Slashdot Mirror
Feature: Where is Integration Going?
The following was written by Slashdot Reader Michael Crawford
Integration:Where is it Going?One trend that has recently been gaining a great deal of momentum is integration. Odds are that if someone has invented a useful piece of hardware someone else has plans to build it into either a CPU, a motherboard or a graphics card. Recent motherboards have built in sound, graphics, modems and other useful components. While Graphics cards are trying desperately to differentiate their products by including options like duel monitors and TV-out capabilities and graphics chipsets are adding features as fast as the features can be conceived. The graphics chipsets are even moving into areas that have traditionally been the CPU^Òs job such as geometry and lighting effects. While CPU designers such as Cyrix are pushing new chips that will eventually contain all of the basic computer components, including large portions of the motherboard and graphics chipsets, on a single chip. The eventual goal of this processes is called PCOAC(PC On A Chip). A single chip that performs all of a computer^Òs basic functions. This is the ultimate goal of the current attempts at integration. Why is this important you ask? It is important because over the next three years the computer industry will be drastically changed as a result of integration. Entire segments of the market such as motherboard chipsets and networking devices will vanish as these components are integrated into other, larger components. Watch for more and more soon to be obsolete companies moving out of their traditional markets in an attempt to survive. An excellent example is VIA^Òs acquisition of Cyrix. As PCOAC becomes standard the vast majority of VIA^Òs current business(low-end chipsets) will be rendered a obsolete. Thus they are in effect trying to escape to high ground before the flood arrives that will spell the end of most of their current competitors. Now you might ask why integrate?
The integration has two primary intended effects. One, the card, board or CPU in question gains the added feature that makes it more attractive to customers. Two, the use of less chips saves on system cost and power consumption. This is why current highly integrated systems tend to be intended for either the low-end of the desktop market or the laptop market. These systems are selling well and there are already plans to further the process, but the benefits are not without risks.
The main risks include that the systems that are currently available tend to lack performance when compared to more traditional systems and they are also more difficult to customize or upgrade. Because many different components are contained within a single chip the possibility of upgrading or changing the components is non-existent. Manufacturers also tend to believe that because a particular piece of hardware is already in the system the manufacturer does not need to make provisions for that component to be upgraded later. This is seen in recent motherboards without AGP slots. This loss of upgradeability could lead to consumers being forced to buy more computers than before. Computers already become obsolete quickly, without upgradeability the situation could become much worse. What I expect each segment of the computer industry to do as PCOAC approaches.
Things to remember, these are predictions that may only become true over several years, most of these predictions apply to the low and mid-range computer market. As a general rule the high-end will be effected last and least.
Motherboards/chipsetsExpect more and more components being built into each board. In a market where being different is good, look for the trend towards integration to be strong. At first these boards will be intended for the low end but early signs of the trend continuing into mid-range systems are already apparent. Once the motherboard makers run out of features to add they will start trying to increase the quality of features. Within the year relatively high-performance graphics and sound cards will be built into many boards. eventually the motherboard makers will lose influence over the industry while the chipset makers all but disappear.
Commentary: In the long run the motherboard manufacturers will do fine, they will have less control over their boards but they make a component not easily replaced and thus will exist for the foreseeable future. The independent chipset designers have much bigger problems. For the short term they will follow the trend like everyone else but in the long run they may cease to exist in their current form. They face several main challenges. They are a small relatively cheaply produced component trying to integrate very complex and expensive components such as video cards. This is a situation that is very similar to a small fish trying to swallow a big fish. They are also being held behind the technology curve by Intel (still the dominate player in the industry). Intel wants control of the chipsets and uses its patents and industry clout to prevent others from making Intel compatible chipsets until years after the new CPU has become available. The combination of being relatively cheap and largely powerless within the industry may lead to the demise of the independent chipset design companies. The chipset^Òs traditional tasks are likely to be integrated into CPU^Òs.
Graphics Cards/chipsets I expect to see more graphics chipset companies buying or being bought by graphics card manufacturers. This gives more control over the use and capabilities of a particular card/chipset combination to the company owning the design. A single company will be able to carry out every step in development and manufacturing and would be capable of locking out competitors. The buyouts allow the company that designs a successful combination to reap all of the benefits without clones dragging prices down.The independent graphics card companies will disappear as the larger companies that design their own chipsets stop selling graphics chipsets to their competitors and are capable of maintaining lower pricing due to entirely in house product development.
Expect the graphics chipset companies to rapidly move into areas that have traditionally been the realm of CPU^Òs as they search for newer and better features to improve performance and differentiate their product. I will not be surprised to see many aspects of the CPU^Òs FPU to be taken over by the graphics chipset. Some of us still remember debating about which math coprocessor (FPU) to add to our system. This could emerge as an option on a graphics card in the next couple years.
Commentary: The graphics chipset design companies will remain relatively safe in the short term but may be in jeopardy in the long term. In the short term the industry is newly emergent (how long ago since the voodoo?) and moving at an incredible pace. New generations of chipsets are coming out faster than one per year. As long as this pace continues upgradability will outweigh the cost savings given by integration. In the long run however the pace of advance is sure to slow rendering the chipset design companies vulnerable. The major CPU companies will eventually try to devoir the chipset design companies. The CPU manufacturers are easily the largest and most influential within the industry and they will probably succeed.
CPUs For the short term the major CPU companies are preoccupied with the current struggle for dominance. They will try to gain the support of the graphics and motherboard makers and will not encroach into other markets until the CPU market stabilizes. The CPU will continue to advance as it has before with an ever increasing influence on FPU performance.In the slightly longer term special CPU^Òs designed for the low end computer market, unlike the K6 which was intended as a top of the line chip but failed in that role or the celeron which for all practical purposes is a P2, will bring integration to new levels. Both Cyrix and Intel have publicly announced plans for a PCOAC design.
Eventually (2 or more years) the CPU will become a single giant conglomeration of components including the motherboard chipset and graphics chipset. It will be in effect the only portion of a computer with any significant control over performance. AKA :PCOAC
Commentary: Many people will say that this will never happen because of the lack of upgradability and that the consumer will not buy it. The problem is that this really will be Much cheaper and Much higher performance. It may never make its way into some markets but it will become the standard. I suspect that the companies that will eventually be left in control are the CPU companies. This is because they are the biggest and are in the best situation to carryout the integration. Two excellent examples are Intel and Cyrix. Intel is just huge, they are already in every market discussed above. Intel makes chips, chipsets, graphics chipset, motherboards (including highly integrated ones) and numerous other networking and specialty devices. They have publicly stated that they are developing PCOAC designs and they have the money and designers to do it. This is a perfect situation for Intel. If successful they would have almost Complete control of their computer. (we all know Intel would like that). Cyrix is in almost as good a situation as Intel. Cyrix has recently been bought by VIA. Giving them a new shot at the CPU market with new tools. Between VIA and Cyrix, Cyrix has access to almost everything it needs to pull of the ultimate feat of integration first. Via makes chipsets (including highly integrated ones) Cyrix makes CPUs (including highly integrated ones) the only thing lacking is a competitive graphics chipset program but that could be acquired. Where PCOAC is concerned Cyrix has the lead in both time and technology. It already has chips like the MediaGX that are the current best examples of integrated CPU^Òs (it contains many memory and graphics functions) and they have more on the way. Only time will tell if Cyrix will be able to stay ahead of Intel in this department. It might pull it off assuming Intel is too distracted with AMD to develop PCOAC seriously. AMD might be the exception where integration is concerned. They have excellent CPU designs and have expressed some interest in PCOAC but lack the experience. They only recently began chipset and motherboard design and have no graphics or other programs to speak of. They might be able to pull it off but will need to either invest in some new companies (see my AMD predictions) or form some partnerships within the industry (something AMD has done a lot of). It is unlikely AMD will be the first or second to make a PCOAC. Don^Òt forget that there may be others out there working on PCOAC besides the big three CPU manufacturers. Recent rumors imply that both Transmeta and ATI are moving in that direction and it is a safe bet that IDT and Rise will make a go at it. As more and more companies try to get in on the CPU business expect the market to change drastically. Only time will tell.
The Rest For the foreseeable future RAM and hard drives will be unaffected. Networking components will become just another feature on either a motherboard or a CPU. Most other components will be integratedCommentary: If it is small and a chip it will as a general rule be integrated. If it is not a chip (ex: Hard drives) it is probably impossible to integrate. Many little companies will vanish when they are rendered obsolete by highly integrated components
You can read more of Michael's writing at his homepage.
I don't think Mr. Crawford is totally correct in all of his assumptions. I do agree that there will be more integration, but not to the extent that Crawford predicts. Chipsets will probably move to the CPU as will video, and the RAM will also move there in the next few years. One of the largest performance hits in a CPU is not having the needed data in the cache. When you compare the access time of a cache to main memory, you get about a 4:1 ratio. Four cache accesses in the time of one main memory access. But move all of the RAM onto the CPU and that penalty is avoided. So, I guess I see system integration in the future, but I don't agree with Mr. Crawfords assumtions.
This all sounds great except for one thing.
These PCOAC are going to need larger die sizes for manufacturing and this leads to less chips on a given wafer. Since no Fab that I know about has near 100% yields this means that companies will get less viable chips off of a given a wafer. Since having even one transistor wrong on a chip causes it to fail ie it just a hunk of silicon now.
I don't see these chips being cheap at all
I'm expecting now that some company starts to create general purpose vetorial co-processor for graphical and accoustic activities.
Then, all that proprietary sound and graphics boards will die!
This heralds one of two options.
1) The disposable computer. In this scenario, Joe Schmoe goes to his local computer store, sells his soul to AOL for 3 years, and has a computer. After a year, he finds that it's obsolete (he wanted to play Unreal II-"the slideshow") and wants one that preforms better. If it's hard to upgrade at all, he'll toss it, and talk to AOL about a new computer. (the indentured servitude program).
2) The console computer. Joe Schmoe isn't afraid of this one. It looks just like the N64 he used to own. The PCOAC fits into the slot on top, the (solid state)hard drive slides into a slot in back. A DVD drive is located in front, next to a couple slots for memory upgrades. Why, they even have the OS on a rom that's conveniently soldered to the hard drive! Hey Joe, do you want a 30 gig that comes with windows or linux?
I'll miss my big, noisy tower.
"The the leap from PCM to wavetable MIDI was not as great. . ." ."
I think you mean, "The leap from Frequency Modulation Synthesis to Wavetable Synthesis was not as great. .
And I'm not sure I agree. When was the last time you heard an SB 1.5 playing music? Man, that was horrid.
Wireless Bus (drool)
Plug a tranciever into a transmission hub, and let short range digital radio handle things...
Wait a min... how many unique digital id's would we need if everyone started doing that?
...that the spelling and grammar of critics are often worse that what they are judging? Examine the statement - "Really, some effort should be made, to edit pieces four, spelling and grammarr." Does that make any sense to anyone? Has anyone heard about the "speck in your brother's eye and the board in your own" (paraphrased)? If you are going to be so rude as to judge someone for their writing, the LEAST you could do is write your critique properly! Otherwise, you'll look the fool you appear to be... that is unless you are purposely trying to look like a fool. If so, you have succeeded.
...that the spelling and grammar of critics are often worse than what they are judging? Examine the statement - "Really, some effort should be made, to edit pieces four, spelling and grammarr." Does that make any sense to anyone? Has anyone heard about the "speck in your brother's eye and the board in your own" (paraphrased)? If you are going to be so rude as to judge someone for their writing, the LEAST you could do is write your critique properly! Otherwise, you'll look the fool you appear to be... that is unless you are purposely trying to look like a fool. If so, you have succeeded.
Actually, the integrated m/b seems to be targeted at a specific market to compete with appliances like Web TV. By reducing the price - and these computers have come WAY down in price - computer manufacturers are able to retain market share that would otherwise be lost to the $200 single purpose Internet PC and such. Simultaniously, workstations with more. Typically, I'll build an inexpensive, integrated boxen for a home user who'll do little more than browse the internet and do homework as well as single purpose office computers, and a high performance modular system for serious gamers and other activities requiring more specialized cards...
Every screwdriver shop buys piles and piles of mobos. I've bought several myself. Do you really think this is all gonna die off in favor of crap from Packard-Bell and fellow-travelers?
I honestly hope that future hardware companies take integration to heart. After all, if one places most of the functions of all that hardware onto a single chip, so that the motherboard essentially ends up: chip, firewire port, usb port, ethernet port, dimm slots, video & sound in out, Computer manufacturing would become far more enviromentally friendly.
Integration will work marvels with component based hardware, and distributed storage. All I want for Xmass is a slew of cheap terminals, and a central data server. If PC on a chip will get me there for $100 USD (or better yet $100 CND) beam me up Scotty.
We (the company I work for) have already felt this trend. We are currently developing our own 3d video card using the 3dlabs chip set. All was going well until 3dlabs decided that they want to market there own video cards. We were cut off and left out to dry. No more tech support. We will be lucky if we can even buy the chips.
What is the difference between a $400 PC and a $200 Dreamcast? Not much. The PSII will blur the lines even more, and the next generation Nintendo will erase the lines altogether.
Next thing you know, we'll be getting computers with the case welded shut.
:)
Wow, Apple really was ahead of their time!
Daniel
Really, some effort should be made, to edit
pieces four, spelling and grammmerr.
Effect is a noun, affect is a verb, generally. (There are other definitions of these words, of course. I am only referring to how they are used in this essay.) Also, what ever happened to ASCII? Every apostrophe is a catastrophy. You see, if the spelling and the grammar are correct then educated people won't be distracted by the mistakes and will be free to focus on the content.
As an example, remember the ide controller in your isa system? Remember how cheap the serial card was? Back in the day, needed a new 16550, you just got a new serial board. The next generation uart (we'll pretend here) is the usb's async controller... whoops, you can't get that UNLESS you buy a new motherboard. On my new pentium boards, if the floppy controller dies, I have to disable the onboard stuff, and plug in an old isa controller (we all know floppies are passe, so it's not like this is necessary). I am positive that if I cared to look I could not find on the market anywhere an isa/pci ide/floppy controller. This is so ubiquitous a motherboard feature these days the only solution is motherboard replacement.
Before you know it, memory manufactures will be in cahoots with the mobo manufactures to build mobo's with all the ram ON BOARD!
Why would they do that you ask? Well just like the former paragraph, it's so that when a compenent fails, or an upgrade is necessary you are buying a new system! this is the ultimate goal of the wintel hegemony. The commodity pc (the ultimate refinement of lets say, the imac and the eMachine) is a box with Everything On A Chip and a flashy box. Performance is not an issue! Microsoft has shown us that consumers don't by performance, they buy GLITZ. Just like iMacs. Just like flash Packard Bells, and eMachines. Winmodems and integrated A/V on board -- a performance nightmare! (At least the imac's performance doesn't suck... comparatively.) Once all pc's are totally commodity, prices may seem low, but profit margins are high (for the reasons in the article). In house control means cheaper machines with the added bonus of being LOCKED into one manufactures UPGRADE PATH.
Of course this path is one way and one big step at a time... Don't like your old PC? Buy a new one! With new stain removing power! or Flame-broiled taste! whatever the latest market lingo is.
Anyhow, enough ranting
Read Heinlein's 1953 Revolt in 2100, now more than ever.
That I'd like to see. "Pistols or daggers, sir?"
(Sorry, I don't usually spell-flame (like I have any right to), but I could not resist. ;)
Boobies never hurt anyone. - Sherry Glaser.
The die size issue is a problem today, but will likely be beaten as fab technology improves. one way around the defect problem is to produce all CPUs to be PC on a chip. If some non-critical componant has a defect (say the DAC for the sound part), just disable that feature and sell it in a market that doesn't want it anyway. Of course if the feature is critical to any operation, it's still junk (Unless they make an earring out of it).
IBM also has an extensive library of ASIC cores. Choose you favorite parts from http://www.chips.ibm.com/techlib/products/asics/co res.html and build your own PCOAC!
First of all, I don't really believe that VIA/Cyrix has the funding, staff or reputation to take be a contender in the forseeable future.
Second, my impression of the Cyrix acquisition was that VIA took the parts of Cyrix that National was going to chop: specifically the x86 pin compatible line. I believe that National has held onto the MediaGX line, which means VIA probably only got their hands on the 686/MII lines -- none of the "integration" products. No doubt these cores could be helpful to them in developing a new SOAC, but they still have a ways to go.
Anyone know any hard facts about what VIA bought from National?
--Lenny
Sometime in the near future...
Gosh, I sure would like newer, better video features... I guess it's time to buy a new computer.
Later
Hmmm.... I need better sound capabilities... I guess it's time to buy a new computer.
What we need is PCOAFC - PC on a few chips. Make it moduler so it's easy to upgrade. Make it cheap so I can buy it. Make it fast so I can run the next iteration of Quake.
Well, i'll step up and give you my opinion that you'll be in luck. Basically, as he stated in the article itself, this will become the dominant for low end users, first time pc buyers, etc. Yes, it'll creep into the middle market - come on, realistically, companies like cheap, and that's what this'll offer. So most paper pushing people in an office will have an integrated board. But for the developers, the die hard gamers, the graphics sector, yes, modular will exist.
I tend to think it will be a lot like modems. Currently, you can buy an internal modem, but often you are getting a glorified pos. (That Blue Light Special v, 90 modem comes to mind...) When you buy an external modem, you have an amount of guarantee that it's a bit more reliable. They make the same pieces, but they have to up the quality, because they can't pass the buck if a component breaks. Same thing will happen for the types of computers our segment of the market buys. Same pieces more or less, but at the higher end of both the quality and price scale, to offset the 'expense' of being modular.
Alot of the functionality needed for things like sounds/video, is going to move to specialized processors, whether or not they get integrated into the CPU. They are going to get consolidated down, to smaller/faster processors. And when new functions appear, they'll be implemented, with software. So the only time to upgrade, will be to get a faster processor.
The down side, is that since the software to drive the functions is basically another driver layer, that means that most likely, there will be a base level at which systems will have decent functionality, with relatively good stability. Then any feature level above that, the systems will become increasingly less stable, in a trade off for the added features. I can hardly wait.
"Politicians are interested in people. Not that this is always a virtue. Fleas are interested in dogs." P.J. O'Rourke
I don't agree with a lot of Crawford's crystal ball gazing, it's attempting to extrapolate much too far into the future in too much detail.
I see that a lot of posters have suggested an alternate view, i.e. that because of consumer choice issues we'll see still more pluggable modules, i.e. the complete opposite of the PCOAC.
I think the market will continue to support the current diversity. PCOAC for low-end 'commodity' PC's and embedded applications such as set top boxes, and a wide range of compatible, function-specific modules for custom, high-end applications.
The configurability issues will likely be settled by standard. Odd that one poster quoted Bill Joy as championing hardware integration. Isn't it more likely that he's rather see all those chip manufacturers licensing JINI for auto-setup capabilities. Then you can plug in all your cards and chips and they'll negotiate their configuration directly with each other and the OS. In other words, son of plug'n'play.
By using a standard but flexible protocol like JINI this would remove the responsibility for troublesome plug'n'play from the OS altogether (hooray!). The hardware would decide and the OS would just do as it's told. The hardware could even instruct the OS precisely which driver to use and where to get the latest version.
That's the way it should work ideally. Just like plug'n'play does today, but controlled directly from the hardware.
Let's face it, Scotty or Geordie LaForge would be pretty lost in a crisis if they couldn't cobble together a repair to the engines out of transporter components because the whole of the USS Enterprise ran on a single ship!
In my opinion this is what the market wants: the flexibility to combine function-specific multi-vendor hardware in whatever combination we choose. On the other hand Michael Crawford's vision allows only for one or two massive hardware manufacturers putting out a homogenous product line of inflexible PC devices. Assuming a free market, the minute that happened, more hardware companies would spring up to fill the demand for pluggable components.
Consciousness is not what it thinks it is
Thought exists only as an abstraction
I don't forsee a time when you will be unable to upgrade. Sure, your $300 computer won't have an AGP slot and might only have 1 PCI slot, but the mid to high-end computers will still have room for expansion. Why? Because there are other things besides upgrades. Are they going to put the modem, nic, tv tuner, video capture card, SCSII controller, and the next great invention all on the CPU? I doubt it. There will always be specialty expansions that not everyone wants, but enough people will to keep the slots around.
Using Microsoft software is like having unprotect sex.
Bite the hand.
What is the difference between a $400 PC and a $200 Dreamcast? Not much. The PSII will blur the lines even more, and the next generation Nintendo will erase the lines altogether.
Give that Anonymous Coward a cigar! :)
Game consoles have been tremendously sucessful at penetrating the home market, especially compared with the PC. That's because of the low cost and the ease-of-use you get when there's zero configurability. They are also marketed to a single task: Playing games.
Most (l)users don't understand the concept of a general-purpose device. Describe a Turing machine to most people, then tell them that on a fundamental level that's all a computer is, and they'll look at you like you're nuts. Even telling people that Nintendos and PCs are based on the same technology is likely to get a few blank stares.
Sell something as an appliance and people will pay hundreds of dollars for it, even though it only performs one task. Try to sell someone one of many similar devices, with prices all over the four-figure range, none of which will do anything unless you buy more stuff (software) to go with it, and you'll confuse the hell out of people.
It's the question of the killer app. For PCs it was the spreadsheet, followed by other office programs. For game consoles it was, well, games. For modems it was the internet.
If you can sell someone the killer app, without also trying to sell them a general-purpose device, you'll have a much easier time marketing it.
Which computer do you think the average person wants to buy...
This one:
PIII-550 MHz, 128MB RAM, 17 GB HD, TNT2 Ultra, 56k Modem
Or this one:
Word Processor, Spreadsheet, Desktop publishing, Games, Internet
The first is just technobabble. The second is something they can actually use. The fact that they may in fact be the same machine (or not) is beside the point.
If the big companies start integrating everything, and selling the applications, especially "Internet", people will be more likely to buy them than some part list.
"It's what BigCo is selling, so it must be a good computer! It does everything we need. And it only cost us $399!"
come on, realistically, companies like cheap
A couple places I know of have a "Technicians shall not open the case" policy. Even though the computers are standard stuff, the cases might as well as be welded shut.
Thinking about this, it made sense. I suppose they're paying the body-shop $75/hour for the desktop techs, it's just a dumb move blow 2-3 hours (a few hundred bucks ~= new elcheap PC) of labor trying upgrade an older PC that's only going to be replaced in a year or so.
--
Business. Numbers. Money. People. Computer World.
Not many people know this but Seiko Epson have developed prototype computers that are simply one piece of glass! All the semiconductor cores and other circuitry where etched into the reverse side of the LCD glass. (We're talkin' severly low power consumption here folks) Currently the only one product shipping using this technology, Seiko Epson's Rupiter, a DOS based wrist PC. Seiko Epson has also developed but never released a flat panel color LCD computer with their 133MHz system-on-a- chip core, memory, and other necessary eletronics all on the reverse side of the LCD glass. Seiko Epson says that the yields right were too crappy to release it but are continuing development, this time on the reverse side of plastic LCDs. Some other records in integration Seiko Epson holds are for its Card-PC (same as MicroModule systems) and their recently released (but only in Japan) GPS handheld PC (a tad larger than a classic M$ PalmPC but with detailed maps of Japan and wireless web and email access). Can't wait to see what they do next.
- Huang Bao Lin
Is natural, especially since everyone wants a palm sized(or smaller) computer that can run all of their favorite apps. Smaller is better in this market, and to get small you need to get integrated. Perhaps this is just one small step in the direction of Human, Computer integration. First, the CPU itself was made of several seperate boards, then the boards became a chip. The FPU was then integrated onto the chip, now it is time to integrate the whole system on the chip. Soon, the chip will integrate into the operator(imagine the death of the keyboard). Just imagine having "Intel Inside" stamped on your head, remember the "simpsons" commercial for Intel. Maybe it wasn't meant to be funny.
"Excuse me, I seem to have suffered a fatal exception, could you reboot me?"
-Master Switch, one more element in the machine
I think there's a difference there; to integrate similar functionallity, or dissimilar ones. To integrate the cache, the DMA and interrupt controllers is another thing than integrating a network och grafic card, into the processor. The first is good, the second is evil while it leads to a uppgrade-all-or-nothing case. Anyway, if this comes true, and NIC is a part of the processor/Chip Set/Graphic card, how do I ever add one additional NIC to build a firewall/bridge?
--The knowledge that you are an idiot, is what distinguishes you from one.
Well put. If Michael feels like a sequel, a wee piece on the social aspects of pcoac would be well worth reading. If the author doesn't want to handle it, I'm sure someone else will.
One thing...
In the long run the motherboard manufacturers will do fine
I respectfully disagree. In the short term they may well survive, but the market for mobos is, I fear, about a decade away from extinction.
Elucidation will follow if anyone gives a shit about my opinion (not recommended).
>The integration has two primary intended effects.
;-).
>One, the card, board or CPU in question gains the
>added feature that makes it more attractive to
>customers. Two, the use of less chips saves on
>system cost and power consumption.
I believe there's a third, even more important feature gained by system-on-a-chip integration: consumer level reliability and configuration (i.e., none).
Dramatic increases in system integration is one of Bill Joy's (of Sun, Java) favorite angles on where the computer industry is coming from and what lies in the future. (I may be missing some of his points but basically) he suggests that the incredible economic leaps in technology have occurred around the introduction of the CPU (brain on a chip) and the motherboard (computer system on a single board). He predicts the next huge gain in cost and reliability to occur with the system on a chip (think PicoJava chip references here
Anyway, the integration is definitely a major factor in computing progress, but the key social impact will be from the high reliability and freedom from configuration issues due to having complete systems on a chip.
Cheers,
= Joe =
P.s., Bill Joy is a great speaker if you ever get a chance to attend one of his presentations!-)
FYI, here's a Bill Joy talk in which he references
o y.html
"Systems on a Chip":
http://www.javasoft.com/features/1999/07/bill.j
Cheers,
= Joe =
>The integration has two primary intended effects.
;-).
>One, the card, board or CPU in question gains the
>added feature that makes it more attractive to >customers. Two, the use of less chips saves on >system cost and power consumption.
I believe there's a third, even more important feature gained by system-on-a-chip integration: consumer level reliability and configuration (i.e., none).
Dramatic increases in system integration is one of Bill Joy's (of Sun, Java) favorite angles on where the computer industry is coming from and what lies in the future. (I may be missing some of his points but basically) he suggests that the incredible economic leaps in technology have occurred around the introduction of the CPU (brain on a chip) and the motherboard (computer system on a single board). He predicts the next huge gain in cost and reliability to occur with the system on a chip (think PicoJava chip references here
Anyway, the integration is definitely a major factor in computing progress, but the key social impact will be from the high reliability and freedom from configuration issues due to having complete systems on a chip.
Cheers,
= Joe =
P.s., Bill Joy is a great speaker if you ever get a chance to attend one of his presentations!-)
> Unless all expandability moved outside the case. Maybe a really (really really really) high-speed external bus standard will come about. So you will then start seeing external 3D graphics adapters and SCSI adpaters.
Right on! Make everything external, use busses that don't have wimpy limits on the number of devices attached to them (ahem ISA, PCI, and SCSI) , and you'll have wonderful expandability. As someone who never has enough card slots or drive bays (let alone IRQs!) in a case to run all my peripherals at once, I really look forward to the day when this will happen.
Maybe Firewire/ISO-1394 will be acceptable for this purpose in a few generations, when they've upped the transfer rate some more.
Lots of benefits on the dumb consumer end too... Can sell smaller boxes because end users won't need all the gimicks. Virtually eliminates installation and upgrade issues (as long as there aren't stupid problems like termination in SCSI). Lowers cost of the cheapest computers, since you only buy what you need. Etc, etc.
-- Div.But my grandest creation, as history will tell,
But my grandest creation, as history will tell,
Was Firefrorefiddle, the Fiend of the Fell.
once you had to buy one turntable, one radio/amplifier, one K7 player all in separate boxes. Soon 3 in 1 boxes start appearing and now with cd we already have 4 in 1 stereos. It also happend to tv/vcr, some models integrate the two.
But you still can buy every peace in separate. I guess the same will happen to computer. And even the ones with integrated stuff could still be upgradable, simply when you connect a external video/sound/etc board the internal could be turned off.
--
"take the red pill and you stay in wonderland and I'll show you how deep the rabitt hole goes"
[]'s Victor Bogado da Silva Lins
^[:wq
Can you imagine a world where everything's on the CPU? "Whoops! That SCSI controller died. Better go buy a new CPU for that server..." That would just suck!
Ah, but your fault tolerant Transmeta CPU will circumvent the problem and reprogram the sound circuits to take care of SCSI.
I read all these comments on how everyone one wants to be able to buy the latest sound card and not have to upgrade the entire computer. How much do you pay for a decent sound card? $150 US? How much do you pay for a decent processor $150 US?
The logistics of seperately making and then integrating all of the items in a modern PC drives the cost way up. Parts have to be placed on printed circuit boards (which are fairly expensive to make), then the whole shebang heated to flow the solder (which is basically lead. Have you seen the environmental regulations for using hazardous materials like lead?), all the while the slightest bump will completely destroy the board. Expensive connectory have to be attached (check you favorite electronics shop for the cost of and ISA slot connector).
If all of the components were in the same IC, possibly even from the same die, cost will take a nosedive, because the manufacturing process is ONE step--make the IC. So you upgrade your sound? You get a faster processor, better network card and scsi controller, and the lastest holographic generator to boot. All for the same price that you would put out for a sound card today.
Now, tell me, how can that be bad?
Aah, change is good. -- Rafiki
Yeah, but it ain't easy. -- Simba
One minor point I haven't seen mentioned is this: properly packaged and cooled, the chip itself is much less likely to croak than it's external connections. More integration means fewer external connections.
Tim
"Whoops! That FPU died. Better go buy a new CPU for that server..."
Increasing integration is nothing new, RAM, FPU, cache controllers and DMA controllers all used to be seperate, and are now included on the CPU.
We are already seeing some CPU/Chipset integration. For portable systems, Intel has their Mobile Dixon and Deschutes Processor Packages which have the 82371EB PIIX4E PCI chipset residing on the same card as the processor, which then connects to the motherboard. Now this is not a true one chip solution, but if it crossed over to the desktop market we could see motherboards that do not have these chipsets on them.
Neomagic makes a video/audio combination chipset also found in portable systems. It is the NM2200 chipset. Now this chipset is not a super speedy 3D solution, or a wizbang audio solution, but for corporate users, it does all you need.
Across the Enterprise, integrated components make more sense; they are less expensive, and easier to implement with standardization of drivers, and add the ability to push a standard software image to all the systems, and not have to worry about who has what video card. If there is a problem with the component, replace the motherboard and go on with your life. If you purchase your systems from a descent OEM, you will have a 3 year warranty. Sure, if the motherboard goes out after 3 years, it is not cheap to replace, but most large companies turn over their systems by then.
At home, I prefer non integrated components. I like being able to upgrade my system component by component as faster/bigger/cooler components come out. Non-integrated components are not going to go away, but they will be used by the consumer, not by business.
But then you have to deal with the fact that there will be a LOT more cables around your computer/desk/floor, and won't that be a safety/stability hazard? Trip on something, and not only do you break your neck, but you also disrupted your new Firewire HDs operation. Whoops. There goes everything you worked on for the past few weeks! I wouldn't mind a fast firewire drive, but i'd prefer it to be inside a protective, stable chassis. Need drive bays? Be a real geek and get a full tower :)
I agree with this; I have issues with integration.
.)
We here at work buy computers from a company called gateway, maybe you heard of them . . . I have NO PROBLEM with this standard here at my place of employment, it simplifies my life, esp. when there is only 2 other real employees in my department and an inept teenager who got his job because his mom is the president's secretary.
Do I sound bitter? Sorry . . .
Gateway just came out with these computers that are about, oh, 3 or 4 inches thick. The whole computer, monitor and all. Flat screens and integrated componants.
This is a neat concept and all, but do I want them to order one? Fuck no. Maybe one, for me, just to dink with. It's got some slots inside, i've been told (by their salesman, so you take that salt too . .
So if the flatscreen goes out, like this monitor next to me, then we got ourselves a $2000 replacement, assuming the warrenty isn't effective still . . .
I dunno, integration scares me, especially after seeing the innards of a packard bell . . .
still shaking
Dan
Exactly what I'm thinking. Hmm, well, that 4MB video card that came with my computer isn't good enough for Quake 3. I plan to go out and buy a TNT2 Ultra. But in this brave new world of integration, I'd have to get a PIII-550 (do you really think they'd stick a TNT2 on anything less?), a SB-Live!, an even newer (but not much different) 10/100 card... forget it. After buying my new PCOAC, I don't have enough to get Q3.
I doubt this will ever come to servers, though. Graphics and sound (the to big components of integration right now) aren't of much use on a server. I think that most administrators would stay away from integrated controllers, as well- for just the reason you mention.
Personally, I'd never, ever, get integrated components on a motherboard. I wouldn't trust the makers of my motherboard with making my video- I'd rather they spend that time on improving the bus architecture (sp?). I like being able to take the old SB16 ISA card (which I think is plenty good sound) from my old 486 and stick it into the new bare-bones K6-2 350 I just got. It's the same reason I dislike laptops- no facility for upgrading (or rather, very little).
Next thing you know, we'll be getting computers with the case welded shut. Sorry, you can't get inside this box. Oh, you want better video? Well, you can get this new computer for only $500. Nope, no getting a new video card for $150. After that, you'll get a solid-state black box with a slot for whatever removable media is the standard. Or better yet, just stick it on the bottom of the monitor! That way, you can't even get that 17" you've had your eye on without also getting that extra baggage they call "integration."
I hope this day never, ever, comes. I'll fight this kind of "integration" to the end. It's almost like MS "integrating" IE into Windows.
-Matt Stegmanmas9483@ksu.edu
-Matthead
"Whoops! That FPU died. Better go buy a new CPU for that server..."
Though I may counter some of original gripes by saying this, there are some integrations that make good sense. Moving things like the FPU and cache onto the same die as the CPU (or into the CPU itself) can certainly help performance.
However, wouldn't you loose the performance gains of cards that can think on their own? Maybe I'm out of date (or out of clues) but can you get "bus mastering" SCSI card benefits if everything's in the CPU?
And this leads to another subtlety. In the context of the original article, does "integration" equate to "on the die" or "in the CPU"? Does it really matter, and what are the pros and cons of each?
I saw this integration trend coming a while back. I used to tell my buddies, "Some day, we'll just have a motherboard with one huge socket. And we'll be using CPU's this big." as I held my hands up to approximate a waffle-sized square. They would chuckle at me.
I see this for the masses. However, I can't see any power user liking this. Take gamers, for example. Aren't most high-end sound and 3D chipsets designed by smaller independent companies? I don't think Intel could put out a high-quality graphics capable board.
Wouldn't this leave a small niche for traditional motherboards and componnets? We might pay a slightly higher price, but they would still be around. Wouldn't they?
(Please tell me they would be!)
Can you imagine a world where everything's on the CPU? "Whoops! That SCSI controller died. Better go buy a new CPU for that server..." That would just suck!
Unless all expandability moved outside the case. Maybe a really (really really really) high-speed external bus standard will come about. So you will then start seeing external 3D graphics adapters and SCSI adpaters.
Hmmm.....
At my previous job (large multinational, 400k+ employees), we were attempting to produce a DVD-ROM system-on-a-chip, integrating the CPU, decoder, servo controller, (some) memory, etc. Even using the a 0.18u process on our own fabs, it always came out cheaper to do 2+ chips.
You see, the cost of a chip this size (some of these integrated chips are huge) is largely driven by the yield, and yield goes down exponentially above a certain size. The bigger the chip, the more bad chips your're gonna make.
Our conclusion was that integration only makes sense if you HAVE to do it, for whatever reason (performance, pin out, ram size constraints), but it always turns out more expensive.
Ok, I have to add my 2 cents to this issue. I work for a major pc service company and I see way too many integrated, piece of junk, off the shelf pc's these days. I thought integration was dying once people realized packard bells suck, but I was way off the mark. I actually replace motherboards with pentium 90's and 100's stuck in them because the video card, sound card, ad nauseum went bad. It is incomprehensible to me why this is done to begin with, but its only because of a simple card that it is being replaced anyway. I've also noticed that most people react negatively toward the manufacturer when I tell them that if this was a "component" system this would have taken about 3 minutes to fix. I love the look on their faces when they ask about upgrading. "You can't really upgrade this machine sir..." What grand emotion is evoked by those simple words. "I paid $3000 for this thing 3 years ago, and you tell me I can't upgrade???" No. And AOL is so simple, it turns people into idiots. Enough of my bitching...