Re: [Pan-users] what I need to do to "jump ship" (Re: To explain why I'm unwilling to update my g lib/gtk+/etc libs ATM…)

[Duncan]

SciFi posted on Thu, 28 Jul 2011 06:03:11 +0000 as excerpted:

> I am recalling a few details I've had in my long history when dealing
> with other people's PC machines especially the owner/friend of a house I
> was renting (he had a brain aneurysm and passed away several years ago).
> His PC was full of lil interface chips made mostly by VIA.
> Some chips were from other makers.
> Usually a BIOS upgrade will take care of all those chips as well but I
> found many lil firmware-upgrade pgms on the vendor's websites depending
> on the part-numbers on those chips.
> (That's why I say EVERYTHING has firmware in it these days.)
> I had to take-apart his machine to inspect them by eyeball (verify
> part-numbers etc).
> And sure-'nuf most of his chips were quite back-level.
> What's more, the descriptions of the upgrades were mostly spot-on with
> the problems he was having.
> So the "fixes" were not in win-xp (at the time)
> nor in the BIOS upgrades but in the chips themselves.
> I had logged every single itty-bitty thing I did,
> esp'ly how to dig-thru the vendor's websites etc,
> but of course nowadays those notes are all lost.
> I couldn't explain to anyone just how much work I have LOST since no-one
> would deal with the situations I was in (medically etc).
> 
> Anyway, for my current project of "jumping off Apple's ship",
> I need that level detail on whatever hardware I'm to be buying because I
> do not want to be in that kind of situation.

It's not /that/ level of detail, but FWIW...

I've been extremely happy with the tyan s2885 mobo I bought back in 2003, 
which is still going today.  I'm shooting for a decade! =:^)  Dual-socket 
Opteron, certified with several Linux distros of the era, and with actual 
Linux support (I emailed them about a couple things).  They even supplied 
a sensors.conf file for the board!

And what greatly impressed me was that while a lot of vendors forced 
users to by new hardware to get BIOS upgrades for the dual-cores when 
they came out, Tyan supplied the necessary BIOS updates for many of their 
systems! =:^)

The manual (and web-site info, pre-purchase) came with the model number 
of all of the major onboard chips and subsystems, too.

Not all Tyan's products are so well Linux supported but I believe some of 
them still are, and finding that sort of info is FAR easier than taking 
the cover off a machine pre-purchase to look at not only the supplied 
chip models (as Tyan listed on-site and in-manual), but apparently, 
stepping-numbers and/or production serials, as well (NOT supplied on-site 
in-manual, but from your description the level you were going to).

That board was an AMD primary chipset, pre-ATI-purchase.  I'm not sure 
how they work that now (post-ATI-merger), but back then, AMD would come 
out with the first chipsets supporting a major new CPU family, and about 
two years later, ATI/nVidia/Via would follow.  AMD's chipsets never had 
the absolute latest features (I only got onboard USB-1, for instance, 
when USB-2 was out and shipping on some non-Opteron boards already, but 
the only chipset for Opterons at the time was AMD's), but the *WERE* 
solidly supported by all the OSs and widely available in boards from 
various manufacturers, because as I said, that tended to be the only way 
to support new AMD CPUs at the time.

But regardless of GENERAL chipset support, again, not all board vendors 
bothered providing dual-core BIOS updates.  Tyan did, for many (tho not 
all) of their Opteron products.

Of course, part of that higher level of support comes from the grade of 
board it was.  It was an over $400 board, even shopping thru 
pricewatch.com.  But somewhat more generically, it was a server-class 
board (dual-sockets do tend to be so), coming at a server-class price, 
but also with server-class support!  Still, compared to the competition, 
Tyan was far and away the best choice and best Linux support out there, 
even at the $400-$500 board quality level.

(This was BEFORE MS had a full release amd64/x86_64 OS out there, tho 
they had betas.  Despite that fact, of the four boards that fit my 
general requirements that I researched, one vendor had no web presence at 
all that I or google could find, a second had some presence but IIRC it 
was Chinese-only, and the first runner-up, MSI, had an English web site 
but all their board docs, BIOS updates, etc, were in MS executable 
format, probably self-extracting-exe, but I didn't bother checking!  I 
scratched around on the MSI site and wrote them an email, telling them 
exactly why they lost my purchase, that MS didn't even have an OS 
released yet for the platform, yet all their files were MS executables.  
Not only that, how were even MS users to know that a cracker hadn't 
gotten access and replaced all the pdfs, zip files, etc, with 
executables!  Whether my individual mail made a difference or not I don't 
know, but a few months later, the MSI site was back to standard PDFs, zip-
files, etc, the only executables to be seen were MS platform hardware-
bundled-software, so SOMETHING made a difference!  Meanwhile, I wrote to 
Tyan explaining why they got my purchase, as well. =:^)

FWIW, if you're in the market for something closer to a full system, Tyan 
sells those, too.  In fact, I believe they're really a full-system vendor 
that designs and makes available their own server-class mobos too, as a 
sideline since they're already designing them for their own systems.

Meanwhile, Intel /generally/ has very good Linux support, at least.  They 
have a number of mainline Linux kernel developers on their payroll, and 
they apparently use Linux for their early development and hardware-bring-
ups for stuff like USB-3, too.  As such, the mainline Linux kernel had 
USB-3 support shipping before any other platform and before hardware was 
publicly shipping as well (altho they were making limited engineering 
samples available to the big players, MS, Dell, IBM, HP, Apple, Lenovo, 
etc, by the time mainline support was complete, but preliminary support 
was in even before that!).

I found out just how good Intel's hardware, acpi, etc, integration could 
be, when I bought my netbook.  Generation 1.5 (or arguably 1.8), I got 
one of the original Acer Aspire One machines.  This was before MS got 
their grubby hands on things and ruined the whole netbook scene, but 
after netbooks had started shipping with reasonable sized traditional 
drives (my requirements included a standard SATA connection and at least 
100 gig drive capacity, this was one of the first models that had both; 
it also had a "just right for me" 9" display, larger than the early 7" 
models but still far more portable than the 10-12" models that soon were 
being called netbooks, too).  It was before they had a purpose-designed 
netbook chipset, so the chipset still used more power than the later ones 
did, but by the same token, it missed out on Intel's most significant 
recent missed-support in terms of Linux, the gma500 poulsbo graphics 
introduced with gen-2 netbooks.  The problem was that Intel outsourced 
the pinetop graphics from another vendor (I could lookup the name but 
it's not important info for this post) *WITHOUT* ensuring that they had 
sufficient contracted permissions to release open drivers or specs to 
develop them.  This was a HUGE mistake, but it's the exception that 
proves the rule, that Intel is generally very good working with Linux, 
both in specs and support, and in directly sponsoring kernel developers 
as well.

Anyway, MS was already encroaching and apparently had a contract with 
Acer not to release Linux-based AA1s in the US, so I had to order mine 
from Canada, since I was NOT going to buy MS -- I'd have rather skipped 
the purchase entirely.  And I'm glad I got it while I could, given that 
as I explained, six months or so later, pretty much all that was 
available was poulsbo based systems, which are still a Linux graphics 
problem today, altho the kernel 3.0 gma500 module is said to have good 2D 
support now, but still extremely buggy OpenGL/3D.  (However, none-other 
than the man who was for many years named as Linus' #2, Alan Cox, is the 
primary gma500 developer, I'm sure because after Intel realized the bind 
they put themselves in, they wanted the best possible to try to get them 
out of it!)

But get it with Linux I did, altho only Linpus Linux, a Fedora clone 
somewhat popular in Taiwan and China, but not a great choice for the 
western world.  Still, it was functional, and got me thru the bad-bios-
flash incident on my main machine (I had bad memory that corrupted the 
flash process).  But now it runs Gentoo, built in a 32-bit chroot on my 
main machine, the Tyan. =:^)

But nearly all (all that I'm aware of, with the gma500 exception, but I 
learned many years ago the wisdom of being wary of claims using words 
like "all", without qualification, since just one exception no matter how 
narrow then disproves the claim) the rest of Intel's recent technology is 
mainline Linux supported, straight out of the box.  Among other things, 
this includes the RAID hardware as found on some of Intel's chipsets.

Anyway, the point I intended to make about the netbook was that the 
chipset, etc, are all Intel, and 100% supported by mainline Linux, 
including all the acpi functionality (function keys for display 
brightness, toggling, volume control, etc, all "just work"), full speed-
step support, problem free suspend (to RAM) and hibernate (to disk), 
etc.  Until I bought it, all the Linux machines I'd used could generally 
have such features eventually made to work, but it always required custom 
configurations, etc, and that's one thing that the Tyan still doesn't 
always manage to recover from, hibernate (it doesn't suspend-to-RAM at 
all, presently, the problem with hibernate tho is slow to properly 
activate disks, which are still stabilizing when the system tries to 
restore so it looks to Linux like the disks got changed out over the 
hibernate and it kicks them out of the RAID!).

Meanwhile, Via is definitely notorious for the problems they have with 
Linux... or actually, in general, tho given the MS Windows-side black-
box, generally the only way people know there's a problem there is when 
they compare performance to decent hardware -- it's only on Linux (and 
probably the BSDs that run on the hardware, if they run on it at all) 
that it's obvious the hardware is limping alone, IO failure's and retries 
or operating several notches down from the rated DMA speed or in PIO 
instead of DMA mode, etc.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman