Asus Z790 Apex XMP issues

[kvarq]

Asus Z790 Apex
13900KF SP 103 (P113/E83/MC73)

G.Skill 2x16GB 6400MHz C32 F5-6400J3239G16GX2-TZ5K (M die)

It seems that the Z690 XMP issues are still persisting this round as well...
XMP I or II activated seem to not be ok for games (these are crashing for some reason), ALTHOUGH GSAT passed overnight for long hours or TM5 with anta's profiles passed as well
Did not touch the IA/SA voltages, just XMP is enabled...

[kvarq]

akgis_ wrote:
Those voltages are the default ones, you only get the XMP voltages next time you restart the system from saved settings in BIOS.

You should check the voltages in Windows in HWinfo for example like I said.

see my karhu screen from the post above, those are the voltages applied after enabling the XMP 🙂

also those are the voltages in bios as well, shown in this screen from this post

[iBruceypoo]

Yea but Mr. kvarq had a bad experience with his Z690 Apex and now it's happening once again with Z790 Apex - that ain't fair. 😞

[Silent_Scone]

iBruceypoo wrote:
You should not have to be tweaking separate individual voltages at this time. The 6400 kit should run with XMP l settings, period.
.

This is objectively incorrect, Bruce. No one CPU will be the same and not all will follow auto rules or predefined settings beyond reference code. This is why the UEFI allows for manual adjustment in the first place. If what you were saying were true, why would the majority of rails be made adjustable for the user? It's as though you've created a pseudo trip point in your mind where you feel things should "simply work". Sadly, this isn't how overclocking works. One would have to have an extremely large sample size pool to form this opinion, and even then it's not an absolute guarantee.

Pointing individuals on the right path is more important than telling them what does and doesn't work. It comes down to an alignment issue that needs to be dialled out. Of course, the more we increase the frequency the slimmer the margins become - but this doesn't mean lower speeds are impervious to instability and do not require adjustment.

kvarq wrote:
worst case scenario I'll fill in manually all the XMP stuff (timings and frequency), that was the only workaround back then - it was (and still is) weird that with XMP enabled passed hours of testing with TM5 or GSAT but in games not stable whatsoever
no one could tell why, this "solution" (dropping XMP and switching to manual) worked for most of people
with the default voltages (including the SA voltage 0.89V and MC voltage 1.154V) the system passed hours and hours GSAT overnight and TM5 with anta's profiles...



As I stated, for the MC voltage I tried 1.30 and 1.33 V and crashes are happening faster... I am trying as suggested to test with a lower voltage 1.2-ish V, below 1.3V
Did not find in earlier posts a suggestion for the SA voltage, but I assume it should be 1.2-1.3V maybe even higher - I remember when I had the Z490 board that each bios had its "own" SA preferred voltage for 4000C16 - sometimes worked with 1.28V and sometimes even 1.36V. Back then things were pretty much clear, if the ram passed GSAT there was like 99% chance to be rock sWolid in gaming. Since Z690 this changed.

I am still stubborn to have XMP enabled this time 🙂

The System Agent voltage is less sensitive on 13th gen, but you may find better stability by tuning it.

Have you ran the MC SP test via the AI Tweaker page?


Try running Karhu Ram Test and see if you can produce any errors. If you're able to pass at XMP1 or XMP2 in GSAT for an hour or more, there may be another issue. I would strongly recommend taking the system to an integrator and seeing if they can best diagnose the problem for you.

[kvarq]

Silent_Scone@ROG wrote:
CPU System Agent is still sensitive this gen, so there's some of that to take into consideration, too.

The board will do 6400 just fine.

Silent_Scone@ROG wrote:

The System Agent voltage is less sensitive on 13th gen, but you may find better stability by tuning it.

based on the latest statement, is it ok to leave SA voltage auto then?

[kvarq]

worst case scenario I'll fill in manually all the XMP stuff (timings and frequency), that was the only workaround back then - it was (and still is) weird that with XMP enabled passed hours of testing with TM5 or GSAT but in games not stable whatsoever
no one could tell why, this "solution" (dropping XMP and switching to manual) worked for most of people
with the default voltages (including the SA voltage 0.89V and MC voltage 1.154V) the system passed hours and hours GSAT overnight and TM5 with anta's profiles...

Silent_Scone@ROG wrote:
Please see the sticky in my previous post. No vendors guarantee overclocking.
Please see previous suggestions for these rails. You will need to tune them methodically. Another alternative is taking the PC to a system integrator and getting them to dial it in for you.

As I stated, for the MC voltage I tried 1.30 and 1.33 V and crashes are happening faster... I am trying as suggested to test with a lower voltage 1.2-ish V, below 1.3V
Did not find in earlier posts a suggestion for the SA voltage, but I assume it should be 1.2-1.3V maybe even higher - I remember when I had the Z490 board that each bios had its "own" SA preferred voltage for 4000C16 - sometimes worked with 1.28V and sometimes even 1.36V. Back then things were pretty much clear, if the ram passed GSAT there was like 99% chance to be rock solid in gaming. Since Z690 this changed.

I am still stubborn to have XMP enabled this time 🙂

[kvarq]

worst case scenario I'll fill in manually all the XMP stuff (timings and frequency), that was the only workaround back then - it was (and still is) weird that with XMP enabled passed hours of testing with TM5 or GSAT but in games not stable whatsoever
no one could tell why, this "solution" (dropping XMP and switching to manual) worked for most of people
with the default voltages (including the SA voltage 0.89V and MC voltage 1.154V) the system passed hours and hours GSAT overnight and TM5 with anta's profiles...

Silent_Scone@ROG wrote:
Please see the sticky in my previous post. No vendors guarantee overclocking.
Please see previous suggestions for these rails. You will need to tune them methodically. Another alternative is taking the PC to a system integrator and getting them to dial it in for you.

As I stated, for the MC voltage I tried 1.30 and 1.33 V and crashes are happening faster... I am trying as suggested to test with a lower voltage 1.2-ish V, below 1.3V
Did not find in earlier posts a suggestion for the SA voltage, but I assume it should be 1.2-1.3V maybe even higher - I remember when I had the Z490 board that each bios had its "own" SA preferred voltage for 4000C16 - sometimes worked with 1.28V and sometimes even 1.36V. Back then things were pretty much clear, if the ram passed GSAT there was like 99% chance to be rock solid in gaming. Since Z690 this changed.

I am still stubborn to have XMP enabled this time 🙂

[kvarq]

My MC SP varies between 77 and 68, depends on the run


It's far from being consistent, I thought that is dependent on the voltage tunings after, but then I left everything on auto (default-default, so not even XMP enabled), and after each run is shown different a different SP.

[kvarq]

Silent_Scone@ROG wrote:
CPU System Agent is still sensitive this gen, so there's some of that to take into consideration, too.

The board will do 6400 just fine.

Silent_Scone@ROG wrote:

The System Agent voltage is less sensitive on 13th gen, but you may find better stability by tuning it.

based on the latest statement, is it ok to leave SA voltage auto then?

anyways, I ran an hour karhu as well (to be honest I trust more GSAT or even TM5 with anta's profiles) and did not show any error


I think any "integrator" will test the system using synthetic tests only, most likely using GSAT or similar - I don't think they will dial the MC SA voltages etc...
I spent alreay 900 Eur on this board, why to spend another hundreds more for something that "will not do"?

The XMP issue in gaming is an Asus Z690 thing - apparently also for Z790.
Testing the systems with synthetic tests (including 3D mark etc) might show that these are "rock solid", but in games is different, for some reason many are sensitive when XMP is enabled, crashing to desktop. No one figured it out so far. This is something that Asus technicians should consider first, not focusing on the OC only (yeah, many of them are OCers unfortunately).

[kvarq]

Silent_Scone@ROG wrote:
CPU System Agent is still sensitive this gen, so there's some of that to take into consideration, too.

The board will do 6400 just fine.

Silent_Scone@ROG wrote:

The System Agent voltage is less sensitive on 13th gen, but you may find better stability by tuning it.

based on the latest statement, is it ok to leave SA voltage auto then?

anyways, I ran an hour karhu as well (to be honest I trust more GSAT or even TM5 with anta's profiles) and did not show any error


I think any "integrator" will test the system using synthetic tests only, most likely using GSAT or similar - I don't think they will dial the MC SA voltages etc, just enabling XMP. And what's the point dialing them, since each bios version requires different values?
I spent alreay 900 Eur on this board, why to spend another hundreds more for something that "will not do"?

The XMP issue in gaming is an Asus Z690 thing - apparently also for Z790.
Testing the systems with synthetic tests (including 3D mark etc) might show that these are "rock solid", but in games is different, for some reason many are sensitive when XMP is enabled, crashing to desktop. No one figured it out so far. This is something that Asus technicians should consider first, not focusing on the OC only (yeah, many of them are OCers unfortunately).

[kvarq]

iBruceypoo wrote:
Yea but Mr. kvarq had a bad experience with his Z690 Apex and now it's happening once again with Z790 Apex - that ain't fair. 😞

Silent_Scone@ROG wrote:
This is objectively incorrect, Bruce. No one CPU will be the same and not all will follow auto rules or predefined settings beyond reference code. This is why the UEFI allows for manual adjustment in the first place. If what you were saying were true, why would the majority of rails be made adjustable for the user? It's as though you've created a pseudo trip point in your mind where you feel things should "simply work". Sadly, this isn't how overclocking works. One would have to have an extremely large sample size pool to form this opinion, and even then it's not an absolute guarantee.

Pointing individuals on the right path is more important than telling them what does and doesn't work. It comes down to an alignment issue that needs to be dialled out. Of course, the more we increase the frequency the slimmer the margins become - but this doesn't mean lower speeds are impervious to instability and do not require adjustment.

Silent_Scone@ROG wrote:
Please see the sticky in my previous post. No vendors guarantee overclocking.

In a way I might agree with Silent_Scone's statements, but then why the expensive boards are advertised that "can do 8000MHz+"?
If they sell something expensive, then they should deliver what they promised - to quote our good fellow Bruce.

In the end, let's face it, it's not a fact that they sell, but just an... "opportunity".
If even a low/middle XMP is not working, "you're not lucky, your CPU is crap". Not fair I think.