On 17 Apr 2005, at 20:16, Jonathan McDowell wrote:

On Sun, Apr 17, 2005 at 05:40:24PM +0100, Wayne Stallwood wrote:

...

While we are being all nostalgic, does anybody else remember the PPC640 laptop that ran off D cell batteries and had an (non backlit) mono LCD screen, I remember desperately wanting one of those when they came out.

First PC I ever owned; twin 720k 3.5" disc drives and an internal 2400 baud modem.

Mine was an Amstrad PC1640 with hercules graphics

On Sunday 17 April 2005 9:28 pm, Tim Green wrote:

IBM PC-XT with 5.25" floppy (low density of 360KB per disk) and a whopping 20MB hard disk. The screen was IBM CGA (not much better than a ZX Spectrum). In its later years I ran a hooky copy of Windows 3.0 without a mouse (excellent way to learn the keyboard short cuts!).

Mine was a strange Atari PC3, more or less 100% IBM PC compatible. Amber EGA display and twin 5 1/4" drives. 8088 (like an 8086 but with an 8 bit bus) and an external 3 1/2" drive on the same sort of bus they provided on Atari ST's (it was actually an Atari ST drive)

It came with GEM and some associated office applications, never did get windows to run on it properly.

First Unixy experience was BSD on an Amiga 1200 (heavily modified with 030 processor, SCSI CDROM and a (massive for the time) 32MB of RAM) I still have it in the loft and keep threatening to bring it to a meet.

On Wed, 2005-04-20 at 10:16 +0100, Adam Bower wrote:

On Sun, Apr 17, 2005 at 11:39:06PM +0100, Wayne Stallwood wrote:

...

First Unixy experience was BSD on an Amiga 1200 (heavily modified with 030 processor, SCSI CDROM and a (massive for the time) 32MB of RAM) I still have it in the loft and keep threatening to bring it to a meet.

Ooh, please do. Even though I recently sold all my Amiga stuff I'm still a great fan of the amiga. Unfortunately my Amiga 4000 only had an EC030 so couldn't run any type of Unix or Linux, BSD etc.

FWIW my first PC was a Pentium II 233, with 64 megs of Ram and a 4 gig disk which I purchased in 1997 so I could get online. The main reason I moved to Linux was because Windows was such a pile of crap compared to my experiences with the Amiga and Linux seemed like the best logical progression. Even now i'm not a fan of PC hardware at all, but then there doesn't seem to be any really "good" architectures around (at least, not that I can afford).

Agreed, I really wish there was a sensibly priced PPC architecture around, but I guess the sheer volumes involved in the x86 market keep the prices so low that anything else is going to seem expensive by comparison. I was reading a piece in the current PC Pro magazine about electricity usage and it started to focus my attention on just how wasteful the x86 architecture is in terms of power. If Intel and AMD would concentrate a bit more on getting the extra performance without the need for heavy duty cooling life would be a lot easier in terms of being cheaper to run, quieter, longer battery life in laptops, etc., etc.. With AMD being reasonably strong in the market, at least visibly, it can be easy to forget the term Wintel - Linux may well work on a lot of different architectures, but x86 is the one most used, and that is significantly as a result of it being the primary (and now only) platform for Windows. Is there nowhere its legacy doesn't infect?!

On the Unixy experience line, my first was AIX on an IBM RS6000 iirc. I got the responsibility of administering the boxes because I had an interest in Linux and had just bought a copy of Caldera OpenLinux 1.0. Massive beasts of machines they were, but even though they were ageing at the time an Exchange server setup was expected to need 4 brand new Pentium Pro 200 x86 servers to cope (not sure whether they were dual processor or not).

My first PC was an IBM L40 SX laptop, which I finally gave in and purchased on the IBM internal purchase scheme so I could run OS/2 - which was the only way I was going to use an x86 machine instead of my Amiga because I considered Windows a toy that had no place in serious usage - after all I was used to be able to put files into drawers (Amiga speak), and this poxy Windows Program Manager thing was merely a pretty launch pad for your applications and nothing more. This machine was directly responsible for getting me into Linux though as I spotted an internal IBM forum on getting Linux running on an L40 (there is still an option in the kernel compile specifically for the L40 I think). I followed that for a while, but didn't have an easy means of getting a copy - well, I probably did if I'd have investigated far enough and had the time to experiment a bit more, but my machine was already creaking at the seems with OS/2 and an enforced requirement of IBM DOS 5 and Windows 3.1 so it wasn't practical to try installing until about 4 years later when I got a spare machine - that was around 1996 ish.

On Wed, 2005-04-20 at 22:27 +0100, Wayne Stallwood wrote:

On Wednesday 20 April 2005 1:50 pm, Paul Tansom wrote:

...

I was reading a piece in the current PC Pro magazine about electricity usage and it started to focus my attention on just how wasteful the x86 architecture is in terms of power. If Intel and AMD would concentrate a bit more on getting the extra performance without the need for heavy duty cooling life would be a lot easier in terms of being cheaper to run, quieter, longer battery life in laptops, etc., etc..

It gets worse in climates where Air Conditioning is used in most office spaces. 100% of the energy a CPU uses is turned into heat output...heat which in turn has to be removed by running the Air Con harder and longer. Air conditioning is not 100% efficient so for every 50 Watts of power you pump into a CPU you waste another 50 Watts plus trying to cool the office.

The P4M is a nice little chip (well as nice as it gets with X86 it seems) you can get respectable performance with less than 25 Watts (for just the CPU mind) Compare that to the current revision of P4 desktop chips than can require more than 100 Watts and it's obvious where the office desktop market should be heading.

I have seen the suggested availability of P4M desktop systems (or at least the availability of the components required to build them) Two barriers are cost and a general public that (thanks to chip manufacturer marketing) is sold well and truly up the garden path of MHz myth.

Don't I know it. I have a single small room as my office (working for myself as I do). In it I have way too much kit for one person, although there is reason for all of it. In winter it is pleasantly warm, in summer it is way too hot. On one hand I could look into air conditioning to cool the room, but it would make more sense to reduce the amount of heat being generated. When you start thinking along these lines you get into the debate of costs vs savings balanced against practicality and benefit for the environment.

Looking at my setup and ignoring those machines that only get switched on when needed (Mac, Amiga, RiscPC, other old 8 bit machines) I have 4 servers and 3 desktops! One desktop is a family one and not in the office, so that can be ignored. Of the other two, one is my primary Linux box and the other is a Windows one for testing (I support Windows unfortunately). On the server side I have the firewall and a DMZ box hosting some mailing lists and websites (as backup for my external server. Inside the firewall there's the main mail server (with ldap, dns and a few other bits too) and a development server (web mainly, but also acts as file and print sharing to). The two internal boxes backup to each other. In technical IT practises this makes sense, but for one person the power usage is just plain silly. I've started to weigh up the pros and cons of consolidating onto a single box with UML or Xen systems to reduce power and heat!

A quick look into power supplies that have intelligent fans, for example, found one that looked good (listed as fanless which is incorrect, but it does spin down when not needed). That costs around £90, so how many years would it take to repay the cost in terms of power savings!

On another side, when you look at recycling old PC kit which is arguably less energy efficient, is it better to scrap it and use something that is cheaper to run. Then again the current range of CPUs are real power hogs (compare around 8W for a 486 to 120W for a top end P4).

Being environmentally friendly is hard work and a confusing area, however good that may be - and necessary in the long run.

...

With AMD being reasonably strong in the market, at least visibly, it can be easy to forget the term Wintel - Linux may well work on a lot of different architectures, but x86 is the one most used, and that is significantly as a result of it being the primary (and now only) platform for Windows. Is there nowhere its legacy doesn't infect?!

I loved Alpha, not sure who exactly to blame for it no longer existing. At least MS tried to support it as a platform (for a while)

It's pretty much unsustainable to support more than one arch when dealing with closed source Operating Systems and Applications. A few of the big players will start only offering binaries for the most prevalent platform and like sheep the rest will follow, before long driver support would also vanish for the less popular platforms.

Even right in the middle of the availability of NT for Alpha I can honestly say that I would have never considered using it because the application and driver availability just wasn't there.

It's the effect of market forces unfortunately, and the cheapest up front cost will almost always win over technical excellence (Betamax vs VHS) and on going running costs don't usually get considered (cheap printers, expensive cartridges). I rather liked the RISC idea and always had a soft spot for the ARM chips (although I've only recently acquired any machines that use them) - although part of that may be their home grown heritage (although they are now Intel in much the same way as old time software legend Ultimate Play the Game are now part of Microsoft!).

Wayne Stallwood wrote:

Tell me about it, I feel guilty about leaving this AMD 64 monster running 24/7, but I need to access it remotely (as do others). I wish it could clock itself down to say 300MHz when it is idle (and perhaps then shut down or idle the CPU and PSU fans) but at least under Linux there seems to be no way of achieving that.

Check out speedfreq. It will let you set the policy by hand, or adjust it dynamically, and let you monitor it. For example, this is what happens when I stop reading email and kick off a compile:

mak@yoda$ speedfreq -m New policy: dynamic 1000-2200 MHz CPU new speed 1000 MHz, 96.0784% idle ... CPU new speed 1800 MHz, 87.7551% idle CPU new speed 1800 MHz, 0% idle CPU new speed 1800 MHz, 0% idle CPU new speed 1800 MHz, 0% idle CPU new speed 2000 MHz, 0% idle CPU new speed 2000 MHz, 0% idle

I always use the dynamic policy, and as a result, in its usual idle state my AMD64 3500+ (in an ASUS A8V Deluxe) uses less power and generates less heat than the P4 it replaced. And provides more performance when I need it, and can run 64-bit should I so desire, but without forcing me to. Lovely.

If you look at the "AMD Athlon™ 64 Processor Power and Thermal Data Sheet" on amd.com you'll see it lists P-states, supported frequencies with documented "Thermal Design Power" usage, per model. For example, at maximum p-state my cpu lists 89W, at the min p-state 22W; that's quite a difference. When I chose my cpu it was based on the supported p-states.

I guess it's market forces, people buy computers based on bogus performance figures and hardly anyone seems to consider running and environmental costs.

This is changing, driven by the common desire for quiet computers. If you're into that, check out the forum on www.silentpcreview.com.

Also, if you want to experiment with the energy costs of your various bits of hardware, you're welcome to borrow my power meter; it can be quite illuminating (CRTs really are power monsters), and gives you a different perspective.

-- Martijn

On Friday 22 April 2005 10:05 am, you wrote:

Check out speedfreq. It will let you set the policy by hand, or adjust it dynamically, and let you monitor it.

ooooh Cheers for that, looks very interesting, I saw CPUFreqD in the SuSE package selection but it only seems to support some mobile chips. Your one looks more interesting.

I always use the dynamic policy, and as a result, in its usual idle state my AMD64 3500+ (in an ASUS A8V Deluxe) uses less power and generates less heat than the P4 it replaced. And provides more performance when I need it, and can run 64-bit should I so desire, but without forcing me to. Lovely.

That looks pretty close to what I want. Still need to do something about the fans. The PSU has thermal control fans (2 of them) but is a bit optimistic about fan speed. The front case fan is disconnected because with the rear case fan and the twin fan PSU I draw enough air past the drives to not need it. The rear case fan and CPU fan are supposed to be running under the Cool 'n; Quiet thing my board has but seem to just run at full speed (ACPI isn't aware of them) all in all the machine is louder than I would like.

You must of had a similar problem with your case Martijn (assuming you have all the fans connected) I scoped it at the last kit meet and it is the same (very expensive but lovely) Cooler Master case my previous machine was housed in. From memory that had 4 case fans (plus PSU, CPU etc)

Also, if you want to experiment with the energy costs of your various bits of hardware, you're welcome to borrow my power meter; it can be quite illuminating (CRTs really are power monsters), and gives you a different perspective.

Thanks for the offer but I have one here, it's one of those rainy day things I keep meaning to get around to.

On Sunday 24 April 2005 5:24 pm, you wrote:

My plan is to scratch-build a case and fix its various issues, but that's a project that hasn't started yet. I'll let you know if that ever becomes a reality.

It's crossed my mind a few times too, I have scratch built cases for MiniITX and Flex-ITX systems, but they tend to be a lot easier because the sizes of the boards give you a lot of flexability and the DC-DC PSU's tend to be very small.

If I was going to build an ATX (or BTX if it ever takes off) system then I would probably end up doing something similar to the Hush systems but in a tower format (I need more space for drives and cards than the Hush boxes allow) The only problem with Heat Pipe technology is that you would have to construct new pipes everytime you changed the position of the CPU (say by upgrading the Mainboard).

If I was going to build a more conventional forced airflow case then I would look pretty closely at what the big names are doing (Large heatsink with ducted air from a large but slow case fan) as that in my mind is a far more elegant solution than the little screamer stuck on top of the heatsink that we are usually stuck with. The reason you don't see ducting on OEM systems is again because the position of the CPU varies between Mainboards and architectures.

I was looking in the back of a iMac G5 yesterday and they use a sort of fan assisted convection system which seems to provide some sort of balance between a fairly power hungry CPU and what is actually a pretty quiet system.

On Wednesday 20 April 2005 10:16 am, Adam Bower wrote:

Ooh, please do. Even though I recently sold all my Amiga stuff I'm still a great fan of the amiga. Unfortunately my Amiga 4000 only had an EC030 so couldn't run any type of Unix or Linux, BSD etc.

All right you've convinced me. Next rainy day I'll dig it out and blow the cobwebs off.

On Sun, 17 Apr 2005, Wayne Stallwood wrote:

On Sunday 17 April 2005 12:19 pm, Bob Dove wrote:

Argh ! They were the ones that had the PSU for the machine built into the monitor weren't they ?

That's right. I've got one inherited from my brother. It used to run an early GUI called GEM. Though it had DOS-6 last time I looked. It was upgraded with a thing called a hard-card - an ISA card with a 20M hard drive for motherboards without an IDE socket. It worked surprisingly well. At some point I should wack it into an old motherboard with an ISA slot and see if it boots.

Michael Morton

==========================================================<

School of Information Systems | Everything is linear if University of East Anglia | plotted on log-log with Norwich | a fat magic marker.

==========================================================<