Optimal settings Z790 Tomahawk Wifi Max i9 14900k

M

MigraineFilms

Astral Fridge Magnet
Joined
Jul 24, 2024
Messages
233
Hi all, as I was filling other threads with questions, here is a dedicated thread to my question.

Can you help me getting my i9 under control, it keeps thermally throttling even under very low power settings, and I dont fully grasp the undervolt settings.

My system:
i9 14900k
MAG Z790 Tomahawk Wifi max
Deepcool LT720 AIO
Contact Frame
DDR5 Trident Z 6400
Samsung 990 Pro
RTX 4080 Super
Corsair Shift RM1000x

So far I've done this:

Lowered PL1 and PL2 to 175 watts
Lite Load 9 (AUTO)
Voltage adaptive/offset -0.125 (screenshots still say -0.110, changed this, still throttles)

It can run for hours perfectly fine, benchmarks all doing great, cinebench R23 multiple runs, Prime95 30 minutes heat/stress test and then when I take a screenshot, HWiNFO says its thermally throttling without showing any high temperatures. HWiNFO also refreshes slowly, even on 1000ms, since the latest update.

Any advice is much appreciated

P.S: I do not experience any instability in the past 6 months I owned this computer, besides a sporadic lag in the audio, there are no crashes, app hangs, game failures, BSOD's or any of that kind.

Edit: All the information is a bit overwhelming and I'm feeling I'm getting lost in all the numbers a bit
 

Attachments

  • BIOS1.jpg
    BIOS1.jpg
    218.1 KB · Views: 311
  • IDLE.jpg
    IDLE.jpg
    1.4 MB · Views: 302
  • LOADThrottle.jpg
    LOADThrottle.jpg
    2 MB · Views: 252
  • System2.jpg
    System2.jpg
    337.5 KB · Views: 224
  • System1.jpg
    System1.jpg
    353.6 KB · Views: 220
  • BIOS4.jpg
    BIOS4.jpg
    229.2 KB · Views: 215
  • BIOS3.jpg
    BIOS3.jpg
    208.8 KB · Views: 321
Sort by date Sort by votes
sascha117715db02f2 said:
Have you tried default values on fast throttling/overTemp protection? Meaning 100°C on both as you seem to have set them to 95. I noticed that on my board/CPU combo (z690 Tomahawk/i7-14700KF), when I had those set to 95, I would also get a thermal throttling entry in HWInfo even though no temp-max values were anywhere near TJMax. Resetting those two to 100°C made that go away.

Also: What happens when you reset everything to optimized defaults and then employ Intel's recommended power-settings for your CPU? I'd start with their "Performance" preset, not sure if they recommend that or the "Extreme" one for the i9, but, whatever. Go with "Intel Default" preset in the BIOS, then double check PL1/2 and ICCmax that the preset produces and, if needed, adjust those to match the Intel recommendations from their spec-sheet.

Might also want to play around with enabling/disabling CEP. I have those disabled on my board.

I'd do what citay recommends in his power-settings thread and start from there and without manual undervolting and see what happens with temps, voltages and power-draw. Then I'd start lowering CPU Lite Load one setting at a time until the PC crashes during stress tests. Then raise CPU LL back up by a step or two and stress test again for stability. In my case (and I would imagine it being as "bad" or even worse on an i9), the default CPU LL setting ("12" now in my BIOS) or the even scarier "Intel Default" setting for CPU LL ("15") result in worryingly high voltages/temps/power draw. I used to be able to go as low as "4" before I started updating my BIOS to 0x125 and then 0x129 and I am now sitting at "6". Without touching anything else related to voltages (other than enabling undervoltage protection), and with Intel's recommended settings of PL1/2 at 253W and ICCmax 307, this results in my VIDs barely breaking 1.4V under the highest of loads.

If you're worried about high voltages/voltage spikes: AFAIK, the 0x129 microcode should be preventing those (I think even Buildzoid conceded that much.. :D) and as long as you don't go crazy on ICCmax and don't use the default-setting for CPU LL, you should be able keep the board from sending excessive voltage to your chip. You also shouldn't have to reduce your PLs to 175W (on an i9... lol) - especially when you're sitting at 307/308A ICCmax. Even my puny i7 will not exceed 200W of power-draw in CB23 with 307A, even though I have both my PLs set to 253W. In my system, I have to raise ICCmax to ..err.. 350 or even 360A to make the chip run into the 253W powerlimit instead of into the current-limit.

S.
Click to expand...
Ive tuned it as much that I got roughly the same benchmark scores at pl 200, and keeping the temps at 75 degrees.

My i9 happily, and I mean HAPPILY, sucks up every single extra watt it has to raise the clock speeds to max, for barely any extra performance.

I will try disabling IA CEP and setting the throttle limit back to 100. If it doesnt throttle then, it means it just goes a bit above 95, which is also a comforting thought.
 
Upvote 0 Downvote
MigraineFilms said:
My i9 happily, and I mean HAPPILY, sucks up every single extra watt it has to raise the clock speeds to max, for barely any extra performance.
Click to expand...
Yeah... same with the i7 if I let if off the leash. I've seen up to 350W of power-draw and well over 1.5V VIDs on this thing with the "right" settings... yikes!
Although... allowing *a bit* more craziness with the power-draw/temps (but no "real" overclocking), will raise my CB23 score from ~33,500 to ~35,500, but I'm not sure that means anything outside of all-core benchmarking.. :)

S.
 
Upvote 0 Downvote
persiz154202d8 said:
If you want more consistent results, download Benchmate and run the benchmarks on higher priority from the settings. Benchmate runs some processes that make higher and more consistent results.
Click to expand...
Benchmate has the exact same problem: 37719 score, 38992 when I kill all background apps.

sascha117715db02f2 said:
Have you tried default values on fast throttling/overTemp protection? Meaning 100°C on both as you seem to have set them to 95. I noticed that on my board/CPU combo (z690 Tomahawk/i7-14700KF), when I had those set to 95, I would also get a thermal throttling entry in HWInfo even though no temp-max values were anywhere near TJMax. Resetting those two to 100°C made that go away.
Click to expand...
hey, ran 20 minutes 253 watt R23, no thermal throttling! P-core 7 gets 92 degrees though, peak, average 86. Thats 20 degrees higher than p-core 1 who sits on 72 degrees. this is most certainly a IHS issue

Also, the R23 benchmarks are about 8% faster when going from PL200 to PL253, which in my opinion (and real world scenarios) is not worth the extra heat and voltage. It's very, VERY pleasant to know that the thermal throttle triggers are from between 95/100 and not 100+ celsius.

All in all, my current setup, with undervolt, at PL200 (R23 score of 37500), is fine for me. I can raise it further if needed, and it runs faster than the i7 14700k. Thats all I expected from the i9. As long as its healthy
 
Upvote 0 Downvote
I set iccmax to 400, instead of 307. The performance limit reasons disappear for IA Cores, but not for Ring.
Surprisingly, R23 went up to 37500, at PL1/2 200, which is the same as PL 253 and Load Lite 9 (lowest stable for me).
Keeping iccmax at 307 results in 36900 R23 score.
 
Upvote 0 Downvote
MigraineFilms said:
I set iccmax to 400, instead of 307. The performance limit reasons disappear for IA Cores, but not for Ring.
Surprisingly, R23 went up to 37500, at PL1/2 200, which is the same as PL 253 and Load Lite 9 (lowest stable for me).
Keeping iccmax at 307 results in 36900 R23 score.
Click to expand...
I'm not sure I can wrap my head around what you are saying unless it goes like this:

IccMax 307
PL1/2 253
Frequency looks like a sawtooth plot, as processor constantly shoots for the moon and then throttles. Rinse and Repeat. [The hare]
Throttling occurs due to IccMax

IccMax 400
PL1/2 200
Frequency is steady, and slightly above the average frequency of the previous settings. [The tortoise]
Throttling occurs due to PL2.

Is that what what we are seeing? Assuming all other settings are the same.
It would be nice to know max temp and average effective frequency. Although, what we really need is average E-Core effective frequency and average P-Core... But you have to make those as custom HWInfo sensors.
 
Upvote 0 Downvote
FlyingScot said:
I'm not sure I can wrap my head around what you are saying unless it goes like this:

IccMax 307
PL1/2 253
Frequency looks like a sawtooth plot, as processor constantly shoots for the moon and then throttles. Rinse and Repeat. [The hare]
Throttling occurs due to IccMax

IccMax 400
PL1/2 200
Frequency is steady, and slightly above the average frequency of the previous settings. [The tortoise]
Throttling occurs due to PL2.

Is that what what we are seeing? Assuming all other settings are the same.
It would be nice to know max temp and average effective frequency. Although, what we really need is average E-Core effective frequency and average P-Core... But you have to make those as custom HWInfo sensors.
Click to expand...
Maybe these screenshots will help.

With PL @ 200, the P-core frequency is about 300 to 400mhz lower, compared to @ 253 (will run these tests too, soon)
In OCCT, you can see that AC/DC40 LLC6 on my motherboard provides damn near identical VID/V-Core readings.
In R23, the readings are all over the place due to the constant dropping of load on the CPU between renders.

ACDC 40-40 -0,100V LLC6 400ICCMAX PL200 R23.jpg

ACDC 40-40 -0,100V LLC6 400ICCMAX PL200 OCCT.jpg


EDIT:

Tested it with PL 253 iccmax 400, R23 of 39K

ACDC 40-40 -0,100V LLC6 400ICCMAX PL253 OCCT.jpg
ACDC 40-40 -0,100V LLC6 400ICCMAX PL253 R23.jpg


Thats less than 5% increase from 200 to 253 watt, with an added 12 degrees to the CPU. The performance limit triggers are gone but the frequency doesnt increase (it always hovers between 5,2 and 5,3). So, it might trigger the performance limit warning at 307, but it doesnt seem to have any noticable effect
 
Last edited:
Upvote 0 Downvote
Well, there's thermal throttling where one of the cores just enters it for a split-second, but the temperatures otherwise being non-critical, and there's thermal throttling where everything is in the red pretty much. So depending which one you're faced with, either you have a reason to react, or you may be able to live with the occasional throttling bit being set which is not putting your cooling etc. in a lot of trouble.
 
Upvote 0 Downvote
Its just one core for a fraction of a second, not even fast enough to measure the actual temperature of the core (which was 87 according to HWiNFO).
It's just nagging at the back of my head, I don't like it.

On the other hand, it rendered 10 seconds of 4K heavy chroma keyed and collor corrected footage in about 10 seconds too. It's insane
 
Upvote 0 Downvote
So, I tried some new settings with the 0x12B microcode update:

Enhanced Turbo (MSI) Enabled
ICC Max 400
PL 253
VR Voltage Limit 1,400

My cores now boost to 5,6 and 6 ghz when running benchmarks, they did not do that before!

Edit: makes no difference in benchmark scores though! Exactly the same as generic PL 253, ICCmax 307, and only 6% higher than PL200
 
Last edited:
Upvote 0 Downvote
charonme said:
does the new update remove the C1E Support enabled/disabled feature from bios?
Click to expand...
no, and I havent had any lower scores in R23 either, had no impact whatsoever.

I take those benchmarks from tech sites with a HUGE grain of salt, because if scores are 500 points lower, that can be caused by SO MANY different factors. It's only a good comparison when done with the exact same system, exact same software updates, exact same windows features installed and exact same background processes running. So much stuff can influence the scores by several 100's of points, its not an indication of "worse" CPU performance
 
Upvote 0 Downvote
MigraineFilms said:
HUGE grain of salt
Click to expand...
yeah I completely discard what they say regarding the performance drop as they didn't even ensure they used the same settings. For all we know they might well have used some "defaults" and their motherboard manufacturer may have changed those auto defaults in the bios update

But what about the C1E Support option in bios? did it disappear also in the MSI update or is that only an asus thing?
 
Upvote 0 Downvote
charonme said:
yeah I completely discard what they say regarding the performance drop as they didn't even ensure they used the same settings. For all we know they might well have used some "defaults" and their motherboard manufacturer may have changed those auto defaults in the bios update

But what about the C1E Support option in bios? did it disappear also in the MSI update or is that only an asus thing?
Click to expand...
with my bloated PC constantly pinging servers and updating stuff, I get higher scores in R15 than wccftech has with both single and multi core, with all background processes eliminated I'm just 200 points away from their R23 score, even still, I'd rather have a working CPU than those 500 extra points. It's been a "bad" purchase anyway.

C1E is still in the BIOS, not removed for me. Even if it WAS removed, it can be enabled with QuickCPU I think
 
Upvote 0 Downvote
Back
Top