dynamic PX with the stock BOOST performance.
HWINFO logging allows to show the dynamics of frequency change, from 4850MHz (base for PX MID) to 5000MHz (peak) on 4 threads.
Why can't CO do that? CO always works at standard voltage, it affects the telemetry values of the processor (cheats), due to which it achieves an increase in frequency. That is, you can get a significant increase in frequency, but the processor will not become cooler.
Dynamic PX works a little differently. My research showed that for Zen 3 silicon the voltage sweet spot is in a completely different range. The voltages should be lower, much lower than AMD suggests. Also the lower temperature of the silicon allows it to run more stable due to reduced static leakage currents. No self-heating which starts a chain reaction (temperature - > add voltage -> temperature). Second, I use completely different bias values for telemetry, which allows to go beyond the limitations of CO.
Another advantage of the dynamic profile is that the frequency will be lowered if necessary. This avoids the crash.
Stock
4775 MHz (avg)
1.469V
~68C
CTR non-dynamic PX
4850 MHz (avg)
1.375V
~60C
CTR dynamic PX
4950 MHz (avg)
1.375V
~60C
Release - next week.
I will also try to publish information about future plans ;)
Comments
This seems to be a thing now... A friend of mine on a Asus B550-E had to reduce RAM speed exactly the same way as you had to after upgrading from an older Bios to one with AGESA 1.2.0.1 patch A . He's also running a 5900. I also have a funny behaviour with my 5950 on the same board... Setting a fclk of 1900 for 3800MHz RAM speed, it does not boot. Going full yolo with fclk 2000 it will boot but produce the occasional WHEA. I gave up trying to understand it :) Still, going back to 3733CL14, everything is very stable.
Shinigami_de
2021-04-19 17:11:50 +0000 UTCI had set mine at 3733 from the start and never thought anything else of it. It is some older Trident B-Die from 4-5 years ago. But Yeah, 3733 manual was more stable than XMP 3600 . Just ordered some Royal CL14 32GB because I am getting to where that 16GB is beginning to be a limitation finally.
James Tiller
2021-04-16 06:15:19 +0000 UTCMine gives WHEA over 3600. Viper Steel 4133, currently clocked at 3600 CL14 How have you set up CTR 2.1 for your 5900x? results please ?
Jahanzeb Jahangir
2021-04-14 12:24:52 +0000 UTCI wasn't getting consistent results for the last few weeks on my 5900, I ran hours long linpack and testmem5 stability tests. I switched to the XMP profile to see if that helped and still I would have tests that stopped lower than normal or CCX1 would fail and CCX2 would keep on going. It was like I smoked my silicon or something. What fixed it for me was setting my Uclck and Memclck down to 3733 from 3800. Now my chip is staying stable where it should be. It's like having a new chip, since I really thought mine was damaged and nothing else helped. I have no idea why I didn't try turning the ram down a notch sooner, I tested using 3 of the best kits I could find - Gskill F4-3800C14D-16GTZN a set of F4-4000C15D- 16GVK and a set of Teamgroup 4500CL18, but always at 3800Mhz, I even upgraded motherboards from an Asus tuf x570 + to an Aorus master x570 and was about to order a new cpu. So if anyone is having OC results that are really inconsistent I hope this helps them. It seems that you can have problems from too fast a ram frequency even at 3800Mhz on premium B-die and passing thorough stress testing.
Tom DePaolo
2021-04-14 02:51:08 +0000 UTCBTW where is the high res wallpaper to be had..?
famich2005
2021-04-13 17:28:19 +0000 UTCI have set the BIOS according to the guide. A good point seems to be the article of Yuriy about CTR and overclocking. The better the MB, the better are the results. / Phase alternating, switching/.
famich2005
2021-04-13 17:24:33 +0000 UTChave PBO2 CU -30 set all individual cores - single threaded workload scores are a lot better than CTR. But if like even 3 threads are under load then sides are turned and PBO2 with its high temps cant keep up. (tried to even use negative offset to lower voltage... had effect on temps but not on perf scores) (my cpu is bronze-gold depending bios settings :D so its little relative) (btw i have same board asus B550-e)
tellimused
2021-04-13 13:16:05 +0000 UTCMy CPU is 5900x "golden sample" according to CTR Phoenix testing, ASUS B550 Strix E MB. I have not used CTR for tuning for over a month and the gap has narrowed , but is still there.
famich2005
2021-04-13 09:59:58 +0000 UTCIMHO,without the curve optimizer CTR cannot move forward, as I am still able to pruduce better results on my 5900x with PBO2 overclocking. Running at -26 allcore curve.
famich2005
2021-04-13 09:58:42 +0000 UTCNo, this motherboard has static OC. It has nothing to do with CTR.
Yuri Bubliy
2021-04-10 11:38:38 +0000 UTCIs this only for the Dark Hero that has dynamic OC?
GregT
2021-04-10 11:35:24 +0000 UTCPATRON FOR 3 MONTHS 但是刚刚才学会用这个网站
2021-04-10 08:26:39 +0000 UTCWow!
Larry Hope
2021-04-09 18:35:46 +0000 UTCSo impressed with how far you have taken this! super looking forward to this as well. 💪
Doug Wilkinson
2021-04-09 17:29:44 +0000 UTCLoving the software so far. Can't wait to try out the new release!
Kryptonic
2021-04-09 16:29:26 +0000 UTC
