Optimal settings Z790 Tomahawk Wifi Max i9 14900k

[MigraineFilms]

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

 

[MigraineFilms]

Just updated my BIOS to the 0x129 microcode update, ran a bunch of benchmarks.
Cinebench R23 still triggers the thermal throttle
20 Minutes OCCT is excellent, averaging at 77 degrees celcius
Cinebench 24 runs fine, higher score than before, maxing out at 88 degrees, no throttle triggers.

 

[FlyingScot]

MigraineFilm, will you double check me here while I try to summarize where our testing has taken us:

Point 1: When using the Intel Defaults (either with 0x123?, 0x125 or 0x129) we see some undesirable behaviors, e.g. higher observed temps and temp spikes at the individual core level.
Point 2: Given point 1, it’s fair to assume that 0x129 has not cured us of high voltage spikes, only reduced their size.
Point 3: If you make any attempts to address point 1 and 2, you get dumped into MSI Defaults, which may negate some, or all, of the benefits of Intel‘s “patch.”
Point 4: In regards to the severity of point 1 and 2, it does appear to have something to do with Intel’s auto-boosting algorithms. And in your case (14900K) there is a boatload of them.
Point 5: In reference to point 4, down-clocking via manual core ratios appears to significantly alleviate the observations in point 1.
Point 6: Cooling a 14900K (or any other Raptor Lake) continues to be a challenge. One has to assume that slightly lower voltage spikes at potentially higher temperatures (see point 1) is not going to reduce degradation as much as if both voltages and temps were lower.

I think that‘s an accurate representation of where we‘re at with your tests.
===========================================
Answer from MigraineFilm...

This is correct, here is some added information:

*Both with Power Level 1/2 set at either 200 or 253, those spikes will still occur and mostly with P-core 5 and 7. (numbered from 0-7, so it is actually core 6 and 8)

• *When setting the P-core offset ratio in the BIOS, a warning pops up that it will disable the Intel Recommended setting. It is presumably safe to do it with Intel Extreme Tuning Utility.
• 
  
• *Cooling is perfectly doable at 200, 253 or even 300 watt (albeit at almost 100 degrees) with an AIO, the heat occurs when Intel tries to boost the CPU because it thinks it can with low temperatures. Those boosts do absolutely nothing in in long run benchmarks or workloads, my renders don't speed up by 40% just because its boosts a single core to a higher limit for 2 seconds.

I will await response from MSI tech support to know for sure IF MSI settings DISABLE the 0x129 patch

AC/DC set to a lower spec, Lite Load set to a lower mode or manual undervolting all have the same result. It's running 75 degrees at PL200, 83 at PL253, and sometimes peaks at 92 degrees for a second. As those peaks happen with both Power Limits (200 or 253), I think it must be a turbo boost that gets enabled. Those spikes happen in cinebench, rarely to never in Prime95 maximum heat test.

I might test the "stock" frequencies next week with ALL boosts disabled (except for thermal velocity boost, as I gather that is not so much a boost but throttle function).

Bottom line: the i9 14900k is absolutely not worth the money. Yes, it runs faster than the i7 even with lower power limits and frequencies, but its the "K" model, which needs to be slowed down to run properly. I paid for a K model to make it run FASTER.

 

[FlyingScot]

Just to further clarify the point for any newcomers reading these posts, especially if you have a 14th gen i7 or i9, CiTay’s power tuning guide found here primarily utilizes CPU Lite Load and PL1 and PL2 to reduce overall power usage and therefore temperatures. And while this approach does have a direct effect on voltages, it may only get you partway to your goals. That’s because unfortunately PL1 and PL2 and IccMax are not effective tools at taming single core and dual core boosting behavior. Even at a PL1 of 125W, you can still see 1.6V 6+GHz boosts.

So, in addition to CiTay’s guide, you will need to think separately about “peak voltage management.” This can be achieved by either manually limiting core frequencies via manual core ratios, or by deactivating Turbo Boost 3.0, (ABT if you have it) and making adjustments to the TVB/eTVB settings. Presumably, if you take the latter approach, you might also be able to stay with the Intel Defaults mode. If you find that this latter method only gets you so far, then you could augment it with a lower max temp via changing TjMax. And, again, all of this is separate from any undervolting efforts you make via either CPU Lite Load or Manual voltage offsets.

 

[citay]

Yes, good old "Intel Turbo Boost Max Technology 3.0", i added a paragraph in my guide about it now too. We're very much managing Intel's mess now, hopefully future BIOS versions can take some of that work off our hands again. We're seeing the end result of years of going in the wrong direction for high-end CPUs (well, their choice of directions was not that manifold, but they just went too far in one direction). This could've easily happened with GPUs too, not too much difference there on the high-end parts.

 

[FlyingScot]

Yes, good old "Intel Turbo Boost Max Technology 3.0", i added a paragraph in my guide about it now too. We're very much managing Intel's mess now, hopefully future BIOS versions can take some of that work off our hands again. We're seeing the end result of years of going in the wrong direction for high-end CPUs (well, their choice of directions was not that manifold, but they just went too far in one direction). This could've easily happened with GPUs too, not too much difference there on the high-end parts.

Never a truer word has been spoken/written.
CiTay, I think you just hit the proverbial nail with the proverbial hammer. Bravo!

 

[MigraineFilms]

So, I've been working for quite a while and keep encountering thermal throttle warnings.

When stress testing with Prime95, everything stays around 85/88 degrees.
OCCT, 80 degrees
2 hours of Handbrake, 80 degrees
A single run of Cinebench R23, it thermally throttles at 95 degrees (my TjMAX), and P-core 7 gets throttled (the boost core, albeit its not boosting).
Premiere Pro renders, 80 degrees
Media Encoder render, P-core 7 throttles at 95 degrees.

Even with Lite Load mode 5 and LLC5, it thermally throttles. PLL at 235 instead of 253, it thermally throttles.
I do not understand why this is happening, at all. Its normal according to Intel, but this cannot be benefitial to the lifespan of the CPU

Edit: I hate this thing. PL200, still thermal throttle triggers

 

[FlyingScot]

So, I've been working for quite a while and keep encountering thermal throttle warnings.

When stress testing with Prime95, everything stays around 85/88 degrees.
OCCT, 80 degrees
2 hours of Handbrake, 80 degrees
A single run of Cinebench R23, it thermally throttles at 95 degrees (my TjMAX), and P-core 7 gets throttled (the boost core, albeit its not boosting).
Premiere Pro renders, 80 degrees
Media Encoder render, P-core 7 throttles at 95 degrees.

Even with Lite Load mode 5 and LLC5, it thermally throttles. PLL at 235 instead of 253, it thermally throttles.
I do not understand why this is happening, at all. Its normal according to Intel, but this cannot be benefitial to the lifespan of the CPU

Edit: I hate this thing. PL200, still thermal throttle

You can customize TVB with certain temp thresholds, so perhaps you could use this feature to rein in that P-Core 7 behavior, or TVB boosting in general. I don’t have access to that menu (i9 exclusive), so I’m no help. However, OCTVB is discussed here: https://skatterbencher.com/intel-overclocking-thermal-velocity-boost/ and I do see it discussed in the MSI Z790 BIOS manual.

I guess this is only relevant if TVB/eTVB is actually doing something.

 

[charonme]

I have TVB ratio clipping (reducing clocks at high temperatures) working on i7 too, I think the opportunistic overclocking at low temps was the type of TVB that was supposed to be available on i9 only although some people with i7s report it works for them too. I try to use the clipping as a kind of "soft throttle" before the temperature reaches the final throttle point and drastically cuts down performance for a few seconds

 

[FlyingScot]

I have TVB ratio clipping (reducing clocks at high temperatures) working on i7 too, I think the opportunistic overclocking at low temps was the type of TVB that was supposed to be available on i9 only although some people with i7s report it works for them too. I try to use the clipping as a kind of "soft throttle" before the temperature reaches the final throttle point and drastically cuts down performance for a few seconds

I can't speak for the Z790 platform, but when I was researching an oddity on my Z590 platform, I found out that early BIOSes exposed the TVB settings for i5 and i7 CPUs. This was apparently an oversight and TVB options were removed in newer BIOSes (for non-i9's). I mention this finding because you may run into this situation when you upgrade from your old BIOS code.

 

[MigraineFilms]

I tried some more settings, locking P and E core ratio's, lowering the wattage by another 50, still thermally throttles.

I will reseat my CPU, readjust the contact frame and refit the AIO with new cooling paste, lets see what that does. The last thing I can think if, is that the contact frame isnt tight enough allowing for a very, very, very, very tiny bit of extra room between the CPU and the AIO, causing air bubbles in the cooling paste.
This thermal throttling issue is getting worse with time, and that is not a common sign of the degredation I believe.

If that fails, I'll RMA the thing.

 

[MigraineFilms]

So, I tried running the benchmarks with HWMonitor and XTU running too, those apps monitor a consistent 85 degrees all over the board, no thermal throttle warnings at all, but HWiNFO monitors peaks of 91 with thermal throttling. I dont get any of this anymore.

 

[Arctucas]

So, I tried running the benchmarks with HWMonitor and XTU running too, those apps monitor a consistent 85 degrees all over the board, no thermal throttle warnings at all, but HWiNFO monitors peaks of 91 with thermal throttling. I dont get any of this anymore.

Have tried different versions of HWiNFO?

Have you reached out to the author of HWiNFO?

 

[charonme]

there are many thermal sensors in the cpu, I think there might be even 2 in each of the cores themselves, and what gets reported to external software are post-processed values from these many sensors. So thermal throttling inside the CPU can be triggered by a high value from some of these, but that specific value might not get reported to the outside software. Also the overtemperature might occur too quickly for a short period of time that it doesn't get reported to anything readable by software

 

[MigraineFilms]

I tried the following settings:

AC/DC 15
LLC Mode 5
-0.100 undervolt.

Lo and behold! No more thermal throttling, temperatures lowered with 3 degrees overall, and after 20 minutes of R23 and 5 minutes of R15, no thermal throttle warnings.
Also, scores went up a little.

Downside is: my idle voltage does not drop below 1,2. It would go as low as 0.780, and this will cause my idle temp to be around 42 average instead of 36.

 

[charonme]

idle voltage does not drop below 1,2

that's strange, are your C-states enabled? Do you have balanced power profile in windows instead of "performance"?
what's your reported vcore in load if you limit the Pcores to 5.5GHz?

I'm trying to estimate what would your LLC5 AC/DC 15/15 -0.1V be equivalent on my LLC5 44/44 and so far it looks like I'd have to go below -0.18, but that's extremely unstable on mine

 

[MigraineFilms]

that's strange, are your C-states enabled? Do you have balanced power profile in windows instead of "performance"?
what's your reported vcore in load if you limit the Pcores to 5.5GHz?

I'm trying to estimate what would your LLC5 AC/DC 15/15 -0.1V be equivalent on my LLC5 44/44 and so far it looks like I'd have to go below -0.18, but that's extremely unstable on mine

C states enabled, C10, balanced power profile, idle vcore doesnt drop, wattage stays around 40/50 instead of 25.

My VID at 55x multiplier is 1,180, actualy speed is 5200mhz during load, does not seem to decrease R23 scores (39k)

 

[Arctucas]

I tried the following settings:

AC/DC 15
LLC Mode 5
-0.100 undervolt.

Lo and behold! No more thermal throttling, temperatures lowered with 3 degrees overall, and after 20 minutes of R23 and 5 minutes of R15, no thermal throttle warnings.
Also, scores went up a little.

Downside is: my idle voltage does not drop below 1,2. It would go as low as 0.780, and this will cause my idle temp to be around 42 average instead of 36.

View attachment 192564

You seem to have several performance limit reasons?

 

[MigraineFilms]

You seem to have several performance limit reasons?

Yes, power limit and current, cant be bothered to tune that further, as my benchmark scores are fine.
Those would probably disappear if I would raise iccmax from 307 to 350 or 400, but Im not comfortable doing that due to the stability issues
Also, I turned off intel turbo max 3.0, which also triggers a performance limit warning.

 

[MigraineFilms]

I'm so done with this, how does it do 30 minutes of R23 just fine, and the day after, it just throttles again. It's such a dumb thing. I'll just leave it at this until it fails, I'll never buy a 150+ watt CPU again, it's just insanely inefficient and not worth the hassle or 10% performance boost at all.

 

[FlyingScot]

If you lower your E-core ratio 1 or 2 ticks, you’ll get your peak VIDs down. Your P-Cores might behave better. The change isn’t likely to be dramatic, though. Just something I spotted. E-cores 19 - 23 have the highest VIDs, which impacts the whole CPU.