Why Apple is betting on Light Peak with Intel: a love story - Page 3

This is an interconnect protocol, and physical system. It doesn't matter what information moves over it.

Remember that when you are talking about digital, it's all 1's and 0's. That's really about it.

What this does is to provide a very fast way of moving those 1's and 0's across a cable to somewhere else.

As far as the protocols for whatever is being moved, they will remain the same.

USB will still be USB at the bus level, and so will FW, SATA

That's the beauty of this.

If another protocol is invented meanwhile, this can be used to transport it. That's never been the case before.

So if at first, only USB and FW are moving through this, but later, SATA is desired, it can be done. as can any other protocol being used.

There are several levels to all transport types. One level is the very low level software protocols used for basic interfacing with whatever it functions with, such as the software in the computer bus. Then there is the level that is the standard, whatever that may be, FW, USB etc. Then there is the physical level, which is the actual components and their interfaces with the hardware.

This appears to have the lowest level and the physical level, but leaves the middle level to be used by the various standards.

So that is an interesting story, but it reads WAY to far into the corporate relationships. I hate to break it to you, but while Intel and Apple have a good working relationship, there is no way either one of them is going to far out of the way for the other. Lets read this for what it is.

Apple's side: They want new and innovative form factors. For anyone who has ever designed a PCB, you know that port connectors suck. They take up lots of space, and severely limit board layout and dimensions, and they are expensive. If you can eliminate connectors, board get smaller cheaper and more flexible. If Apple can eliminate ports without eliminating functionality, that is a huge win for future industrial designs.

Intel: They get in at the ground level of a new interconnect, and they already have an industry backer lined up to used it. What this means for them is simple, chip sales. Intel makes a host chip or chipset that uses Lightpeak, and suddenly everyone wants Lightpeak and they are selling chips like gangbusters.

So no, this isn't an elaborate love story of companies developing over years. It is a unique opportunity where market conditions have evolved to give an Apple/Intel team an edge. Apple leverages their market leading design skill, and Intel leverages their market leading chipset skills. Both parties win with the support of the other.

All this stuff about legacy ports, Atom processes, ARM, etc is all unrelated. Business is about opportunities moving forward not those from the past.

-JM

My Quadra 950 from mid 1992 had Ethernet. It was, I think, the first computer other than workstations and others of their ilk to get Ethernet as a standard feature.

Other models began to get Ethernet over the next few years until 1997 when virtually every Mac had it.

This was well ahead of the PC world where it took until later for a substantial number to get it.

The problem with PCI adoption on the PC was that it was done improperly. Because so many people still had 8 bit model boards and other old cheap junk that they didn't want to throw away, most manufacturers kept building machines with the obsolete ISA bus for years afterwards in the same machine. In the PC world, there is too much concern with backwards compatibility. It's one reason why PC's have so many problems.

This caused all sorts of difficulty.

One of the problems the PCI bus was supposed to ameliorate was the problems of "interrupts", which anyone here who used those older PC's is familiar with.

Interrupts, for those who don't know what they are, was an old primitive scheme for plug-in cards. The old ISA bus needed to have each card use its own interrupt, so that led to many problems where more than one card wanted the same one. You could change which one cards used, but there was a limited number available. When you ran out, well, something had to give. It caused many problems. PCI does it differently, so there aren't those problems.

But when a new PCI machine also had ISA slots, there were problems that affected the entire machine.

Also, for those who remember, MS was trying to emulate the Mac - plug it in, and it just works. So they came out with "Plug-N-Play". That worked only with the new PCI machines.

But if you had ISA slots, and just one card plugged in the ISA, well, because of the interrupts, and timing problems, Plug-N-Play wouldn't work.

Later on, in an interview, Gates was asked about Plug-N-Play, and his response was that; "It didn't work as well as we thought it would."

When Apple moved from the NuBus to PCI, there was speculation as to whether Apple would include any NuBus slots. They didn't. So while for about a year after the changeover there were fewer cards for the new Macs, there were none of the headaches the PC industry experienced for years afterwards.

There's nothing like making a clean break.

Actually, we DO like reading about other people's life stories. That's how we become a community of people, rather than just a big bunch of screen names.

When you know more about someone, you can understand why they feel the way they do. That's important when interpreting someone's remarks.

You know, like your friends.

It's called replying to someone else who thinks Dan made up all the tech history in his articles.

Care to revise the history of Microsoft with me? I had a person say that I was drinking Kool-Aid for believing the lies of Apple of how Win98 crashed all the time. That loser probably is only 12 years old and never personally used Win98.

Thank you for saving me the trouble.

You're kidding right?

Apple went to the PowerPC because the 68060 did not live up to expectations and they went Intel when the dedicated chip market made the PowerPC a supply boon doggle.

I should point out that well written programs take a lot to break. World Builder a 1986 program case in point; the program didn't totally break until the switch to Intel in 2006 some 20 years later. Hypercard was another old kept going until the end program of similar age.

The real reason most programs break is not because CPU changes but because so many programmers are sloppy about writing programs, take risky shortcuts, or both.

Thanks, that clarifies a lot of my questions. It sounded like that was the case in some of the discussions, but it was hard for me to tell for sure. Does that mean that the new Intel controller chip at each end of the line has enough smarts to do all the emulating of other protocols? I assume that would be necessary to emulate more sophisticated connections like FW with its peer-to-peer architecture?

(I think you were actually answering my other post though. Any thoughts on the possibility that this system has the optical encoder/decoder built into the plug to avoid the user making & breaking optical connections?)

Fascinating. A couple of years ago a Verizon FIOS installation tech told me that the biggest remaining bottleneck they were trying to solve was getting fiber inside the home. He made it sound like the biggest problem with that was the limited bend radius of fiber which would cause a lot of difficulty for home users to handle themselves and which would in turn mean a lot of botched installations and expense. If they could find a break through on the cable flexibility they expected fiber in the home to become practical and open up the possibility of significant speed improvements. If what you say about the Corning break through is part of this, then it might be a very big deal indeed.

It also sounded like the other major hang up was the delicacy of optical connections as you mentioned. It looks to me from the Intel web page that maybe this Light Peak system keeps the optical connections within the cable plugs such that they are made at the factory and the user only makes traditional copper connections. If so, that would seem to solve the connection problem quite nicely. Maybe I am just imagining that though, any thoughts?

The need for fiber in the home for significant improvements to FIOS was told to me by a FIOS installation tech a couple of years ago. He said there was some room for improvement in FIOS speed with the current system, but to really reach its full potential they were working on practical ways to bring fiber from the network interface terminal on the outside of the house all the way to each router, TV interface box, etc. The costs and difficulties for home users to deal with optical connections and the stiffness of fibers seemed to be the main problems. If this Light Peak really has solved those problems, then it would be just the break through needed. If it can even carry an ethernet signal, then there is no reason to stop the fiber at the router - just go all the way to each computer.

Not for me. Wireless may be nice for some things, but wired will always be faster and more reliable.

I haven't yet read Intel's specs. But their interface is involved in getting the light back into electrical form. Whether they have decided to add controller chips for any protocols, I don't know.

FIOS is nowhere near providing any kind of speed that would saturate in-house CAT5e/6, there's no reason to drop the copper until they do. There may be the untapped bandwidth, but the limits of in-house wiring is not why they aren't giving you a faster link.

Well all I can say for sure is that the current connectors are all-optical end-to-end. This is a problem because it's very easy to get reflections off the end of the cable if they don't meet really nice and flat. That's why TOSLINK cables are so tightly connected.

In the case of the computer cabling I've used it's typically two fibres ending in this big plastic block. The connector is surprisingly large, oddly shaped and flimsy. It's terrible industrial design. I can imagine a dual-TOSLINK connector though, one that would fit firmly enough and still be easy to work with. In fact, it would seem the existing FW connectors would have all the right qualities.

BTW, you know where the FW connector came from? It's adapted from the Game Boy. Yes, that's true, right from Michael Teener's mouth.

Maury

I couldn't agree more. Sadly, however, this article contains nothing of the sort. Do you see any "context" other than bland generalities about unrelated topics, or "history" about anything related to THIS topic? I see neither. After three pages we're left with a single block of text about the actual topic, one that merely regurgitates the existing press feed. Can you find a single statement about Light Peak that isn't copied from another source?

If I want the Wikipedia, I'll go to the Wikipedia. At least that tends to be on-topic (I should know, I probably wrote it).

Maury

Finally found a useful image:

http://www.engadget.com/2009/09/26/e...ntel-could-be/

Note that the connector shown in the Intel press documents is visible on the inside of the case, just to the left of the Apple logo (pasted on top). The controller has several fibres leading from it, and one of those curls around into the topomost of the four ports. That port is also connected to a small PCB that clearly also includes power.

So I think it's safe to say this is light-to-the-motherboard, AND supplies power. Looks like an adaptation of the FW connector, too.

Maury

If there's a power cord, there might as well be a Light Peak optical fiber inside it, with the whole thing connecting the laptop to the monitor.

It's apparently an adaption of the USB connector.

Sweeter words have never been spoken!

It should be noted that interrupts and cad addresses aren't in and of themselves big deals, but the users had to be capable of understanding configuration issues. An engineer can easily get such cards up and running on a specific system. The problem is systems change and users really don't have the background.

As to Apple I suspect that one reason we see a limited number of machines supporting slots, from them, is that slots add to warranty issues. Like it or not if you leave the slots out of a consumer grade machine there are far fewer ways for the user to screw up his machine. Especially with rugged I/O through connectors like USB.

Dave

The other problem PCs had with interrupts was that there weren't enough of them. The way Apple handled it solved the problem.

We constantly get this call for slots, but people rarely use them when they have them. The consumer Macs are being aimed at a group that never used the slots.

Apple knows that they're not going to get 70% of the PC business, so they can afford to ignore those who want slots. There are a vast sea of people out there who will move to a Mac, or not, solely because of the OS, and the way the machine looks.

This is true for people who are demanding a higher performing machine.

There's a bit of a limit though. With these USB sticks for memory or networking that stick out 2", that makes me very nervous, so I end up feeling like I have to go cautiously, like stepping around broken glass. The length basically acts like a lever.

Good points. Personally speaking, the only reason I would upgrade a current machine is to upgrade the graphics card, and I think I'm more an exception than a rule.

Some of those cards never go away. I still work on and configure machines at work using ISA cards.

I agree with the idea that they are being aimed at consumers but the very lack of slots eliminates them from use in non consummer applications. Even with slots the Mac Pro is way to large for many commercial applications.

A slot or two may mean very little for a machine marketed to consummers but it is often required for commercial sales. Even in a commercial environment the slots don't always get used but the people maintaining the machine ls like to keep the mix of hardware to a minimal. Thus swapping around computers often is just a software change and sometimes a card change or two.

In any event the lack of a slot takes Apple out of the running for many organizations as they see it as a requirement. Since effectively there is little difference between a consummer machine and a corporate machine the lack of a low cost slotted machine is a huge problem for Apple when it comes to commercial sales.

Yes this is true from the standpoint of the consummer. It is not however the case when it comes to winning corporate sales. Right or wrong many expect a slot or two

Would Apple win those sales if it had such hardware. I can't say but there was a lot if frustration with MS that could have opened more doors if Apple had an economical model. With Windows 7 that door will Close a little bit.

What is true? That people will move to Apple hardware if the performance is terrible? I think not. In fact Windows 7 running on modern eight thread processors could easily revers that trend because the one thing people won't put up with is really bad performance. Especially when in Snow Leopard we have an OS that was explicitly desgned to leverage the new generation of processors. If Apple can't deliver hardware to leverage that software they are screwed. Why deal with a company that delivers half solutions?



Dave

Very few people care about slots anymore. Even audio guys have moved on.
Corporations love the whole 'sealed box' approach, since it minimizes support costs.

The only good reason to buy a consumer/prosumer machine with slots in 2009 is to get Light Peak in 2010.

but people / Corporations want to have there own screen.

and the mini is too weak / over priced and the mac pro is over the top / over priced as well.

and the you need a video card for many things as well and I don't see taking over pci-e bus for stuff on board.

What do most people in a corporation do? Sure, there will be some that need powerful computers. However, outside of IT, engineering and the creative media people, it looks to me that most people just need to run Office, email and web browsing type tasks. Some jobs only need a virtual TTY terminal, which only requires all the processing power of a freaking graphing calculator.

Heck, I just retired a dual 500MHz Xeon only because the hard drive was acting up, before that, it was doing everything the user needed. A Mac mini could have done what the old machine did just fine, but with far less electricity consumption. The user simply didn't need anything fancy.

I've not denied the desire for a slot from some groups, be they business or not. But as I've also said, Apple can afford to ignore them.

The fact is that Apple is interested in doing what they want to do. They can do that because in those areas, they are apparently doing as well as they want, and expect to.

We can say that there are groups that won't buy the machines because of that, and that's correct. But, so what? If Apple doesn't care, that's that.

I really think that we have to forget this. Apple has a plan for years in advance.

We've been calling for an xMac for years, myself included, if you remember. I even had a design I submitted to my friends there.

Lets not kid ourselves. Everything we've said here has been thought of at Apple. They've very likely done the math.

They aren't interested in pursuing business if business wants them to move from their plan. Business is actually responding. Use in large corporations has more than doubled in the past three years. It's a small number, but it's been increasing.

If business will come to Apple, then they are obviously happy to let it.

Meanwhile 30% of consumer computers bought in the US in the past 100 days have been Mac laptops. If you include the rest, it's likely close to 35%.

That's quite a number. If their revisions will be somewhat cheaper, that number could rise.

What more do people here want? That's success.

Just because people with technical demands aren't getting what they want doesn't mean that Apple is doing the wrong thing.

Since there are pros who do use the iMac line, it shows that it's good enough for much of that segment as well.

I've called for a graphics slot in iMacs since the first white flat models came out. But it is what it is. I've bought two 24" for my family, and they love them.

I'd like to see an i7 model with a couple of slots, but Apple isn't interested, and so that's that. We can argue until the cows come home.

The truth is that businesses don't need slots.

Really, what are they going to put in one?

Business doesn't run 3D software. They don't play 3D games. Most computers are on the desks of secretaries.

The graphics on these models is more that adequate. How much RAM do they need? Not much. HDD's? Big enough.

It's a myth about what business NEEDS. Yes, what some of them WANT is a different story. And that's really just the IT department.

But the truth is that the hardware is less important than the OS. It's the OS that's holding business back.

When all the software they need is available for the Mac, and the backend operations can run on Macs with the home written applications they use, then switching will be easier, and more will do it.

But until then, it will be a struggle. Only in the past two years have we seen some significant movement from major software makers for business, moving their apps over. This includes heavy duty network management software and the like.

But those very big apps that have been developed by IT staff over time isn't going to be ported over any time soon. And for many large corporations, that's the straw that breaks the camel's back. Without that, many operations can't be done on Macs.

Virtualization software, and Bootcamp are not enough for those companies.

but the mini is a bad buy for it's hardware at it's price and only 1gb of ram??

also the hdd need to be easier to get to. I don't corporations sending out a hd that may have data on it to some out side corporation to get fixed.

The real advantage of lightpeak:

http://www.ranum.com/security/comput...polis-2005.pdf

Slide 33...

"— If PCs for example, had 1) hard drive interface (instead of IDE, SATA, SCSI, etc) 1) video interface and 1) network interface we could reduce kernel code size of Linux by 30%"

Motherboard complexity is related.

Right now you might have a case that can have up to four drives. Maybe a user will want three DVD's and a single big HD, or maybe three HD's and a a single DVD, etc. Since you don't know, you need to put on four of each connector, for a total of eight. That's in spite of the fact that you can only physically add four drives.

With a new connector that runs both, you need four connectors total. So the MB space used up goes down. And since the new connector is much smaller, it goes down even more. So now you start to get some real space savings in the case.

Maury

I thought the kernel was modular so you only load in the code that's needed.