1usmus
patreon
CTR 2.1 RC5 hotfix (ver. 16) tomorrow!
🕑 Added 2021-04-26 07:11:56 +0000 UTCI have reviewed your reports and requests in detail and an update will be ready tomorrow!
Also tomorrow I will tell you about further plans. They are there and they are ambitious ;)
Comments
Yuri Bubliy
Hey!I lowered the base frequency so that the profiles switch are softer. The resulting frequency is independent of the base frequency you see in the profile. The maximum frequency is controlled only by the OB. Curve Optimizer values you see in the diagnostics should be used if you are not going to use CTR.
Sisay
Welcome back. After a month of break I sat down to the new version :) Compared to RC5, at 1050mv there are much lower values in calculation p1 (4150 to 4075). I also do not understand this curve optimize, I calculate it, do some tests on the clock around 3950mhz and set the OB. Is it for 3950mhz? Should I stick values for curve optimize in the bios?
Yuri Bubliy
MIX = FAST, AVX = SAFE. If you read the manual, it states that the profile base frequency is not the final frequency. Even if you use 3800MHz for the PX, the maximum boost will not change. The base frequency exists only for soft switching of the profile.
Tanveer Jugpal
No more fast preset for PX, numbers are the lowest I've gotten, but, I will just run with what it gave me.
84wayz
These Gigabyte boards do be having some serious vdroop. I feel your pain. I have an x570 aorus master and the vdroop keeps my ideal settings between turbo and extreme also. I've found that these llc settings bring my vdroop to the ideal range and i've been running it without any issues.
Peter Liu
OK Thanks. I'll keep voltages under 1.375V.
Yuri Bubliy
Thank you, guys for your trust, I appreciate it.
DeeJay
Totally with you Brandon and Doug, this simply is cutting edge and I appreciate Yuri's efforts to further optimize those mass silicon products which still face a lot of variation: both in in terms of quality (also a result from current pandemic shortages and manufacturers trying to speed up the production), and board partners who interpret specifications differently, make mistakes and are just one part of the chain of failures that can impact quality visible to us customers. For me, CTR is a genuine attempt to even fix some things that manufacturers pushed under the carpet, and so a good investment even if 2.1 still bit embryonic. No matter the term, we all know this is in (very active) development, have accepted terms that say things may change any time, so by no means people expect being close to a stable release (and by the way, no software is bug-free ever).
Yuri Bubliy
The main condition is not to use voltages higher than 1.375V. Let's try Extreme.
Andrea Squadrani
i'm on Zen 2 too, CTR for our CPU is a must, AMD stock settings are absolutly insane in terms of temperature/power consumption and performance, with CTR i get much better performance with much low heat on the CPU, we will never have the 1-click diagnostics, but hey, it's just a little bit of extra work (and for me, fun)
Peter Liu
My Vdrop is around 5%-6% even with Turbo setting. Is it safe to push to Extreme or Ultra Extreme? My board: Gigabyte B550I AORUS PRO AX.
Derève
I'm on the starting block =)
Gleison Rodrigues
waiting 2
Ricky Campbell
waiting
famich2005
And that at 1.31 V , at 1.3 both CCX stable on OCZ, Cinebench. That CPU was labeled by CTR as silver , another RC as platinum etc etc ?...????
famich2005
Well, support, sure. My point being that all this “RC” are merely alpha versions and it may take long time till the final release is ready. This time I have had luck with 5950x and yesterday after minor tweaking I was able to boost up to 5050 MHz in Cyberpunk.
LC FX
Happy to lend a hand.
Yuri Bubliy
I think yes, I am still testing the build... :(
Alberto Braghieri
Sorry Roy, my limit, I'm quite lost.
LC FX
Tomorrow now?
Doug Wilkinson
CTR is already in a "good enough" place for some of us already, and we've logged many hours gaming and productivity with CTR running without crashes and with cooler and quieter systems. I too have multiple machines and configs over here (zen2 and zen3) - each one behaves differently and CTR gives all better performance per watt vs stock or PBO+CO, which is the goal of the project. So for many of us things are working well already. I appreciate the breaking changes since this means development is moving forward and I get to provide feedback on it. If a version breaks i go back to the previous that worked well.
Patrick Bowman
I think Roy has a ignorant (this is a term not an insult) view on what Alpha, Beta, and Release Candidates are. The quality of the build does not define the category state change. These are all defined terms that tell where in the cycle a project is. Generally speaking: Alpha: There's going to be bugs. There's going to be new features. Features will change frequently. Beta: There's going to be bugs. New features are usually prohibited. Some features may change. Release Candidate: No new features. No features changing in core design. Bug's that are severe enough will stop the release. Upon proper testing over time it will advance to Silver or Gold master.
Doug Wilkinson
i just want to say i feel exactly the same way! I am supporting Yuri more than CTR. Im happy to support him because i really appreciate this project. I'm also happy i get to test something bleeding edge since like you, i'm going to end up tinkering no matter what. I have a Zen2 processor and there is no curve optimizer so CTR has been really fun to test with.
Brandon Kneeland
This latest update im hitting higher frequencies at lower temps than i ever have and I never crash its not perfect it does take some thought on your end and to follow the instructions but for my sample size of 1 im very impressed max stable PBO + CO i ever would get is 4850 with high voltage using max LLC sometimes I can hit 4950+ stable with this release with less voltage and lower temps and before that 4875 I always love playing with new tech and always wanted a way to OC without losing the boosting effects on ryzen its not perfect but its the only project im aware of in this space im glad we can all air grievances but some of this feels unwarranted to me its also been my understanding that this is a beta and we are paying to support to have access to it and help to test. (Having the right LLC stable dram OC and turning off PBO were critical for good results) Thanks for everything Yuri!
Michael
Can u give cb20 multicore example scores and frequencies for stock and ctr, please?
Roy Berry
Alberto; I find it hard to believe that my experiences are so quickly dismissed. I have never had CB bench as high as pre tuning with CTR on any 3 of my systems (ver 2, and 2.1) - both Zen 2 and Zen 3 processors.. I have never pushed PBO+CO yet on the Zen 3 systems. @Yuri - I understand you spend many hours on the project, and it's a great project, it's one of the reasons I am a Patreon subscriber, EVEN THOUGH I can't make full use of it. I never mentioned max performance - I never mentioned mythical frequencies - I just said I cannot get "better" performance with CTR than without (no PBO+CO); and that I can relate to these other guys - the software doesn't feel, nor resemble a "RC" product, and to give them shit about "software dev" when this is a clear fact is just callous.
Yuri Bubliy
Roy, tell me what's bothering you?
Yuri Bubliy
Regardless of how it is called, I spend more than 14 hours every day on the project. Our team's testing may differ from the results some users get. Some processors have no reserve, some users don't read the manual, and so on. I would also like to remind you that the CTR project is primarily about getting maximum energy efficiency. It's about not using AMD's high voltage or mythical frequency with stretching. The weighty argument is performance relative to temperature.
Tanveer Jugpal
Same
Alberto Braghieri
We can for sure debate whether or not it's an Alpha/Beta or RC, but sincerly I find hard to believe that CTR have had ZERO runs higher than stock settings. In my experience the results are close to the ones of PBO+CO e in some cases even better (Cinebench 20 about 8700 vs 9000 with lower temperatures) but for sure far better than stock.
Yuri Bubliy
This month 16 patches for CTR were created, users receive only 2. These are the most stable versions and innovations. I try as much as possible to save users from using software of improper quality.
Yuri Bubliy
You can use your settings if they are stable, I have nothing against it.
Yuri Bubliy
New version new diagnostic :) This is the rule for all releases. This will allow you to get rid of the mistakes that were made earlier.
Roy Berry
I hate to join in, but these guys are right - there has been zero releases that have been "release candidate" - they have all been beta at best. and before anyone misunderstands our (my Position) we're here as supporters - and believe in what is being done; and love the outlook of what is coming. but we should get less "ground breaking programming changes" when there isn't even solid base (that doesn't change rapidly). this isn't RC(X) territory, this is Alpha/Beta Territory. for example, I have had ZERO runs on my 5800x/5900x that have benched higher than stock on my system; and that's only counting the runs that don't crash; that's not a program that's working; and not a program that I could agree would be RC - even discord is confused at some of the issues that are circulating.
Yuri Bubliy
At the moment my B550I has an inverted LLC. This is a gross mistake made by ASUS programmers. The second reason is the temperature of the processor. The higher the voltages the lower the frequency will be. The default voltages are the preset from me. They have the best power efficiency.
Norman P. Aquino
Yes, it's called a "release candidate" for a reason.
Nickolay Semendyaev
Probably because of cpu temp. Higher temp = lower clocks.
Ywc
Got it. Thanks. Another quick question. Is it safe to put in the current values from older version or should i do a full diagnostics all over again.
Yuri Bubliy
New version new life. Old - delete.
Yuri Bubliy
do you imagine early access or development process differently?
famich2005
All those hotfixes , patches...
Sisay
I have some strange anomalies. Asus x570 - calculate p1 on llc2 VID: 1100mV Vdroop: 6% CCX # 1: 4275MHz CCX # 2: 4175MHz (real cpu voltage is 1019mv in stress) for llc3 calculate p1 VID: 1100mV Vdroop: 3% CCX # 1: 4250MHz CCX # 2: 4150MHz (real cpu voltage is 1050mv in stress) Why ctr calculate lower mhz in llc3? Setting up manually and testing in prime95 (smallest fft) with llc3 I have a stable 4350/4225 The second thing is Ob, I still don't get it. With llc2 and 4275/4175 settings, I set ob = 8 and + 50mhz, while cinebench see 4350/4275 (it's jet + 75mhz) and i am confused :) another thing is prime95. With llc2 and ob = 8, it also boosts by 75mhz and prime bursts with errors. However, with llc3 and on = 8, even more underclocking (I have less about -50mhz, despite the fact that the clocking as I wrote above and so calculated lower than with llc2) Help
Ywc
Yup. I understand. Just that I have been scrolling through old posts on patreon regarding how to update to newer version but I can't find any info. Like do I have to re-tune everything? How do I transfer my old settings to the new one etc.
Norman P. Aquino
It's not out yet. Yuri will post it here once it becomes available.
Norman P. Aquino
Smallest FFTs result in the same errors. Two or three "workers" also stop in less than 15 seconds.
Ywc
Me too. Minus 25mhz always stable for me.
Ywc
Question: how to update to the upcoming v16 from v15? Can't find any information about upgrading to new ver in the guide.
Gabriel
ha I thought it was just me!
Danut Rotaru
If my mem is stable with 1.05 vcore and vddg ccd 0.950 and vddg iod 1.0 and vddp of just 0.880 will this impact max OB in CTR ( saw the recommendations for vsoc of at least 1.15 vcoreand IOD 1.05 )
Gabriel
Thanks I will play with it now. I am loving that RC5 gave me the best L3 performance ever but also the lowest latency I ever got with my 4x16GB setup at 58.2. With PBO+CO and identical settings for RAM I was usually hovering between 60.5 and 62 Plus compared to the previous PBO+CO settings I was able to get temps down about 20C for the same CB20 multi performance. Impressive!
Yuri Bubliy
Set the new voltage value and press "Calculate P1" button. It is not necessary to perform a complete diagnostics again If it is easier for you (for understanding), you can convert OB to CO, just divide the OB value by 3.5-4 1 CO = 3-5mV 1 OB = 1mV
Gabriel
One more question, my P1 calculated with only 1050 in my 5950..if I wanted to calculate with more voltage, adding the voltage to the profile then clicking Calculate will be as "accurate" as when doing a full diagnostic? Thanks for all the help!
Yuri Bubliy
It is difficult for me to explain how the processor calculates frequency values in real time, so I suggest you not to think about it, it is too complicated. You simply adjust the maximum frequency by increasing the OB or decreasing it in steps = 8 Only PX requires additional configuration (P1/P2 are set up perfectly)
Yuri Bubliy
OB 100 does not equal 100MHz. It is a relative value , sometimes +OB 8 equals +25 MHz , sometimes + 50OB equals +25 MHz . I suggest you use OB step = 8 to find the maximum values.
Gabriel
Ok great but in this case, instead of a lower CO number we actually add a higher value in order to trick the CPU to run higher, correct?
Yuri Bubliy
CO tricks the processor and tells it "you are consuming low current, you can increase the frequency". The processor says "yes I will". Only in this case it is not the CPU that calculates the maximum frequency, but the CTR.
Gabriel
Ok so adding OB there, wont make me hit over the OB limit in my example (4900mhz with limit 100), maxing at 5000 no matter what I enter in OB? Sorry, just trying to understand how all of them work in relation to the other. It seems OB will just push higher for the same amount of voltage than a lower OB setting..if I understood. Thanks
Yuri Bubliy
There is a concept of frequency in relation to voltages. OB is the change in the proportion of frequency relative to the current voltage.
Gabriel
I am confused because it literally says OB High(mV) on that parameter though.
Yuri Bubliy
OB does not add voltage, by any means, it only biases the SMU calculations (frequency per current voltageу) to get a higher frequency.
Gabriel
I am trying to understand the new OB option..so if I am at 4900 with OB Limit +100....does adding more mV to the OB mV option push the CPU over the 5000Mhz max or will that OB limit still prevails? Just trying to understand the relation between the OB Limit, the OB mV and the Mhz speed entry.
Sisay
in my case, I always have to decrease from 25 to 50mhz in relation to what the CTR set so that there are no errors in the prime95 smallest FFts test
David Ibbotson
Silly bug I found, when I double click on the icon in the tray it opens up my resolution screen changes.
LUis Ramírez
Is there a chance to set PX, P1 and P2 to a "normal use/gaming" scenarios and have another profile for undervolting? This way we don't have to override profile values when changing from one scenario to another.
Yuri Bubliy
Small FFTs - not supported. This is an unrealistic task for the real use of the processor. Even if you do scientific calculations on your processor. Only Smallest FFTs supported (high power/heat/CPU stress). This is enough to make sure the system is 100% stable. This is the test on which OB testing is based. That is, no additional testing is required.
Norman P. Aquino
I've tried lowering the clocks under P1 but I still get the errors.
Norman P. Aquino
Small FFTs -- maximum power/heat/CPU stress.
PewWORKS TV
Man you have a godly DIAMOND SAMPLE lmao
chrisi Trog
I feel like the8auer from Germany without extra cooling system 🤣
Yuri Bubliy
Sometimes you can see this funny visual bug :) I've seen already 65657MHz 😂
chrisi Trog
Ah okay. Yuri answer in the post before, abnormal sensor.
chrisi Trog
I go away yesterday and leave my pc on , come back and see highest frequency 6018mhz. Wtf. is it normal on a 5800x? in games myhighest result was 5046
Yuri Bubliy
what test are you using?
Norman P. Aquino
Prime 95 gives me errors (some threads stop working after a few minutes in P1 mode). I didn't experience this in previous CTR versions.
Yuri Bubliy
Any monitoring program will be a problem for CTR in most cases. There is a lot of software now with dirty sub-optimal code and their mode of operation looks selfish. If these programs don't communicate correctly with the SMU, it can cause a slow CTR response. You may also see abnormal sensor readings (most of the time this is a visual bug). + I will prepare an additional question and answer material.
Michael
Hi, any info about the hotfix content? I had some strange readings regarding power in monitoring tools and an overheating event causing reboot before ctr reacted.
Dizziee
I’m ready to fine tune 👨🎓
