[Discuss] [POSSIBLE SPAM] Re: [POSSIBLE SPAM] Re: I need a recommendation for a Linux-compatible high-end PC builder

[Deryk Barker (Honorary Colleague)]

From: Discuss [discuss-bounces at vlug.org] on behalf of Alan W. Irwin [irwin at beluga.phys.uvic.ca]
Sent: Tuesday, January 09, 2018 7:07 PM
To: discuss at vlug.org
Subject: [POSSIBLE SPAM]  Re: [Discuss] [POSSIBLE SPAM] Re: I need a recommendation for a Linux-compatible high-end PC builder

>Thanks, Deryk and Murray for your responses to my question.  More
>below in context.

Ditto

On 2018-01-09 17:20-0000 Deryk Barker (Honorary Colleague) wrote:
...
> Aside from one of the two DIMMs being faulty everything "just
worked", booted Mint from a USB stick and installed it in about 10
minutes.

>That's a great Linux compatibility result.  What did you do about
>Linux compatibility?  Did newegg.ca guarantee that, did you do some
>due diligence by checking out each component for their Linux
>compatibility, or did you cross your fingers and hope for the best?

No guarantee from NewEgg, but Intel specifically mentions linux in their web description of the Skull Canyon (still can't get used to that silly name), comments on various vendor sites (newegg, amazon etc) from people running various distros. Mint wasn't specifically mentioned, but I figured WTH.

The memory and SSD were both on the Intel SK compatibility list.

I've used this strategy before: last year (gosh, no, 2016 is now the year BEFORE last) we ordered a Zotac SFF machine also from newegg and I did similar compatibility checks online Once again no problems whatsoever getting Mint installed and running.

Before that, especially with laptops, I'd take a live distro on a USB stick and ask to boot it. Oddly enough I never got a "no", even at Future Shop. 

@Murray: same questions about Linux compatibility strategy (with Mother Computers
in your case) to you....

I'll mention that my (and my wife's) previous machines both came from Mother. In my case I ordered a MB that they assured me at least one other customer was running linux on. NO problems with that and for my wife a couple of months later, we simply got a slightly cut-down version of the same machine. (Her old machine still runs, BTW, but has been reduced to a DVD-ripping station.) 
 
> This is to replace my 4-year old machine, a quad-core i5 with 16GB,
which started to get hardware-related errors just before Xmas.

>Well, I do get the sense that my still-working 10-yr-old box is a bit
of an anomaly, and your experience (failure after 4 years) is more
usual.  Nevertheless, I still hope my next box will last at least
another 10 years simply so I can put off doing this exercise again as
long as possible and especially to reduce electronic trash.

I suspect 10 years probably is anomalous,  but if it keeps working and does the job...

> BTW the machine is a SFF and about the size of a paperback book.

Does that small form factor mean you cannot use an optical drive or do
they somehow squeeze it in?

No, you have to use an external optical drive. And they do tend to be a bit rubbish these days (I can't recall the last time I saw one for sale which did NOT take its power from the USB port). 

I still use my 10 (12?) year-old external USB drive, which is far more reliable at ripping CDs than anything we've had since.

However, for ripping DVDs the drive in my wife's old machine is better than that and also than the - supposedly identical - drive in my old machine.

>Of course, the really good news about the SFF that you bought is it
>virtually guarantees that your i7 doesn't consume a lot of power
>(small size would compromise the cooling that would be necessary if
>the i7 had high power consumption) so that will likely mean you save
>some small but still noticeable bucks on the power cost, and your
>computer likely is very quiet (due to the lack of need for hyper cooling).

45w max apparently. And, although it does have a fan (which my wife's Zotac and my Fit-pc Mintbox, (which is probably 5-6 years old, runs my disk and print server and also runs a web site for the Lakehill Lawn Bowling Club in a VMware VM running debian) do not. They both use even less power, but then her machine has an i3 and my Mintbox (I think) an i5.

>Consideration of the quality of my decade old hardware compared to
>what I can buy now leads to the next much more general question for
>you since you presumably have a lot of expertise in computers.  What
>do you and your computer science colleagues think of Moore's law these
>days?  Is it permanently over?

>In the decade after the pentium-133 (my first computer) was introduced
>in the mid 90's, cpu clock rates seemed to follow that law, but then
>they appeared to hit a speed barrier, and clock rates haven't changed
>that much since, e.g., my 10-yr-old computer has comparable clock
>rates to what you can buy today.  Of course, the computer hardware
>manufacturers are still making big strides in memory speed; memory,
>HDD, and SDD size; power consumption; and number of cores.
>Neverthess, the industry still seems to be in a bit of a holding
>pattern right now in terms of the overall quality of PC's.  Is
>everyone waiting for the introduction of quantum computers to start
>achieving enormous speed increases again?

Well, in the three years (and a half!) since I retired I can't claim to have kept up: I couldn't, for example tell you the difference between the microarchitecture of a 6th and 7th generation i7. Whereas once I could have gone into tedious detail.

Moore's Law was never really about clock rates, that was a by product, if you like. What Moore originally said, back in 1965, was that the packing density of transistors in chips would roughly double every 18 months.

The increase in memory capacities etc was largely a consequence of this, and I suspect there are physical limitations on die size, but then I've been reading predictions about the imminent demise of Moore's Law since the early 1990s at least.

But memory densities to keep increasing and the price dropping. The 32GB I bought with the new system cost $492. When I bought by 386, in March 1990, RAM was $200/MB (I used to enjoy making my students gasp at this), so my massive 4MB cost me $800.

Which means that the memory on my new system, aside from being considerably faster, cost less by a factor of  about 13,000 (which does NOT represent a halving every 18 months, but still...)

Clock speeds, OTOH, have not increased, as you observe, in over a decade.

There is a simple reason for this: heat. As the clock speed increases the amount of heat produced increases, but not linearly: the faster the clock the more of the power consumed goes to generating heat.

Couple this with the increasing complexity of chip design and all of the things needed to make them as fast as possible: pipelines, out-of-order execution, speculative execution, etc all of which require controlling and coordiinating, which was also starting to consume an increasing amount of chip real estate.

The two main responses to this were a) multi-core chips and b) Very Long Instruction Word machines (the Intel Itanic, sorry Itanium, is, despite their claims to the contrary, an example).

Both require significant parallelism to provide speed ups, one difference being that VLIW machines require the compilers to work hard to make the parallelism explicit, so that the logic for doing this is not down to the CPU (scoreboarding and register-renaming anyone?). Which may be one reason why they haven't caught on.

As for Quantum Computing. I first heard a talk about this at a conference in 1987. I'm not holding my breath; and while QC may well speed up some applications mightily, it ain't gonna run your browser or word processor any faster.

deryk
-------
Deryk Barker, Faculty Emeritus
Computer Science Department
Camosun College, Victoria BC