Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 11 Feb 2025 23:45:59 -0800
From:      Mark Millard <marklmi@yahoo.com>
To:        Poul-Henning Kamp <phk@phk.freebsd.dk>, freebsd-hackers <freebsd-hackers@freebsd.org>
Subject:   RE: First impressions: Lenovo T14s with Qualcomm Snapdragon ARM64 CPU [a "Worldstone" M4 MAX comparison]
Message-ID:  <83C23D68-8C78-40E8-B6DA-DE5B4ADAB515@yahoo.com>
References:  <83C23D68-8C78-40E8-B6DA-DE5B4ADAB515.ref@yahoo.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Poul-Henning Kamp <phk_at_phk.freebsd.dk> wrote on
Date: Fri, 15 Nov 2024 20:52:03 UTC :

> Those of you how know me, know that I am not a big fan of the X86
> architecture, which I think is a bad mess, mangled by market power
> considerations, rather than the CPU architecture this world actually
> needs, in particular in terms of performance/energy ratio.
> 
> So when my regular HW-pusher had a T14s G6 with Qualcomm Snapdragon
> ARM64 CPU, for only EUR1000 + VAT, and I couldn't resist.
> 
> . . .
> 
> Worldstone:
> -----------
> 
> >>> World built in 3210 seconds, ncpu: 12, make -j12

What follows is an FYI based on some testing that I
did. I do not claim that it makes for a well controlled
comparison/contrast to the details of what contributed
to your Worldstone figure. But I've tried to be fairly
explicit about various points that might contribute to
distinctions.

I just did an experiment on a MacBook Pro M4 MAX with
128 GiByte of RAM that was running FreeBSD under
Parallels, with 14 cores assigned (of 16) and 104 GiByte
of RAM assigned (of the 128 GiBytes). The media is the
same as I use on the Windows Dev Kit 2023 and RPi5: a
USB3ish adapter for U.2, with Optane U.2 media. macOS
looks to treat that as USB 3.1 Gen 2 (10Mbps) context.

The buildworld's are from scratch (rm -fr used first)
but of the same source as for what the system booted.
(So no bootstrap toolchain was built.) The system doing
the build is running a non-debug build. (more details
later.)

Non-debug targeted build:

World build completed on Tue Feb 11 22:22:35 PST 2025
World built in 724 seconds, ncpu: 14, make -j21

Debug targeted build:

World build completed on Tue Feb 11 22:47:04 PST 2025
World built in 747 seconds, ncpu: 14, make -j21

The system it was running had been set up to target
Cortex-A76 via -mcpu= use and such and that is also
what was built.

Also, ssh over wifi was in use to get into FreeBSD
and operate it: No serial console slowdowns involved.

I used the Amphetamine application to be sure that
macOS would not have the system sleep during the
Parallels run. The system had external power and
was set up with "On power adapter" assigned "High
Power" status.

In other words, things were set up biased to save
time rather than to save power.

As for the kernel's from scratch builds . . .

Non-debug targeted build:

Kernel build for GENERIC-NODBG-CA76 completed on Tue Feb 11 22:23:33 PST 2025
Kernel(s)  GENERIC-NODBG-CA76 built in 58 seconds, ncpu: 14, make -j21

Debug targeted build:

Kernel build for GENERIC-DBG-CA76 completed on Tue Feb 11 22:48:00 PST 2025
Kernel(s)  GENERIC-DBG-CA76 built in 56 seconds, ncpu: 14, make -j21

> ... using 24 Wh of energy, measured with a proper power-meter.

I do not have a measurement of that, unfortunately.


===
Mark Millard
marklmi at yahoo.com




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?83C23D68-8C78-40E8-B6DA-DE5B4ADAB515>