Dell Latitude 5491 and 5591

Numpads have effectively killed 15" laptops for me. To go one further, removal of separate touchpad buttons at the bottom of the touchpads have killed many other laptops (precisions/xps/thinkpads for example) for me.

The Latitude 74x0 is the only sane choice.

I hope they make a minimal bezel version with sufficient cooling for the quad cores. (I'm not sure if the 7490 is as quiet and cool as previous 74x0s with dual cores).

 

The 7490 is effectively silent in my wife's use as a (pretty heavily-used) general-use machine, with a SATA SSD, which is comparable to her E7440 (which was still working fine for her, but died.)

For a light development machine when I've borrowed it, and some casual gaming, the fan definitely comes on occasionally, but nothing like the 7820HQ 5470 or the new 5491.

I wouldn't put employer code on her machine (although I'm glad my present employer allows it on personal machines at all, given how long it took them to get me a proper Linux machine rather than a Mac or Windows box) and she doesn't want a Linux dual boot environment on it, so I can't compare directly -- but I would not expect either the SATA SSD nor the 15W quad cores are fast enough for my professional use (large Java applications with a ton of legacy code.)

Now that I've gone 6-core, I wouldn't want to go back to even a full-power 4 -- the 5491 even with the undervolting and capped turbo boost wattage throttling drops my build times by about 1/3 (not quite as much as one would hope given 2 more cores AND a pretty good pre-core speed bump, but it's still doing really well for a generational bump.)

 

That's good to hear. I would expect these 6 core machines to improve the thermals in the upcoming generations.

Sorry to hear about your legacy Java workload though. (Sorry, couldn't resist.)

 

I don't name employers here, but I definitely miss the ease of developing on an (essentially) all-modern-microservices architectures -- although 4 more hardware threads would have helped with running a 30+ service constellation in other ways.

On the other hand, I'm much happier with a lot of other things at this place, and I'm a geezer who's been doing Java for a _long_ time (since the Microsoft JDK on 1.1; my first work desktop was a Pentium II, I think a 400mhz one) so having to refresh my memory about OSGi isn't the worst thing.

 

@mr_handy
WRT the bottom panel, are the tabs more liable to break in the new generation chassis (5480 and upwards)? Meaning, are they using that hook and tab system (which always results in broken hooks for me) or are they using something more akin to the older style (E5450, E5470 I assume) tabs?

 

I haven't opened any of the 5x80/7x80 machines to compare to; I don't think I broke any tabs as it still fits well but I can't be 100% sure. It is a much more curved bottom cover than the flat bottom plate on my 5470.

 

We finally get to see the juicy innards of the 5491 with dGPU. Mainboard is indeed different, with the relocated southbridge occluding the SATA connection (which explains why the dGPU forces one into a M.2 SSD).

https://www.notebookcheck.net/Dell-Latitude-5491-8750H-MX130-Touchscreen-Laptop-Review.315432.0.html

It doesn't have any "better" of a cooling solution, unlike the Precision 35x0 line that is based on the H series quad core Latitude 5xxx line. The Precision gets dual heatpipes. In their review model, the area near the battery charger circuitry gets really hot under full system load.

 

Further undervolting testing (and keep in mind that the thresholds will vary between processors): -145 mV is stable, but still sees throttling on all 12 threads, and -155mV crashed.

 

Looks like NBR also did another unannounced review, this time of the 5591.

https://www.notebookcheck.net/Dell-Latitude-5591-8750H-MX130-Touchscreen-Laptop-Review.319956.0.html

In this case, it definitely gets the dual heatpipe arrangement of the Precision 3520, but it doesn't do much for heat, noise, and performance vs the 5491. It also has the MX130 while the nearly identical Precision 3530 gets a P600 (MX150).

However, it does do something I wish the 54xx line would do: when no 2.5" drive is present, a larger battery is able to be utilized.

 

This is really a general Linux point, not specific to the 5491 but I figured out one cause of my throttling: undervolting (and changes to the RAPL power limits) get lost when the machine sleeps.

I copied the following from /etc/local.d/local.start to /etc/pm/sleep.d/wake.sh and now it re-sets the undervolting and wattage limits after waking up.

Code:

#!/bin/bash
/usr/local/undervolt/undervolt.py --core -145 --cache -145 --analogio -90 --uncore -145 --gpu -90
echo 41000000 > /sys/devices/virtual/powercap/intel-rapl/intel-rapl:0/constraint_0_power_limit_uw
echo 49000000 > /sys/devices/virtual/powercap/intel-rapl/intel-rapl:0/constraint_1_power_limit_uw