AMD Curve Optimizer any guides / experience

[juanyunis]

@harm9963 wow, thank you so much for the settings I think I have a winner, Best CPU-Z score I have ever had, great Cinebench R20 and R23 scores, voltage CPU Core SVI never goes beyond 1.469V) and temperatures are okish, never reaching 90C which is OK for Zen 3 CPU (According to AMD Technicians) and not thermal throttling.

I think you might have a better cooling solution and your ambient temperature is lower, I'm thinking of getting the Artic Liquid Freezer II 360mm. Could you tell me what you have? Thanks!

1. Voltage: Auto
2. LLC: Auto
3. 200/200/145 for PPT/TDC/EDC
4. Curve -30
5. Scalar: Auto
6. Max Boost Override: +50Mhz
7. Platform Thermal: Auto

 

[mtcn77]

You see! Now, that is a good score! If you look, you are at EDC max and 75% PPT. Keep dialing it in, I'm sure there is more to come.
PS: at this setting there is zero difference between PPT:200 and PPT:250 - both are limited by EDC which is a good thing in order to overcome temperature instabilities.

I think I have something nice

 

[juanyunis]

Yes!! I understand now, however according to @Zach_01 is not good for the chip to have the same TDC/EDC. What do you think? I could dial it in an have it the same, and HW info will show 100% for PPT/TDC/EDC.

 

[mtcn77]

According to AMD Technicians

I wouldn't trust them. That skin head is no good in comparison to @The Stilt .

 

[harm9963]

@harm9963 wow, thank you so much for the settings I think I have a winner, Best CPU-Z score I have ever had, great Cinebench R20 and R23 scores, voltage CPU Core SVI never goes beyond 1.469V) and temperatures are okish, never reaching 90C which is OK for Zen 3 CPU (According to AMD Technicians) and not thermal throttling.

I think you might have a better cooling solution and your ambient temperature is lower, I'm thinking of getting the Artic Liquid Freezer II 360mm. Could you tell me what you have? Thanks!

1. Voltage: Auto
2. LLC: Auto
3. 200/200/145 for PPT/TDC/EDC
4. Curve -30
5. Scalar: Auto
6. Max Boost Override: +50Mhz
7. Platform Thermal: Auto

View attachment 183490 View attachment 183491
View attachment 183492 View attachment 183493

Artic Liquid Freezer II 360mm , if you can find one in stock.

Arctic Liquid Freezer II 420 FTW​

 

[mtcn77]

@Zach_01 is not good for the chip to have the same TDC/EDC.

Zach knows his beef. TDC is there to save you from having to worry about TjMax. If you feel safe - the temperature is not rising to PTTL that is - feel free to overlook it, since it is a thermal throttle point in the FIT scale.

So, one thing I had been adamantly quarreling about was EDC vs PPT since EDC more precisely tunes to a set level. However that is against people's expectations, it starts and continues too steady. If you are one of those people that want the cheap thrills, you can limit via PPT. It will provide a nice boost curve and as the cpu heats up you will notice it pushing back a little bit. However I find it crude with this level of control, but if you are running short stretches, you will notice a bump in your performance if you can balance it out. Say, your consumption is 180, so try PPT:180 and EDC some higher value than your current 125 median.

 

[jesdals]

Im using the Arctic Liquid Freezer II 360 as well - but please not that there is a new mounting mekanismen and it give problems on some boards.

 

[juanyunis]

@jesdals i think i might have to go with EK new 360 AIO, it's on par with Liquid Freezer, Artic's one has the fan that cools VRM, but I think for our X570 AORUS master is not really important according to what Steve from GN said.

 

[jesdals]

Leaving the curve settings (-10/-5) I get almost the same GPU-z result as Juan

will try to see if I ca do negative 30 Negative 30 cant boot un my setup - will test negative 20 for stability

If stable it looks promising

 

[juanyunis]

are you using my settings? if you are, then i can confirm that your ambient temperature and your cooling solution is giving you so much gains, specially ST scores and @harm9963 gets more gains because he has the 420mm one, probably good ambient temperature too.

 

[jesdals]

are you using my settings? if you are, then i can confirm that your ambient temperature and your cooling solution is giving you so much gains, specially ST scores and @harm9963 gets more gains because he has the 420mm one, probably good ambient temperature too.

I kept my loadline settings and the 500MHz target (wishfull thinking) but since I couldnt boot with negative 30 - I went with negative 20 - if stable I give negative 25 a try. Doing the benchmark I saw above 4500MHz all core speeds

 

[juanyunis]

Funny enough I just tried -35 same settings (Could it be a golden/platinum sample I have?) I used and it booted and ran all benchmarks, and i got incredible performance, for example Cinebench R20 got 11490 and R23 got 29107, so far my best results yet. CPU-Z same ST but MT got 10 points more.

I think for stability purposes and to be on the safe side, I'm gonna leave my settings like this:

1. Voltage: auto
2. LLC: Auto
3. For PBO, PPT: 200, TDC: 137, EDC: 145, Curve -30, Scalar: Auto, Max Boost Override: 0 Mhz, Platform Thermal: Auto

I get about same performance in ST/MT and same Cinebench results, but I think it should be stable.

 

[jesdals]

-20 seems stable, temps are high with these settings

That 700 mark is just not possible

 

[juanyunis]

@jesdals that's incredible. What settings?

 

[jesdals]

The same as before

but no 700 today

 

[juanyunis]

I think I'm gonna wait for a better thermal solution lol, it feels like my 5950X doesn't like my ambient temperature and 240MM AIO.

I made a little experiment just for fun lol, I opened the case, put fans to max and added an external fan pointing to the CPU and my MT score is up using my latest settings (at some point they reached over 13200 but didn't save it lol), so yeah looks like my lower performance compared to you is because of thermals lol.

--------------------------EDIT------------------------------------

Another update, I limited my PPT to 180 (down from 200) and adjusted the TDC to 125 (down from 137), increased Max Boost Override to 500Mhz (thanks @jesdals) and also set the Platform Thermal to 82C, my CPU-Z scores are about the same (685 ST and 13K MT), Cinebench R20 dropped a little bit to 11350, R23 got more points (weird) 29220 and also some games that increased the temperature like crazy (Assassin's Creed Odyssey) now are being restricted to a max of 82C (It was previously running at 87C~90C).

Most cores go beyond 5Ghz, some of them reached 5.1Ghz, 5.17Ghz and 5.25GHZ.

I think this is my sweet spot for my ambient temperature which is 27C with no AC (Feels like 30C - Accuweather says that today's range was 26~32C) with a high humidity (78%).

When a heavy MT application/process is running, the CPU doesn't go beyond 80C, and when some rogue game/app/process tries to increase temp, the limit I put will throttle the processor to avoid going beyond 82C. This should keep my CPU happy and avoid future degradation.

 

[harm9963]

The same as before
View attachment 183586
but no 700 today

Final tweak 195 195 125 -30 , PBO 1st 2nd AMD OC 3rd XRF combination lost it, found it again by luck ! forgot the order ,man that sucks , auto auto auto , disable ,enable

Peak means very little , race between a rabbit and a tortoise ends with a moral: "Slow and steady wins the race" , plus temps were 73c in 10 min test, but one kickup at 77c .

One more test ,before I go to sleep

 

[mtcn77]

Peak means very little

Like I said, EDC peaks will mean very little as long as the cpu temperature is yet to normalize from a cold boot. You get a nice boost of "unobtanium", but pay for it down the line with slowing down. Overclocking beyond your cpu cooling requirements is good because you get plenty of current in the short term which enables all your target performance benchmarks without specifically getting to 100% stability. It is like the casual "ice bucket" that record holders use.

 

[AMD718]

Just adding mine for reference. This is my first pass at PBO CO on my 5950, Aorus B550 Master

PPT=200, TDC=200, EDC=150
Scalar @ 4x
CPU +200 Mhz
PBO Thermal Throttle = 80c
[CO]
Core 0 Negative, 15
Core 1 Negative, 5
Core 2 Negative, 15
Core 3 Negative, 15
Core 4 Negative, 15
Core 5 Negative, 15
Core 6 Negative, 10
Core 7 Negative, 15
Core 8 Negative, 10
Core 9 Negative, 15
Core 10 Negative, 15
Core 11 Negative, 5
Core 12 Negative, 15
Core 13 Negative, 15
Core 14 Negative, 15
Core 15 Negative, 15

I feel I can do much better but I've wanted to make sure this is 100% rock solid stable before loosening up negative offsets. So, it's been a week with dozens of benchmarks, tests, OCCT, etc. and I haven't had a single crash, reboot WHEA, or anything. I'm feeling good about these settings as a stable baseline I can progress from.

Cooler is NH-D15

 

[thesmokingman]

Why are you using a positive offset then trimming with negative voltage (another -offset) in the curve optimizer? It seems a bit redundant. In either case ya both are pushing 200w so I'm not surprised by the perf.

 

[juanyunis]

Why are you using a positive offset then trimming with negative voltage (another -offset) in the curve optimizer? It seems a bit redundant. In either case ya both are pushing 200w so I'm not surprised by the perf.

were you referring to my previous tests? if you are, then my reason was that leaving the voltage in auto causes high spikes, making my system reach the thermal limit at the time (90C), when i switched to normal with a positive offset (found that info in a reddit post) i found that the clocks were good, performance was pretty good and my temperatures were about 10C cooler. (I know it sounds stupid, maybe a BIOS bug but it worked) however i have moved on from that a found a better tune for my system and ambient temperature.

 

[thesmokingman]

were you referring to my previous tests? if you are, then my reason was that leaving the voltage in auto causes high spikes, making my system reach the thermal limit at the time (90C), when i switched to normal with a positive offset (found that info in a reddit post) i found that the clocks were good, performance was pretty good and my temperatures were about 10C cooler. (I know it sounds stupid, maybe a BIOS bug but it worked) however i have moved on from that a found a better tune for my system and ambient temperature.

Yea, I was. I'd just add that be careful with the settings cuz often they can overlap. Also, these high spikes... did you mean spikes in temp? Are you watching the main temp sensor and not die average?

 

[juanyunis]

Yea, I was. I'd just add that be careful with the settings cuz often they can overlap. Also, these high spikes... did you mean spikes in temp? Are you watching the main temp sensor and not die average?

Oh I'm currently using this:

Voltage: Auto
LLC: Auto
PBO: 180 PPT, 125 TDC, 145 EDC, -25 Curve, Auto Scalar, 200Mhz Max Boost Override and 80C Platform Thermal Limit

This is giving awesome performance and keeping my voltages under control and the temps too.

Yes, spikes in temperature, I was watching CCX temps.

 

[thesmokingman]

Oh I'm currently using this:

Voltage: Auto
LLC: Auto
PBO: 180 PPT, 125 TDC, 145 EDC, -25 Curve, Auto Scalar, 200Mhz Max Boost Override and 80C Platform Thermal Limit

This is giving awesome performance and keeping my voltages under control and the temps too.

Yes, spikes in temperature, I was watching CCX temps.

Ah that's why. That is NORMAL behavior. The individual die temps will light up as it switches contexts. I find die average more relevant.

Also we should realize that the for those with dual CCD chips, you have 3 sensors a main sensor show all core temps and your two CCD die temps. The main temp sensor will fluctuate like mad all the time because you're seeing 12-16 temps all lumped into one sensor. And given how fast the cpu switches contexts, you can see why this is not an accurate reflection of the cpu temps overall.

 

[juanyunis]

Ah that's why. That is NORMAL behavior. The individual die temps will light up as it switches contexts. I find die average more relevant.

Also we should realize that the for those with dual CCD chips, you have 3 sensors a main sensor show all core temps and your two CCD die temps. The main temp sensor will fluctuate like mad all the time because you're seeing 12-16 temps all lumped into one sensor. And given how fast the cpu switches contexts, you can see why this is not an accurate reflection of the cpu temps overall.

I'm confused lol.

Which temp sensor in hw info should we use for people with dual CCX?

I was looking at both CCX temp sensors when doing my tests.

And the spikes where while doing benchmarks, before I could get 89C while starting running CPU-Z, and then drop to 80C all cores. When I changed to normal voltage and positive offset the spikes were upto 77C and all core about 70C. But @mthembo and @Zach_01 told me that I was getting very high voltages over 1.5V and even tho temps were below 80C those voltages are not safe.