Overclock.net banner

1 - 20 of 37 Posts

·
Registered
Joined
·
2,286 Posts
Discussion Starter #1 (Edited)
Edit (add):

On the Crosshair VII with the latest 2203 BIOS you can control the multiplier for PE 3 and I'm assuming PE 4 from within BIOS by adjusting EDC under PBO options after setting it to manual. However, while it allows you to adjust EDC without being ignored it will also cap EDC to what is set in BIOS while in Windows. So if you want to increase EDC with Ryzen Master from within Windows you won't be able to adjust it past what is set in BIOS.

Edit 2:

I have just found that setting EDC to 134 in BIOS translated into a cap of 140 when using Ryzen Master in Windows. I haven't explored beyond this so I don't know if 140 is the floor or if there is a certain value that will be added to what is set in BIOS when using a value below 145. 145 is what PE 3 would normally use.

Edit the third:

Ok I've now found that although Ryzen Master will display a max value for EDC to be 140 (above the 134 I set in BIOS) it will not actually change that value above the 134 value set in BIOS. So it does appear that what is set in BIOS is the cap. Even though Ryzen Master may show a higher value for the max.


Original post below:

A little preamble: I was searching for a way to land between where the Performance Enhancers in BIOS were putting me -- something like a Level 3 1/2. My Level 3 multiplier was usually 41~41.3 and my Level 4 could be anywhere from 42.5~43.5. As you can imagine I wanted to be somewhere between those and be able to land there more consistently. Elmor and The Stilt both gave the guidance of changing the EDC value in BIOS under the PBO settings, but I've found those values won't apply when using a Performance Enhancer so I couldn't manipulate Level 3 or Level 4 that way. Both also said they would look into why that might be so I want to thank them for their help. In the meantime I have found that Ryzen Master works perfectly for changing the EDC value without needing to reboot and it can be done reactively after seeing at which multiplier your CPU boots at.

Below I will be posting some screenshots of Ryzen Master and a terrible description of my methodology.

So Level 3 boots at an EDC value of 145 (and I believe Level 4 boots at 160(actually it appears Level 4 boots at 168 or sometimes 1) and this scenario below my CPU booted with a multiplier of 41 using Level 3. Because the CPU boots at a different multiplier at times I don't believe a specific EDC value equates to a specific multiplier. How I think of it is since EDC can range from 140 - 168 making adjustments in increments of 4 you should see a new multiplier value after each adjustment. Since Level 3 boots at 145 it is using the multiplier that would fall between the 144-147 range. If you change EDC to 140 the mutiplier will drop .25 from its boot value which you will be able to see in the photos I'll post below. Aslo, a very crude text "table".

Edit:

With much cooler (idle at 21*C) temps I have found that my CPU went from a multiplier of 42.5 at an EDC of 164 to a multiplier of 43 and in a separate instance 43.5 at an EDC of 168. Separate instances of adjusting the EDC from 164 to 168 created the unexpected adjustments up to multipliers at 43 and 43.5 (the highest jump to 43.5 dropped back itself to 43). Both of those increments are significantly above the previously observed steps of .25 seen previously at warmer temperatures. I will also add that the larger step in the multiplier was benchmark stable with these cooler temperatures. The CPU seems to weigh start-up temperatures more heavily than previously observed in earlier testing. All other multiplier steps with EDC increments (140-160) were in-line with previous observed behaviors. In summation: an increase to the EDC value, at its highest end, adding 4; resulted in a multiplier increase of .5 and in another instance 1. Temperature, being the observed value, was significantly different from previous testing.

Starting from a multiplier of 41 with PE Level 3 (this is a sliding scale that will change):
EDC // Multiplier
140~143 // 40.75
144~147 // 41
148~151 // 41.3
152~155 // 41.5
156~159 // 41.75
160~163 // 42
164~167 // 42.3
168 // 42.5

And now some pictures.
 

Attachments

·
Registered
Joined
·
2,286 Posts
Discussion Starter #2 (Edited)
Now after you've made your EDC adjustment you can close Ryzen Master and you don't need to reboot.


I know some folks would prefer to make this change in BIOS but I'll remind you that this method allows you to make adjustments after you see what your CPU multiplier boots at and also it doesn't work in BIOS currently.


Please note that I don't believe there is a specific EDC value to give a specific multiplier and that it works more like a sliding scale. As such -- if your CPU boots with a higher Level 3 multiplier (ie 41.5 instead of 41) going up to an EDC value of 168 will probably put you at a multiplier of 43 instead of 42.5 as in the example I posted.
 

·
Registered
Joined
·
2,286 Posts
Discussion Starter #3
If this helps someone I hope you'll let me know.
 

·
Registered
Joined
·
202 Posts
This is interesting but the downside is that I can run my 4.25Ghz at 1.4125V with manual voltage while using this makes then similar to yours at 1.48v
 

·
Registered
Joined
·
2,286 Posts
Discussion Starter #5
This is interesting but the downside is that I can run my 4.25Ghz at 1.4125V with manual voltage while using this makes then similar to yours at 1.48v

That's odd. My CPU only goes to 1.38V when I set it up to run at 4.25GHz.
 

·
Registered
Joined
·
202 Posts
That's odd. My CPU only goes to 1.38V when I set it up to run at 4.25GHz.
Manually or using PE3?

PE3 it runs higher than needed, unless you know a way to control the voltage than offset? Because with offset it just gets stuck on like 4.15Ghz at 1.3v without moving when I offset too low
 

·
Registered
Joined
·
2,286 Posts
Discussion Starter #7
Manually or using PE3?

PE3 it runs higher than needed, unless you know a way to control the voltage than offset? Because with offset it just gets stuck on like 4.15Ghz at 1.3v without moving when I offset too low
Sorry I should have clarified. Using PE3 with additional EDC tweaking. And that is with a Core + offset of .0250.

I don't use PE4 because it will boot at 43 or higher sometimes.
 

·
Registered
Joined
·
2,286 Posts
Discussion Starter #8
An edit has been added to OP for new observed behavior.
 

·
Meddling user
Joined
·
7,423 Posts
Going to link this in ROG C7H OP, PE section :) .

You may find changing Sense Mi Skew changes boosting, but as your skewing tCTL, CPU temperature readings would also be incorrect.
 

·
Registered
Joined
·
2,286 Posts
Discussion Starter #10
Going to link this in ROG C7H OP, PE section :) .

You may find changing Sense Mi Skew changes boosting, but as your skewing tCTL, CPU temperature readings would also be incorrect.
Could you elaborate more on tCTL skewing? And I don't know values for Sense Mi Skew so I have it disabled. Can you give guidance on that too please?
 

·
Meddling user
Joined
·
7,423 Posts
Sorry for delayed response :eek: .

Use ASUS Turbo V and you'll see the default offset, IIRC it's 272. I have dabbled with it occasionally, so my memory is hazing on which way tweak results on what kinda skew. If you try a small change (ie 268 vs 278), you should be able to see affect.

Be aware using this skewing also means over temperature protection on CPU may not at correct temp, due to the skewing. So if cooling fails you could be looking at CPU going to excessive real temp, resulting in damage, etc.
 

·
Registered
Joined
·
2,286 Posts
Discussion Starter #12
Sorry for delayed response :eek: .

Use ASUS Turbo V and you'll see the default offset, IIRC it's 272. I have dabbled with it occasionally, so my memory is hazing on which way tweak results on what kinda skew. If you try a small change (ie 268 vs 278), you should be able to see affect.

Be aware using this skewing also means over temperature protection on CPU may not at correct temp, due to the skewing. So if cooling fails you could be looking at CPU going to excessive real temp, resulting in damage, etc.
No, worries. I didn't immediately recognize Tctl when I first read your response, but it clicked in my head after a peak at HWiNFO.
 

·
Registered
Joined
·
2,286 Posts
Discussion Starter #14
Awesome info, I'll be trying this out, start low and work my way up...
Thanks for sharing.
Did you have a chance to try this out with your Strix board? I'm curious to know if it's the same.
 

·
Registered
Joined
·
462 Posts
I am on asus prime x470 pro , if i set pbo enabled and scalar x10 it will give me 168a limit as i lowering the scalar the edc limit is reduced. But as far as i can see the temprature and the sillicon are the things that gives the real limits of multi core boost.

It seems for me that my cpu (i think depends on silicon) can keep 4125mhz all core at heavy loads until the temprature reach 50c. Above 50c the cpu starts to drop multi. As the cpu getting hotter the multiplier is getting down.
s
So i can't understand how changing edc from ryzen master could be useful for something. is something that i miss ? why to use pe and dont use pbo scalar ?
 

Attachments

·
DRAM Obsessor
Joined
·
1,033 Posts
I am on asus prime x470 pro , if i set pbo enabled and scalar x10 it will give me 168a limit as i lowering the scalar the edc limit is reduced. But as far as i can see the temprature and the sillicon are the things that gives the real limits of multi core boost.

It seems for me that my cpu (i think depends on silicon) can keep 4125mhz all core at heavy loads until the temprature reach 50c. Above 50c the cpu starts to drop multi. As the cpu getting hotter the multiplier is getting down.
s
So i can't understand how changing edc from ryzen master could be useful for something. is something that i miss ? why to use pe and dont use pbo scalar ?

Well, for a couple of reasons...In the new bios update there is no longer a scalar, or at least on the C7H there isnt and also PE locks the all core multiplier and it doesnt actively downclock during load and PE also goes beyond limits if there is sufficient thermal headroom to do so, although I dont see many having much thermal headroom above 4.4ghz. He changes EDC to trick the multiplier values that are locked at POST. There is some algorithm or something that looks at the thermals and determines the all core multiplier and the single, 2, 4, and 6 core multiplier and once in the OS the only way to go from a 41.5x all core multiplier to 43.5x would be to do what he is suggesting since the multiplier will not go above that value once in the OS otherwise. The differences Ive seen in PE3 and PE4 besides how aggressive the voltages are is that with PE3 Single Core multiplier seems to be favored over the all core multiplier. At PE3 I can run my bclk to 106.6mhz resulting in 4640mhz Single core and roughly 4.425ghz all core but on PE4 at 104.8mhz the all core is much more aggressive with its multiplier as I will hit 4558mhz all core as well as the single core. It does take a lot more voltage and to get teh 43.5x multiplier you have to be running around 10c idle temps on the cpu. Its much easier to get the 43.5x multiplier on PE3. I just use PE3 for single core performance when I need it and PE4 for multicore performance. PBO does not have the same characteristics and has limitations during load and just cant reach the performance that PE3 and 4 can but PE is harder to maintain stability so each has their pros and cons.
 

·
Registered
Joined
·
462 Posts
Well, for a couple of reasons...In the new bios update there is no longer a scalar, or at least on the C7H there isnt and also PE locks the all core multiplier and it doesnt actively downclock during load and PE also goes beyond limits if there is sufficient thermal headroom to do so, although I dont see many having much thermal headroom above 4.4ghz. He changes EDC to trick the multiplier values that are locked at POST. There is some algorithm or something that looks at the thermals and determines the all core multiplier and the single, 2, 4, and 6 core multiplier and once in the OS the only way to go from a 41.5x all core multiplier to 43.5x would be to do what he is suggesting since the multiplier will not go above that value once in the OS otherwise. The differences Ive seen in PE3 and PE4 besides how aggressive the voltages are is that with PE3 Single Core multiplier seems to be favored over the all core multiplier. At PE3 I can run my bclk to 106.6mhz resulting in 4640mhz Single core and roughly 4.425ghz all core but on PE4 at 104.8mhz the all core is much more aggressive with its multiplier as I will hit 4558mhz all core as well as the single core. It does take a lot more voltage and to get teh 43.5x multiplier you have to be running around 10c idle temps on the cpu. Its much easier to get the 43.5x multiplier on PE3. I just use PE3 for single core performance when I need it and PE4 for multicore performance. PBO does not have the same characteristics and has limitations during load and just cant reach the performance that PE3 and 4 can but PE is harder to maintain stability so each has their pros and cons.
Thanks you made it clearer to me,
 

·
Registered
Joined
·
2,286 Posts
Discussion Starter #18
Thanks you made it clearer to me,
I actually upgraded from the X470 Prime Pro to the Crosshair VII to get the extra Performance Enhancer Level 3 and 4. Those are the levels that "lock" the multiplier. All the ROG boards have the extra levels, but The Stilt said the B450 boards don't have the VRM to run it.
 

·
Registered
Joined
·
224 Posts
Well, for a couple of reasons...In the new bios update there is no longer a scalar, or at least on the C7H there isnt and also PE locks the all core multiplier and it doesnt actively downclock during load and PE also goes beyond limits if there is sufficient thermal headroom to do so, although I dont see many having much thermal headroom above 4.4ghz. He changes EDC to trick the multiplier values that are locked at POST. There is some algorithm or something that looks at the thermals and determines the all core multiplier and the single, 2, 4, and 6 core multiplier and once in the OS the only way to go from a 41.5x all core multiplier to 43.5x would be to do what he is suggesting since the multiplier will not go above that value once in the OS otherwise. The differences Ive seen in PE3 and PE4 besides how aggressive the voltages are is that with PE3 Single Core multiplier seems to be favored over the all core multiplier. At PE3 I can run my bclk to 106.6mhz resulting in 4640mhz Single core and roughly 4.425ghz all core but on PE4 at 104.8mhz the all core is much more aggressive with its multiplier as I will hit 4558mhz all core as well as the single core. It does take a lot more voltage and to get teh 43.5x multiplier you have to be running around 10c idle temps on the cpu. Its much easier to get the 43.5x multiplier on PE3. I just use PE3 for single core performance when I need it and PE4 for multicore performance. PBO does not have the same characteristics and has limitations during load and just cant reach the performance that PE3 and 4 can but PE is harder to maintain stability so each has their pros and cons.
This is very interesting to read. Are you able to share with us the thermal condition/cooling setup of this system? Also did you say you're using +0.025vcore offset on both pbo3 and 4?

I have a custom loop with overkill capacity, and now I have a better idea how I should approach my PBO tuning, but I seem to have been getting diminishing returns on my benchmarks despite using these overpowered configs such as pbo3/4.

I have heard using - offset is better for PBO3/4 by allowing greater headroom for boost, but my config does not seem to reflect that.

Sent from my SM-N950U using Tapatalk
 

·
Registered
Joined
·
2,286 Posts
Discussion Starter #20
This is very interesting to read. Are you able to share with us the thermal condition/cooling setup of this system? Also did you say you're using +0.025vcore offset on both pbo3 and 4?

I have a custom loop with overkill capacity, and now I have a better idea how I should approach my PBO tuning, but I seem to have been getting diminishing returns on my benchmarks despite using these overpowered configs such as pbo3/4.

I have heard using - offset is better for PBO3/4 by allowing greater headroom for boost, but my config does not seem to reflect that.

Sent from my SM-N950U using Tapatalk
If you have strong cooling then you'll want to add a BCLK increase to get more out of your CPU. Manipulating PE 3/4 will only get you so much until you adjust BCLK also. I've used PE 3 with a BCLK of 104 to reach 4.4GHz on all cores. It required very low ambient temps though -- cold weather with the window open.

If you want something more everyday then using PE 3 with BCLK 102 will allow you to get to 4.3GHz all core.
 
1 - 20 of 37 Posts
Top