Precision T3650 thermal throttling Xeon W-1370P

I have some/ slight doubts about this low profile heatsink/fan combo for tower,
Fan is ranked at 40 CFM which is fine
but heatsink' fin is roughly 38x25mm in size with 92 fins
(I've estimated sizing from a downloaded picture)
which brings total surface to 38x25x92 (x2 for both sides) = 175K (with obstructed airflow by say 25%-30% by 3 heat-pipes over small fins)

https://www.ebay.co.uk/itm/234404330202

This 0PWYKH SFF heatsink (I have in my case, I saw somewhere it's labeled as 80W) has FAN ranked at 43 CFM (I've purchased separately 7cm fan because second-hand heatsink came with very dusty one having troubles with bearings) and has 57 fins of average size 70x58mm leading to 70x58x57 (x2 for both sides) = 463K (plus direct connection to base and two heat-pipes obstructing only 7% of airflow) and it's blowing directly into VR heatsink.
There are 3-heatpipe versions of it too (from older Optiplex 9020 models as long as you can adjust mounting holes to larger socket sizing)

Generally this premium 175K index heatsnik/fan assembly seems to be materially worse than older assembly having 463K index (although older CPUs weren't that heat/power efficient like recent generations) despite having one extra heatpipe (although there might be better contact between fins and copper heatpipe).

I was referring to this 125w cooler kit, same with the cooler posted by other member.

Edit:  Yours looks similar to a 65w cooler from a Precision SFF.

1000W PSU...not power supply issue....  it is BIOS TPL issue....  I have used a tools called  ThrottleStop to adjust the Turbo time limit to unlimit, now all core can run up to 4.08Ghz now.... I am not dare to adjust other value, as I worry it may burn the processor......

Unless you change to water cooling or replace the case and cooler, thermal throttling will not improve much.

https://www.dell.com/community/Precision-Fixed-Workstations/Precision-3650-change-case/m-p/8153964#M9890

Yes, I was referring to 93XV1 product too, my picture attached to previous post was about SFF heatsink (since I've Precision 3450, not Tower 3650).
But I also do understand there limitations is coming from for 3650, power supply sits right above the CPU so you can't install proper tower heatsink there, and this low-profile is rated at 125W, it does it's job but simply can't go much higher than that as described here:
https://linustechtips.com/topic/1408494-overheating-brand-new-dell-precision-tower/

By the way - my i7-11700 starts thermal-throttling at 90 degrees C (as most likely applicable to all Gen11 CPU line), not 80 as someone has complained below.

Also picture attached to one of my previous posts shows sustainable frequency on all cores at 3.3 Ghz (above base but below turbo-boost 2.0 and 3.0) and relatively low temperature (70 'C) - at the same time Package power sits right at 65W (as per Intel TDP specification for this CPU). So it's not thermally-throttling but Power-throttling.
I've run stress-test on a single core/thread and got it up to sustainable 4.2 Ghz (again - low temp at 80 'C on that individual core but at the same time 64.5W on same core) and it peaked for around 10 seconds to 4.8 Ghz.

So it looks like HW cap limitation implemented by vendor which can be avoided if not via BIOS but at least via Intel's XTU or ThrottleStop (can't use either of these since enabling Hyper-V and non-persistent Sandbox as I'm often running multiple VMs for various testing). Thankfully there's no need for now as I'm ok with these PC performance results anyways.

I would have to disagree, there's no evidence of thermal-throttling in @tigerwife case, it rather points to power-throttling (although temperature is result of higher power/amps running through conductors, so those two are correlated).

Thermal-throttling normally occurs at 90 'C , not 80,
while power throttling by default (depending on CPU) is triggered (for my i7 CPU) at around 140W (+/-) for turbo-boost mode with fall back to 65W for sustainable workload (as per CPU TDP on Intel specification, there are K versions for higher TDP).

Thermal-throttling can be worked on with better cooling (e.g. improved airflow via extra fans or going liquid) - Power-throttling is a bit more dangerous since other electronic components on motherboard may not be designed to handle higher currents (and missing sensor-control logic to deal with surges) potentially leading to permanent/irreversible damage/failures (which won't be covered by warranty).

You will find that thermal throttling is inevitable, but if the room temperature is low, the clock can be increased to 4.2GHz, on the contrary it will only maintain 3.8GHz or lower.

i5 11500 all core should able to run at 4.2Ghz

why t3650 will clock down to 3.7 only then?

have got latest bios and thermal solution upgraded already

it is clearly that it have problem in dell BIOS for TLP handling if it is not thermal issue....

dell should allow user adjust turbo boost time limit in bios at least...

While running stress-test on just 2 cores I'm getting the same sustainable limit of 65 W and higher frequencies and there are more detailed reasons on enhanced information section of  HWInfo64 (marked with red and green):

It clearly states Electrical design related to power draw.

I'm sure it does run at 4.2 Ghz for some short time (10-20 seconds?)
Then it falls back to lower Power Limit (not thermal).
I'll repeat the test with clearer indications towards power throttling on my i7-11700 below (similar picture was already shown in one of previous posts):

As you can see (marked in magenta [1] ) there's Power Limit Exceeded event on both: per core and for a whole package (and no thermal-throttling event flags triggered).

On green part [2] it shows current CPU package approx 65W and max 142 Watts (20 seconds while CPU was running at 4.9 Ghz or x49 multiplier)

Part marked as [3] shows historic (5 minutes) load where in the beginning (nearly 20 seconds) it runs at 142 W / 4.8 Ghz then switches to x33 ratio with 3.3 Ghz & sustainable 65 W of power drain (as per Intel TDP specification for 11700 CPU)

Now about the most likely reason why this limitation was imposed at these levels (whatever those are).
Probably OEM manufacturer for Dell hasn't designed motherboard for higher currents/power drain and it will simply fry electronic components (characteristics degrade rapidly at higher temperatures and exceeding certain limits causes this changes to be irreversible or less and less reversable over time) causing Dell to have a nightmare with warranty cases.

I'm quite sure it's not a problem in Dell BIOS, it is by design (capabilities of electronic components and PCB conductors used in assembly).

There are separate tests you can run with CPU-Z - like limiting threads (say just for two Cores incase if you run AVX-512 tests since there are only two AVX512 execution units on this CPU) - you can get higher sustainable frequencies on less cores (although it will most likely be limited to nearly 65 Watts for sustainable workload anyways). Perhaps it might be worth reading Turbo-Boost documentation and intel recommendations on how these things actually work. I never was under impression what Turbo-Boost warrants max frequency on all cores for prolonged time - until seeing this thread.

The BIOS must have some trick..... while CPU on Asus / Asrock / MSI motherboard,  processor can all core running at turbo boost mode at infinity time..... but not on dell......

it is not overclock... just running turbo boost mode only

don't know why DELL not doing the same job as other brand.....

In order to protect the system, the BIOS cannot allow users to customize the CPU settings, otherwise they can already be adjusted with TXU (as shown in the figure).

Discussion is good, testing is great.  However, don't lose the focus of this post.  OP was asking for assistance and I quote  "No ideas? Would love to reduce temperatures and throttling"

For the temperature part, I found that CPU temp could be lower by upgrading a couple of fans.  On a similar system, it was done and achieved around 12° lower .  Liquid cool options are not viable due to limited space and associated costs.

For the throttling, if it is still persisted after the CPU temp being reduced, then it's safe to say that the cause was of power throttling.  By then, OP have to weight between the benefit of achieving goal of performance vs. the risk of system damage by power modification.

If change case is considered, a full tower workstation such as 5820/7820 would be better choice.

If you read details you might notice what most of my latest/corresponding replies (tests etc.) were intended for TW.. (who's discussing slightly different problem but very correlated at the same time) not jp_j..