Improving Computer Form Factors?

eschasi asks: "Recently we've been seeing some trend towards smaller footprint machines like the new iMac or the tiny PC system shown in this recent article. All these tiny systems have scalability or quality problems for me, and I don't think it has to be that way. But all the solutions I see require discarding much of what we think is standard to desktop/deskside mechanical design for PCs. Apple has been able to do with much more freedom becuase they own the whole process. PC makers, however, don't. The last major improvement to PC motherboard/case design was the ATX design, and that was evolutionary rather than revolutionary. I think that major improvement can still be made in an evolutionary way. I want both worlds: I want a small footprint; I want it in a premium system; I want it to have enough room for a pair of hard drives, a 5.25" external slot, and a 3.5" external slot; and I want it using largely off-the-shelf components, and I don't want to have to re-invent ATX and PCI and ya-da-day to get it!" Do you feel the same way? How would you improve the current crop of PC form factors?

"IMHO large chunks of the problem could be solved by taking advantage of both sides of the motherboard. Put the CPU and most of the misc. chips on one side, put the expansion slots and RAM on the other. The case design would have to change but that's quite doable.

Using both sides of the motherboard does more than reduce footprint by half while returning to full expandability. You also get the benefit of having the RAM sit directly opposite the CPU, reducing trace length problems and permitting RAM bus speed increases. Other timing problems might also be reduced.

I've left aside the rest of my extensive arguements for why this is or isn't a good idea; what I'm interested in is this:

• Do other people see the same drawbacks with small-footprint systems?
• What seems to be out there on the drawing board for post-ATX systems?"

Krow's comments: Form factor and cute design is all great and lovely but I want to know what can be done to reduce electrical usage for the average geek's home hosting farm collection. Even small usage reductions add up.

Comments

cPCI Cards

[fiori]

I'd love not to need to open the case, move all the cables around, and try not to slice my fingers on the case when changing or adding a card. I'd like to see the PCI and AGP internal slots replaced with cPCI cards that slide in on rails from the outside.

Re:cPCI Cards

[fobbman]

Why stop at cards? I could also see hard drives and CD-ROM drives going this route as well. For those who have Dell laptops (maybe others do this, too. I'm just mentioning them because I have one) look how they have the hard drive installed. Remove one screw and it slides out of the side, easily replaced or swapped.

Also, it's been mentioned before but bring back the USB hub in the monitor base. Means I don't have to run long USB cables for my mouse and keyboard. And for that matter, how about speaker outputs and microphone inputs? Heck, IBM for awhile had units that had the CD-ROM drive in the monitor base. Really wouldn't take up that much real estate beneath a CRT that already has a lot of dead space beneath it.

Re:cPCI Cards

[IntlHarvester]

I could also see hard drives and CD-ROM drives going this route as well.

A few years ago, Intel and Microsoft came up with a spec called Device Bay that would allow end users to easily slide-in USB and 1394 drives.

Never caught on with OEMs though. Probably because of the lack of integrated 1394 controllers, but also because in the white-box market every penny counts, and for things like the iPaq desktop, the vendor would prefer that you buy proprietary parts.

Re:cPCI Cards

[Sawbones]

It's a plug for the Evil Empire, but their natural keyboards have a small usb hub built in now. They've also got an array of "Internet Buttons" along the top that are basically useless to me, but I've been using my USB mouse through it just fine.

Also, the HP machines at the EE department of my alma matterhad headphone and mic ports in them. It may not be as convenient as if they were in the monitor, but then again it's a smaller sting to replace a keyboard than a monitor if things start to flake out.

Re:cPCI Cards

[4of12]

Personally, I'd like it if the whole PCI bus was changed from electrical to some kind of optical connector.

Something in a connector kind of like what you find on newer consumer electronic audio components these days.

I know that optical BW has got to be sufficient, AFAICT the only issues are the cost of putting the converters on motherboards and the latency of these devices.

Then, the CPU and RAM would be in a single small quiet cube with a small power supply and fan and a handful of optical connectors.

These would then connect to CD, DVD, IDE drives, monitors, keyboards, boxes that convert to RJ-45 for copper Ethernet, to USB, IEEE1394 or whatever else in the way of legacy connected devices are still around.

Then, you could keep the noisy disks in a utility room where they belong and the desktop would be a handful of small quiet cubes with kbd,LCD,mouse instead of this large hulk of a PC case that needs to enclose a populated M/B with wide ribbon connectors, CD drives, disks and the whole doghouse.

Just dreaming.

Re:cPCI Cards

[seanadams.com]

Remove one screw and it slides out of the side, easily replaced or swapped.

...or DROPPED!

I've seen a Dell hard disk destroyed in exactly this way. Granted, the owner should have noticed the screw missing, but still, I'd say Dell is mostly to blame for having the hard disk held in by a single, tiny screw.

I'll take a Powerbook any day:

- Batteries and CDROM are securely held in their compartments by a spring-loaded latch. It's almost impossible to accidentally unlatch them, and if you do, who cares, it's just the battery.

- The hard disk, memory, and 802.11 card are stowed safely underneath the keyboard. It takes about two seconds to flip down the keyboard, 20 seconds to install an Airport card, five minutes to install memory, and five minutes to swap the HD.

- *much* sleeker/thinner styling than the Dells, whether Titanium or Bronze.

Vaios are my favorite PC laptop, but that's not saying much. I can't stand the keyboards on the Vaios. Why does nearly every PC latop maker think they're doing us a favor by putting the arrow keys where the shift key ought to be?

Re:cPCI Cards

[seanadams.com]

Then, you could keep the noisy disks in a utility room where they belong and the desktop would be a handful of small quiet cubes

There are some products out there which let you extend DVI using fiber optics (100 yards) or electronic repeaters (50 yards). Take a look at www.gefen.com. You could put your PC off in the closet and have just a perfectly silent monitor+keyboard in your office.

The fiber equipment is ridiculously expensive - about $3K, but the electrical repeaters are probably a bit more affordable.

You can't do this with VGA due to the distortion and ghosting, but it works flawlessly with DVI. Also the spec on DVI is an interesting read - it was actually designed with optical transmission in mind. The data rates are insane, so it's split across four channels (clock + 3 colors, IIRC) - the total data rate for a 1600x1024 display is something like 4Gbps.

Re:cPCI Cards

[PhotoGuy]

SGI's O2 computers were cool like that. Everything, including power supply, hard drives, etc., popped out in a slidable tray. When you popped 'em all out, it was pretty much an empty shell.

It would definitely be a nice trend.

I also like the early Macs (not sure about the current ones), where you could remove all components with tabs and such, no screwdriver required.

-me

Where do you want to go today?

[Renraku]

The reason ATX has been the last major change in PC form factors in a while is because it was about the last thing we could really need from it. However, I do agree that there should be smaller or larger form factors available, ATX simplifies the whole process by allowing case/power supply builders to create a bulk supply of the same thing. If you want something different, build it yourself, or pay someone to build it for you. It goes back to my old equation. In order to invent something, your need has to overcome the time needed to invent it, and the resources used in creating it.

Re:Where do you want to go today?

[rblancarte]

One of the largest problems with the ATX design however (at least for the argument of this article) is the size of the motherboard itself (no pun intended). I mean, even if you said - let me build my own case, you are limited by the size of the MoBo you have.

From what I am reading from this article is that the writer wants to pursue as whole new motherboard design, lets call it MAT - for Mini AT. With this MAT design, you would be able to still have the current expandabiliy of the desktop, but have it in a smaller system.

Look at the bulk of current systems - you have a side mounted MoBo. If MAT kept this, BUT was able to use this double sided printed - you could take much of the depth of you case out. The height might not change much. Or you could keep the same depth, but change the height, amd move much of the stuff (like the power supply) to the front where you just vacated much of the MoBo. This would result in a case that is slightly wider than what we see now, shorter, and still just as deep.

What I envision when I am thinking about what the auther is talking about is to take the MoBo and do a top mounted design (you will see why in a bit). The processor and most of the chips would reside on the side of the MoBo that faces UP. You would mount your cards on the other side (probably same size cards). Your power supply and drives could be mounted at the bottom of the system, under the cards. The main issue would be venting - and with that, I would propose vents along the TOP back of the system, with the MoBo doing most of the channeling of air. It would direct air in a way that it would have to go to the front of the case, OVER the MoBo and then out the back - this would HOPEFULLY displace the rising air off of the CPU and other chips. You could use a number of fans to input air from the lower side of the system. Overall, you would end up with a wider, but possibly shorter and not nearly a long front to back.

This would be a very interesting project to undertake. But while making a case would probably be easy to do OR get done, making a MoBo that would fit to these standards is probably beyond the range of most people who are reading this.

-RonB

hmm

[nomadic]

I think the main problem is cooling; you can jam everything into a surprisingly small volume, you just run into heating problems. What I don't understand is why don't they make laptop-sized desktops? Just make a notebook computer without a monitor; small footprint and it should be relatively cheap (considering that the display is one of the most expensive components in a laptop).

I know what I'd get rid of...

[codetalker]

RIBBON CABLES. I hate the things. I would much prefer something where you just plug the drive in like it were a card and off it goes.

Re:I know what I'd get rid of...

[x136]

This will be possible as soon as drive manufacturers get around to building hard drives and the like with a native (or at least semi-native) FireWire (aka IEEE1394) interface. Most FireWire cards I have seen have that mystery internal connector. FireWire might even provide enough power to power the drive itself. How does daisy-chaining internal drives with one cable sound?

I haven't read much into Serial ATA, but that sounds promising also.

Other Form Factors

[sachmet]

There are other form factors, such as NTX, that one could build a small computer from. The major problem lies in mass-produced small footprint cases; these can be difficult to come by. Someone pointed me to this cheap chassis at GoogleGear. It might be worth looking in to.

Big Cases.

[Heem]

I still like the old, big towers that are about the height of your desk. you can fit so many goodies in there, who cares if you really need them or not, you still CAN have them. Why does everything have to be so small these days? Little tiny cars,pager sized cell phones, cell phone size computers, small PC cases, iMac's and other similar computers (gateway profile etc, basically laptops on sticks). No thanks, not for this guy anyway. I'll keep my big truck, my clunky cell phone, and my 3 foot tall computer case.

Uhhh... we already have a small standard.

[Phosphor3k]

Its called MicroATX and FlexATX. One of which(i cant remember) the FV24 Shuttle system uses. The problem is manufacturers not correctly implimenting the standard, makes for hard interchangability of parts.

Drawback - They Have no "Killer App"

[Ieshan]

With the market like it is in the countries where people will be buying computers, the "footprint" system doesn't make up for in style what it drops in price. Currently, the people who are buying computers are either looking for power or affordability, not a suave appearance.

You might say this isn't true, but the majority of computer users aren't "geeks". Since the computer isn't so widespread that it's in every room of the house - usually, there's one or two an "affluent" household - people can hide them away in offices which aren't going to be seen by your houseguests and the like, making the "style" thing unnecessary. Furthermore, the consumer has gotten so used to the grey-box case that it's still considered "in style".

That's not to say that someday in the future, the smaller, chic computer won't become more widespread. I think the problem is that the consumer isn't willing to spend more AND sacrifice power to do it, and currently, that's what they've got to do.

Simi-OT What about the Inverse?

[AnalogBoy]

mofoinasciiartistwannabetroll...er HI! me again. Just wanted to ask the inverse question. What if you want your computer to be big, frightening looking, with lots of blinkinlights?

I mean, granted, a modern minitower is nice.. but two or three lights just arent enough.

I therefore start the campaign for more blinkinlights.

Sun Systems are nice. Large (You have the "Refrigerator Cabinet", "End Table" and "Ottoman" form factors). Adequate blinking lights (ESPECIALLY on 6500's!)..

I want my computer to have lots of lights. I don't care if they do anything. I want my keyboard to sit in the middle of a console with a lot of dials, gauges, digital readouts, switches, buttons, knobs, and things that go "PING!" (a ping light WOULD be good, to think of it.) I want to see my network utilization on a graph led. I want to see my processor usage there, too. I want to see a red light come on each time some PFY discovers WinNuke, or someone tries to NIMDA me. I want to have one of those covered red buttons that may or may not launch a thermonuclear device. I want screens that go "zeeeeeeeerp, zeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeerp" as text scrolls across it.

I already have three monitors on my desk. Wires strewn everywhere. But what else do i have? An Ultra 5, A beanie baby, and a dell optiplex. And a Dell Keyboard (at least i makes those satisfying *click*'s.). I WANT SWITCHES, dammit.

soo... any suggestions?

[notices steam rising from ears]

err.. teehee..

Re:Simi-OT What about the Inverse?

[geekoid]

find an old Irwin Allen computer prop(those where all those computer wqith blinking lights a spinning tapes in shows like 6million dollar man, Lost in space, etc...) then stick a PC in it. Be sure to take pictures of the people who first see your "killer" box..

Re:Simi-OT What about the Inverse?

[BeBoxer]

At one of my old jobs, they had a "console" for an old mainframe. An Amdahl IIRC. Anyway, it was basically the size of a desk with a dashboard of lights and switches along the back of the desk. It was easily big enough to use as a normal desk if you wanted to, as it had a ton of flat space on top. But the best part was it had an honest-to-god mechanical odometer in the dash part which counted CPU hours (I think.) Anyway, I always wanted to liberate it and take it home, as it sat there unused for years because no one wanted to move it. But alas, it disappeared before I had the chance. Would've made a great desk. Oh well. At least I have my Sun 3/160 end tables.

Re:Simi-OT What about the Inverse?

[AnalogBoy]

You know, between that knob, some normal analog switches, randomly blinking lights, LED displays, a dremel, and a few flat panels, my cube at work is going to be undergoing a major renovation soon. :)

"See this knob? I turn it, and our CRM goes faster." "This switch disconnects all users" "This button redirects all bandwith to my workstation" "This blinking red light is.. just a blinking red light" "Oh.. the blinking blue light? It means one of the executives is donwnloading porn. yes, i know its on solid."

I like the suggestion someone else made.. Analog VU meters for Bandwith.. Except i'd want some form of steam actuator on it, for when bandwith approaches maximum.

Attention chassis manufacturers: HANDLES

[mojotooth]

I'm with the submitter that smaller form factors are nice. However one thing that I know they can control right here and now is the presence of HANDLES on the cases.

I'm an avid gamer. I had to buy a third-party (albeit excellent) product from CaseAce to help me carry my box to LAN parties etc. And I got a flat-screen monitor that was easier to carry than my 17" monitor.

But if computer manufacturers would just put a stout handle built in to the case, that would go a long way towards making me happy. Same thing for monitor builders, although I understand the problem is different for them.

For instance, look at the Gamecube. Arguably as much power as the other gaming platforms, but much easier for portage purposes. Why can't computer makers take a hint here?

Stackable components

[TheFlyingGoat]

What about changing the way the AGP/PCI/ISA cards fit onto the motherboard? SBC's use cards that stack onto one side of the board. If this was done, you'd save about 4" of height. Just have the AGP stack on one side of the motherboard and the PIC cards stack on the other. Cooling would be a factor, but a single fan would push air more directly over the components, so I think it would work out ok. Laptop sized hard drives could become industry standard in time, and hopefully at some point floppy drives won't be used anymore (I dont have one in my system). Finally, CD/DVD drives could either be made smaller by making them slot-loading, or by using external drives (ala SCSI or USB). I guess that leaves the power supply, which I don't know much about, but it seems like there's a lot of wasted space in that little metal box. :) My computer, using a stacking concept, small footprint motherboard and an external CDR/DVD would probably be able to fit inside a 5"x5"x5" cube.

Old School

[Golias]

Some of the features that have been inherited from the old IBM's and Compaqs should probably be reconsidered.

For a lot of systems that were made and sold Back In The Day, the power supply was, wisely, in a separate box from the mobo. The CPU is not the only heat source on a PC... Setting the power supply behind your desk, away from the case, might make a small system a lot easier to do.

Also, we have been keeping the keyboard as a separate component ever since the old 8086 days... but is that always the best way to go?

It seems to me that the perfect reduced-footprint destop PC design would look a little like a laptop PC with no screen, a nicer keyboard, and no touchpad... perhaps with a cable output for bridging to an optional stand-alone box for PCI expansion cards (for those who want the flexibility).

The end result: a latter-day C-64. Ahh, nostalgia...

Few things stick out in my mind

[joeflies]

Most of my things don't have to do with the board form factor but rather irks about cases that mount the board.

1) Improve the layout to facillitate air flow. Perhaps if there was a way to vent the cpu, vid card, and power supply and atmospher cool the rest, then perhaps it could be done with fewer fans.

2) Make front-mounted slots a standard so that adding front mounted devices such as the Creative EAX, USB and Firewire ports, headphone jack, could all be done without making custom modifications or using up an external drive bay.

3) Edge-mounted cabling? Nothing bugs me more than having to unplug all of my IDE drives to change ram.

4) How about let's all de-evolve into s-bus computers form factor, then scale by adding cpu-self contained boards (what was that, the compaq 386?) that plug in the bus.

5) screwless drive bays?

6) how about a 1U alternative designed for the home?. I'd think more home appliances things with WiFI could be made with a equipment rack-mounted system if it didn't take up so much room.

Board design doesn't work that way.

[pagercam2]

The idea of puting all the chips on one side and slots on the other, while it sounds easy is a real nightmare for HW designers. Many of these chips have 500+ pins (PGA) or 1000+ ball (BGA) these are all comming from a chip less than 2" on a side (PGA) and 1"+ for BGA and all those wires need to go somewhere, the PGA devices have to have holes in the board and can only get wires out from underneath, by going between pins and often only 2 wires fit so another wire has to use another layer. The problem is layers increase cost much faster than additional area, so the boards being bigger than they have to be saves money. There have been specialty technologies developed to get higer denisties, but these are majorly expensive and not realistic for consumer level products. The most impressive I've heard of is IBM super-computer technology, which used a 57 layer board, getting 8 layers to line up is expensive and there is a 30% scrap rate (IIRC) they had to make 100+ boards at 57 layers to get just a couple of working boards. The best way to make smaller machines is to go multi board and stack, the intel processor modules go along way in this direction as do the PCI adapter board that have one slot that you plug a board into that has multiple PCI slots. The heat is going to be the bigest problem no matter how hard you try, smaller means less air and as clock speeds increase less air isn't the right answer. Water cooling, or the electric coolers could help, but its always a how much do you want to pay to get small, fast etc...

Look at consoles

[Apreche]

but not the XBox, it doesn't solve any size problems. I like the idea that I don't have to open the computer to change cards. But I would also like to not have to fiddle with ribbon cables. I want all my drives 5.25 and 3.5 to be hot swap. Just slide in and out while the computer is on, and they get auto configured like USB. Also nix the PS/2, serial, and paralell ports. I want like 8 USBs instead of 2. And on RAID boards with 4 IDE plugs put them all next to each other. Not two in the normal spot and two over by the PCI slots.

Putting stuff on the other side of the board has advantages and disadvantages, but however you do it a shrink in board size will rock. You will no longer need a full tower case to get a bunch of drives in. Mid tower cases will be able to hold as many drives as full towers do now if you shrink the board.

As for cooling, look at other small computers, like console systems. Especially the GameCube. It has one fant blowing air in on one side, out on the other. Because of it's small size, it becomes easier to cool, not harder. You have less fans, moving more air.

Legos are the answer

[nsample]

Really. Legos.

I was given the idea by looking at the original Google server in the basement of the Stanford CS department. Its case is built from Legos (or, maybe Duplos in reality... they're pretty big).

But why not start with a small form factor for the mobo+busses+limited drivespace. Then, anything you want to add on gets clicked together with another piece. Need two external 5 1/4 bays? Use a 2H piece. Build it as you see fit.

The real issue at that point is a standard bus architecture to bridge the components, but I think the flexibility potential is immense. Got too many devices? Click in an additional 1H powersupply. Need to move your CDRW between boxes? It's a "snap" (insert "hardy-har-har" here).

The drawbacks are of course heat and redundancy. Each module whould have to be self contained at the start, but I can even envision a series of holes bridging components as well... like those old hamster tubes worked. Need more flow because you've got a stack of 43 devices? Add an extra fan module somewhere along the path.

Anyway, it's not here yet, nor will it likely ever be, since it's not mass-marketable (I think). But, it would allow flexibility to grow, a small footprint and size for home users, and massive physical component compatability.

Legos were always the answer when I was 12, too. Some things never change.

Two sided doesn't work

[afidel]

The number of layers required to make the board two sided would be so high that the added expense would definitely be noticable. Like 3 TIMES more, at least. Just going from 4 to 6 layer pcb design makes manufacturing more expensive, and also increases defect rates, further increasing the costs or the board. I don't think most slashdotter's (or pc users in general) will want $300-$400 system boards.

My thoughts...

[cmowire]

There already is a further evolutionary move towards smaller PCs. Just not for premium users.

The MicroATX/FlexATX/WhateverATX form factors are the latest steps towards a smaller PC. You can reuse most of the ATX parts for a smaller system. Of course, the problem is that most people who like premium computers want lots of slots, so that doesn't happen.

So the market of people who want small PCs who aren't willing to buy an already made low-profile PC from someone, who want good parts, is too small to be useful.

Now, eventually motherboards will get smaller for the simple reason that as speeds go up, traces need to be shorter. And when there isn't room for improvement, parts will be integrated. Think of disk controllers, serial controllers, etc.

The main reason why you don't like the Shuttle low-profile system is probably video, which is right now too much of a moving target. Eventually, video will settle down and people will be able to get great 2D/3D for a $25 chip. At least, until they find a new area to push video cards with.

Now, it probably WOULD be possible to attach components to both sides of a motherboard. But you wouldn't want any user servicible parts on the bottom, because that would increase the difficulty of upgrades. This won't give you a perfect reduction in motherboard size simply because the motherboard layout is subject to a lot of finicky tollerences that mean that certain parts need to be in certain places.

I suspect that the ATX form factor could use some changes. I'd like to see explicit support for front-mounted jacks, better management of the LED/Speaker/Etc jumpers, standardized front-mounted ethernet activity lights for rackmount, etc. But remember that the ATX form factor is best extended over time with compatable changes instead of going through massive changes all at once. That's the beauty of FlexATX/MicroATX -- they didn't require massive retooling of assembley lines to the new standard, and you can always stick a FlexATX or MicroATX motherboard into a regular ATX case, in a pinch.

Why I don't want a laptop

[huie]

Laptops have everything integrated and to upgrade anything, if possible, you have to buy it from the manufacturer (namely the CD/DVD/whatever).

The Shuttle SV24 allows you to use one PCI card, a 5-1/4" bay and your choice of processors (okay, maybe not _any_ processor, but still, you have more choice than with most laptops). You're not stuck with the relatively low-res display that the laptop has and you can use your preferred monitor, keyboard and mouse. It's also nice to have the keyboard disconneced from the display so you can get a more ergonomic setup.

Yeah, if you've got a USB keyboard and mouse, you don't really have to worry about that, but then you'll need to get a hub. And you can plug your external monitor into the laptop, but why spend the money on a laptop anyway? For the built-in UPS/battery?

No, Tom's Hardware has it right- the only problem with the Shuttle design is the integrated graphics (and possibly the audio too), but for most people, that's a pretty reasonable compromise for a small unobtrusive system.

Yeah, I'd like to wait and see what might come out with the nVidia GeForce chipset in the Flex-ATX form factor.

Interestingly, notice how no one cares about where to put a floppy drive? :)

Re:Other Form Factors - NLX not NTX

[WebCowboy]

I think yyou mean NLX form factor and not NTX. In any case, you make a valid point. These flexATX boards are a nice idea, however they are a hack job of a standard (ATX) that wasn't designed with compactness in mind (in my opinion). Shrinking the ATX form factor comprimises expandability--typically a tiny ATX board has only one PCI slot and integrated everything. Also, the small flexATX cases strike me as not very serviceable (cramped, awkward layout, cooling issues and so on).

The NLX form factor was designed for compatness, serviceability and flexibility in mind. The mainboard itself is very small and has no expansion slots--rather it has a card edge that plugs into a special slot on a backplane. The backplane contains the mainboard slot plus one or more PCI slots (the only drawback I see is that I havent found an NLX backplane that has an AGP slot--other than that it would be an ideal form factor!). The typical mini-desktop might only have one or two PCI slots, but with this form factor largger cases or industrial racks could have a backplane with many more slots (expandability).

Also, since the mainboard plugs INTO a slot rather than having slots with cards in them it makes expandability/serviceability MUCH easier--it you need to expand memory, upgrade or replace the motherboard, etc. you simply unlatch the board and pull it out the back of the case, without removing any cards or disassembling the chassis. The prime drawback, as sachmet mentions, is the lack of mainboards and cases sold in this form factor. Consequently they fetch larger prices. Why this is is beyond me--I guess it's the inertia of maintaining a modicum of compatibilityh with the 20 year old PC form factor...

Swappable hard drive etc.

[HawaiianMayan]

It's not specifically about a small footprint. It's about improving the usability of the hardware.

Apple improves usability by making it "just work" out of the box for 90% of computer users. But obviously Slashdot power-users are in the other 10% ;) So how can we improve hardware usability for people who work inside and expand their systems.

DeviceBay was a spec for hard drives and other components that would slide in and out of the chassis, and connect using FireWire or USB (so you could swap things on-the-fly). It was mostly for rack-mounter server farms, but would have made everyone's life easier. Unfortunately, it never went anywhere.

Back in the early 90s Apple and a few design firms were playing with ideas for a computer that looked like a rail, or a backplane. Components, cards, drives, were hung on the rail or slotted into the backplane.

Manufacturers (besides Apple) don't really seem to care about the usability of the box, though. If they did they would have ripped off the PowerMac G4's side door well and pronto (Dell's attempt at it can most charitably be described as "half-assed").

Apple's take on things

[kelv]

It is worth nothing that the reson that apple has been producing such small machines (from the Mac Plus, Mac Classic days to the current iMacs) is that they realsied that 99% of all computer users NEVER want to open the case.

As far as I can see this is a fact that the rest of the PC industry has never caught onto properly. If you sensibly match all the components in a machine then most users will never need or want to change a thing.

This is different to the /. crowd who want to do this all the time but we are hardly a random sample are we.

Kelv!

why not?

[moosesocks]

First, there are a few limitations to this that should be discussed.

CPU The cpu needs room to breathe. Athlons and p4s output lots of heat. The heatsink cannot be obstructed... Period! An obstructed cpu is an accessability pain, as well.

Drive bays 1 internal 3.5 bay and 1external 3.5 and 1 5.25 would be plenty. Very few users have 2 hard drives (most geeks do, but this is geared twoards end users. If they do, they can sacrifice their floppy drive, and possibly buy a usb fdd). Adding an extra internal bay wastes precious space... Of course, there could be exceptions made to the specification for some cases with more bays. Anyhow, firewire drives are getting dirt-cheap nowindays...

Now, for my suggestions

45 degree expansion slots. Probably my most radical argument. It's pretty self explanitory: PCI and AGP slots are on a 45 degree angle, thus reducing the amount of space needed. This is assuming that the motherboard has 3ish slots to begin with. It's not as efficent for pcs with only 1 or 2 slots.

"Unspecified hole" On the rear ports template, place a rectanglar "hole" there, where the motherboard maker can place any extraneous ports it wants (s-video, extra usb or firewire ports, video capture, scsi.. etc).

Rear template The rear template needs more connectors then a regular pc. This is becasue these compact pcs will have few expansion cards in them. This is where ATX really falls on its face. The ports it should have are
4 usb
2 firewire
2 ps/2
1 ethernet/modem (hole big enough to fit either/or)
2 serial/video
1 parallel (legacy, my friend, legacy)
4 sound (for s/pdif or rear channel)
Game port?? This seems rather extraneous with the rise of USB joysticks. Perhaps a bracket should be provided instead. Any die-hard gamers will have a dedicated sound card, anyway...

CPU at the BOTTOM of the board Place all the cool-running components at the top of the case, under the drive cage and PSU, and put the cpu and the connectors at the bottom. Sheesh. Didn't the ATX people think of this... It's also a shorter fall when your heatsink falls off. Possibly put the IDE connectors at the bottom too??

ATX power supply It ain't broke, it seems to work fine, the units aren't horifficly big, either. An idea would be to have super-compact units accept d/c power with an external a/c adapter...

Non-conductive motherboard mount points. It's quite annoying when you're installing a motherboard, and it requires washers at the mount points, in order to function. This adds about $0.25 to the manufacturing process.

Drive rails. Every case should have these. Sure, it's a luxury... but, it's a nice one. So are thumbscrews, and removable motherboard trays, etc...

All in all, we have a case that is signifigantly smaller then normal ATX cases, yet not too small to loose it's functionality... I got a bit carried away, and it looks like some of my specifications would possibly add size.... oh bother! Did i forget something?

Both sides of the board.

[wowbagger]

You don't want to use both sides of the motherboard - it makes production a nightmare.

Consider a single sided board:
Solder screening deposits solder paste on board
Pick and Place machines place the parts on the board. They are held in place by the surface tension of the solder paste.
IR reflow melts the solder. Any misalignment of a part is handled by the surface tension of the liquid solder.
Board cools.
Smoke test.
Ship it.

Now, consider a two sided board:
Screen solder on backside.
Pick and place parts, using small daubs of glue to hold the parts on the board (surface tension won't hold a part against gravity).
IR reflow backside. Because parts are not free-floating, any misalignment stays.
Flip board over.
Screen solder on top of board
Pick and Place parts.
IR reflow. Solder on bottom of board melts too, hence the glue.
Smoke test.
Swear, since the parts on the bottom of board are out of alignment and board doesn't work.

Cappuccino & Espresso anyone??

[TMLink]

Have we forgotten about these two beauties already? It's already been done. Don't know if there has been another updated version of these to come back out, but I assume you can still get the Cappuccino machine.

Great little machine, if you've got the money for it, and have a reason for wanting it...(easy machine to haul around for gaming sessions). But this machine's exactly what you're looking for.

Who needs expansion?

[Alomex]

Expandability is highly overrated. At work, where I'm a "power user", the only "expansion" of eight PCs in the last six years was extra memory.
In all other cases by the time I need a bigger hard drive, I also need a faster CPU, more memory and a better monitor.

At home, where the usage is less demanding (web surfing), I upgraded modems twice, replaced one hard drive, and added memory. This is of seven computers I had.

All in all, IMHO expandability is something that the average user does not need (beyond swapping components), and the advanced user thinks will need but doesn't either (a few devoted hackers excepted).

Roll your own

[RainbowSix]

I've posted this before, but if you want a use-specific case, you often have to do it yourself or pay $$ for a custom job.

If you are looking for something small and easy to move around yet is upgradable, consider doing something like this:
my server

Built into a small briefcase, I can carry it to school and have it running without even opening it. Running linux I can ssh in and do everything I need. It runs a generic Slot1 mobo with 3 pci slots so you can conceivably throw in your favorite PCI Radeon or GF2, a 1 gig PIII, 3 slots worth of RAM, and have a pretty good machine.

Wah wah

[Graymalkin]

What is it about the slashdot crowd that wants everything at the cost of nothing. Maybe they get it from using Linux far too long. You can either get efficiency or you can get scalability with a very small range between the two. For efficiency look at gaming consoles, they sacrifice the ability to scale in order to have a tight nit efficient system. Another entrant into this catagory is/was Cyrix's Media GX chip and Intel's 810 chipset. The cost and waste of the system is lowered by putting more components in the same packaging. On the side of scalability look at the Mac 9600 with it's 12 memory sockets and 6 PCI slots. Of course just about any PC or workstation class system falls into the scalable classification, the 9600 is just an example of sheer expandibility. The marrige of these two is something like the G4 Cube which is hated in many circles. It was small and fairly efficient yet had the ability to be upgraded a little bit. However it came at the cost of not being able to use widely abvailable standard sized expansion card (a video card upgrade costs beaucoup cash because you can only get it from Apple). Slashdot folks want something that can fit a half dozen components into off the shelf yet be compact and efficient. It isn't going to happen unless somebody releases a system with a "computer on a chip" plugs directly into a backplane that links it to other components. Even then people would bitch because the COC components weren't up to their expectations.

2-sided motherboard

[markmoss]

IMHO large chunks of the problem could be solved by taking advantage of both sides of the motherboard. Put the CPU and most of the misc. chips on one side, put the expansion slots and RAM on the other.

First, many people would say that you should get rid of the card slots and hang any add-ons outside the box on the USB or Firewire boxes. I'm not going to address that -- it's OK for those who won't do much with their 'puters, and it would save $10 to $30 in parts, but I want those slots!

The change to the case is rather minor: tall standoffs for MB mounting, make the case an inch or two thicker but smaller in other dimensions. There would be resistance from people who were worried about being able to get a new MB to fit that case when needed, but this hasn't stopped case evolution before. And I do see a trend towards shorter, more cube-like towers. However, there are some real manufacturing & physical difficulties. They're solvable, but might run the costs up:

1. The CPU heatsink has to change considerably, because you won't want to waste the two+ inches of space needed for a conventional heat-sink/fan. How about a liquid cooling bag using that side of the case as the radiator? Does that cost a lot more?

2. Most motherboards presently use bus and RAM connectors with wave-soldered through-hole pins. That is, the MB end of the connector is male pins, which go through holes on the board, then they are soldered all at once by pumping liquid solder up to the bottom of the board as it goes down a conveyor. You cannot put the CPU socket through this, and it's probably not a good idea to put the chipset's giant IC's through the wave either. (Small capacitors and resistors are OK on the wave-solder side, and it's quite common to find them on the bottom of a MB.)

3. Also, if you are using through-hole connectors and a board not much bigger than required to hold the connectors, where do you put the CPU and other big chips that don't conflict with a through-hole pin?

The solution to #2 and #3 is obviously to change to surface-mount (SMT) connectors. They cost more now, although that would change if more MB manufacturers ordered them. But at present they are also more trouble-prone than the through-hole parts. There's been more than 60 years of work done on improving wave-solder yields, and about 20 on SMT, so we get more defects in SMT, and the testers aren't as good at finding them.

Finally, bus connectors take quite a lot of force when inserting and removing boards. There's no chance of through-hole connectors pulling loose from that, because the holes in the board were plated with copper, and the solder goes all the way through and broadens out like a rivet at each end. It's not going anywhere unless the solder breaks at one end and the copper separates from the board everywhere else; I have no idea of the breaking strength of the pin/solder connection, because if I hook a hydraulic puller onto a through-hole connector either the connector body or the fiberglass-epoxy board will break first.

SMT has a lot less inherent strength; the pads are just laminated onto the top of the board, and can pull off. So the connectors have to also have pegs or screws that go through the board and are fastened on the other side. In a PCI connector, you can only put pegs at the end, so the connector body has to be rigid enough that the middle won't pull up. This is another reason (besides lower production volumes) that SMT connectors are more costly -- extra plastic is expensive.

OTOH, we put the through-hole connectors in by hand and the SMT by machine. All the issues are potentially solvable (sometimes by spending more on parts and machines and less on labor), and I expect that in another 10 years pure SMT will become cheaper than the present SMT/throughhole hybrid technologies. It just hasn't happened yet, and I have no idea what PC's will look like by the time SMT bus connectors do become common.

Think on the Side of the Box

[SatelliteBoy]

Changing a cliché:

One of the big problems of modern PC design is cooling. Perhaps the answer is in plain sight - the side of the case.

Just provide a thermal path from the CPU to the side of the case. This may not be easy, but is solvable.

Ideas:

Heat pipes. Overclockers have used them, spacecraft use them - they're proven. A heat pipe with rounded ends and mounts with rounded receptacles and a bit of thermal paste. Conduct heat to the side of the case, and allow for variations of the CPU location.

Thermal strap. Some spacecraft use these, too. They conduct heat well. They are not as multidimentionally adjustable as above, but can be worked out.

At least one side of the case would have to be thermally conductive. Aluminum or copper would suffice. The case side could then be made flat or ridged. The "side" could also be the upper panel of the case, provided you don't really want a big, heavy monitor on top of it. Just imagine a sculpted, polished copper case sitting on top of your desk. A pain to clean, put purty.

The idea takes the concept of shrinking the PC, and also answers another problem - noise. One can eliminate a fan (or more, extend the idea to other components) and gain coolness.

I think I may start designing a case for myself.

Jonathan

[cpt kangarooski]

I remember the 'Jonathan' design prototypes Apple developed in the 80's. Essentially the computer would consist of modules that plugged into a common backplane, extending the plane as it went.

It looked a lot like a row of books standing on a desk. Working out the various buses would be an interesting challenge, (just put everything onto fiber optics?) but it could be very small or large, depending on user tastes and what hardware they'd acquired. And parts could easily be reused or replaced.

Naturally, being a really cool idea, nothing came of it. (particularly since Apple was considering opening the standard to everyone)

There's some pictures in the book "Apple Design" but I haven't seen any online.