Asus Z490 stuff

[bscool]

from that I would conclude that my performance difference comes from RTLs, not IOLs, see screenshots on page before.

But how do you change IOLs without also affecting RTLs anyway??? I always thought they were tied by a certain factor, depending on speed and timings.

Yes, setting MRC Fast boot to disabled "conserves" the trained RTLs/IOLs for next boot, but that is an ugly solution.

Also I have noticed that stability also varies from boot to boot (training to training), from say karhu error at 7% vs karhu error at 1000%. So ideally I would want to conserve the 1000% state, but for that I need to reboot once first before getting back into the bios and setting MRC Fast boot to disabled. chicken and egg...


I've had 2 CPUs, 4 memory kits and 1 board and behavior always was the same. So is my board shot in a very weird way?

To start use 21 21(or Auto) and not 25 25.

Then when you boot and get IO of say 15 and 14. Change them to 7 7.

Then change RTL from 73 73 to 65 66

(15-7=8 73-8=65 14-7=7 73-7=66 example).

This is just an example as your actual # will be different.

If you do this correctly you will get the same performance as you were using your 25 25 and it boot and will be stable IF your CPU, RAM and IMC are able to do it. You are really asking a lot(4700 17 17 17 1t) for only having 2 CPUS and 4 kits of mem. That is a level few can run daily and be stable. It will require top "binned" CPU IMC, RAM, mobo and cooling.

Here is a guide if you need more info, examples and pics. https://www.hardwareluxx.de/communit...tipps.1230518/

scroll down to the RTLs and IO-Ls section

"The trickiest are these values. On [AUTO] a lot of boards train a lot of crap. You have to remedy the situation manually, which is sometimes not easy for the professionals. You should definitely have tested the setting for stability before touching these values. This is the only way to ensure that potential errors come from the RTLs and IO-Ls.

Here is an example of automatically (poorly) trained RTLs and IO-Ls: After manual correction, you can see that not only the latency has improved, but also the write and copy values. What values ​​do I now enter for the RTLs and IO-Ls? good question. In a nutshell: trail and error! There are a few rules to keep in mind to help you succeed faster.

First of all, you should only change these values ​​together.
So you can't just change RTL CHA D0 R0 and see if it works. It will not. Training will fail and the PC will not boot.
So you have to touch all the values ​​directly.

Goal:
the IO-Ls should be as equal and low as possible for both channels.
Values ​​of 6 or 7 for the IO-Ls are always practical.

The RTLs strongly depend on the selected clock and the CL value (primary timing)."

[martine-dee]

Next question from me. Can M12E and i9-10900K control display without a graphics card? How?

Got to ask before leave my current box without its 1080 Ti.

[bscool]

Next question from me. Can M12E and i9-10900K control display without a graphics card? How?

Got to ask before leave my current box without its 1080 Ti.

Nope

[GemueseMonster]

snip

thank you for all that explanation.

I was going off the assumption that getting first error in karhu at 1000% is better than getting an error at 7%. Therefore I assumed training differences boot to boot would mean stability differences.

But in reality, as I've just learned, the difference between passing 7% and 1000% is just RNG, even at statically controlled dimm temps of 39C max.

At higher clocks higher percentages of test completion will be reached faster, which would lead you to believe your setup is sable faster. But percentage does not mean anything alone. Also needs a certain amount of time passed it seems.

So like test for 10000% karhu or one hour, whichever takes longer, to be really sure. Also means the main selling point of karhu over hci pro, that it is faster, is worthless.

[martine-dee]

Thanks a lot everyone! I am already running this system on stock, and ready to give it some test run before returning to cable management, bios etc. etc.

One question. Now that I fell back to using the mobo's integrated sound (M12E), I again face the problem that the sound is very screetchy while the graphics is rendered. This might get better once I move the video card into vertical position, but I am not yet there. Is there a way to mitigate that noise?

Update: One workaround is to attach the speakers to the stereo output on the chassis, which goes to HD Audio. Now, I still need the i/o panel for 5.1.

[Ezveedub]

Im just fully guessing here but try to turn off windows "Fast start up"

https://www.asus.com/us/support/FAQ/1031533/

And right below the setting make sure sleep and hibernate are also unchecked.. Give that a shot....

I did this setting in Windows 10 v2004 and now it seems to have resolved the random power ons once it's shutdown. The PSU was swapped out from an EVGA B3 850 to a Corsair AX860 and it still had power ons 30 seconds or so after shutting down. Ever since applying just the disable Fast Start only, this issue is gone.

[GanjaHouse]

I have just formatted my system and now I have all the updated drivers.
But there are certain games that my rtx2080ti does not have a 99% load gpu and varies between 89/98%
The cpu has a 5.1ghz oc but it appears as 5,080ghz

Any ideas why my gpu doesn't work 99%?

Resolution is 2K Native, Gsync On

[Falkentyne]

I have just formatted my system and now I have all the updated drivers.
But there are certain games that my rtx2080ti does not have a 99% load gpu and varies between 89/98%
The cpu has a 5.1ghz oc but it appears as 5,080ghz

Any ideas why my gpu doesn't work 99%?

Resolution is 2K Native, Gsync On

Is your FPS uncapped?
Your GPU will not work at 100% if

1) it doesn't need to (it's rendering up the maximum the engine allows).
2) you're using a FPS capping or vertical sync method.
3) the CPU is holding back the rendering ifrom the GPU for some reason (I believe there were some issues with RDR2 and framerates higher than 144 fps, with certain amounts of cores or threads specifically enabled?).

[GanjaHouse]

I don't use any kind of uncapped FPS.
My gpu uses 89/95% load only, never 99%

2. I don't use vertical sync and gsync is activated in full screen.
3. Right now I don't remember but I use the same oc in the cpu I haven't modified anything and the use of the cores seems normal.

[btrach144]

Have you tried other games? Some games scale better than others.