Slashdot Mirror
How Cheap Can A PC Be?
geoff lane writes "Ballmer wants a $100 computer. OK, can we build a reasonable PC for just $100 and a copy of Linux? The rules are: It's assumed that a monitor, keyboard and mouse are already available. Ethernet connectivity must be provided. All components must already have Linux support. All components must be new and currently available. The result must be electrically safe for the home. Is it possible?"
he wants something windows only and to sell windows-lite for $40 for it.
Why should the hardware profits be sacrificed to support high software prices?
Perhaps Windows should be cheaper to support high hardware prices. Cheaper software might also reduce piracy since the it would be more affordable.
ShoutingMan.com
You can go to the average garbage dump and find at least one computer that will run something like Debian without a GUI. If you're lucky, you might find a Pentium II or faster, and be able to run something like DamnSmallLinux. Chances are, you'll be able to find a monitor, keyboard, and mouse there too. That accomplishes the task for $00.00.
eclecti.cc
Absolutely; They're sold by a company named "used".
Seriously though, do we really need a $100 disposible pc when there are so many functional used machines stacking up in corporate closets?
-Chris
--an unbreakable toy is useful for breaking other toys--
My IBM PC300PL is worth about 100 bucks. It's got 288MB, a 40GB drive, a 40XCDRW, an Intel P3-450 and a free Ethernet card even though it's already built in to the MoBo. The problem is NOT NOT NOT NOT the hardware it's that Steve Balmer wants to sell you a PC that needs at least twice the hardware as that. If MS just gave us a secure efficient version of W2K we could all have 100 dollar PCs.
Only if you want to run today's bloated software (even open source), which as far as word processing goes doesn't do much now that a well-developed product 10 or 12 years ago didn't.
A 100 dollar computer, hell a 50 dollar computer doesn't seem out of reach if it doesn't have to run all of today's windows and linux apps, but only has to be capable of running more svelte applications which do the same things.
"PCs are not selling to the lower end of the population in China and India. People buying machines there are relatively affluent. So...should the prices be lower? Not really. Until government and situational factors reduce piracy...those people...don't pay," Ballmer said. (article clipping)
Now an open letter to Ballmer
Ballmer
Shouldn't people in the lower end of the population spend their money on something a little more worthwhile then a computer.
Maybe just maybe they could spend that money on their family Before purchasing such a luxury item as a computer. Of course I am not going to be naive and say they don't need a computer for some reason. But to say that I want money from the lower end of the China/India population is selfish, Specially when they have better things to spend it on..
I don't do business with your company on those rash comments. I get by without using your software. Sorry if you feel that I am not being fare.
Not saying I haven't pirated your software before, instead of attacking me you're attacking someone who couldn't even pay you if they wanted to is just harsh. Oh and by the way I used your software to learn about and then go into computers so in a indirect way your company benefitted from it.. So the very thing that you are against has kept your company afloat, by customer awareness.
I no longer use any pirated software from your company. I get by with alternate platforms (Mac, Linux)
Daniel
And that's a whole hell of a lot harder.
There are no trails. There are no trees out here.
You can get a Linksys wireless router for about $70. It's a machine with 16M of memory, 4M of flash, and a 125 or 200MHz chip. It also comes with a hub, a wired Ethernet, WiFi, and a power supply. So, that shows you can ship a lot of hardware for fairly little money.
Replace WiFi with a simple VGA controller and give it a couple of USB ports and a little more flash instead of the hub and you would end up, at roughly the same price, with a usable personal computer that could run a light X11 desktop and some useful apps (browser, word processor, etc.). If you add a CF slot, people even have removable storage.
Another choice is the standalone file server appliance, also for under $100 AFAIK; it already has the USB port and also runs Linux.
And some of the game consoles also show it can be done, if you get the volume high enough.
Most of the answers are along the lines of, "I can't find the parts at that price in this catalog or that store". I don't think that was the question.
Some other comments have focused on whether what Balmer said was reasonable. Interesting topic, but that isn't the question either.
Some other comments have said, "Yes, get a used one." That still isn't the question.
The question is: Could we spec out a PC that, in volume, could sell for $100 and run Linux?
An interesting twist on the question: Can we consider it "a PC" (for purposes of this question) if it doesn't have an Intel-compatible processor? Say, a StrongARM CPU? (Note that the criterion was that it run Linux; well, Linux runs on a wide variety of CPUs.)
In the old days of mini-computers, sellers would charge more for the minicomputer version of software than for the PC version, even when less people were using the PC version (ie. there was no volume discount argument). The reason they could get away with this was that people who'd paid for a $10k computer would balk less at paying more for the software.
Turning this around, while MS charges a fraction of the cost of a new PC, people are prepared to see it as a relatively insignificant expense (eg here in NZ, I'd pay probably NZD1K retain for a computer (inc monitor etc and WinXP)) and WinXP is only say NZD200 of this.
If however the computer price came down to say NZD400, of which WinXP was half that, then I'd have a much harder time brushing the WinXP cost under the carpet.
Lower PC costs will force lower software prices.
Now I have RTFA, but Ballmer probably has it in his head that people will pay NZD1K for a computer and if the hardware costs only NZD200 then he can put NZD800 in his pocket. People are not as dumb as that.
Engineering is the art of compromise.
[laughing] My first PC was a 2-floppy, 2MHz XT, with Herc mono graphics. WordPerfect 5.0 was crisp, even running off a floppy. After a dedicated word blender, it was heaven. And when I replaced that with a 12MHz 286 (also with Herc mono, but it had a HD, and WP5.1 along with various other apps of the day), it was, like, WOW!! Everything ran like the wind. Well, Ventura Publisher 2.0 took a while to load, but it ran fine. I still have the 286, and in a pinch... it still does everything I can't live without.
:)
:D
Nowadays... we struggle to get decent performance out of machines THOUSANDS of times faster than those relics.
BTW I'm writing this on a P3-550, somewhat slower than the average of what's now found on the curb. (Methinks I need to look at a better class of curbside.
But I still use WP5.1 every day.
~REZ~ #43301. Who'd fake being me anyway?
Look, a sub $101 computer isn't rocket science. There are landfills full of say 500Mhz and below machines...
:) $99m &cate gory=51110&item=5133297107&rd=1
A 400Mhz machine, even a 166Mhz machine is suffice to run lots of stuff...
Face it, we all use to use them...
A 400 Mhz machine with 128mb RAM is quite a lot of machine for what the average person wants it for:
1. Word processing
2. Calculator
3. Web browser
4. Lousy paint program
A majority of cycles are wasted with the user sitting there..
Here's an old Dell that meets your lofty needs
http://cgi.ebay.com/ws/eBayISAPI.dll?ViewIte
For $200 you could get the keyboard, mouse and LCD monitor all in the nice form of a portable computer. Be it 500Mhz or so, Linux will run just fine.
What the hell does everyone need a 1Ghz or 2Ghz spec'd machine for? It produces tons of heat, typically noise too and eats up tons of electric with that huge power supply you all want...
I saw an apple 2 gs go for 35 bux on ebay last month.. bastard sniped me at the last minute. It came with a monitor, 5.25 and 3.5" disks, a dot matrix printer, and a box of software. That computer can do word processing, spreadsheets, AND play hundreds of classic games! (aztek, threshold, wizardry, wolfenstein, karateka, sigh I love apple ]['s)
We sell pentium 2 and 3 cpu computers to employees for $75 at my company when they get swapped out. These computers are able to run all modern business software, browsers and email. They just don't have the speed and snappiness that we are all used to. Everyone wants flat panels and small form factor PC's these days, so they just sell of these old computers and do some wacky accounting magic to write it off or depreciate it or god knows what.
New $100 computer? Only if you are a manufacturer. Used $100 computer? totally do-able.
music lover since 1969
Hey:
Its great to ask this question, and I'm all for cheap hardware. But...given that hardware must be manufactured, consume raw materials etc. I would expect that the floor cost for hardware should _never_ go as low as the floor cost of software - especially after you get past some R&D point for both.
Can you say "monopoly"? It seems much clearer to me that software ought to have some fully commodified components and that the OS ought to be that component. Given that the world of software has (intelligently) landed on layered architectures, we'd expect to be spending money at the higher layers and have ever increasing commodification at the lower layers. Again...can you say monopoly?
Now...I"m not arguing that hardware should NOT fall under this rule, but....well....some costs associated with hardware are a given, and those costs will forever be higher than the "given" costs of software.
Just my 2cents.
I know there are a million of these sites out there, but I've actually purchased stuff from these guys before (besides the fact they are local for me, no shipping!) Here's one for $100, Dell Tower that's a P3 550, 128mb, and a 10gig drive. They even have a _6_ month warranty for any issues that may arise. You can also upgrade the memory for another 25 or so. That'll run a lot of flavor's of Linux for CHEAP.
Throw in a network cable for half a buck's worth of parts.
Total cost BEFORE cd-burner/dvd-player*:
Motherboard: $35
128MB RAM: $20
4.3GB HD: $14
Case w/ power supply: $22
Floppy drive: $6.50
Ethernet cable: $0.50
Total: $98
Linux: Free, in both senses of the word
Look on Steve Ballmer's Face when he reads this on
Um, Microsoft, when you get the license cost of Windows down to $1.99, you too can play this game.
*internet cafe's don't need CD players on every machine.
Knowledge is how to play a game, intelligence is how to win, wisdom is knowing what game to play.
Uh, no.
32MB of RAM? Are you *kidding* me? Even my minimal setup (X.Org + Fluxbox 0.9 + Firefox 1.0 + rxvt) is using 221MB as reported by free, with one instance of each running. (Not counting caches, buffers, etc). I'd consider 256MB the bare mimumum, MAYBE 192 in an emergency, but X alone uses 59MB. 32MB might have been enough to run XFree 3.x and fvwm or windowmaker, but it just doesn't come close for even a semi-modern desktop.
(I'm running Slackware 10.0 btw)
TODO: Something witty here...
One IDE channel. No floppy, serial, parallel, or PS/2 ports. Kill IrDA support.
The problem is that virtually none of this saves money.
Legacy support costs virtually nothing. The only expensive parts are the connectors (the interfaces are all integrated: if you want IDE at *all*, you basically get everything else) and you can just put them on a pin header if you want.
IrDA, serial, and PS/2 are all the same thing - parallel, floppy, and even IDE are usually supported on one chip. They're so cheap that there's no point not to put them on. For one thing, they're useful enough to the people testing the board that they earn their keep just that way.
The CPU and RAM chips could be soldered onto the board. Bundle it with a cheap mass-market OEM hard drive, a case with a 40W power brick, and you've got a PC.
RAM prices fluctuate too much for this to be succesful. CPU integration makes sense, although again, the price drops quickly enough for you to be left with a platform that's far overpriced in just a few months. Keep in mind, that's one of the main reasons you don't integrate the CPU and memory - price concerns.
For one thing, in the time it takes the system to get to market, the board will be a bit overpriced/underpowered for its price point. Systems that have socketed CPUs/memory are viable on the market for a long enough period for people to sell off their supplies.
The way you make a cheap motherboard is to only use the integrated peripherals in the southbridge, and then volume, volume, volume.
yeh... i saw that...
i'm _certain_ that the production price on that is about $70 for the hardware. plus another 50-or-so for the msft tax. still leaves a nice heatlhy 35% gross margin.
via is really giving the geode line a run for its money though, and i think that theyve got the better SoC tech right now.
although, i think that amd's manufacturing advantage could crush via.
the real shame here is transmeta - they would a perfect fit for this type of a device, but they're:
1) too expensive.
2) financially insolvent.
and as for the pricepoint for the linux... its actually amazing - sometimes, the linux devices are more expensive, for identical hardware, and have a higher GPM.
really though, these low-end devices are more than enough for 90% of the computing publics needs.
... hi bingo
I was in India over the summer, and I visited one of my cousins who sells PCs from a small shop in Bangalore. Most people over there buy $200-$400 PCs, because that equates to 10,000 to 20,000 rupees. Farmers and laborers (75% of Indian population) make 50 rupees per day, so only the upper class city dwellers can afford PCs. Still, 25% of 1 billion is 250 million people. Would you pay $100-$200 for software on a $200-$400 PC? No. So free or pirated software rules in India, and will continue to do so for the forseeable future.
It's not reasonable for Ballmer to expect Indians (or others in the developing world) to pay $100 for a copy of XP, unless he can magically make the average Indian earn $40,000 per year rather than $3000. Also, keep in mind that in India, electricity costs more and a UPS is mandatory, so funds available to purchase hardware and software are less.
Even at $200-$400, hardware costs far outweigh labor costs in India. In the US, computers under $200 are not even worth the time to sell them or fix them, given that any qualified PC tech costs $60-$80/hour. So all the sub-$200 class PC components get junked.
This leads to an interesting business opportunity. If there was an efficient means of accumulating all the junk components, they could be shipped by sea container to India, where PC techs could sort out and re-sell the working parts. It costs about $4000 to ship a 20x10x10 ft. container to India from the US. So, you'd need to collect twenty $200 PCs or the equivalent to cover shipping costs. Since once the parts get to India, labor is almost free, the only other cost is gathering the components together in the US. A large corporation might have the means to do this cheaply, however. Maybe large corps should partner with Indian salvage companies to get rid of their old computers. They might make some money rather than paying to have them disposed, and also, Indians could get their $100 PCs for checking Hotmail.
Via already puts the CPU package on the mobo; it saves PCB space and power leakage. We're not going for a powerful system; just a cheap one.
The Via systems with integrated CPU chips are more expensive than socketed counterparts, partially for the reasons I stated. They have to make them more expensive because they make less of them - they make less because their shelf life is shorter.
Note that there are two sets of Via integrated CPU boards - there's the Eden set, which exist for homebrew PVR and somewhat of the embedded system market. But they have a higher feature set - this justifies the fact that they're underpowered. There was also an older one which sold disastrously (Syntax is the only one I knew that actually sold it - the S8601MP and similars).
The point is that Via can't expect to sell them very cheap, because if they've got the CPU and memory integrated, then the product depreciates much faster than the board without an integrated CPU. It should also be noted that if you buy the chips in solderable packages, you're buying more specialty parts rather than commodity, though if you're the chip manufacturer, it's not such a big deal.
And yes, you do save on the connectors.
For a company which buys connectors in volume, you save nothing on connectors. Even when I've bought in very small volumes, I've been able to pull DB9s down to under $1 per. In the larger volumes, you can get it much cheaper. In fact, if you make things other than the motherboard, then they're virtually free, as it's just a fluid item. Unless you do it for space concerns, the cost of a DB9 is just too cheap to even bother not putting on. Which is, of course, why it still exists on most things.
One other thing is the fact that the LPC port is very useful for debugging and board testing/prep, as well as hardware monitoring. There's enough of a demand for low-cost motherboards that if you could save money by dropping the legacy ports, you would.
Even the several legacy-free motherboards have the legacy ports on the board.
The board could also be small with no legacy stuff -- smaller than ITX form factor.
You'd be better off, cost wise, sticking with the largest market - micro-ATX. Volume, volume, volume rules all.
If the volume is high enough, you can design a southbridge that doesn't have the legacy support.
That's the problem - it's not the volume that's important, it's the volume*margins - that is, the profit. If you're starting off by saying "this is going to be a super-cheap platform", your margins will start off being thin, and as the platform ages, either the demand falls off, or your margins get thinner as you drop the price. Now, half the problem is you still need to pay for the development cost.
I don't think you're going to see someone come out and design something like this to be cheap. It wouldn't become profitable. We're almost at the point where the commodity items are near that price point as it is.
Basically, I don't think you're going to shave much cost off from any of the methods you're suggesting. You might be able to get it started on a size argument, but I think the small form factor PCs have shown that that design can command a premium, and so they'll charge it.
One other concern that you do fail to note is that the smaller the motherboard, usually the more expensive its design, regardless of how simple it is - you simply crowd the design, and the routing becomes very, very complicated. I have no doubt that the nano-ITX boards took a few iterations to get the signal integrity reliable enough.
I think VIA might do it - but I think you'll see far more of a commodity PC than you would expect. If it's specialty, it's expensive - if it's commodity, you don't have to pay the development, and risk having stock without any demand.
If we're going to start plugging things, then I'll seize the opportunity and throw in a plug for FreeCycle. FreeCycle is a great way to get a good used computer (or anything else) for zero cost, and also an easy way to clear out all your old junk by giving it away to local people who find it useful. No packing or shipping hassle, since the recipient typically will come by to pick it up, and you'll earn more karma that way then you ever will posting to Slashdot. :^)
I don't care if it's 90,000 hectares. That lake was not my doing.
Funny, I've got a machine that was designed just that way, right down to the blinking email LED. It's called an Audrey, it was made by 3com, and it failed miserably in the market.
People want a "real" computer for the same reason that they want "upgradable" computers; futureproofing, whether they need it or not.
--saint