[Discuss] I need a recommendation for a Linux-compatible high-end PC builder

[Alan W. Irwin]

On 2018-01-28 12:47-0800 Adam Parkin wrote:

>>> Typically every new system has a speed gain compared to older systems
>>> because of faster cpu's, faster memory, faster HDD throughput, etc. So
>>> did you actually measure that SSD gave you any advantage by copying
>>> your system from SSD to HDD and comparing speeds between HDD alone and
>>> SSD + HDD for the same system and same hardware?
>
> For my main home laptop (which I'm writing this email on) it's had a
> Samsung Evo SSD in it for about 2 years now as the only drive.  I've had
> zero issues with it, and the performance difference was night and day going
> from the original drive.  Originally it had a spinning disk in it (TBF, I
> *think* it was only a 5400 RPM drive, so not the best of spinning disk
> technology), but in terms of "single change upgrades" going from the
> spinning disk to SSD on this thing is probably the biggest single
> improvement I've seen in a machine I've owned (not just load times, but
> general responsiveness).  Before the SSD I was ready to buy a new machine
> because it was so slow, and after the SSD I've been completely happy with
> the machine and it's the only thing I've replaced/changed on it.
>
> Full disclosure: it's not a Linux-based system (Macbook), and is an
> "average", day-to-day use laptop (so not "dream" machine by any stretch,
> only 4GB of RAM & an I5 processor).

Hi Adam:

That appears to be a clear argument that my mental model of how RAM is
used to mitigate drive use is not applicable to Mac OS X for some
unknown reason, but I am still pretty sure that mental model is
correct for the Linux case.

To explore that Linux world a bit more here are the numbers I get
from the "free" app for my own 4GB system:

              total       used       free     shared    buffers cached
Mem:       4049020    3451228     597792      75696     384756 1134688
-/+ buffers/cache:    1931784    2117236
Swap:     15625212    1378960   14246252

The sum of "buffers" and "cached" is the RAM used for file needs and
is included in the "used" value in the first row, and the "used"
values in the second and third rows indicate how address space needs
are shared between RAM (second row) and drive (third row).
(To learn more about interpreting "free" results see
<https://serverfault.com/questions/23433/in-linux-what-is-the-difference-between-buffers-and-cache-reported-by-the-f>.)

So from the above snapshot of memory use on my system it appears to me
the kernel is making pretty good judgements about which fraction of
physical RAM will be used for address space needs and which fraction
will be used for file needs, and if I increased memory by ~
1.4GB there would be no need of swap space (the drive area used to
store little-used address space needs) at all, and any increase in RAM
over that number (e.g., the 64GB I am planning to buy for my new
system) would add directly to the file space needs that could be
stored in RAM.

The above result shows there are no noticeable memory leaks and no
clear evidence of memory starvation for my current system.  But also
there is nothing I have seen so far that contradicts the general
conclusion that adding RAM is going to make the SSH versus HDD
performance issues less and less relevant at least in the Linux case,
and if you accept that general principle, then it is a matter of
argument whether 64GB will completely remove (aside from first use
since boot) any practical advantages of SSD's compared to HDD's.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________