Dell Pro Max 16/18 Plus (2025 model) pre-release discussion — MB16250, MB18250

[Easa]

I dont think so, these CPUs behave like this in laptops, it corresponds to the Tjmax. I belive that @AL123 observed the same behaviour. 

 

[MyPC8MyBrain]

38 minutes ago, Easa said:

these CPUs behave like this in laptops

ooh wow 105c TJmax 🤯
that cant be good for the silicon die or overall performance.

ignore Intel attempt to cover up their cheap edge silicon die slices, that much heat is still excessive and indicates one of two things, either cooling is insufficient for the 285HX in that chassis or the cpu needs repasing and reseating,

another user experiences similar temps is not an indicator for this being ok! 
it just means you both have similar issue.

last round i went through 6 replacement before they finally sent a unit with proper cpu,
before that i was idling at 90+ out of the box and was ready to just give up on dell,
with a good cpu i was idling out of the box 2-4 degrees above ambient.
(its documented somewhere in the 7670/7770 owners thread, and that cpu was no where near the 285HX efficiency)

[Easa]

3 hours ago, MyPC8MyBrain said:

ooh wow 105c TJmax 🤯
that cant be good for the silicon die or overall performance.

ignore Intel attempt to cover up their cheap edge silicon die slices, that much heat is still excessive and indicates one of two things, either cooling is insufficient for the 285HX in that chassis or the cpu needs repasing and reseating,

another user experiences similar temps is not an indicator for this being ok! 
it just means you both have similar issue.

last round i went through 6 replacement before they finally sent a unit with proper cpu,
before that i was idling at 90+ out of the box and was ready to just give up on dell,
with a good cpu i was idling out of the box 2-4 degrees above ambient.
(its documented somewhere in the 7670/7770 owners thread, and that cpu was no where near the 285HX efficiency)

 

 

Excuse me, but where have you been for the last decade?

 

1. Laptop CPUs frequently go all the way to the TjMax since Coffee Lake came. Boost to the max => Throttle => Keep sustained levels. Its not pretty, but it is like this for almost a decade. Phone SoC chips behave the same way, consoles do too, as do Macs. Precisions behaved the same way, too (with TjMax at 100°C). On the PMP16/18, this spike is there for about a second, then it drops down to 80-90s. Keep in mind that the heat transfer from these delicate lithographies is difficult, even with a vapor chamber. Overall perfomance? That overall performance, which is currently the best amongst all mobile workstations ever designed, almost on par with gaming laptops that have way more power budget, and at the same time better than both its direct competitors?
    
2. Do not mistake heat for temperature, its not the same thing. Efficiency has nothing to do with the operating temperature either. The cooling system in this machine is far superior to anything that Dell has ever produced. It is able to cool down the machine whilst keeping the fans more or less relaxed. There is no issue. All the review units that I was able to source the data from behave the same, even the NotebookCheck´s PMP16 and PMP18. I have no itentions of repasting the unit at the moment, if someone here will beat my benchmark results significantly after a repaste, that would change my mind. 
    
3. Idling at 90+° out of the box is not normal, but lets stay out of the fairy tales please. No modern Intel CPU, not even desktop, will idle at 2-4°C above ambient, unless the ambient is 30°C (if you live in tropical climate then you may be right). I am typing this from a workstation with 285K, cooled by a quad radiator system, and even this idles at 34°C (coolant at 27.8°C, ambient 21°C). In laptop it is simply impossible, unless it has been in a cold for very long and you have applied liquid metal. Pro Max Plus 18 idles at 38-40°C , and I am 100% sure that none of the old Raptor Lake Precisions will idle at lower values with the stock TIM. 

P.S.: It might be about time to stop judging thermal behaviour of a highly advanced semiconductor device against the boiling point of water, just because it was the standard back in the day. Yes, modern CPUs run hot, hotter than we would like, but thats the way it is. To cool a 150W CPU to a "nice" temperatures you need a pretty beefy heatsink even in a desktop.   

[Aaron44126]

I agree, don't worry if the temp is pushing a bit over 100 °C for a bit. Modern CPUs are well-tuned to throttle themselves if they get too hot, and it is expected that they will get hot under load, because running them close to the throttling point is how you get the best performance.

 

(Not just an Intel thing. My MacBook also hits around 104 °C under full load. I've run video encode jobs that last for days and it's never been a problem.)

 

Repasting might help raise the frequency that you can run the CPU at while hovering at the thermal throttling point, and it might help lower the not full load running temperature (really only useful to calm down the fans), if the replacement thermal transfer material or application is actually notably better.

 

The time to worry if is the system is getting unstable under load. (Random lockup would be the most likely symptom, but you could also get apps crashing, BSOD, ...)

[yslalan]

12 hours ago, Easa said:

I am enclosing the comparison between the Dell Pro Max Plus 18 and the ThinkPad P16 Gen3 with the same configuration. You may as well ignore the PL/TDP data, as on the PMP18 it means theoretical maximums of the platform, whereas on the TP P16 G3 it is couting on the platform power / thermal limitations already. Both systems were tested with clean OS and stock power settings, same conditions 1:1, it was just a misunderstanding. 

 

The result looks great. Is it possible to share your GPU vBIOS, I will try to flash into my PM16P, when I am able to reach to my device next month. I want to do a comparision between the stock 140w and 175w vBIOS on PM16P.

 

[yslalan]

9 hours ago, MyPC8MyBrain said:

what you're describing sounds more like badly pasted CPU, repast it before you kill the silicon die

The truth is, no matter how thermally efficient the paste is, the scheduler will still drive the CPU/GPU to their maximum PL and hit Tjmax because of the limited heat capacity of a thin heatsink. That’s completely normal on laptops, especially high-performance platforms with aggressive power allocation. These short bursts of high temperature don’t cause any real issues; they’re there to provide the instantaneous performance you need.

[MyPC8MyBrain]

My personal experience with multiple high-end Dell Precisions confirms this extreme variation:

6 Bad Units: Three 7670s and three 7770s all idled at 90∘C+ and throttled easily.

1 Golden Unit: The final replacement unit, with the same model and same flagship CPU, ran at Ambient+2∘C at idle—proving that a low-voltage, low-heat chip does exist in the bin, but that the average chip is a thermal disaster.

The high temperatures we’re seeing aren’t just an unavoidable side-effect of efficiency — they’re largely the result of wider binning tolerances. In other words, the chips span a much broader quality range, which leads to higher heat output on the weaker bins. Intel smooths that out by raising the allowed thermal ceiling (higher Tjmax), so everything appears normal under sustained load. I can dive deeper into my theory, but that’s the short version.
 

7 hours ago, Easa said:

Excuse me, but where have you been for the last decade?

I’ve been busy dealing with life since 2022. I stopped tracking the fine-grain CPU landscape right around the time 13th gen showed up. And to make matters worse, Intel introduced ‘Undervolt Protection,’ which locked out ThrottleStop and shut the door on the one tool that made laptop tuning predictable.
 

So yes — I missed some of the incremental changes, but the fundamentals haven’t shifted: hotter silicon, wider bins, higher Tjmax, and fewer user controls.
 

  

5 hours ago, yslalan said:

The truth is, no matter how thermally efficient the paste is, the scheduler will still drive the CPU/GPU to their maximum PL and hit Tjmax because of the limited heat capacity of a thin heatsink.

True — with a distinction.
Yes, the scheduler will push the CPU/GPU to Tjmax regardless of paste quality when the heatsink’s thermal capacity is this limited. But that doesn’t mean the paste is irrelevant. Your operational thermal buffer matters.
 

If the system is already idling in the 80–90°C range because the interface material isn’t performing well, then half of your thermal window is already gone before the real workload even starts. That window should ideally begin only a few degrees above ambient, not 40–50°C higher.
 

Starting that close to Tjmax means you hit the ceiling almost instantly, which forces the system into aggressive throttling long before it should.

[MyPC8MyBrain]

21 hours ago, Easa said:

lets stay out of the fairy tales please.

  
Wooohat… you don’t believe in fairy tales? 😮
Then maybe you’ll believe the i9-12950HX running cooler and more efficiently than your brand-new i9-285HX?"
 

(Images of my own system below — since you think numbers like these belong in storybooks.)

Surface temps, idle temps, and CB23 results were all posted in this forum long before the current generation even existed — nothing new, nothing fabricated.
 

On 3/10/2023 at 3:29 PM, MyPC8MyBrain said:

 

 

 

On 12/16/2022 at 9:34 AM, MyPC8MyBrain said:

On 12/4/2022 at 4:20 AM, MyPC8MyBrain said:

i am still tinkering with my new 7770, allot of mixed feelings so far,

one very interesting tidbit i found out I think every 7X77 owner should be aware of,

in theory and on paper iGpu is suppose to run a more efficient shift for mundane chores,

for most parts it has been this way in previous systems i owned, i run day to day activities over the integrated gpu, when i need more gpu processing power i run the applications assigned to the dGpu, this has always been the assumption at least for me,

 

i also know now why i was getting better performance with bios "Hybrid Mode" off in both the 7670 and now with the 7770, that's because running both iGpu and dGpu at the same time adds 10c degrees to the overall idle temps!

having a iGpu on the 7X77 seem to achieve the opposite of the desired effect, 
it uses more power and needlessly increase system heat by 10 degrees Celsius,
it is much more efficient to run directly off the dGpu ("Hybrid Mode" toggle OFF),

 

my current 7770 idles at 38-39c 😉
CB23 run video https://a.uguu.se/tcBwWBH.mp4 
(on Cool Power Plan in bios, on power saver in windows,
if i switch to ultimate performance in windows the score will be over 24k, but i idle 10c hotter)

more to come...
 

 

None of this was “luck of the draw.”
It took 3 months, 6 replacement units, and refusing to accept canned responses from entry-level techs who weren’t even aware of these thermal behaviors. I escalated repeatedly, every time with technical data they couldn’t refute. Only after that did the correct unit arrive.
 

Had that final system not performed the way it should out of the box, I was ready to walk away from Dell altogether — and I’ve been buying Precision systems since the M60 era in ’02–’03.
This isn’t my first rodeo.
 

So no — these numbers aren’t fairy tales.
They’re what happens when you understand the platform, strip out the inefficiencies, and hold Dell accountable for delivering a properly binned unit.

 

[Easa]

1 hour ago, MyPC8MyBrain said:

  
Wooohat… you don’t believe in fairy tales? 😮
Then maybe you’ll believe the i9-12950HX running cooler and more efficiently than your brand-new i9-285HX?"
 

(Images of my own system below — since you think numbers like these belong in storybooks.)

Surface temps, idle temps, and CB23 results were all posted in this forum long before the current generation even existed — nothing new, nothing fabricated.
 

 

None of this was “luck of the draw.”
It took 3 months, 6 replacement units, and refusing to accept canned responses from entry-level techs who weren’t even aware of these thermal behaviors. I escalated repeatedly, every time with technical data they couldn’t refute. Only after that did the correct unit arrive.
 

Had that final system not performed the way it should out of the box, I was ready to walk away from Dell altogether — and I’ve been buying Precision systems since the M60 era in ’02–’03.
This isn’t my first rodeo.
 

So no — these numbers aren’t fairy tales.
They’re what happens when you understand the platform, strip out the inefficiencies, and hold Dell accountable for delivering a properly binned unit.

 

 

No, I do not, and you still have to disprove that fairy tale of yours. 

 

Exuse me, but do you actually understand what the term ambient means? Because unless you are in a tropical climate, where room temperature is 31-32°C, then no, you are not idling 2°C over ambient. And even this was reached, by your own words:

 

"(on Cool Power Plan in bios, on power saver in windows,
if i switch to ultimate performance in windows the score will be over 24k, but i idle 10c hotter)"

 

So thats that. But yes, lets all compare our meaningless idle values whilst we cripple our system in every way we can. Or it just might be that you do not know what idle is. And you still have to show the actually meaningful value - the core temperatures during the CB23 rendering phase. Not the 1 minute HWInfo window with power draw topped at 26W. This power draw, to be precise:

 

The second thing you, apparently, have no idea about, is efficiency. Under no circumstances will the i9-12950HX be more efficient than Ultra9 285HX. It has nothing to do with its temperature. Efficiency is measured by comparing the value of input to the value of output, applied here it means electrical power in => compute power out. And the old Alder Lake CPU will not hold a candle to the Arrow Lake CPU, whatever PL you cap it at. They have almost the same PL1/PL2 values, yet the Ultra scores more than 50%, even when you hit Ultimate Performance. Thats efficiency. Brought in by the advancements of lithography and architecture. 

 

 

 

 

 

[MyPC8MyBrain]

  

5 hours ago, Easa said:

unless you are in a tropical climate

is Vegas "tropical" enough?
 

On 12/9/2022 at 10:34 PM, MyPC8MyBrain said:

 

These weren’t cherry-picked images. I simply went back to my older posts and selected the most recent ones I had shared. You’re free to read the thread from my first post onward, there are plenty of screenshots covering idle, load, and various test conditions.
 

As for ambient: yes, I understand exactly what it means. Those measurements were taken under specific conditions that I already documented at the time, BIOS Cool mode, Windows Power Saver, and a controlled environment. When switching to Ultimate Performance, I explicitly noted the +10°C increase. None of this was hidden or misrepresented.
 

Regarding efficiency: I never claimed architectural efficiency of Alder Lake exceeds Arrow Lake. My point was strictly about thermal behaviour under mobile constraints. Architectural efficiency on paper and real-world thermal behaviour in a confined chassis are not the same thing, and they often diverge.
 

Your interpretation mixes two separate discussions:

Silicon-level efficiency (input power -> compute output)

Thermal behaviour and sustained performance in a mobile cooling budget
 

The former favors newer architecture.
The latter is heavily dependent on bin quality, power limits, chassis design, and thermal headroom — which is the context of my earlier comments.
 

So let’s keep the discussion technical and grounded in actual mobile behaviour rather than assumptions about what someone ‘does or doesn’t understand.

[yslalan]

Dell finally released the QCM AI100 NPU configuration for the PM16P, also available to see the card in the manual:

The 3rd fan size is compressed due to the wider daughter board

[Easa]

The laptop has been working fine for a month, since this day, I have experienced this weird issue. Without a clear link to any activity or app, while browsing or watching a video, on battery or plugged in, the screen freezes, goes black (without backlit), then flashes about two or three times and stays black with backlit on. Like when MUX switching. However, after this, the laptop either freezes or continues to work, but without image. Have to force reset. Switching off Advanced Optimus in the BIOS seems to do the trick, but I would like to keep my battery life. Also, there was no issue for a month. The drivers are the same. I have experienced this several times during the last few hours.

 

Each time it happens, there is this entry in the eventlog, exactly like this one:

 

 

Anybody has experienced this behaviour ? Thank you. 

 

[MyPC8MyBrain]

This is classic Dell Advanced Optimus + ADL/Meteor Lake iGPU + Nvidia BW-series Pro GPU instability.
Your user's symptoms line up perfectly with the known failure mode of Dell’s 2024–2025 Pro Max / Precision redesigns.

textbook Advanced Optimus failure the pattern is too exact to be coincidence

This is NOT a touchpad event. That Win32k entry is a side-effect It appears every single time the display pipeline resets.
It is the last successfully logged event before the GPU chain collapses.

1. Disable Panel Self Refresh (PSR) on the Intel GPU
Intel Graphics Command Center -> System ->Power ->Disable Panel Self Refresh

2. Force-panel fixed refresh rate (disable dynamic refresh) Set the panel to a fixed 120Hz, not variable.
Windows Display Settings ->Advanced Display
Turn off VRR/Adaptive Sync.

3. Update these three components (in order):
Nvidia Studio/Enterprise driver (not Game Ready)
Intel GFX from Intel’s site, not Dell
Dell BIOS + EC package

4. If issue persists: leave Advanced Optimus OFF
Dell’s implementation is unstable on MTL/BW right now.
This is not user-specific — it’s widespread.

5. Optional but effective In Nvidia Control Panel:
Manage Display Mode ->Prefer discrete GPU
(but leave MUX off so you keep battery life)
This reduces the number of switching events.

[Easa]

9 hours ago, MyPC8MyBrain said:

This is classic Dell Advanced Optimus + ADL/Meteor Lake iGPU + Nvidia BW-series Pro GPU instability.
Your user's symptoms line up perfectly with the known failure mode of Dell’s 2024–2025 Pro Max / Precision redesigns.

textbook Advanced Optimus failure the pattern is too exact to be coincidence

This is NOT a touchpad event. That Win32k entry is a side-effect It appears every single time the display pipeline resets.
It is the last successfully logged event before the GPU chain collapses.

1. Disable Panel Self Refresh (PSR) on the Intel GPU
Intel Graphics Command Center -> System ->Power ->Disable Panel Self Refresh

2. Force-panel fixed refresh rate (disable dynamic refresh) Set the panel to a fixed 120Hz, not variable.
Windows Display Settings ->Advanced Display
Turn off VRR/Adaptive Sync.

3. Update these three components (in order):
Nvidia Studio/Enterprise driver (not Game Ready)
Intel GFX from Intel’s site, not Dell
Dell BIOS + EC package

4. If issue persists: leave Advanced Optimus OFF
Dell’s implementation is unstable on MTL/BW right now.
This is not user-specific — it’s widespread.

5. Optional but effective In Nvidia Control Panel:
Manage Display Mode ->Prefer discrete GPU
(but leave MUX off so you keep battery life)
This reduces the number of switching events.

 

Thank you for the suggestions. I knew that it has something to do with the iGPU driver, that the touchpad entry is not really the culprit. However, I dont know why is there no entry about iGPU / Display driver failure.

1. These systems have Dell iGPU drivers that install without the Command Center. If you install the CC by yourself (through MS Store for instance) it results in a bugged optimus behaviour, where the machine keeps powering the dGPU on and off. The last iGPU driver that Dell provides is 32.0.101.6987 from 29.7.2025. 
    
2. There is no dynamic refresh rate on this machine (sorry, its in Slovak language, but you see the setting): 
   
    
3. Nvidia RTX Pro driver updated to the latest version as of this morning (581.80), the regular way with a clean install, not using DDU or anything similar. Intel GFX, I am going to try it, but as I have said, it causes other problems. Dell BIOS and EC package was updated about a week ago to the latest version, which is still the last available. 
    
4. If the issue persists, I will have to run dGPU only, yeah. 
    
5. In the Nvidia Control panel, the options are Automatic, Optimus and dGPU only. What exactly is difference between the Automatic and Optimus setting? AFAIK the Optimus always goes through iGPU, even when running dGPU, and the Automatic is like an on-the-fly MUX Switch. Or am I wrong?
   
   

 

At the end of the day, the question still stands. Why has the issue manifested only after a month? I am using the machine in the same way every single day, and there was no recent alterations to the drivers or the configuration. The only thing that I have done differently, is that I have played Witcher III while docked, for like 10 minutes, to test temperatures. 

[MyPC8MyBrain]

The timing isn’t random, and it doesn’t require a driver update to break.
Advanced Optimus failures often build up over time because the instability sits in the runtime switching logic, not in the installed driver files.

Let me break down your points directly.

 

 

1. “Why is there no iGPU / Display driver failure entry?”

Because when the Advanced Optimus handshake collapses, the failure happens below the Windows logging layer.
If TDR recovery fails at the firmware/EC level, Windows never gets a chance to record the actual iGPU fault.

The Win32k “touchpad QA” entry you keep seeing is just the last subsystem that manages to log anything before the display pipeline dies.

No iGPU error log doesn’t mean no iGPU failure, it means the driver never got to write one.

 

 

2. Command Center vs non-Command Center Intel driver

You’re correct that adding CC on top of Dell’s trimmed package causes power-state oscillation.
But that issue is separate from what you’re seeing here.
The black-screen behavior is tied to AO switching, not CC.

 

 

3. Dynamic refresh rate: you’re right, it’s not available on this panel

Correct. This panel only exposes fixed refresh modes.
The instability isn’t coming from VRR; it’s coming from PSR and low-power iGPU states during AO transitions.

 

4. Nvidia Automatic vs Optimus

Your understanding is close, but here’s the precise difference:

Optimus -> iGPU is always the display controller; dGPU only renders.
Automatic -> Nvidia’s runtime decides whether to bypass the iGPU (MUX-like handoff), depending on load and power state.

Automatic is not true hardware MUX switching; it’s AO logic deciding when to engage a quasi-direct pipeline.
That transition is exactly where these freezes happen.

 

5. “Why did this start after a month with no changes?”

This is the key misunderstanding.

Advanced Optimus failures often start after accumulated runtime state drift, not changes.

 

These triggers are well-documented on multiple machines:
* After a heavy dGPU session (like your Witcher 3 test).
* After long uptime or multiple sleep/wake cycles.
* After powering the dGPU during docked use.
* After EC power-state desync.
* After a PSR -> non-PSR transition under load.
* After the iGPU hits a low-power retention state and fails to reassert scan-out.

Your Witcher 3 test is exactly the kind of load-state that destabilizes the AO handshake.
It doesn’t matter that it lasted only 10 minutes, you forced:

* dGPU full power
* iGPU low power
* docked display chain
* AO auto-switch upon undock or mode change

That alone can push the platform into an inconsistent state that only resolves with a full reboot or, if unlucky, gets stuck in the failure mode you’re describing.

This is why disabling Advanced Optimus immediately stabilizes the system.

 

 

Bottom Line

Nothing “changed.”
The platform simply hit the known weak point:

Meteor Lake iGPU + Blackwell Pro + Dell’s AO implementation + recent power-state transitions = exactly the pipeline collapse you’re seeing.
 

You didn’t do anything wrong and there is no single “event” that needs to occur for AO to break.
It’s a fragile runtime switching system, and once it destabilizes, the failures start appearing in clusters.
 

If you want stability and battery life, then:

* keep PSR disabled
* update Intel + Nvidia in that order
* avoid using Automatic mode
* or leave AO off entirely
 

But don’t chase the “why now?” angle,
the cause is the design, not your usage.

 

 

[ebourg]

Quote

Let me break down your points directly.

 

Was this message IA generated? 🤔

[MyPC8MyBrain]

oops... you got me, but i'm AI not IA 😅
kidding aside, real person generated not AI.

[MyPC8MyBrain]

On 11/16/2025 at 9:27 AM, win32asmguy said:

I returned the Pro Max 18 Plus and opted for a Legion 9i Gen 10 instead.

Smart move i will most likely follow suit as well!

Given the current market landscape, the Legion 9i Gen 10 really is the only logical replacement for what used to be Dell’s flagship class, and it proves how redundant and outdated ISV only configurations have become.
 

I’d love to see a proper head-to-head comparison in a non legacy, non OpenGL workflow. Better yet, even bring in a legacy ISV certified task and run them side-by-side. That would make the reality obvious very quickly.
 

What Dell seems to overlook is that the workstation market has changed.
ISV only configurations made sense 15-20 years ago, but today they’re a niche. Most modern workflows don’t benefit from ISV drivers at all, many are actually slowed down by them.
 

That’s why systems like the Legion 9i exist: they offer workstation class performance without forcing you into an ISV locked GPU stack that’s irrelevant for 90% of real-world users.
 

If Dell wants to stay competitive in the high-end mobile space, they need to bring back configuration flexibility, ISV when it’s needed, GeForce when it’s not. Forcing everyone into one certification path doesn’t reflect the reality of today’s workloads.

Good job Dell, the operation was a success, but the patient didn’t survive.