cancel
Showing results for 
Search instead for 
Did you mean: 

mapping Intel XTU to BIOS settings for undervolting Coffee Lake on ASUS Prime Z370-A

menaquinone
Level 8
Hi. I'm attempting to undervolt my Coffee Lake CPU Intel Core i7-8700K (8th generation) which is on the ASUS Prime Z370-A motherboard on Windows 7 (I bought this configuration thanks to info I've seen on phoronix) and I was wondering how to transfer the settings from Intel® Extreme Tuning Utility (aka Intel® XTU) to the ASUS BIOS settings...

More specifically, in my case, I'm mainly interested in two Intel XTU settings, how they map to ASUS BIOS settings(for Prime Z370-A):
1) Core Voltage Offset (in Intel XTU) which when changed also changes in tandem the setting Cache Voltage Offset
and
2) Processor Graphics Voltage Offset which also changes Processor Graphics Media Voltage Offset

In my tries, it seems as if the following mapping is NOT quite accurate:
1) Core Voltage Offset and Cache Voltage Offset (in Intel XTU) -> mapping to -> CPU Core/Cache Voltage - Offset Mode (in BIOS settings)
and
2) Processor Graphics Voltage Offset and Processor Graphics Media Voltage Offset -> mapping to -> CPU Graphics Voltage - Offset Mode (in BIOS settings)
The reason I say that it's not, is because I get 3-4 degrees Celsius higher temperature with the BIOS settings than with the Intel XTU settings. Also, the Wattage draw at the socket is 14 Watt higher with the BIOS settings.

My guess is, that I'm missing some setting which also needs to be offset lower. Or, simply Intel XTU is using different ways of applying offset which affects more components that way (seen by looking at CPU HWMonitor's IA Offset, GT Offset and LLC/Ring Offset which are modified by Intel XTU but not by BIOS's offset settings)

Let me show some screenshots to ease this.

Here are Intel XTU and BIOS settings when I'm using ONLY Intel XTU to undervolt:
In idle:
85105
85106
85107
85108
85109
85110
85111
85112
85113
85114
HWMonitor in idle:
85115
Here's how it looks like in load:
85116
85117
(that's 190W constant consumption at the socket, this way with Intel XTU settings)


Here are Intel XTU and BIOS settings when I'm using ONLY BIOS settings to undervolt(thus Intel XTU settings are reset to their defaults to not interfere):

in BIOS (compared to previous BIOS screenshots from above) this is what's changed(note: I did reboot once with this changes so that 0.912V will be in effect and thus detected correctly):
85118
here's idle:
85119
I can't attach more than 15 pics per post(due to forum rules, it seems), so the other two will be in next post.

Note that even though ASUS and Intel don't provide a graphics driver for Windows 7, I was able to successfully use win7_8th_i3_i5_Driver_2.0.rar which I've found on biostar's site. Apparently some code modifications had to be done (the .inf hacks were not working for me, found on other sites) and thus re-sign the drivers for them to actually not display Code 43(like the .inf hacks would do for me, for DEV id. 3E92)

I mention that because I thought that the graphics driver being installed was important in order to know for sure that my undervolt is stable, for example using PowerMax to run a graphics test while also Prime95 was running, to test both CPU and GPU at the same time and see they don't crash.
4,930 Views
5 REPLIES 5

menaquinone
Level 8
continuation(adding the 2 pics which I couldn't add to OP)

Here are Intel XTU and BIOS settings when I'm using ONLY BIOS settings to undervolt(thus Intel XTU settings are reset to their defaults to not interfere):

HWMonitor idle:
85120
HWMonitor in load:
85121
(the socket Wattage here(with BIOS settings for undervolting) is 204 Watts, up by 14 Watts compared to the Intel XTU way of undervolting)

The load refers to using Prime95 with these settings:
85122
sha256 and version of Prime95 used(if it matters):
cde7e4c490f060e303a58d3cdc85b6e47b9133e84381ef5805651a1c6ab3fb3d
p95v298b6.win64.zip

What I found puzzling to note is that VCORE and VIDs when using Intel XTU are both at around 1V(approximately - implied),
but when using BIOS settings only VCORE is at 1V, the VIDs are at 1.3V in idle and only 1.2V when in load. But still, I'm guessing if they were 1V then that 14 Watt difference wouldn't exist, my guess.

Also to note: I don't know how to set the LLC/Ring Offset(seen in HWMonitor) by using BIOS settings. Maybe this is the one that if set would get rid of the extra 14 Watts?

Side note: Processor Cache Ratio is wrongly detected by Intel XTU as 44x(or maybe it shows only the upper limit there? and thus I'm misreading it), when in fact it is 40x (aka 4000Mhz NB Frequency, seen by me in CPU-Z and HWInfo) because BIOS settings Ring Down Bin is Auto(aka Enabled) and Cache MIN/MAX Ratio are both on Auto. (true at the time true screenshots were taken)

I was able to apply undervolting in a completely different(and surprising to me) way. But this way the voltages seem to be reported more accurately(eg. in HWMonitor) and I'm reaching the same temperatures and power usage(at socket) as with the previous Intel XTU settings(which was based on negative offsets alone).

All I had to do is set(in BIOS):
IA DC Load Line to 0.01
IA AC Load Line to 0.01
and
CPU Core/Cache Voltage - Offset Mode
to +0.050 (yes, plus) [+0.035 was unstable - error in the first 2 minutes of Prime95 maxium stress)
whereby:
CPU Graphics Voltage - Offset Mode
remained at -0.130 (yes, minus still, because GT AC/DC Loadline are unchanged to 3.1 mOhms (and I can't find a BIOS option to change them) - reported by HWInfo)

Now the reported HWMonitor voltages are stable and not changing much. Too lazy to get screenshots, besides I thought there'd be more traffic here, heh. But, here it is(this info) for completion.

undervolting a 8700k was never a thing since deliding can get almost 20c drop on load temps.

theres no independent cache voltage setting as far as i now.only vcore and llc.
there are other voltages that also attach to cache like vccio but that is more used in ram xmp or overclocking along with vccsa.if you just enabled xmp and didnt touch those then chances are they are set too high and that will raise cpu temps a little too.

as for the mOhoms settings on gigabyte 1=0.01 mOhms on asus its 0.01=0.01 mOhoms so the description in the bios was wrong.

tl;dr: it was the battery, it had no current.

For some reason (could be coincidence, but I don't know why I doubt it!), ever since I've used the new settings above, after leaving the computer disconnected from power socket for 8+ hours (just like before-using the new settings btw), then plugging it in and turning it on, the BIOS settings, the time&date and the administrator password are all gone(I didn't have user password set). Date is set to 2016. So, this is just as if the CMOS battery lost all power at some point during that time, or was removed or something similar! But I find it odd that this only started happening now after using the new IA AC/DC Load Line settings with the slight +0.060 offset (I've had to increase it a bit due to another Prime95 error in the first 4mins, but it does seem stable now!).

Since then, I've tried shutting it down, unplugging computer from power socket, pressed power button to drain capacitors, wait a few seconds, replug, turn on - but it still had the date/time and BIOS settings. Perhaps the CMOS battery now, still had enough charge to keep the settings(and clock). Then, it really is an odd coincidence and my CMOS batter is a gonner(only had this computer for about 2 years and a couple of months). So if I leave it again for 8+ hours I should expect the battery to be drained again...

Is this batter actually just a battery or is it a rechargeable battery ? on ASUS Prime Z370-A.

According to this thread the admin password should've been saved and not lost after removing CMOS battery for 24 hours.
I have removed the cmos battery and left it out for over 24 hours, the boot menu was reset but, the password was still saved.

The password is designed to survive resets and power loss.

Maybe by "power loss" they don't mean CMOS battery power loss(at the same time when there's no power connected to the board via PSU).

I suppose I could test if it's the settings, by using F5 to restore all settings to defaults, and then shutdown, unplug, wait 8+ hours! Maybe 24 hours just to be extra super sure! And if the settings aren't lost, then by god wtf... :)) then maybe BIOS is resetting them at startup due to unstable settings? but this has never happened before: I'd still be allowed to see and change my settings even when undervolting was unstable enough to not even POST (it would show me a POST after a few tries)

Side note: confirmed that restoring settings from a profile doesn't also restore the BIOS Logo Display - it should be Disabled but it's Auto! (F5 sets it to Auto and loading a profile doesn't reset it to the value that it was when that profile was saved). This little BIOS bugs are quite damaging the reputation of ASUS in my mind, gotta say. Attention to details (or lack of it) tends to do that to me, maybe I'm alone.

Confirmed lost all settings again after about 10-20 minutes of being unplugged from socket! Next try is with Load Optimized Defaults(ie. F5).
Confirmed lost all again even with default settings.
I thus checked the battery it's says:
+
KTS
LITHIUM BATTERY
JAPAN STD
CR2032
and it had 2.87-2.90 V and 0.00 A, replaced it with new which has 2.96V, 0.33-0.37A
thus it really was a coincidence, my old battery really was dead.