AMD Curve Optimizer any guides / experience

[mtcn77]

lower the LLC 1

LLC isn't incrementally ordered, it is at its peak at 1.
Have fun ruining your chip.

far W250 T200 E130 i best

This won't work. You set EDC lower than TDC which should be the other way around. Also, TDC won't kick in until PTTL temperature, afaik. So far, good job circumventing safety protocols.

 

[harm9963]

LLC isn't incrementally ordered, it is at its peak at 1.
Have fun ruining your chip.


This won't work. You set EDC lower than TDC which should be the other way around. Also, TDC won't kick in until PTTL temperature, afaik. So far, good job circumventing safety protocols.

Ryzen Dram Calculator , power supply setting calls for 3 or 4, so please show some documentation to back up your assertions please

 

[mtcn77]

to back up your assertions

Not an assertion, 5 is default while 8 is the least level setting. If you are doing this with incidental knowledge, I cannot help you.

 

[jesdals]

Final and best result for boost - W250T200E130 rest the same as above

Lost the router because of the cold - so had to stop testing - but did break 5,3GHz

Giving more than T=200 did not bring any boost performance and E=130, sweet spot some where around these values

 

[harm9963]

Not an assertion, 5 is default while 8 is the least level setting. If you are doing this with incidental knowledge, I cannot help you.

http://imgur.com/a/U562wHh

Not an assertion, 5 is default while 8 is the least level setting. If you are doing this with incidental knowledge, I cannot help you.

 

[Zach_01]

View attachment 184192
with the 6900xt my chipset is never that cold in normal condition

I know what you mean. The place of the chipset is terrible...
In this screenshot your chipset temp is around 30C. This temp is followed by BIOS to control PCH fan. You can confirm this temp inside BIOS SmartFan5 and Gigabyte SIV.
What you see as a stand alone HWiNFO chipset temp is something else. Maybe some voltage controller/regulator.

 

[harm9963]

200 140 215 / -30 stock clock / using Dram Calculator for Ryzen , power supply settings .

 

[mtcn77]

http://imgur.com/a/U562wHh

The last time usmus was mentioned, the Stilt warned against it, now that he is gone, be wary of that.

 

[thesmokingman]

Final and best result for boost - W250T200E130 rest the same as above
View attachment 184195
Lost the router because of the cold - so had to stop testing - but did break 5,3GHz

Giving more than T=200 did not bring any boost performance and E=130, sweet spot some where around these values

You need to update to the latest hwinfo beta. You're missing some stats most importantly the power draw.

LLC isn't incrementally ordered, it is at its peak at 1.
Have fun ruining your chip.


This won't work. You set EDC lower than TDC which should be the other way around. Also, TDC won't kick in until PTTL temperature, afaik. So far, good job circumventing safety protocols.

They're new to this so they're going thru that phase... lol.

Ryzen Dram Calculator , power supply setting calls for 3 or 4, so please show some documentation to back up your assertions please

That doesn't mean you should follow it now does it? Dram calc doesn't calc anything, it's based on a set of assumptions.

 

[SesioN]

Hey guys,

can it be that I have such a bad CPU?? Probably bronze?

I've the following specs:

MB: Gigabyte X570 AORUS MASTER v1.2
CPU: 5950X
RAM: G.SKill TridentZ 32GB Kit DDR4-3200 CL14 (Running in XMP profile currently)
Cooling: AiO ARCTIC Liquid Freezer II 360 (In general very well cooled pc-case with many good fans and proper air-flow concept)
Bios: F32a
(same results on previous official one)

I tested like all possible settings.

Curve optimizer all cores -5 leads to instant blue screen when windows is loading. No matter what settings I try.

Example settings:

Vcore: auto
LLT: auto
PBO: 200/140/140
boost override +50 Mhz
curve optimizer: all core -5

I've tested all combinations. No matter what I blue screen or not stable or totally hot.
I can do like curve optimizer all core -2 then it doesn't blue screen. But on CPU-Z Benchmark quickly goes over 80c

 

[jesdals]

It could be memory issue - try to set your memory to CL 16 with more lose settings and se if that change anything. What kind of modules are there on your memory - I am running 4x8gb Samsung B-die, that why my settings demand 1.52V on memory. You can se my Artic Coolers behavior in the pages above - but 80c boost is nothing special. Bios F32a is that one of the upcomming betas? could also be why its buggy - try F31

 

[SesioN]

It could be memory issue - try to set your memory to CL 16 with more lose settings and se if that change anything. What kind of modules are there on your memory - I am running 4x8gb Samsung B-die, that why my settings demand 1.52V on memory. You can se my Artic Coolers behavior in the pages above - but 80c boost is nothing special

Those are also samsung B-die. I'm testing now with CL16.

So using curve optimizer can make the memory unstable? If so, how to counter it? just put more voltage on the memory?

 

[jesdals]

I think its more a genral stability issue - but what is your Infinity settings - you can se my bios settings above. But running 16GB modules is more tricky than 8gb. But if your running 3200MHz try setting voltage to 1.35 manually in bios - it may help running above that speed might take some more juice - at 1900MHz Infinity/3800MHz my memory need 1.52v at my settings. Please vil out your System Specs in the profile on Techpowerup,

 

[SesioN]

I think its more a genral stability issue - but what is your Infinity settings - you can se my bios settings above. But running 16GB modules is more tricky than 8gb. But if your running 3200MHz try setting voltage to 1.35 manually in bios - it may help running above that speed might take some more juice - at 1900MHz Infinity/3800MHz my memory need 1.52v at my settings. Please vil out your System Specs in the profile on Techpowerup,

I guess it's not the memory.

I've tested 1.5V on memory with 3000 Mhz CL16... WHEA BLUE SCREEN with optimizer all core negative 10

Also running infinity fabric 1:1 with memory, so 3000 Mhz = 1500

It must be the worst CPU ever, i'll just send it back..

eg:
PBO: 200/140/140
curve: -10 all core
boost +0
memory: 1.5v 3000 CL16

= blue screen

Curve: -2 all core = no blue screen

 

[jesdals]

1.5 might be to much at 3000MHz memory runs hot - but try the 3200 at 1.4 and see if it still crashes - but also did it behave the same with F31 bios?

 

[SesioN]

1.5 might be to much at 3000MHz memory runs hot - but try the 3200 at 1.4 and see if it still crashes - but also did it behave the same with F31 bios?

F31 same results.

Just tested following settings:

PBO: 200/200/130
curve: -10 all core
boost +100 Mhz
memory: 1.4v 3200 CL16

= windows doesn't even boot as it crashed and restarted mid-boot..

However, curve -8 booted.. So I guess I have some really bad cores?

Here results with those settings

Here one more... screenshot = mittle of cpu-z benchmark multicore

All other settings auto
PBO: 200/200/130
curve: -8 all core
boost +100 Mhz
memory: 1.4v 3200 CL16

 

[harm9963]

F31 same results.

Just tested following settings:

PBO: 200/200/130
curve: -10 all core
boost +100 Mhz
memory: 1.4v 3200 CL16

= windows doesn't even boot as it crashed and restarted mid-boot..

However, curve -8 booted.. So I guess I have some really bad cores?

Here results with those settings

View attachment 184383

Here one more... screenshot = mittle of cpu-z benchmark multicore

All other settings auto
PBO: 200/200/130
curve: -8 all core
boost +100 Mhz
memory: 1.4v 3200 CL16

View attachment 184384

Pull your MB battery out and RAM, reinstall , had the same problem a while back .

 

[SesioN]

Pull your MB battery out and RAM, reinstall , had the same problem a while back .

Sadly no change, -10 on curve always results in instant instability..

Also another issue I've noticed:
- First time I turn on the bios (before applying any changes whatsoever) it runs "smooth", eg each keyboard input results in instant action in bios
- After initial configuration, when I come back to bios, the bios is totally lagged, eg slow to input (looks like delayed/low fps etc)

Do you guys also have this?

 

[harm9963]

Sadly no change, -10 on curve always results in instant instability..

Also another issue I've noticed:
- First time I turn on the bios (before applying any changes whatsoever) it runs "smooth", eg each keyboard input results in instant action in bios
- After initial configuration, when I come back to bios, the bios is totally lagged, eg slow to input (looks like delayed/low fps etc)

Do you guys also have this?

Might have to wait for a new BIOS , or reinstalled window and chipset , AGESA V2 PI 1.2.0.0 since I got my 5950X, 4 BETA BIOS updates, plus two with my old 5800X .

 

[juanyunis]

Sadly no change, -10 on curve always results in instant instability..

Also another issue I've noticed:
- First time I turn on the bios (before applying any changes whatsoever) it runs "smooth", eg each keyboard input results in instant action in bios
- After initial configuration, when I come back to bios, the bios is totally lagged, eg slow to input (looks like delayed/low fps etc)

Do you guys also have this?

I do have that issue with lagged bios. It happens when you turn off CSM. A workaround is to press Ctrl + Alt + F6, this will lower the resolution and make it faster.

It is weird that you are getting such crappy results. I used to have a MSI X570 Gaming Plus, and that board was not able to properly run my 5950X and my 4x8GB RAM (I had it stuck at 3400Mh CL, wouldn't post beyond that),,
any change on PBO resulted in huge losses in performance in some multicore apps like cinebench, getting all core clocks of 546Mhz. That's why I switched to the X570 Aorus Master and my ram is running at 3800Mhz CL16.

I tried the F32a bios, it made my master able to boot beyond 1900 FCLK but with whea errors, also voltages were increased to cpu compared to F31.

You should share screenshot of your bios settings, also run PBO like this 210 / 140 / 145, scalar auto, curve -5, max boost override 0, platform thermal limit 90. Use cpu voltage in auto and cpu llc in auto.

Also it might be your windows settings, disable fast startup in power and sleep settings, and if you use gigabyte fast boot disable that.

 

[SesioN]

I do have that issue with lagged bios. It happens when you turn off CSM. A workaround is to press Ctrl + Alt + F6, this will lower the resolution and make it faster.

It is weird that you are getting such crappy results. I used to have a MSI X570 Gaming Plus, and that board was not able to properly run my 5950X and my 4x8GB RAM (I had it stuck at 3400Mh CL, wouldn't post beyond that),,
any change on PBO resulted in huge losses in performance in some multicore apps like cinebench, getting all core clocks of 546Mhz. That's why I switched to the X570 Aorus Master and my ram is running at 3800Mhz CL16.

I tried the F32a bios, it made my master able to boot beyond 1900 FCLK but with whea errors, also voltages were increased to cpu compared to F31.

You should share screenshot of your bios settings, also run PBO like this 210 / 140 / 145, scalar auto, curve -5, max boost override 0, platform thermal limit 90. Use cpu voltage in auto and cpu llc in auto.

Also it might be your windows settings, disable fast startup in power and sleep settings, and if you use gigabyte fast boot disable that.

With the settings that you proposed.

CPU-Z Result after test:

While test is running multicore:

 

[juanyunis]

So is it stable with -10? It is, try increasing max boost override to 50Mhz and test again, and if it does work, then go back to 0Mhz and test -15.

 

[SesioN]

So is it stable with -10? It is, try increasing max boost override to 50Mhz and test again, and if it does work, then go back to 0Mhz and test -15.

Nope, that was with curve -5, as you asked me to.

On -10 I get a blue screen.

EDIT:
With your proposed settings but with boost override 100 Mhz, same result, no difference between override 0 or 100 (Still curve -5)

 

[juanyunis]

So now you have to test core by core to see which one is giving you the bsod. In PBO curve optimizer put per core and put -5 from core 1 to core 15, for core 0 put -10. If it does boost try doing the same for next core.

 

[SesioN]

So now you have to test core by core to see which one is giving you the bsod. In PBO curve optimizer put per core and put -5 from core 1 to core 15, for core 0 put -10. If it does boost try doing the same for next core.

Yeah I figured that already. Thanks allot for trying to help, really appreciated!

But I don't think this is the way to go, I will just send this CPU back (The company that I ordered it at has a 30 days return policy with no questions asked) and will get another one.

I cba to pay so much money (925€) and have a cpu that gets bsod at curve -10...