Overclock.net banner
4841 - 4860 of 4999 Posts

·
Registered
Joined
·
434 Posts
Has anyone played with this BIOS parameter?

"Synch ACDC Loadline with VRM Loadline"

I think the description is inverted.
I think, in fact, it syncs the VRM Loadline with ACDC loadline...

EDITED.:

I found out how it works. The description is correct, but AC/DC_LL shall be in auto.
 

·
Optimal Pessimist
Joined
·
3,175 Posts
@Astral85

Fantastic looking rig! And congrats on the card.
 

·
Optimal Pessimist
Joined
·
3,175 Posts
Has anyone played with this BIOS parameter?

"Synch ACDC Loadline with VRM Loadline"

I think the description is inverted.
I think, in fact, it syncs the VRM Loadline with ACDC loadline...

EDITED.:

I found out how it works. The description is correct, but AC/DC_LL shall be in auto.
What would be useful is to sync DCLL to VRM Lloadline.
 

·
Registered
Joined
·
434 Posts
What would be useful is to sync DCLL to VRM Lloadline.
Yes!

And this is a good way to find out the LLC impedance in MB that doesn't have Die-Sense to tune VID=VCore.

Set AC/DC load line in AUTO and "Synch AC/DC Loadline with VRM Loadline" = enable"

After that you go to HW-Info and look to AC/DC...
 

·
Optimal Pessimist
Joined
·
3,175 Posts
Yes!

And this is a good way to find out the LLC impedance in MB that doesn't have Die-Sense to tune VID=VCore.

Set AC/DC load line in AUTO and "Synch AC/DC Loadline with VRM Loadline" = enable"

After that you go to HW-Info and look to AC/DC...
True. But, if I understand it correctly. that would run some pretty high ACLL for anything but LLC6 wouldn't it?

What I did to estimate it (custom HWINFO value) was to take the running average of (AVG VID across cores - Vcore)/(CPU Current) with ACLL=.01, then fine tune it so that reported CPU Power = reported Power from EC. I got best fits of LLC5 = 0.79 mohm and LLC4 = 1.11 mohm
 

·
Registered
Joined
·
292 Posts
Yes!

And this is a good way to find out the LLC impedance in MB that doesn't have Die-Sense to tune VID=VCore.

Set AC/DC load line in AUTO and "Synch AC/DC Loadline with VRM Loadline" = enable"

After that you go to HW-Info and look to AC/DC...
I asked you this in PM before but you say i should not enable it.
 

·
Registered
Joined
·
434 Posts
I asked you this in PM before but you say i should not enable it.
I was trying to understand how it works... the problem is that it set AC and DC with the same value of LLC... And it's not the best way to tune AC_LL.

Even DC_LL need some adjust to make VID=VCore under load.

This option when enabled set AC_LL=DC_LL=LLC (all in milliohm), and theoretically you must have VID=VCore and no droop...
It is a little different in real world, but a good start point.

Besides that, droop is good when you're tuning high frequencies and light loads.

I kept it disabled.
 

·
Registered
Joined
·
434 Posts
True. But, if I understand it correctly. that would run some pretty high ACLL for anything but LLC6 wouldn't it?

What I did to estimate it (custom HWINFO value) was to take the running average of (AVG VID across cores - Vcore)/(CPU Current) with ACLL=.01, then fine tune it so that reported CPU Power = reported Power from EC. I got best fits of LLC5 = 0.79 mohm and LLC4 = 1.11 mohm
These were "exactly" the numbers I found...
LLC#4 = 1.12
LLC#5 = 0.8
 

·
Optimal Pessimist
Joined
·
3,175 Posts

·
Optimal Pessimist
Joined
·
3,175 Posts
I asked you this in PM before but you say i should not enable it.
It can lead to very high voltages if you are not careful. You would have to adjust your VF curve voltage points quite a bit downwards from their default to compensate.
 

·
Registered
Joined
·
434 Posts
It can lead to very high voltages if you are not careful. You would have to adjust your VF curve voltage points quite a bit downwards from their default to compensate.
I don't know if all MB vendor have "VMax Stress" option, but I think in stock it should be enabled as default.

This option allows me to get 56x...

I set adaptive ~1540mv and let this knob cut the voltage and frequency to safe values.

If VMax Stress feel that VID will > 1500mv it reduces the frequency... So 56x only runs in determinate conditions of load and temp.

Another trick I found was set (AC_LL) = ~(DC_LL / 2)
So I use DC_LL = 1,12 and AC_LL = 0.5

2518103
 

·
Optimal Pessimist
Joined
·
3,175 Posts
Another trick I found was set (AC_LL) = ~(DC_LL / 2)
So I use DC_LL = 1,12 and AC_LL = 0.5
I think what you mean to say is you set your ACLL = LLC/2, that is what counts.
This may work for you, but I think that is specific to your chip and what your VF curve is or is adjusted to. I run LLC5 so my DCLL is 0.8 and my ACLL is .16, and my VF curve has minimal adjustments. If I set ACLL = .4, I would have to adust a lot of my VF points down (I already have negative points as is). It is pretty much just a matter of preference, since once you get it adjusted for minimum required for stability, you'll be sending the same VIDs to the VRM.
 

·
Registered
Joined
·
434 Posts
I think what you mean to say is you set your ACLL = LLC/2, that is what counts.
This may work for you, but I think that is specific to your chip and what your VF curve is or is adjusted to. I run LLC5 so my DCLL is 0.8 and my ACLL is .16, and my VF curve has minimal adjustments. If I set ACLL = .4, I would have to adust a lot of my VF points down (I already have negative points as is). It is pretty much just a matter of preference, since once you get it adjusted for minimum required for stability, you'll be sending the same VIDs to the VRM.
You are right, it's not a rule, and depends of a lot of other things.

If I use LLC5 I need less AC_LL or need to decrease voltage in VF points... But can't hit 56x...
 

·
Registered
Joined
·
3,814 Posts
MSI BIOS does not have vmax stress unfortunately.
 
  • Sad
Reactions: RobertoSampaio

·
Registered
Joined
·
434 Posts
MSI BIOS does not have vmax stress unfortunately.
VMax Stress make magic... LOL
You can set 1600mv and 6GHz and CPU will hit maximum possible frequency keeping below 1500mv.

2518137


 

·
Registered
Joined
·
668 Posts
thats what i meant for months..even if u solve the l0.. u can get game crashes that wont trigger whea.

btw disable dx12 and RT on that game if ure on 10900k..
Is it something in DX12 or RT that trigger these crashes?
 

·
Registered
Joined
·
668 Posts
WHEA are corrected errors. Eventually you will hit one that can't be corrected. Or it is some other instability. What are the details of the error (what did it say on the monitor or in the event log)?
Hi Geneo, the fatal error is generated by the game, it's not a Windows error.
 

·
Registered
Joined
·
668 Posts
It happens to me too in my overclocked system: it's either fatal errors or (more frequently in my case) Parity Errors.

By the way, the solution Falkentyne suggested for Minecraft (setting affinity from Task Manager with less threads assigned to the related .exe) seems to work for Ghostrunner too. If I allocate no more than 3 physical cores (for a total of 6 threads) to the process named "Ghostrunner-Win64-Shipping.exe" (not the main executable, the other one basically), I don't get fatal errors or parity errors anymore with DX12 RT Enabled, at least in the infamous third level. As soon as I increase the number of threads/cores errors begin to reappear. What is interesting is that the number of physical cores must not exceed 3 (assuming that I'm correct in identifying them with the even CPUs, 0 2, etc. in Task Manager, the odd ones being the correspondent logical cores): if for example I assign 6 physical cores to the process, resulting again in 6 threads assigned to the process, the game has still issues, even though they occur less frequently.

Of course a simpler solution is to disable RT, as cstkl1 suggested.
Can you show me what that should look like in Task Manager? I disabled a bunch of cores/threads on the game .exe and dropped almost 30 fps...
 

·
Registered
Joined
·
362 Posts
Hi Geneo, the fatal error is generated by the game, it's not a Windows error.
Yes, the error was not corrected by windows (otherwise there would be a WHEA report catching/correcting the error), the error then passes onto the game and crashes it.

Can you show me what that should look like in Task Manager? I disabled a bunch of cores/threads on the game .exe and dropped almost 30 fps...
Set affinity: Task manager, details tab, right click on game.exe, set affinity, tick/untick as prefered (odd numbers should be logical i.e. hyperthreaded "cores", and even ones should be the physical ones).
 
  • Rep+
Reactions: Salve1412

·
Registered
Joined
·
668 Posts
Yes, the error was not corrected by windows (otherwise there would be a WHEA report catching/correcting the error), the error then passes onto the game and crashes it.



Set affinity: Task manager, details tab, right click on game.exe, set affinity, tick/untick as prefered (odd numbers should be logical i.e. hyperthreaded "cores", and even ones should be the physical ones).
Is CPU 0-9 the physical cores and CPU 10-19 the Hyperthreads?
 
4841 - 4860 of 4999 Posts
Top