Dell Precision 7540 and 7740 Owner's Thread

Thank you for the quick reply!

Well, kinda sad, as every laptop Quadro i owned could be pushed at least a little (Now typing from ZBook 15 G1 with K2100M). Anyways, i hope that Dell with stay true to current DGFF design with the next gen. of WS Laptops, as i would like to swap out new GPU by then.

As for the display, it seems like the upgrade path to something decent is open:

http://www.panelook.com/NV156FHM-N4B_BOE_15.6_LCM_overview_34862.html

Everything matches so far, just gotta try it. I have no use for [email protected]", as the laptop will be docked on external display most of the time, and I would rather enhance gaming experience a bit. If Im correct, this is currently the best 30pin eDP IPS on the market, that you can swap there (looking at colour depth, frequency, contrast, response and gamut)

 

Might have been true for Kepler and Maxwell-generation Quadros, but Pascal and Turing (even non-Quadros) have been locked for overclocking. One can push the Pascal Quadros in theory (given a hardware programmer is available), but even the GeForce Turing notebook GPUs are locked down, as of now.

 

Theoretically, i do have access to HW programmer (and a guy who know this stuff) so if anything changes about that lockdown, it would be cool.

About the 7540, it will be an upgrade from the old M4800 / ZBook 17 G1 i use now. It was a difficult choice between P53, this, and ZBook 15 G6, but at the end of the day i think the 7540 is the best allrounder. One thing I´m still not so sure about is the CPU. I would normally opt for a better 6 core and happily settle down with it, namely the i7-9850H, but the possibility of having an unlocked 8 core in 15" laptop is just intriguing. With the proper adjustments and tweaks, including application of liquid metal, i can imagine pretty good results.

Whats the local opinion on having 8 core in 7540? Do we have any owners here, happy or regretting the CPU choice? What about RAM? Is it beneficial going 3200MHz route, or should I just keep it low with standard 2666?

 

Ok, the new heatsink is arrived, and I proceeded with "thermal step 2":
-passed 800 and 1000 sandpaper over heatsink cpu area to better-finish the surface;
-removed the heatsink screws springs and replaced them with iron flat washers (to hand-control the final heatsink pressure over cpu);
-used quality thermal paste (artic mx-4);
-used high quality thermal pads to cover entirely the VRM area;
-throttlestop fine-tuned with active undervolt of ~ 120mV.

The good:
From my original 2-heatpipes no-gpu heatsink, to the new 3-heatpipes full-version heatsink there's an avarege difference of 5-6°c especially in transient load peaks (as supposed). The higher heatsink pressure genereted from the improved tightness of the screws reduce another 4-5°c under full heavy loads bringing the thermal throttling to be negligible and allowing the management of turbo speeds in the range of 4.2-4.4Ghz allcore depending on the cpu load type (AVX, non AVX, etc..).
The bad:
The 2 power threshold of 75W and 107W for long and short terms turbo power are not bypassable even with throttlestop, it seems like they're hardwired inside bios. So, even if thermally the cpu could stay in the 100W area, after the short-term turbo timer preset it cut-down to 75W area, meaning an allcore turbo speed of 3.8-4Ghz depending on the cpu load type. Not a bad result anyway, but definitely annoying.

As important side note, originally I would like to use liquid metal instead of normal thermal paste, but I've got a strange issue: after 5-10 minutes, the notebook shut down and goes in a sort of protection-mode that prevent it to power on, and only with A LOT of repeated hard reset it could power on again (in a first time I think I had screwed up the mainboard). I think that the fact that liquid metal is highly conductive, in some way goes to generate some current-interference that put the machine in protection mode. I read something similar some years ago regarding the incompatibility of liquid metal with some kind of gpu. A pity, since this kind of paste could bring down the temperature another 4-5°c.

The new heatsink with modified screws:


Some results:


Geekbench v5: https://browser.geekbench.com/v5/cpu/335919

Anyway, I think that finally I can say 'this is fast'.

 

Good results, thanks for posting.

Questions:
Were you using Arctic silver MX-4 with the old two-pipe heatsink? (Trying to tease out whether your original 5-6 degree improvement with the 3-pipe heatsink might be just a result of using better paste, or if the heatsink itself makes that much of a difference.)

I haven't seen anyone use the washer trick before but that makes sense to increase/tune the pressure from the heatsink on the CPU. Are those the same screws as before (springs removed and washers added) or did you have to find new screws?

Are you undervolting the CPU as well? (I assume you discussed this in a previous post, I don't remember everything )
[Edit] Nevermind I see you mentioned -120 mV in the post above. Nice.

 

Thanks, it's a pleasure to contribute to forum's life with some infos.

To answer your questions:
-yes, same mx-4 paste also with original heatsink (I repaste cpu the same day I got the notebook)
-yes, same original screws but with removed springs (and new washers added)
-yes, cpu undervolted to -120mV

 

Great work.

On the comment about repeated hard reset, I had something like that happen to me, I think once under wall power and once under battery power.

I was still getting few random windows crashes with my -160mv cpu and my -50 mv igpu (had igpu for testing once at -100mv I think that was bad crashs and possibly the hard resets)
I have since moved it to -150mv and 0 mv igpu ( believe igpu can give some benefit to higher results to cpu scores as the igp draws less, but I think its problematic)

And as you have mentioned about the hard power throttle limits in bios which sucks, but at least with the undervolt and ur repasting you get much higher shorter burst turbo and the all core short burst turbo oc as well, which in some tasks will help, but for longer number crunching we are limited by the bios.

In conclusion it can be assumed that undervolting the cpu and/or igpu can lead to repeated hard resets (since I did not repaste, I can only blame the undervolting)

 

oh on a different note, I have my 3900x running and for the price of the cpu and a x470 board its pretty impressive, paired with a AIO h115i push 2x140mm pull 3x140mm with a bunch of fans, but most will run at 500-600rpm and with running the benchmarks my water will only go up by 1c at times, with the quite setup, if fans on med or full blast I can have the water stay at same temp, but with these cpu's the ambient temp and water temp batter a lot, I can get 50-100mhz all core auto oc on top what I get regularly
R15
Single: 204
Multi: 3145-3175

R20
single:507
multi:7236

https://browser.geekbench.com/v4/cpu/14675973

I am slightly undervaluing it, and having more aggressive vrm settings for bios, to keep things more stable, but I played with all kinds of settings, for the small gain of benchmarks scores the temps can go up really quick, im sure custom water loop would be best, but I do not see the justification in cost for that, temps can dip to high 30c to avg in the 40c's but will spike to 50's during single R15/R20 the temps will go to 59-61C and with multi from 72C-83C depending on the type of loads
This is with a max fan speed for best scores, now doing this on quite will be a little lower, as the water is a little warmer with low rpm speeds, and by the time the water rises up and fans kick in as Im using the radiator fans connected all 5 to the water temp, other case fans are controlled by cpu temp that can jump up and down, though I put a threshold floor
single core boost is a bit weak, I do not see it hit 4.5ghz but between 4.2-4.35 ish sometimes on light loads it will go to 4.5, but not on full load, and multi core does to 4.1-4.2ghz AMD still has room for improvement. This is with latest bios and Arctic silver MX-4 on cpu

Nice to note that my 12 core is above the old threadripper 1950x and in single above some intel chips

Its interesting that with this chip its kinda like the mobile chips, where my slight undervolt or near stock is best results, as auto oc works so well

 

@va123 nice work on the Ryzen, I played a little with a 3800x and a x570 strix, impressive performance/cost ratio. And yes, with this new kind of cpu the bios manual tuning is almost negligible.

Back to the 7740, I also had some power off issue with excessive undervolt, but nothing compared with what happens to the machine using liquid metal (I really can't find a reason if not something related to a strange kind of core-interference).
With the modified thermal management, the performance level could be really steady, the only problem is that on long enough runs the power-throttling kick-in and the system remains on the 75W area. This means, for example, that the Cinbench scores goes from the initial ~1900/4200 to 1725/4000 after 20-30 continuos runs (and in real utilization this could be the case for long time process like rendering).

I think that Dell should have set the 2 power limit threshold to 90/125W instead of 75/107W for this kind of cpu.

 

Thanks

I agree, and not only that, shouldn't they have used some logic circuit to analyzed the temps? I mean that should be the smart way to deal with things and possibly have more different stepings. Because a laptop can be in different temperature environments

That liquid metal issue, is quite the anomaly

Also have you thought about IC Graphite?
I would think this is best case for devices like laptops and the like


Their is similar if not same one at digikey, you can buy a much larger sheet for cheaper, I have some but have not used it on anything and really do not have time to experiment