Lennart, thanks for reply. My remarks below. Steve ----- Original Message ----- From: Lennart Sorensen To: Giles Orr ; GTALUG Talk Cc: Steve Petrie, P.Eng. Sent: Tuesday, March 27, 2018 12:27 PM Subject: Re: [GTALUG] New Deaktop PC -- To Run debian Linux - PCPartPicker Recipe; On Tue, Mar 27, 2018 at 12:14:57PM -0400, Giles Orr via talk wrote:

It's probably too late to convince you of this (as I suspect it would

Well I discovered on my laptop a few years ago that suspend to disk takes a very long time when you put 24GB ram in a laptop and use a 5400rpm disk. An SSD makes it much better of course. *** Nice tip. I'll be sure to use the SSD not only for fast software loading, but also for suspend workspace. -- Len Sorensen

Lennart Sorensen via talk wrote:

100MB/s disk speed (if lucky) and 24000MB ram = at least 240 seconds to write ram to disk.

Resume was equally painful. It was faster to shutdown and boot normally and open my programs again.

That'd be a totally brute-force solution. If the kernel pushed all dirty pages into swap then there should only be the page tables and other core stuff needing to go out, sleep, and be read back in before it could start paging the rest back as needed. Being a tad more aggressive about page-cleaning during normal runtime might even help it suspend faster. I'd thought there was kernel suspend support already, in fact, not just hardware/firmware versions of it, but lack of a nice laptop and being the sort of person who crypts my swap partition mean I've never had the occasion to try it. -- Anthony de Boer

Giles, thanks for reply. My remarks, questions below. Steve ----- Original Message ----- From: Giles Orr To: Steve Petrie, P.Eng. ; GTALUG Talk Sent: Tuesday, March 27, 2018 12:14 PM Subject: Re: [GTALUG] New Deaktop PC -- To Run debian Linux - PCPartPicker Recipe; On 27 March 2018 at 09:08, Steve Petrie, P.Eng. via talk <talk@gtalug.org> wrote: Greetings To GTALUG, At long last I have created a final build recipe for my new desktop PC to run debian Linux. Replacing an ancient Dell PC running WIndows XP: a.. https://ca.pcpartpicker.com/user/Steve_Petrie/saved/#view=krFNNG The configuration is an Asus LGA1151 socket motherboard with Intel i5-6500 CPU and 16 GB DDR4 memory.. A dialup modem remains to be added: US Robotics USR5637 V.92 USB Faxmodem Comments welcome. * * * * * * My plan is to be ordering the parts by Friday 30 March or or Monday 2 April at the latest. The HP PS/2 keyboard and HP PS/2 mouse are already ordered (from PC-Canada) but that order is still on hold, pending addition of items to the order.. It's probably too late to convince you of this (as I suspect it would strongly imply a move to a newer and more expensive processor) but getting USB-C and USB3.1 on the motherboard would be good. You can _probably_ get a PCI-E card later, but I wonder if it will fully support USB-C's relatively complex alt-modes and etc. Of course, it may not support them on the motherboard either, but I think it's worth a try. ***Good points. I will look into a possible motherboard upgrade to add USB-C and USB3.1. I'd also encourage you - fairly strongly - to not get PS/2 accessories. You're already encountering one of the problems, trying to plug two devices into one port - I suspect that won't go well for you. Not only is USB much more versatile, it's much easier to "split," and add more ports. *** *** I know that PS/2 is an ancient interface. I only chose PS/2 because it is my understanding that a PS/2 keyboard and mouse combination is better supported and earlier, during the actual Linux boot process, than a USB keyboard and mouse combination. So PS/2 keyboard and mouse offer more liklihood of functionality when there are Linux boot problems. *** I also understand that USB keyboards and mice, take far more and more complex software drivers to be loaded before keyboard and mouse functionality appears. So a non-bootong Linux may not offer a functional USB keyboard / mouse service, whereas a PS/2 cpmbination might already be alive and functional. *** If I'm wrong on this PS/2 versus USB keyboard / mouse boot-friendliness business, I will happily switch to using USB keyboard / mouse. And finally, I'm in awe of your memory strategy. But I totally agree: max out every slot - when you can afford it. That's the way to go. And 16G is already a decent baseline. *** Confession: Actually, the choice of a single 16 GB memory DIMM came about at the end of an arduous and desperate search for ANY memory available in Canada at reasonable price, that is listed on the Asus motherboard's memory QVL (qualified vendor list). *** But yes, I'm not unhappy to be keeping the full 64GB address space open. *** Is it just a superstition that I read somewhere, that there is there a memory access speed advantage, to using more DIMMs to get the desired total memory size ?? For example, that a dual-DIMM 2x8GB configuration provides faster memory access than a single DIMM 1x16GB ?? -- Giles https://www.gilesorr.com/ gilesorr@gmail.com

On Wed, Mar 28, 2018 at 02:47:26PM +0000, Ivan Avery Frey via talk wrote:

Some motherboards would interleave their memory accesses. I don't know if that's true today.

I don't think anyone makes one today that doesn't. -- Len Sorensen

On Wed, Mar 28, 2018 at 09:50:42AM -0400, Steve Petrie, P.Eng. via talk wrote:

*** I know that PS/2 is an ancient interface. I only chose PS/2 because it is my understanding that a PS/2 keyboard and mouse combination is better supported and earlier, during the actual Linux boot process, than a USB keyboard and mouse combination. So PS/2 keyboard and mouse offer more liklihood of functionality when there are Linux boot problems.

The BIOS emulates PS/2 for USB mouse/keyboard just fine and should not cause any issues. But yes it is slightly more complex to use a USB device. I don't remember the last time I dealt with a PS/2 device. I don't miss them. I don't think I have had a problem with a USB keyboard for a decade. USB mice are never a problem since you never need a mouse unless the system is actually booted normally, so only for a keyboard is PS/2 at all interesting. Well unless you are one of those extreme gamers that care about the potentially lower latency of PS/2 mice over USB mice.

*** Confession: Actually, the choice of a single 16 GB memory DIMM came about at the end of an arduous and desperate search for ANY memory available in Canada at reasonable price, that is listed on the Asus motherboard's memory QVL (qualified vendor list).

*** But yes, I'm not unhappy to be keeping the full 64GB address space open.

*** Is it just a superstition that I read somewhere, that there is there a memory access speed advantage, to using more DIMMs to get the desired total memory size ?? For example, that a dual-DIMM 2x8GB configuration provides faster memory access than a single DIMM 1x16GB ??

It is much faster because all CPUs today have at least two memory channels (some have 6 now). Having equal memory on both channels doubles your memory bandwidth. Now for some uses memory bandwidth isn't that important, for other uses it is. -- Len Sorensen

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hugh wrote:

UEFI firmware usually (but not always) includes a CSM (Compatibility Support Module) to emulate BIOS for the sake of old OSes. If present in the firmware, it usually can be enabled or disabled in the firmware config menus. As far as I know, all stand-alone motherboards have a CSM (Win 10 tablets often do not).

I recently repaired an Asus laptop in which the firmware had no setting for compatibility. I was alarmed, but it worked fine with Ubuntu 17.10, although not the orginal Win10 on the hard drive that came with it. It always booted to the firmware config screen, and wouldn't load the OS (Win10); wouldn't load the .EFI file from the system32 folder. I tried to swap in an old drive for Win7, and Windows booted but refused to behave properly, as I suspected it might -- WiFi drivers didn't exist in Win7 for that model laptop. So I swapped the original non-booting drive back, and mysteriously everything worked again... *Now* the .EFI files loaded, and the laptop was back to normal. I don't much like the new, low-cost hardware. I've also had other problems on an ultra-cheap desktop with minimal firmware settings. Wouldn't dual-boot Ubuntu (probably 16.04) unless I switched it to Compatibility Mode, but then it wouldn't boot Windows. Had to leave detailed instructions for the client on switching in and out of Compatibility Mode... I welcome further off-topic insights, although the list moderators may not be so tolerant. - --Bob. On 2018-03-28 03:04 PM, D. Hugh Redelmeier via talk wrote:

| From: Lennart Sorensen via talk <talk@gtalug.org>

| On Wed, Mar 28, 2018 at 09:50:42AM -0400, Steve Petrie, P.Eng. via talk wrote: | > *** I know that PS/2 is an ancient interface. I only chose PS/2 because it is my understanding that a PS/2 keyboard and mouse combination is better supported and earlier, during the actual Linux boot process, than a USB keyboard and mouse combination. So PS/2 keyboard and mouse offer more liklihood of functionality when there are Linux boot problems. | | The BIOS emulates PS/2 for USB mouse/keyboard just fine and should | not cause any issues. But yes it is slightly more complex to use a | USB device. I don't remember the last time I dealt with a PS/2 device. | I don't miss them. I don't think I have had a problem with a USB keyboard | for a decade. USB mice are never a problem since you never need a mouse | unless the system is actually booted normally, so only for a keyboard is | PS/2 at all interesting. Well unless you are one of those extreme gamers | that care about the potentially lower latency of PS/2 mice over USB mice.

(I may have some of this wrong. These details are considered too complicated to explain to users so we are just supposed to assume everything just works.)

- firmware isn't BIOS these days. In fact, that was a misnomer on the PC since day 1. It's a carry over from the CP/M days on Intel 8080 processors. Now it is called, uhh, firmware. Or UEFI firmware.

- UEFI firmware usually (but not always) includes a CSM (Compatibility Support Module) to emulate BIOS for the sake of old OSes. If present in the firmware, it usually can be enabled or disabled in the firmware config menus. As far as I know, all stand-alone motherboards have a CSM (Win 10 tablets often do not).

- System Management Mode is a little known feature of all Intel and AMD x86 processors since the 486 and the Am486 25 years ago. It allows the firmware to keep some control after booting the OS. The OS has no control over this.

- SMM is used (among other things) to capture USB keyboard and mouse events and turn them into simulated PS/2 events. This is usually optional but default behaviour on motherboards. In effect, this emulation appears to be a hardware feature.

- One other limitation of USB and not PS/2 is that there is a limit to how many keys can be pressed at once (n-key rollover) and sensed properly. The limit is based on how many key events can be recorded in a single USB packet. But the limit is something like 6. Surely more than enough.

So: mouse emulation and BIOS emulation are behind the back of the OS. This means legacy OSes should have no trouble with these features.

I recommend using a USB keyboard and mouse. Unless you are a crazy gamer. --- Talk Mailing List talk@gtalug.org https://gtalug.org/mailman/listinfo/talk

- -- Bob Jonkman <bjonkman@sobac.com> Phone: +1-519-635-9413 SOBAC Microcomputer Services http://sobac.com/sobac/ Software --- Office & Business Automation --- Consulting GnuPG Fngrprnt:04F7 742B 8F54 C40A E115 26C2 B912 89B0 D2CC E5EA -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 Comment: Ensure confidentiality, authenticity, non-repudiability iEYEARECAAYFAlq8AMIACgkQuRKJsNLM5erSYQCgx0xa7BWK8puxukaSfEIeyowv VioAoOVidXAuYha15uCokaVkZkINtbsA =3L7N -----END PGP SIGNATURE-----

D. Hugh Redelmeier via talk wrote:

I recommend using a USB keyboard and mouse. Unless you are a crazy gamer.

Or have a real keyboard from 1989 still going strong. -- Anthony de Boer

Alex, thanks for reply. My remarks below. Steve ----- Original Message ----- From: Alex Volkov via talk To: Steve Petrie, P.Eng. via talk Sent: Tuesday, March 27, 2018 12:23 PM Subject: Re: [GTALUG] New Deaktop PC -- To Run debian Linux - PCPartPicker Recipe; If you're willing to settle for a slightly smaller screen there's LG 23" IPS screen on sale for $150 https://www.newegg.ca/Product/Product.aspx?Item=N82E16824025346 *** Thx for the tip. If you're not planning on adding a video card, you can save some money by getting a case with 300W power supply. *** Good point. Could definitely save some money here. Agree that the power supply is over-specified. But this intentional, being in keeping with the 24_7 service duty ethos for the new PC. And I do want the expandability readily available with this larger power supply. *** According to the PCPartPicker website, the current build recipe needs a maximum of around 170 W. So my reasonong is to allow for expansion bringing a doubling of the maximum power load, and still leave the power supply operating at only 66 percent of its maximum output capacity. *** Also I understand that there can be a significant dropoff in quality and reliability with power supplies of maximum output less than around 550 W. Alex. On 2018-03-27 09:08 AM, Steve Petrie, P.Eng. via talk wrote: Greetings To GTALUG, At long last I have created a final build recipe for my new desktop PC to run debian Linux. Replacing an ancient Dell PC running WIndows XP: a.. https://ca.pcpartpicker.com/user/Steve_Petrie/saved/#view=krFNNG The configuration is an Asus LGA1151 socket motherboard with Intel i5-6500 CPU and 16 GB DDR4 memory.. A dialup modem remains to be added: US Robotics USR5637 V.92 USB Faxmodem Comments welcome. * * * * * * My plan is to be ordering the parts by Friday 30 March or or Monday 2 April at the latest. The HP PS/2 keyboard and HP PS/2 mouse are already ordered (from PC-Canada) but that order is still on hold, pending addition of items to the order.. Steve --- Talk Mailing List talk@gtalug.org https://gtalug.org/mailman/listinfo/talk ------------------------------------------------------------------------------ --- Talk Mailing List talk@gtalug.org https://gtalug.org/mailman/listinfo/talk

On Wed, Mar 28, 2018 at 2:18 PM, D. Hugh Redelmeier via talk < talk@gtalug.org> wrote:

| From: "Steve Petrie, P.Eng. via talk" <talk@gtalug.org>

| At long last I have created a final build recipe for my new desktop PC | to run debian Linux. Replacing an ancient Dell PC running WIndows XP:

The i3-8100 is faster, cheaper, newer and has better graphics hardware than the i5-6500. Why not use it? Or an even faster "8th gen" processor?

Perhaps the motherboards are more expensive. I haven't looked. But they are likely to have a few more modern features.

Not the OP but - - - - one issue I'm seeing - - - my main system (its over 6 years old) has 8 slots for ram. Almost none of the newer mobos have that many. (Just one point for where newer isn't always 'better' (whatever that means! - - grin).) Dee

On Wed, Mar 28, 2018 at 5:35 PM, D. Hugh Redelmeier via talk < talk@gtalug.org> wrote:

| From: o1bigtenor via talk <talk@gtalug.org>

| Not the OP but - - - - one issue I'm seeing - - - my main system (its over | 6 years old) | has 8 slots for ram. Almost none of the newer mobos have that many. | (Just one point for where newer isn't always 'better' (whatever that means! | - - grin).)

My Sun 3/60 has 24 slots. But the largest memory module that it will accept is 1MiB (note: not 1GiB). So more isn't always better :-)

As memory systems get faster, fan-out and signal path length matter more. So 4 is apparently the limit for "unbuffered" memory on one PC memory bus these days.

If you have multiple processors on a motherboard, they probably have a separate bus for each processor so you get 4 slots per processor.

If you have a big server, you use buffered memory (probably with ECC too) and can have lots* of RAM sockets. With buffering, you lose speed but gain fan-out.

Complicating this is the market segmentation games that Intel is playing. Apparently you pay a lot for a Xeon processor that has enough address lines. AMD Epyc is smashing those limits so interesting things might happen.

This supports 16 LRDIMM DDR4 modules, each could be 32GiB (I think). 2TiB in total (that's 16 x 128 GiB, so I'm missing something). <http://b2b.gigabyte.com/Server-Motherboard/MZ31-AR0-rev-10>

In the good old days, bulk dynamic RAM chips were 1-bit wide so if your bus was 32-bits wide, you could drive 32 chips with no fan-out on the data lines. Now RAM chips are typically 8-bits wide, modules are 64-bit wide, and busses are 64-bit wide, so fan-out is a larger problem.

* "lots" is a technical term meaning more than four, but I don't know how many

OK - - - if I had been referring to server boards I wouldn't have commented but as I was talking about desktops - - - well - - - my point still stands. IIRC when I last looked (a couple months now) all the desktop boards were topping out at 4 ram slots. Dee

On Wed, Mar 28, 2018 at 03:59:44PM -0500, o1bigtenor via talk wrote:

Not the OP but - - - - one issue I'm seeing - - - my main system (its over 6 years old) has 8 slots for ram. Almost none of the newer mobos have that many. (Just one point for where newer isn't always 'better' (whatever that means! - - grin).)

Faster memory has made putting that many dimms per channel a problem. New servers have even dropped to two dimms per channel, while they used to allow 3 per channel (of course when they went from 4 to 6 channels at the same time, at least the number of slots stayed the same). When ram was slow, having 4 dimms on one channel was no problem. -- Len Sorensen