Dumping the contents of the ROM chip into ultra slow dynamic RAM instead of swapping the slower ROM chips they were using at the time for speedy static memory(Which. by the way. is seldom used in large amounts due to the fact of that it costs so much more than dynamic memory. but the fact of that the Mac ROM code is only about 6-8MBs makes static memory an excellant choice for small, frequently accessed data like is on most ROMs) chips. like the SRAM used in most CPUs as cache. was probably one of the worst moves Apple has ever done.
</strong><hr></blockquote>
You honestly suggest Apple should have rather put 6-8MB (!) of SRAM onto their mainboards? This would have been enormously expensive. So I wonder how exaclty shadowing the slow ROMs in much faster RAM was such a bad idea.
[quote]<strong>And if you're refering to switching to a standard motherboard chipset across all lines. I think the PC industry has been doing that since about the day there were clones.
</strong><hr></blockquote>
Huh? They aren't even doing this right now. In fact, as of now, most vendors even use proprietary interconnects between their chipset components (HubLink, V-Link, MuTIOL), although this will probably change with the widespread adoption of HT or 3GIO.
[quote]<strong>
Yuck. DVD-R. like CD-R before it. is about the most user unfriendly storage medium ever invented. Apple already had DVD-RAM drives in their G4s.</strong><hr></blockquote>
While I wholeheartedly agree in preferring DVD-RAM, DVD-R is what the most customers seem to be demanding.
[quote]<strong>which. aside from their far superior hard disc like user experience. also costed roughly $750 less than Apple's DVD-R drive at the time. and currently cost $250 less</strong><hr></blockquote>
The media costs more, though, IIRC.
[quote]<strong>
If you were to swap a modern G4's ZIF slot for a Pentium's Socket7.</strong><hr></blockquote>
G4s don't use ZIF sockets any more (and actually, Socket/ is a ZIF socket).
[quote]<strong>and switch Apple's ROM chip for a Wintel BIOS chip. there wouldn't be so much as _one_ functional difference between it and an equivalent Wintel motherboard. quite truly.</strong><hr></blockquote>
Well, since both are mainboard after all, there can hardly be very much variation in the first place.
There are a lot of significant details, though, like GBit ethernet on-board and not on the PCI, same for FireWire, 64bit PCI slots, OpenFirmware rather than real mode BIOS, Target Disk Mode and others.
[quote]<strong>
all that Apple makes now are wintel clones with PPCs and some flavor of UN*X on them.
</strong><hr></blockquote>
You do realize that Apple have their own, proprietary North- and Southbridge? They don't (and couldn't, actually) just go ahead and use chips also found on x86 mainboards.
Rather than spend 2 hours explaining how terribly wrong a certain above post is, can we all just acknowledge that Eric D.V.H. is very, very, wrong about 97% of his post? thanks.
<strong>Rather than spend 2 hours explaining how terribly wrong a certain above post is, can we all just acknowledge that Eric D.V.H. is very, very, wrong about 97% of his post? thanks.
</strong><hr></blockquote>
But I just can't resist.
[quote]Originally posted by Eric D.V.H.
<strong>That's just what they want you to think. in this Api NetWorks/Sibyte whitepaper and this Api NetWorks executive summary. they say that: "The 50-ohm impedance and differential signaling also permit trace lengths up to 24 inches, and they span board interconnects well." meaning that HyperTransport works perfectly well across the shattered pins and traces at the bottom of a PCB card.
They also said that: "There is an ongoing effort with major connector companies to define a variety of connectors that can fit different price/performance parameters. More information on this topic will be available in 2001."
And in a recent AMD whitepaper. they discussed Plug ?n Play in it's session layer.
The final clincher though. is on AMD's online HyperTransport FAQ. where they more cautiously echo the precise words of the Api NetWorks/Sibyte pieces by saying: "A HyperTransport connector has not been endorsed at this time; however, some members of the Consortium are exploring possible use of connectors for different applications."
So why are they pretending that HyperTransport is an intra-board-only standard? darned if I know.</strong><hr></blockquote>
Yawn. PCI-X is here already. 3GIO is around the corner. All you've listed here are ideas that exist only on paper or at best are buried deep in some geeks lab. Do you think this stuff will just pop up overnight? Will these technologies even work, or will impassable problems arise? Can the companies that are creating these designs sucessfully get them to market? Considering all that still needs to be done, then how long will they take to reach the marketplace?
So what you're saying is that Apple should pass on technologies that are in the very near future just to implement stuff that merely exists on paper.
What are you <img src="graemlins/smokin.gif" border="0" alt="[Chilling]" /> , cause I'll try to avoid that batch?
<strong>At this point, I'd rather they did adopt a Wintel-clone motherboard. Their current one is so outdated it's ludicrous.</strong><hr></blockquote>
That's why Apple should use HyperTransport as soon as possible. and before they do.
[quote]Originally posted by TJM:
<strong>Calm down for a moment, and think a bit. What is it that makes a Mac? For me, it's the ergonomics of the hardware and the software, particularly the Mac OS. Does it really matter what's inside the box? Would it run any differently if it had a solid gold case and platinum wiring? It's Apple's attention to detail in the system integration, their industrial design for the cases, and the ease-of-use in the software that make the quality of it. The technology used doesn't really matter. As long as it produces my beloved "Mac experience" they can have gerbils in a treadmill running it, for all I care.</strong><hr></blockquote>
You forget that revolutions like the Macintosh are only possible by using totally new software [i]and[i] hardware. in order to make a uniquely better machine. you need to use uniquely better components.
[quote]Originally posted by TJM:
<strong>It's one thing to use off-beat technology just to "Think Different" when it is superior or equivalent, but much of the Mac's technology at that point was out-of-date and orphaned (i.e. NuBus and ADB). ATI, 3dfx, and nVidia would never have come out with Mac video cards if they were still running NuBus. Thanks to PCI (then AGP) we have access to first-rate video. The list goes on and on.</strong><hr></blockquote>
Ugh. I wasn't implying that Apple should have remained frozen in the 80s . I was saying that. in their prime. those technologies totally outclassed nearly every other solution on the market _at the time_. there were obviously vastly superior. competing solutions back then too. read "Apple bets on Intel's bus". in a circa 1994 issue of MacUser magazine.
[quote]Originally posted by TJM:
<strong>As a "fer instance" on the relative costs, a couple years ago I bought an ATA card for my 7600 because I just couldn't see paying $350 for a 10 GB SCSI drive when I could get a 50 GB ATA drive for $120. Even after $100 for the ATA card, I was still over $100 ahead with 5x capacity. I can get memory for my G3 desktop dirt cheap because it uses PC100.</strong><hr></blockquote>
Yes. but your(And my) machine aren't several times better. hands down. than the best IBM PC on the market. which the Mac used to be back when it used it's own. superior standards.
[quote]Originally posted by TJM:
<strong>So, I appreciate your sentiment - I like Macs because they ARE different . But if different = inferior or different = outrageously expensive (with no clear benefit), you're going to have a hard time broadening your appeal or even staying in business.</strong><hr></blockquote>
Exactly. that's the reason Apple's market share is so low now. Mac sites all over the web are proclaiming that it "Doesn't really matter which platform you use. just whichever you like". they never would have said that back then.
Times have changed and there is no way a company could possibly reinvent the personal computer in both hardware and software as Apple did back in 1984.
Building a platform today is made by taking the best standards around and implementing them as good as possible, not buy spending a few billion dollars on a proprietary design that eventually nobody else will use or support. back in the 80s a garage company could do these things, simply because there were different market and industry situations. Today that is only possible in software anymore.
The first example is probably the adoption of PCI back in 1995 in the PM 9500. that was already about 1 or evne 2 years behind the PC, and Apple chose PCI because it was: superior to Nubus (which was superior to ISA admittably). Instead they could have developped NuBud further and gotten more performance out of it, but they didn't because: It would have cost a lot of R&D and it would have brought with it the same problems that NuBus already had, namely that only very few select companies made products for it.
So yes, PCI was the only reasonable choice back then, and gladly Apple didn't spoil it.
I can tell you one thing: without PCI, we'd be in Atari land today. they kept using old hardware until they were hopelessly lost and now they're dead. But their OS and their machines were among the very best in the mid to late 80s.
Right now there are various new technologies emerging, and as I said, the best thing Apple can do is stay updated and be ready for release of a product incorporating those new techs as soon as they're generally accepted and supported.
Times have changed and there is no way a company could possibly reinvent the personal computer in both hardware and software as Apple did back in 1984.
Building a platform today is made by taking the best standards around and implementing them as good as possible, not buy spending a few billion dollars on a proprietary design that eventually nobody else will use or support. back in the 80s a garage company could do these things, simply because there were different market and industry situations. Today that is only possible in software anymore.</strong><hr></blockquote>
From my other post:
Actually. HyperTransport is backwards compatible too. as you can drive multiple PCI, Infiniband etc. buses. as well as an AGP bus. with a HyperTransport based system. I would also expect a HyperTransport slot to PCI slot adapter. much like the PDS to NuBus adapters of yore.
[quote]Originally posted by G-News:
<strong>The first example is probably the adoption of PCI back in 1995 in the PM 9500. that was already about 1 or evne 2 years behind the PC, and Apple chose PCI because it was: superior to Nubus (which was superior to ISA admittably). Instead they could have developped NuBud further and gotten more performance out of it, but they didn't because: It would have cost a lot of R&D and it would have brought with it the same problems that NuBus already had, namely that only very few select companies made products for it.
So yes, PCI was the only reasonable choice back then, and gladly Apple didn't spoil it.
I can tell you one thing: without PCI, we'd be in Atari land today. they kept using old hardware until they were hopelessly lost and now they're dead. But their OS and their machines were among the very best in the mid to late 80s.</strong><hr></blockquote>
Atari? were you thinking of Amiga? they're still keeping a sort of market in europe. 2-3%. just like Apple in the US. so I guess we are in _Amiga_ land today.
[quote]Originally posted by G-News:
<strong>Right now there are various new technologies emerging, and as I said, the best thing Apple can do is stay updated and be ready for release of a product incorporating those new techs as soon as they're generally accepted and supported.</strong><hr></blockquote>
From my other post as well:
What if it just lies there and rots like USB or VCD? someone needs to take action. and with everyone else as leadbottomed as they are. that someone is Apple.
<strong> You have to differentiate between the past and the recent past. ADB, NuBus and SCSI were selected because they were superior, and in the case of SCSI, already working on other systems.</strong><hr></blockquote>
Yup.
[quote]Originally posted by G-News:
<strong>NuBus was just another standard that emerged out of the first experimenting phase of modern computer design, as was ADB and GeoPort.</strong><hr></blockquote>
All of these were better than their mainstream counterparts.
[quote]Originally posted by G-News:
<strong>DAV definitely doesn't count as I have yet to see a single product that was ever released to the public for that slot.</strong><hr></blockquote>
<strong>DB-15 is functionally identical to a normal VGA port, that was just a proprietary decision, not a quality over cost one.</strong><hr></blockquote>
Yes. but I was trying to point out that Apple has gone on it's own before.
[quote]Originally posted by G-News:
<strong>And for USB, as you probably know, Apple helped it make a breakthrough, but they didn't invent it (Intel did) nor did they introduce it (most PCs actually had USB about a year before the iMac was originally released, just that nobody knew about it.</strong><hr></blockquote>
Adopting HyperTransport now would be kind of half-way between when Apple got USB and when they got SCSI.
[quote]Originally posted by G-News:
<strong>If you look 20 to 10 years back, you'll see that Apple had to develop and introduce new technology, because there was nothing suitable to meet their needs. In the recent past you'll soon notice that Apple just picked the best standards that were already available on other platforms, combined them and took out the hubbles of the road.
Apple's implementation of PCI is wonderful compared to the PC counterpart (both in terms of performance and software support (for example we don't have to fiddle with IRQs).</strong><hr></blockquote>
I agree. and PCI/AGP just don't look like they'll keep meeting Apple's needs for long.
[quote]Originally posted by G-News:
<strong>Apple hasn't been innovating in hardware in terms of developing itself, but they have been innovative in terms of making things actually usable (example USB and FireWire (partially).</strong><hr></blockquote>
Apple is the one that invented FireWire. and I think it was pretty revolutionary(Too bad they waited so long to use it ).
[quote]Originally posted by G-News:
<strong>If Apple adopted 3GIO or similar now, totally replacing PCI, you'd have a system with a new standard that nobody supports until the PCs adopt it too. The days of small companies making extra hardware for the Mac (ie NuBus falvor cards of their products too) are over, companies no longer an afford extra tours.</strong><hr></blockquote>
Actually. HyperTransport is backwards compatible too. as you can drive multiple PCI, Infiniband etc. buses. as well as an AGP bus. with a HyperTransport based system. I would also expect a HyperTransport slot to PCI slot adapter. much like the PDS to NuBus adapters of yore.
[quote]Originally posted by G-News:
<strong>Wake up, the best thing Apple can do now is stay on top of the emerging standards, experiment with them and use the best ones for their machines, as soon as support from most sides is granted.</strong><hr></blockquote>
What if it just lies there and rots like USB or VCD? someone needs to take action. and with everyone else as leadbottomed as they are. that someone is Apple.
[quote]Originally posted by G-News:
<strong>PCI-X, if downwards compatible, is currently the only option that you could implement right away, without awaiting support. But then again we already have 64bit PCI slots that nobody supports either.</strong><hr></blockquote>
Refer to prior comment.
[quote]Originally posted by G-News:
<strong>Apple is not going to adopt 3GIO before they can be absolutely sure, that the companies that make PCI cards now, will continue making 3GIO cards. And that likely isn't going to happen before Intel hasn't got a chipset out that supports 3GIO.</strong><hr></blockquote>
I can think of one kind of company that would gladly make HyperTransport cards. GPU makers. ATI and nVIDIA are already starting to feel the pain from the Kyro II. and I'll bet that they'll start burning the midnight oil to get out HyperTransport based chipsets and cards once the Kyro III comes onto the market.
[quote]Originally posted by G-News:
<strong>Of all the standard hardware Apple has introduced during the last 5 years, name me ONE that wasn't available on other platforms before.
(FireWire is the only barely legal item)</strong><hr></blockquote>
<strong>I'll be sure to send a note to SJ to remind him to consult you before he makes any more strategic moves.</strong><hr></blockquote>
<img src="graemlins/lol.gif" border="0" alt="[Laughing]" /> Steve Jobs didn't even work at Apple when they switched to PCI. I'm not quite sure. but I think it was that dolt Scully.
[quote]Originally posted by TJM:
<strong>At the time, Apple was losing $100s of millions of dollars per quarter, its market share had already plunged to single digits, and it was rapidly on its way to bankruptcy. Going to PCI and ATA and standard memory (PC66 on Beige G3) saved them a ton of money and helped them survive. It was a matter of change or die. The "cachet" of owning a Mac with its unique technologies wasn't going to last long in any event.</strong><hr></blockquote>
<strong>Yawn. PCI-X is here already. 3GIO is around the corner. All you've listed here are ideas that exist only on paper or at best are buried deep in some geeks lab. Do you think this stuff will just pop up overnight? Will these technologies even work, or will impassable problems arise? Can the companies that are creating these designs sucessfully get them to market? Considering all that still needs to be done, then how long will they take to reach the marketplace?</strong><hr></blockquote>
Read that stuff more carefully. Hyper transport IS HERE NOW. it entered the market TWO YEARS AGO. the problems have already arisen and been vanquished in testing OVER TWO YEARS AGO. people are using it in things they bought from these companies. and they say it works JUST FINE.
[quote]Originally posted by Arty50:
<strong>So what you're saying is that Apple should pass on technologies that are in the very near future just to implement stuff that merely exists on paper. </strong><hr></blockquote>
PCI-X and AGP are set to just barely exit the experimental phase this year. 3GIO/Aprahoe actually _is_ only on paper or at best buried deep in some geeks lab. HyperTransport is already gathering momentum across the industry. and needs only a strong supporter to break the barriers. for since it's already on the market. that means Apple can start right now. or already has.
<strong>You honestly suggest Apple should have rather put 6-8MB (!) of SRAM onto their mainboards? This would have been enormously expensive. So I wonder how exaclty shadowing the slow ROMs in much faster RAM was such a bad idea.</strong><hr></blockquote>
It can't have been _that_ expensive. but even if it was. Apple should have used a faster variety of DRAM. in a separate block. directly connected to the CPU. so as to increase access speed to this oft consulted data.
[quote]Originally posted by RazzFazz:
<strong>Huh? They aren't even doing this right now. In fact, as of now, most vendors even use proprietary interconnects between their chipset components (HubLink, V-Link, MuTIOL), although this will probably change with the widespread adoption of HT or 3GIO.</strong><hr></blockquote>
I didn't mean the PC industry as a whole. I meant individual PC vendors like Dell, Compaq, Gateway etc?
They use a lot of the same parts. and pretty much always have.
[quote]Originally posted by RazzFazz:
<strong>While I wholeheartedly agree in preferring DVD-RAM, DVD-R is what the most customers seem to be demanding.
The media costs more, though, IIRC.</strong><hr></blockquote>
<strong>Well, since both are mainboard after all, there can hardly be very much variation in the first place.
There are a lot of significant details, though, like GBit ethernet on-board and not on the PCI, same for FireWire, 64bit PCI slots, OpenFirmware rather than real mode BIOS, Target Disk Mode and others.</strong><hr></blockquote>
<strong>You do realize that Apple have their own, proprietary North- and Southbridge? They don't (and couldn't, actually) just go ahead and use chips also found on x86 mainboards.</strong><hr></blockquote>
I'm not exactly sure on this one either. so you might be right.
You do know that HyperTransport is a bus technology that is designed to replace systems like the current MaxBus, Intel's and AMD's incarnations of their front side busses etc, right?
It's not aimed at peripheral cards, the fact that it can support line lengths to card degrees, pinnage etc only implies that it's feasible to make slots, similar to Slot A and similar, ie for CPUs and other extremely performance sensitive parts of a computer system.
HyperTransport is just going to change the way a CPU can communicate with its peripherals, however by peripherals you have to understand that these are chips like keyboard controllers, PCI busses etc etc, the stuff we have today on our mainboards, not the stuff we plug INTO the mainboards today.
In that way HyperTransport and RapidIO (which are kind of the same thing from another company, with different target applications) are going to be adopted soon, and I wouldn't be surprised if Apple adopted it as one of the first companies out there. However that has nothing to do with 3GIO that we were talking about in the first place, since 3GIO is like PCI, ISA, NuBus, AGP (which is just a PCI derivate) and PCI-X a so called Peripheral Interconnect Bus, and not a system bus or a CPU bus or anything like that.
You are obviously mixing things that don't belong into the same basket at all.
Also wtf do you mean by talking about the Kyro 2?
The Kyro 2 is:
-a PC only modern class graphics GPU
-exists only on AGP bus, as far as I know, PCI is probably possible
-is based on a tile-based rendering system, unlike the (correct me if I'm wrong) polygon based rendering system of Matrox, ATI and nVidia cards.
-is not incredibly slow and, most important of all:
-the company that made those boards just went belly up a few months ago, that's why you can have a Kyro 2 for like 50$ today.
-Kyro 3 is therefore likely never going to show up.
I have no idea what you're mixing there again, but apart from the rendering engine approach, the Kyro boards are cards like ever other card as well.
Chips and connectors on a PCB plate with gold pins that go into an AGP slot, driven by a 2x or 4x AGP slot. No HyperTransport, no RapidIO, no 3GIO and no nothing like that.
last but not least, the link to your DAV card that you were going to tell me about (although it surely existed, I even know there was one or two products for that slot, just that only about 0.004% of all DAV owners ever used that slot), doesn't work: 404 Object Not Found. I think that speaks for itself.
So if your statement is that Apple should adopt HyperTransport (and, more likely also RapidIO with the G5), then my answer is: they surely will.
If your statement is that Apple should accept 3GIO, then my answer is still: they will, once it's ready and widely supported.
If your statement is that Apple should invent a sort of HyperTransport slot, replacing AGP and PCI, then my answer is: not going to happen anytime soon, neither from Apple nor anyone else.
maybe you have seen this: [quote]Has a board-to-board connector for HyperTransport? been defined?
Answer:
A HyperTransport connector has not been endorsed at this time; however, some members of the Consortium are exploring possible use of connectors for different applications.<hr></blockquote> Thus yes, it's feasible, but then again, it would bring the same problems with it that I already mentioned: a connector without a card is just lost money. Apple is a member of the HT consortium, and they're surely looking at it very closely, have actually already implemented some ideas of it into their current motherboards, but this stuff takes time. Maybe (hopefully) with the G5, we'll see more.
last but not least let me show you this image, right off their docu, for some clarification:
edit: I can't believe how much wrong info you seem to have. The ROM is accessed exactly once, and that's during startup, then everything is in the RAM, which is faster in any way. Since the iMac, we even have the biggest part of the ROM on the harddrive, which is fast and inexpensive. I don't know where you're stuck in tech development, but I know I wasn't even born back then.
[ 03-05-2002: Message edited by: G-News ]
And one last thing:
Based upon my current knowledge, the following pict shows the best possible mainboard architecture for future Macs. I'm not expert in these things, but based on what I know so far, this is roughly feasible. Maybe you meant that.
<strong>You can't be ?SNIPPED?viously mixing things that don't belong into the same basket at all.</strong><hr></blockquote>
It does that too. but it also supports intra-board communications between various devices such as graphics accelerators, networking adapters and audio D/A divices. the fact of that HyperTransport features a full cold boot sequence instead of just having all of that pre-programmed into the ROMs of the motherboard, native PnP functionality and native device support. make it obvious that HyperTransport has had hardware support for local bus style slots from the start. needing little more than for the PCBs to be soddered on.
[quote]Originally posted by G-News:
<strong>Also wtf do you mean by talking about the Kyro 2?</strong><hr></blockquote>
I'll get to that.
[quote]Originally posted by G-News:
<strong>-is based on a tile-based rendering system, unlike the (correct me if I'm wrong) polygon based rendering system of Matrox, ATI and nVidia cards.</strong><hr></blockquote>
It's tile based. but the term "Tile" refers to the bitmapped output. not the internal geometry. it works by cutting each output image into tiles. it then analyzes each tile to see if there are any polygons hidden behind other geometry. and thusly not worth rendering. after going over the whole image. it renders it. but it only renders visible geometry. thus needing less data(Most of all textures) to be transferred in from outside the card.
[quote]Originally posted by G-News:
<strong>-the company that made those boards just went belly up a few months ago, that's why you can have a Kyro 2 for like 50$ today.
-Kyro 3 is therefore likely never going to show up.</strong><hr></blockquote>
They didn't go "belly up". they just <a href="http://us.st.com/stonline/press/news/year2002/c1146h.htm"; target="_blank">pulled out of</a> the Kyro deal. although they [i]are[i] looking for a buyer. and it was the company(<a href="http://us.st.com/stonline/index.htm"; target="_blank">ST Micro</a>) that _manufactured_ them.<a href="http://us.st.com/stonline/index.htm"; target="_blank">Imagine Technologies</a>. the company that designs the Kyro(And it's ancestor. the PowerVR) is still going at full steam(If a bit bruised up). and ST Micro said that they're <a href="http://us.st.com/stonline/press/news/year2002/t1152p.htm"; target="_blank">going in for a second bite</a> anyways. so I think that. once they have their new Kyro III chip. Imagine will find _someone_ to manufacture them.
[quote]Originally posted by G-News:
<strong>I have no idea what you're m?SNIPPED?othing like that.</strong><hr></blockquote>
Eesh. the reason I mentioned the Kyro is because it uses bandwidth more efficiently than other chipsets. why does this make ATI and nVidia shiver in their shanks?
Remember how newer GPUs. when running things with higher levels of detail. especially at lower resolutions. the frame rates seem to "Top out" at the lower resolutions until the level of detail is decreased below a certain amount?
This is due to the fact of that higher levels of detail require more data to be piped into the GPU from outside. while the GPU normally goes at full speed. as it renders the output so slowly that the data for the next frame has enough time to flow in. when the resolution is decreased to the point where the GPU can render frames faster than it can get the data for the next frame. it just sits there and "Idles" between frames. thus making an effective "Ceiling". above which the GPU's power goes to waste.
The Kyro on the other hand. due to the fact of that it generally doesn't need as much data to render the same scene. can use more of it's GPU's power at a given time than ATI and nVidia's chips.
The only way to fix this problem(Aside from work-arounds like the Kyro employs). is to increase the speed and width of the bus connecting the GPU to everything else. unless this happens. Imagine and their Kyro are destined to catch up to(And surpass) ATI and nVidia VERY soon. at which point they'll be in it _deep_.
[quote]Originally posted by G-News:
<strong>last but not least, the link to your DAV card th?SNIP?aks for itself.</strong><hr></blockquote>
Yes. I guess that _everyone_ who did video out were only 0.004% of all DAV owners.
[quote]Originally posted by G-News:
<strong>So if your statement is that Apple should adopt HyperTransport (and, more likely also RapidIO with the G5), then my answer is: they surely will.</strong><hr></blockquote>
I hope so.
[quote]Originally posted by G-News:
<strong>If your statement is that Apple should accept 3GIO, then my answer is still: they will, once it's ready and widely supported.</strong><hr></blockquote>
3GIO is vomit.
[quote]Originally posted by G-News:
<strong>If your statement is that Apple should invent a sort of HyperTransport slot, replacing AGP and PCI, then my answer is: not going to happen anytime soon, neither from Apple nor anyone else.
maybe you have seen this: Thus yes, it's feasible, but then again, it would bring the same problems with it that I already mentioned: a connector without a card is just lost money. Apple is a member of the HT consortium, and they're surely looking at it very closely, have actually already implemented some ideas of it into their current motherboards, but this stuff takes time. Maybe (hopefully) with the G5, we'll see more.</strong><hr></blockquote>
What I was saying is that. in my estimation. based off of the facts stated above. the HyperTransport standard already has _full support_ for cold swap alteration of the daisy-chain. and. judging from AMD's use pf the term "endorsed" instead of "defined". that the work on making a HyperTransport slot has almost unquestionably _already been done_ for Apple by one of several companies. so Apple needs only to side with one of them. thus aquiring access to a complete. fully functioning HyperTransport slot at _no cost to them_.
The use of Hypertransport slots now would give Apple a strategic advantage over waiting for the slower 3GIO and RapidIO to come to them is enormous. and I think that Apple would be foolish not to take advantage of it.
[quote]Originally posted by G-News:
<strong>last but not least let me show you this image, right off their docu, for some clarification:</strong><hr></blockquote>
As you can see. HyperTransport is a fully daisy-chained(Like FireWire or SCSI) bus. allowing for switches, bridges, terminating devices, open chains, device to device communications and native devices.
[quote]Originally posted by G-News:
<strong>edit: I can't believe how much wrong info you seem to have. The ROM is accessed exactly once, and that's during startup, then everything is in the RAM, which is faster in any way. Since the iMac, we even have the biggest part of the ROM on the harddrive, which is fast and inexpensive. I don't know where you're stuck in tech development, but I know I wasn't even born back then.</strong><hr></blockquote>
That's just how it works now. the ROM chip contains(Or at least used to contain) more than just boot-up code. Mac ROMs usually contained extremely extremely frequently accessed code. such as the toolbox. thus making rapid and unimpeded access to it's contents vital for the efficient operation of a system.
[quote]Originally posted by G-News:
<strong>And one last thing:
Based upon my current knowledge, the following pict shows the best possible mainboard architecture for future Macs. I'm not expert in these things, but based on what I know so far, this is roughly feasible. Maybe you meant that.</strong><hr></blockquote>
That looks unlikely to ever happen, as this graph shows that the CPU has to support LDT (an older name variant of HT) on chip, while my newer graph shows that the CPUs can run a seperate bus to the HT controller.
Thus, unless Motorola decides to drop RapidIO in future chips in favor of HT, were not going to see a solution like this, but rather like the one I painted above.
As for HT being slottable: I said that it's out of question whether it can be done or not, but the questiom is, whether it would make sense or not.
Apart from a CPU/GPU, I can't think of any part that could possibly exploit 12.8GB bandwidth. And then again, as I said in the beginning, these cards would have to be manufactured and engineered first.
I'd find it likely that Apple would allow ofr one "personality slot" using HT, like they made a Personality slot in the Beige G3s, using a PCI variant.
From what I read, general consensus about HT is that it's not going to replace PCI (&co) as a peripheral interconnect for cards, but that it's going to replace PCI for mainboard applications, something the macs doesnt do anymore since the Beige G3, afaik. (FireWire, Gigabit Ethernet etc are all handled on their own busses, not over an integrated PCI bus, as seen on every PC today.
(Basically, as soon as a PC motherboard has something on board today, that you can have as a PCI card separately, that something is connected to the FSB using a PCI bus, just that it's integrated into the logicboard and not slotted. This creates a 133MB/sec bottleneck for all these PC peripherals, and it's only obvious that AMD and co want to remove that bottleneck, using a new standard, ie HT.
as for the Kyro and it's "render only what's needed" stuff. You may be aware that but nVidia and ATI also have certain features that reduce rendering overhead. In ATI's case it's called Hyper-Z and is aparently highly efficient, in nVidias case, it's write combining technologies and some driver tweaks.
As good as the idea may sound, atm the Kyro still lags way behind the ATI and nVidia offerings, overhead reduced or not. And with the upcoming AGP8x, performance is getting some headroom to go once again.
Last but not least, what remains to be seen is how HT performs REALLY. 12.8GB/sec sounds awfully nice, but it's pure theory and is certainly doing do depend on implementations etc. Maybe the Hammer will provide the first platform to experiment with.
<strong>That looks unlikely to ever happen, as this graph shows that the CPU has to support LDT (an older name variant of HT) on chip, while my newer graph shows that the CPUs can run a seperate bus to the HT controller.</strong><hr></blockquote>
Careful with that vocabulary fella. "name variant" makes it sound too much like LDT is a dead branch-off HyperTransport. for those folks in the audience. LDT was just the developmental codename for HyperTransport.
[quote]Originally posted by G-News:
<strong>Thus, unless Motorola decides to drop RapidIO in future chips in favor of HT, were not going to see a solution like this, but rather like the one I painted above.
As for HT being slottable: I said that it's out of question whether it can be done or not, but the questiom is, whether it would make sense or not.</strong><hr></blockquote>
I think the illustrator just drew it as being on-chip for simplicity's sake. but even if HyperTransport does require the I/O port to be on-chip whenever daisy chaining is used(Which makes absolutely no sense whatsoever). Apple could just preach the gospel of HyperTransport to Motorola. or promise them a RapidIO bridge would make it onto each motherboard.
[quote]Originally posted by G-News:
<strong>Apart from a CPU/GPU, I can't think of any part that could possibly exploit 12.8GB bandwidth. And then again, as I said in the beginning, these cards would have to be manufactured and engineered first.</strong><hr></blockquote>
The newer varients of nearly everything you can think of. 1Gb/10Gb Ethernet, Ultra 640 SCSI, even faster ATM and FibreChannel and many more. as well as new external busses like InfiniBand.
Also. lets not forget that Apple wouldn't be going it _entirely_ alone for all of the future. HyperTransport is an industry wide standard. and Apple would just be leading the way. Apple would. of course. cajole the rest of the computer industry to adopt it. Apple would be alone for a while. but they would be seen by the rest of the industry as an early adopter. instead of a futurist radical like usual. this strategy would temporalily combine the safety of when they started using USB with the tremendous advantage of when the got NuBus.
[quote]Originally posted by G-News:
<strong>I'd find it likely that Apple would allow ofr one "personality slot" using HT, like they made a Personality slot in the Beige G3s, using a PCI variant.
[QUOTE]Originally posted by G-News:
[QB]From what I read, general consensus about HT is that it's not going to replace PCI (&co) as a peripheral interconnect for cards, but that it's going to replace PCI for mainboard applications,</strong><hr></blockquote>
That's because most people don't know what I told you. so they never even consider the possibility.
[quote]Originally posted by G-News:
<strong>as for the Kyro and it's "render only what's needed" stuff. You may be aware that but nVidia and ATI also have certain features that reduce rendering overhead. In ATI's case it's called Hyper-Z and is aparently highly efficient, in nVidias case, it's write combining technologies and some driver tweaks.
As good as the idea may sound, atm the Kyro still lags way behind the ATI and nVidia offerings, overhead reduced or not. And with the upcoming AGP8x, performance is getting some headroom to go once again.</strong><hr></blockquote>
ATI and nVidia's techniques are cheap tricks compared to tile based rendering. once Imagine Tech gets in gear and raises the speed of the Kyro's core to higher levels. there'll be no stopping them. and although nVidia <a href="http://www.theregister.co.uk/content/archive/19962.html"; target="_blank">owns something like that called Gigapixel</a> that it got in it's purchase of (God bless their souls)3DFX. it's been about a year since nVidia <a href="http://www.nvidia.com/view.asp?IO=IO_20010530_5354"; target="_blank">acquired them</a>. and nVidia hasn't said word one about using it. that. combined with ATI and nVidia's legendary lazyness when it comes to radically altering the fundamental design of their chips. prompts me to think that they would probably choose shifting to HyperTransport over increasing the intelligence of their chips.
As for AGP 8X. the thing to remember is that HyperTransport is here _right now_. and has been for a over two years. as well. ATI and nVidia(Especially them) heve had very extensive HyperTransport experience for a quite a while. I'll bet some engineer in nVidia already has a design he/she thought of for a GeForce HyperTransport card taped on their wall(And even if they don't. a soldered on GPU chipset al´a iMac would still totally rock). AGP 8X(And PCI-X) standards are [I]just[I] exiting the _experimental stage_ as we speak. no one on earth has experience with a shipping version. and nearly no one has experience with prototypes. with the fact of that HyperTransport supports slots are perfectly easy. there's no good reason for them to choose anything else.
[quote]Originally posted by G-News:
<strong>Last but not least, what remains to be seen is how HT performs REALLY. 12.8GB/sec sounds awfully nice, but it's pure theory and is certainly doing do depend on implementations etc. Maybe the Hammer will provide the first platform to experiment with.</strong><hr></blockquote>
AMD has kept their promises so far. lets just sit back and hope for the best(As well as harassing Apple to use HT as soon as possible. and AMD to endorse some HT PCB ).
name variant "unequal" (stupid UBB doesn't eat the sign) variant. name variant = variant of a name for the same thing, thus it's perfectly legal to say it like that.
As for HT vs AGP 8x: Both have been in development for a long time and both are not ready for the market aparently. even if HT is taped out and all, there's no product a mortal person can buy today. As we know from things like FireWire, sometimes it takes a long road from taping out to rolling out.
And last but not least: If you give me the choice between a slotted AGP 8x port around 2GB/sec and a soldered HT graphics chip with 12.8GB/sec, I'll take the AGP ANY day. non-upgradeable graphics is one of the biggest sins you can make designing a computer with a long lifespan. Do that for mobiles if you absolutely have to, but even the iMacs are really bad in that case.
I happen to own a Beige G3 with soldered on Rage2+ and I'd give an empire for an extra PCI slot instead of this now totally useless chip.
(especially because the chip is connected via PCI bus anyway)
If HT is ever going to be used for Graphics, CPU's and other quickly ageing devices, then it will absolutely have to be slotted, or it's going to fail. But I'm sure the consortium knows that better than Apple.
HT is going to get us fast motherboards without bottlenecks. With a little luc it's also going to give us extra slots for ultra fast expansion, but it's not going to kill PCI, PCI-X or AGP anytime soon.
Comments
<strong>
Dumping the contents of the ROM chip into ultra slow dynamic RAM instead of swapping the slower ROM chips they were using at the time for speedy static memory(Which. by the way. is seldom used in large amounts due to the fact of that it costs so much more than dynamic memory. but the fact of that the Mac ROM code is only about 6-8MBs makes static memory an excellant choice for small, frequently accessed data like is on most ROMs) chips. like the SRAM used in most CPUs as cache. was probably one of the worst moves Apple has ever done.
</strong><hr></blockquote>
You honestly suggest Apple should have rather put 6-8MB (!) of SRAM onto their mainboards? This would have been enormously expensive. So I wonder how exaclty shadowing the slow ROMs in much faster RAM was such a bad idea.
[quote]<strong>And if you're refering to switching to a standard motherboard chipset across all lines. I think the PC industry has been doing that since about the day there were clones.
</strong><hr></blockquote>
Huh? They aren't even doing this right now. In fact, as of now, most vendors even use proprietary interconnects between their chipset components (HubLink, V-Link, MuTIOL), although this will probably change with the widespread adoption of HT or 3GIO.
[quote]<strong>
Yuck. DVD-R. like CD-R before it. is about the most user unfriendly storage medium ever invented. Apple already had DVD-RAM drives in their G4s.</strong><hr></blockquote>
While I wholeheartedly agree in preferring DVD-RAM, DVD-R is what the most customers seem to be demanding.
[quote]<strong>which. aside from their far superior hard disc like user experience. also costed roughly $750 less than Apple's DVD-R drive at the time. and currently cost $250 less</strong><hr></blockquote>
The media costs more, though, IIRC.
[quote]<strong>
If you were to swap a modern G4's ZIF slot for a Pentium's Socket7.</strong><hr></blockquote>
G4s don't use ZIF sockets any more (and actually, Socket/ is a ZIF socket).
[quote]<strong>and switch Apple's ROM chip for a Wintel BIOS chip. there wouldn't be so much as _one_ functional difference between it and an equivalent Wintel motherboard. quite truly.</strong><hr></blockquote>
Well, since both are mainboard after all, there can hardly be very much variation in the first place.
There are a lot of significant details, though, like GBit ethernet on-board and not on the PCI, same for FireWire, 64bit PCI slots, OpenFirmware rather than real mode BIOS, Target Disk Mode and others.
[quote]<strong>
all that Apple makes now are wintel clones with PPCs and some flavor of UN*X on them.
</strong><hr></blockquote>
You do realize that Apple have their own, proprietary North- and Southbridge? They don't (and couldn't, actually) just go ahead and use chips also found on x86 mainboards.
Bye,
RazzFazz
<strong>Rather than spend 2 hours explaining how terribly wrong a certain above post is, can we all just acknowledge that Eric D.V.H. is very, very, wrong about 97% of his post? thanks.
But I just can't resist.
[quote]Originally posted by Eric D.V.H.
<strong>That's just what they want you to think. in this Api NetWorks/Sibyte whitepaper and this Api NetWorks executive summary. they say that: "The 50-ohm impedance and differential signaling also permit trace lengths up to 24 inches, and they span board interconnects well." meaning that HyperTransport works perfectly well across the shattered pins and traces at the bottom of a PCB card.
They also said that: "There is an ongoing effort with major connector companies to define a variety of connectors that can fit different price/performance parameters. More information on this topic will be available in 2001."
And in a recent AMD whitepaper. they discussed Plug ?n Play in it's session layer.
The final clincher though. is on AMD's online HyperTransport FAQ. where they more cautiously echo the precise words of the Api NetWorks/Sibyte pieces by saying: "A HyperTransport connector has not been endorsed at this time; however, some members of the Consortium are exploring possible use of connectors for different applications."
So why are they pretending that HyperTransport is an intra-board-only standard? darned if I know.</strong><hr></blockquote>
Yawn. PCI-X is here already. 3GIO is around the corner. All you've listed here are ideas that exist only on paper or at best are buried deep in some geeks lab. Do you think this stuff will just pop up overnight? Will these technologies even work, or will impassable problems arise? Can the companies that are creating these designs sucessfully get them to market? Considering all that still needs to be done, then how long will they take to reach the marketplace?
So what you're saying is that Apple should pass on technologies that are in the very near future just to implement stuff that merely exists on paper.
What are you <img src="graemlins/smokin.gif" border="0" alt="[Chilling]" /> , cause I'll try to avoid that batch?
<strong>At this point, I'd rather they did adopt a Wintel-clone motherboard. Their current one is so outdated it's ludicrous.</strong><hr></blockquote>
That's why Apple should use HyperTransport as soon as possible. and before they do.
[quote]Originally posted by TJM:
<strong>Calm down for a moment, and think a bit. What is it that makes a Mac? For me, it's the ergonomics of the hardware and the software, particularly the Mac OS. Does it really matter what's inside the box? Would it run any differently if it had a solid gold case and platinum wiring? It's Apple's attention to detail in the system integration, their industrial design for the cases, and the ease-of-use in the software that make the quality of it. The technology used doesn't really matter. As long as it produces my beloved "Mac experience" they can have gerbils in a treadmill running it, for all I care.</strong><hr></blockquote>
You forget that revolutions like the Macintosh are only possible by using totally new software [i]and[i] hardware. in order to make a uniquely better machine. you need to use uniquely better components.
[quote]Originally posted by TJM:
<strong>It's one thing to use off-beat technology just to "Think Different" when it is superior or equivalent, but much of the Mac's technology at that point was out-of-date and orphaned (i.e. NuBus and ADB). ATI, 3dfx, and nVidia would never have come out with Mac video cards if they were still running NuBus. Thanks to PCI (then AGP) we have access to first-rate video. The list goes on and on.</strong><hr></blockquote>
Ugh. I wasn't implying that Apple should have remained frozen in the 80s
[quote]Originally posted by TJM:
<strong>As a "fer instance" on the relative costs, a couple years ago I bought an ATA card for my 7600 because I just couldn't see paying $350 for a 10 GB SCSI drive when I could get a 50 GB ATA drive for $120. Even after $100 for the ATA card, I was still over $100 ahead with 5x capacity. I can get memory for my G3 desktop dirt cheap because it uses PC100.</strong><hr></blockquote>
Yes. but your(And my) machine aren't several times better. hands down. than the best IBM PC on the market. which the Mac used to be back when it used it's own. superior standards.
[quote]Originally posted by TJM:
<strong>So, I appreciate your sentiment - I like Macs because they ARE different . But if different = inferior or different = outrageously expensive (with no clear benefit), you're going to have a hard time broadening your appeal or even staying in business.</strong><hr></blockquote>
Exactly. that's the reason Apple's market share is so low now. Mac sites all over the web are proclaiming that it "Doesn't really matter which platform you use. just whichever you like". they never would have said that back then.
Eric,
[ 03-05-2002: Message edited by: Eric D.V.H ]</p>
Times have changed and there is no way a company could possibly reinvent the personal computer in both hardware and software as Apple did back in 1984.
Building a platform today is made by taking the best standards around and implementing them as good as possible, not buy spending a few billion dollars on a proprietary design that eventually nobody else will use or support. back in the 80s a garage company could do these things, simply because there were different market and industry situations. Today that is only possible in software anymore.
The first example is probably the adoption of PCI back in 1995 in the PM 9500. that was already about 1 or evne 2 years behind the PC, and Apple chose PCI because it was: superior to Nubus (which was superior to ISA admittably). Instead they could have developped NuBud further and gotten more performance out of it, but they didn't because: It would have cost a lot of R&D and it would have brought with it the same problems that NuBus already had, namely that only very few select companies made products for it.
So yes, PCI was the only reasonable choice back then, and gladly Apple didn't spoil it.
I can tell you one thing: without PCI, we'd be in Atari land today. they kept using old hardware until they were hopelessly lost and now they're dead. But their OS and their machines were among the very best in the mid to late 80s.
Right now there are various new technologies emerging, and as I said, the best thing Apple can do is stay updated and be ready for release of a product incorporating those new techs as soon as they're generally accepted and supported.
G-News
<strong>I say it again eric: WAKE UP.
Times have changed and there is no way a company could possibly reinvent the personal computer in both hardware and software as Apple did back in 1984.
Building a platform today is made by taking the best standards around and implementing them as good as possible, not buy spending a few billion dollars on a proprietary design that eventually nobody else will use or support. back in the 80s a garage company could do these things, simply because there were different market and industry situations. Today that is only possible in software anymore.</strong><hr></blockquote>
From my other post:
Actually. HyperTransport is backwards compatible too. as you can drive multiple PCI, Infiniband etc. buses. as well as an AGP bus. with a HyperTransport based system. I would also expect a HyperTransport slot to PCI slot adapter. much like the PDS to NuBus adapters of yore.
[quote]Originally posted by G-News:
<strong>The first example is probably the adoption of PCI back in 1995 in the PM 9500. that was already about 1 or evne 2 years behind the PC, and Apple chose PCI because it was: superior to Nubus (which was superior to ISA admittably). Instead they could have developped NuBud further and gotten more performance out of it, but they didn't because: It would have cost a lot of R&D and it would have brought with it the same problems that NuBus already had, namely that only very few select companies made products for it.
So yes, PCI was the only reasonable choice back then, and gladly Apple didn't spoil it.
I can tell you one thing: without PCI, we'd be in Atari land today. they kept using old hardware until they were hopelessly lost and now they're dead. But their OS and their machines were among the very best in the mid to late 80s.</strong><hr></blockquote>
Atari? were you thinking of Amiga? they're still keeping a sort of market in europe. 2-3%. just like Apple in the US. so I guess we are in _Amiga_ land today.
[quote]Originally posted by G-News:
<strong>Right now there are various new technologies emerging, and as I said, the best thing Apple can do is stay updated and be ready for release of a product incorporating those new techs as soon as they're generally accepted and supported.</strong><hr></blockquote>
From my other post as well:
What if it just lies there and rots like USB or VCD? someone needs to take action. and with everyone else as leadbottomed as they are. that someone is Apple.
Eric,
<strong> You have to differentiate between the past and the recent past. ADB, NuBus and SCSI were selected because they were superior, and in the case of SCSI, already working on other systems.</strong><hr></blockquote>
Yup.
[quote]Originally posted by G-News:
<strong>NuBus was just another standard that emerged out of the first experimenting phase of modern computer design, as was ADB and GeoPort.</strong><hr></blockquote>
All of these were better than their mainstream counterparts.
[quote]Originally posted by G-News:
<strong>DAV definitely doesn't count as I have yet to see a single product that was ever released to the public for that slot.</strong><hr></blockquote>
Try <a href="http://www.allelec.com/AVIDinfo.html"; target="_blank">this Avid card</a>.
[quote]Originally posted by G-News:
<strong>DB-15 is functionally identical to a normal VGA port, that was just a proprietary decision, not a quality over cost one.</strong><hr></blockquote>
Yes. but I was trying to point out that Apple has gone on it's own before.
[quote]Originally posted by G-News:
<strong>And for USB, as you probably know, Apple helped it make a breakthrough, but they didn't invent it (Intel did) nor did they introduce it (most PCs actually had USB about a year before the iMac was originally released, just that nobody knew about it.</strong><hr></blockquote>
Adopting HyperTransport now would be kind of half-way between when Apple got USB and when they got SCSI.
[quote]Originally posted by G-News:
<strong>If you look 20 to 10 years back, you'll see that Apple had to develop and introduce new technology, because there was nothing suitable to meet their needs. In the recent past you'll soon notice that Apple just picked the best standards that were already available on other platforms, combined them and took out the hubbles of the road.
Apple's implementation of PCI is wonderful compared to the PC counterpart (both in terms of performance and software support (for example we don't have to fiddle with IRQs).</strong><hr></blockquote>
I agree. and PCI/AGP just don't look like they'll keep meeting Apple's needs for long.
[quote]Originally posted by G-News:
<strong>Apple hasn't been innovating in hardware in terms of developing itself, but they have been innovative in terms of making things actually usable (example USB and FireWire (partially).</strong><hr></blockquote>
Apple is the one that invented FireWire. and I think it was pretty revolutionary(Too bad they waited so long to use it
[quote]Originally posted by G-News:
<strong>If Apple adopted 3GIO or similar now, totally replacing PCI, you'd have a system with a new standard that nobody supports until the PCs adopt it too. The days of small companies making extra hardware for the Mac (ie NuBus falvor cards of their products too) are over, companies no longer an afford extra tours.</strong><hr></blockquote>
Actually. HyperTransport is backwards compatible too. as you can drive multiple PCI, Infiniband etc. buses. as well as an AGP bus. with a HyperTransport based system. I would also expect a HyperTransport slot to PCI slot adapter. much like the PDS to NuBus adapters of yore.
[quote]Originally posted by G-News:
<strong>Wake up, the best thing Apple can do now is stay on top of the emerging standards, experiment with them and use the best ones for their machines, as soon as support from most sides is granted.</strong><hr></blockquote>
What if it just lies there and rots like USB or VCD? someone needs to take action. and with everyone else as leadbottomed as they are. that someone is Apple.
[quote]Originally posted by G-News:
<strong>PCI-X, if downwards compatible, is currently the only option that you could implement right away, without awaiting support. But then again we already have 64bit PCI slots that nobody supports either.</strong><hr></blockquote>
Refer to prior comment.
[quote]Originally posted by G-News:
<strong>Apple is not going to adopt 3GIO before they can be absolutely sure, that the companies that make PCI cards now, will continue making 3GIO cards. And that likely isn't going to happen before Intel hasn't got a chipset out that supports 3GIO.</strong><hr></blockquote>
I can think of one kind of company that would gladly make HyperTransport cards. GPU makers. ATI and nVIDIA are already starting to feel the pain from the Kyro II. and I'll bet that they'll start burning the midnight oil to get out HyperTransport based chipsets and cards once the Kyro III comes onto the market.
[quote]Originally posted by G-News:
<strong>Of all the standard hardware Apple has introduced during the last 5 years, name me ONE that wasn't available on other platforms before.
(FireWire is the only barely legal item)</strong><hr></blockquote>
Isn't that what I said?
\t\t\t\t\t\t\t\t\t Eric,
[ 03-05-2002: Message edited by: Eric D.V.H ]</p>
[quote]Originally posted by TJM:
<strong>I'll be sure to send a note to SJ to remind him to consult you before he makes any more strategic moves.</strong><hr></blockquote>
<img src="graemlins/lol.gif" border="0" alt="[Laughing]" /> Steve Jobs didn't even work at Apple when they switched to PCI. I'm not quite sure. but I think it was that dolt Scully.
[quote]Originally posted by TJM:
<strong>At the time, Apple was losing $100s of millions of dollars per quarter, its market share had already plunged to single digits, and it was rapidly on its way to bankruptcy. Going to PCI and ATA and standard memory (PC66 on Beige G3) saved them a ton of money and helped them survive. It was a matter of change or die. The "cachet" of owning a Mac with its unique technologies wasn't going to last long in any event.</strong><hr></blockquote>
And what's our marketshare at now?
\t\t\t Eric,
<strong>Yawn. PCI-X is here already. 3GIO is around the corner. All you've listed here are ideas that exist only on paper or at best are buried deep in some geeks lab. Do you think this stuff will just pop up overnight? Will these technologies even work, or will impassable problems arise? Can the companies that are creating these designs sucessfully get them to market? Considering all that still needs to be done, then how long will they take to reach the marketplace?</strong><hr></blockquote>
Read that stuff more carefully. Hyper transport IS HERE NOW. it entered the market TWO YEARS AGO. the problems have already arisen and been vanquished in testing OVER TWO YEARS AGO. people are using it in things they bought from these companies. and they say it works JUST FINE.
[quote]Originally posted by Arty50:
<strong>So what you're saying is that Apple should pass on technologies that are in the very near future just to implement stuff that merely exists on paper. </strong><hr></blockquote>
PCI-X and AGP are set to just barely exit the experimental phase this year. 3GIO/Aprahoe actually _is_ only on paper or at best buried deep in some geeks lab. HyperTransport is already gathering momentum across the industry. and needs only a strong supporter to break the barriers. for since it's already on the market. that means Apple can start right now. or already has.
Eric,
<strong>You honestly suggest Apple should have rather put 6-8MB (!) of SRAM onto their mainboards? This would have been enormously expensive. So I wonder how exaclty shadowing the slow ROMs in much faster RAM was such a bad idea.</strong><hr></blockquote>
It can't have been _that_ expensive. but even if it was. Apple should have used a faster variety of DRAM. in a separate block. directly connected to the CPU. so as to increase access speed to this oft consulted data.
[quote]Originally posted by RazzFazz:
<strong>Huh? They aren't even doing this right now. In fact, as of now, most vendors even use proprietary interconnects between their chipset components (HubLink, V-Link, MuTIOL), although this will probably change with the widespread adoption of HT or 3GIO.</strong><hr></blockquote>
I didn't mean the PC industry as a whole. I meant individual PC vendors like Dell, Compaq, Gateway etc?
They use a lot of the same parts. and pretty much always have.
[quote]Originally posted by RazzFazz:
<strong>While I wholeheartedly agree in preferring DVD-RAM, DVD-R is what the most customers seem to be demanding.
The media costs more, though, IIRC.</strong><hr></blockquote>
Yeah. fortunately there's <a href="http://www.dvdforum.com/tech-dvdmulti.htm"; target="_blank">DVD-Multi</a>.
[quote]Originally posted by RazzFazz:
<strong>G4s don't use ZIF sockets any more (and actually, Socket/ is a ZIF socket).</strong><hr></blockquote>
I couldn't say either way. <a href="http://developer.apple.com/techpubs/hardware/Developer_Notes/Macintosh_CPUs-G4/PowerMacG4/2Architecture/index.html"; target="_blank">Apple's docs</a> just say "a 300-pin connector".
[quote]Originally posted by RazzFazz:
<strong>Well, since both are mainboard after all, there can hardly be very much variation in the first place.
There are a lot of significant details, though, like GBit ethernet on-board and not on the PCI, same for FireWire, 64bit PCI slots, OpenFirmware rather than real mode BIOS, Target Disk Mode and others.</strong><hr></blockquote>
Hmm? while the <a href="http://developer.apple.com/techpubs/hardware/Developer_Notes/Macintosh_CPUs-G4/PowerMacG4/2Architecture/KeyLargo_I_O_Controller.html"; target="_blank">KeyLargo I/O controller</a> looks suspiciously generic. Apple's docs <a href="http://developer.apple.com/techpubs/hardware/Developer_Notes/Macintosh_CPUs-G4/PowerMacG4/2Architecture/Uni_N_Bridg__Controller.html"; target="_blank">don't mention</a> any companies. so Apple _could_ have had it designed just for themselves.
[quote]Originally posted by RazzFazz:
<strong>You do realize that Apple have their own, proprietary North- and Southbridge? They don't (and couldn't, actually) just go ahead and use chips also found on x86 mainboards.</strong><hr></blockquote>
I'm not exactly sure on this one either. so you might be right.
Eric,
You do know that HyperTransport is a bus technology that is designed to replace systems like the current MaxBus, Intel's and AMD's incarnations of their front side busses etc, right?
It's not aimed at peripheral cards, the fact that it can support line lengths to card degrees, pinnage etc only implies that it's feasible to make slots, similar to Slot A and similar, ie for CPUs and other extremely performance sensitive parts of a computer system.
HyperTransport is just going to change the way a CPU can communicate with its peripherals, however by peripherals you have to understand that these are chips like keyboard controllers, PCI busses etc etc, the stuff we have today on our mainboards, not the stuff we plug INTO the mainboards today.
In that way HyperTransport and RapidIO (which are kind of the same thing from another company, with different target applications) are going to be adopted soon, and I wouldn't be surprised if Apple adopted it as one of the first companies out there. However that has nothing to do with 3GIO that we were talking about in the first place, since 3GIO is like PCI, ISA, NuBus, AGP (which is just a PCI derivate) and PCI-X a so called Peripheral Interconnect Bus, and not a system bus or a CPU bus or anything like that.
You are obviously mixing things that don't belong into the same basket at all.
Also wtf do you mean by talking about the Kyro 2?
The Kyro 2 is:
-a PC only modern class graphics GPU
-exists only on AGP bus, as far as I know, PCI is probably possible
-is based on a tile-based rendering system, unlike the (correct me if I'm wrong) polygon based rendering system of Matrox, ATI and nVidia cards.
-is not incredibly slow and, most important of all:
-the company that made those boards just went belly up a few months ago, that's why you can have a Kyro 2 for like 50$ today.
-Kyro 3 is therefore likely never going to show up.
I have no idea what you're mixing there again, but apart from the rendering engine approach, the Kyro boards are cards like ever other card as well.
Chips and connectors on a PCB plate with gold pins that go into an AGP slot, driven by a 2x or 4x AGP slot. No HyperTransport, no RapidIO, no 3GIO and no nothing like that.
last but not least, the link to your DAV card that you were going to tell me about (although it surely existed, I even know there was one or two products for that slot, just that only about 0.004% of all DAV owners ever used that slot), doesn't work: 404 Object Not Found. I think that speaks for itself.
So if your statement is that Apple should adopt HyperTransport (and, more likely also RapidIO with the G5), then my answer is: they surely will.
If your statement is that Apple should accept 3GIO, then my answer is still: they will, once it's ready and widely supported.
If your statement is that Apple should invent a sort of HyperTransport slot, replacing AGP and PCI, then my answer is: not going to happen anytime soon, neither from Apple nor anyone else.
maybe you have seen this: [quote]Has a board-to-board connector for HyperTransport? been defined?
Answer:
A HyperTransport connector has not been endorsed at this time; however, some members of the Consortium are exploring possible use of connectors for different applications.<hr></blockquote> Thus yes, it's feasible, but then again, it would bring the same problems with it that I already mentioned: a connector without a card is just lost money. Apple is a member of the HT consortium, and they're surely looking at it very closely, have actually already implemented some ideas of it into their current motherboards, but this stuff takes time. Maybe (hopefully) with the G5, we'll see more.
last but not least let me show you this image, right off their docu, for some clarification:
edit: I can't believe how much wrong info you seem to have. The ROM is accessed exactly once, and that's during startup, then everything is in the RAM, which is faster in any way. Since the iMac, we even have the biggest part of the ROM on the harddrive, which is fast and inexpensive. I don't know where you're stuck in tech development, but I know I wasn't even born back then.
[ 03-05-2002: Message edited by: G-News ]
And one last thing:
Based upon my current knowledge, the following pict shows the best possible mainboard architecture for future Macs. I'm not expert in these things, but based on what I know so far, this is roughly feasible. Maybe you meant that.
[ 03-05-2002: Message edited by: G-News ]</p>
[ 03-05-2002: Message edited by: Arty50 ]</p>
<strong>Arrrggghhh, my last post disappeared! Alas though, sanity has been restored.</strong><hr></blockquote>
I always feel like my brain exploded whenever that happens. I'm usually to lazy to write posts in something with autosave like FrameMaker though.
[quote]Originally posted by AirSluf:
<strong>My brain hurts from the relentless length of some of those posts though.</strong><hr></blockquote>
Sorry.
Eric,
[ 03-06-2002: Message edited by: Eric D.V.H ]</p>
<strong>You can't be ?SNIPPED?viously mixing things that don't belong into the same basket at all.</strong><hr></blockquote>
It does that too. but it also supports intra-board communications between various devices such as graphics accelerators, networking adapters and audio D/A divices. the fact of that HyperTransport features a full cold boot sequence instead of just having all of that pre-programmed into the ROMs of the motherboard, native PnP functionality and native device support. make it obvious that HyperTransport has had hardware support for local bus style slots from the start. needing little more than for the PCBs to be soddered on.
[quote]Originally posted by G-News:
<strong>Also wtf do you mean by talking about the Kyro 2?</strong><hr></blockquote>
I'll get to that.
[quote]Originally posted by G-News:
<strong>-is based on a tile-based rendering system, unlike the (correct me if I'm wrong) polygon based rendering system of Matrox, ATI and nVidia cards.</strong><hr></blockquote>
It's tile based. but the term "Tile" refers to the bitmapped output. not the internal geometry. it works by cutting each output image into tiles. it then analyzes each tile to see if there are any polygons hidden behind other geometry. and thusly not worth rendering. after going over the whole image. it renders it. but it only renders visible geometry. thus needing less data(Most of all textures) to be transferred in from outside the card.
[quote]Originally posted by G-News:
<strong>-the company that made those boards just went belly up a few months ago, that's why you can have a Kyro 2 for like 50$ today.
-Kyro 3 is therefore likely never going to show up.</strong><hr></blockquote>
They didn't go "belly up". they just <a href="http://us.st.com/stonline/press/news/year2002/c1146h.htm"; target="_blank">pulled out of</a> the Kyro deal. although they [i]are[i] looking for a buyer. and it was the company(<a href="http://us.st.com/stonline/index.htm"; target="_blank">ST Micro</a>) that _manufactured_ them.<a href="http://us.st.com/stonline/index.htm"; target="_blank">Imagine Technologies</a>. the company that designs the Kyro(And it's ancestor. the PowerVR) is still going at full steam(If a bit bruised up). and ST Micro said that they're <a href="http://us.st.com/stonline/press/news/year2002/t1152p.htm"; target="_blank">going in for a second bite</a> anyways. so I think that. once they have their new Kyro III chip. Imagine will find _someone_ to manufacture them.
[quote]Originally posted by G-News:
<strong>I have no idea what you're m?SNIPPED?othing like that.</strong><hr></blockquote>
Eesh. the reason I mentioned the Kyro is because it uses bandwidth more efficiently than other chipsets. why does this make ATI and nVidia shiver in their shanks?
Remember how newer GPUs. when running things with higher levels of detail. especially at lower resolutions. the frame rates seem to "Top out" at the lower resolutions until the level of detail is decreased below a certain amount?
This is due to the fact of that higher levels of detail require more data to be piped into the GPU from outside. while the GPU normally goes at full speed. as it renders the output so slowly that the data for the next frame has enough time to flow in. when the resolution is decreased to the point where the GPU can render frames faster than it can get the data for the next frame. it just sits there and "Idles" between frames. thus making an effective "Ceiling". above which the GPU's power goes to waste.
The Kyro on the other hand. due to the fact of that it generally doesn't need as much data to render the same scene. can use more of it's GPU's power at a given time than ATI and nVidia's chips.
The only way to fix this problem(Aside from work-arounds like the Kyro employs). is to increase the speed and width of the bus connecting the GPU to everything else. unless this happens. Imagine and their Kyro are destined to catch up to(And surpass) ATI and nVidia VERY soon. at which point they'll be in it _deep_.
[quote]Originally posted by G-News:
<strong>last but not least, the link to your DAV card th?SNIP?aks for itself.</strong><hr></blockquote>
Yes. I guess that _everyone_ who did video out were only 0.004% of all DAV owners.
[quote]Originally posted by G-News:
<strong>So if your statement is that Apple should adopt HyperTransport (and, more likely also RapidIO with the G5), then my answer is: they surely will.</strong><hr></blockquote>
I hope so.
[quote]Originally posted by G-News:
<strong>If your statement is that Apple should accept 3GIO, then my answer is still: they will, once it's ready and widely supported.</strong><hr></blockquote>
3GIO is vomit.
[quote]Originally posted by G-News:
<strong>If your statement is that Apple should invent a sort of HyperTransport slot, replacing AGP and PCI, then my answer is: not going to happen anytime soon, neither from Apple nor anyone else.
maybe you have seen this: Thus yes, it's feasible, but then again, it would bring the same problems with it that I already mentioned: a connector without a card is just lost money. Apple is a member of the HT consortium, and they're surely looking at it very closely, have actually already implemented some ideas of it into their current motherboards, but this stuff takes time. Maybe (hopefully) with the G5, we'll see more.</strong><hr></blockquote>
What I was saying is that. in my estimation. based off of the facts stated above. the HyperTransport standard already has _full support_ for cold swap alteration of the daisy-chain. and. judging from AMD's use pf the term "endorsed" instead of "defined". that the work on making a HyperTransport slot has almost unquestionably _already been done_ for Apple by one of several companies. so Apple needs only to side with one of them. thus aquiring access to a complete. fully functioning HyperTransport slot at _no cost to them_.
The use of Hypertransport slots now would give Apple a strategic advantage over waiting for the slower 3GIO and RapidIO to come to them is enormous. and I think that Apple would be foolish not to take advantage of it.
[quote]Originally posted by G-News:
<strong>last but not least let me show you this image, right off their docu, for some clarification:</strong><hr></blockquote>
That image makes out HyperTransport to be far more centralized than it really is. this image is from <a href="http://www.sibyte.com/mercurian/docs/whitepaper_ldt.pdf"; target="_blank">another one of their docs</a>:
As you can see. HyperTransport is a fully daisy-chained(Like FireWire or SCSI) bus. allowing for switches, bridges, terminating devices, open chains, device to device communications and native devices.
[quote]Originally posted by G-News:
<strong>edit: I can't believe how much wrong info you seem to have. The ROM is accessed exactly once, and that's during startup, then everything is in the RAM, which is faster in any way. Since the iMac, we even have the biggest part of the ROM on the harddrive, which is fast and inexpensive. I don't know where you're stuck in tech development, but I know I wasn't even born back then.</strong><hr></blockquote>
That's just how it works now. the ROM chip contains(Or at least used to contain) more than just boot-up code. Mac ROMs usually contained extremely extremely frequently accessed code. such as the toolbox. thus making rapid and unimpeded access to it's contents vital for the efficient operation of a system.
[quote]Originally posted by G-News:
<strong>And one last thing:
Based upon my current knowledge, the following pict shows the best possible mainboard architecture for future Macs. I'm not expert in these things, but based on what I know so far, this is roughly feasible. Maybe you meant that.</strong><hr></blockquote>
No. I meant something like this:
\t\t\t\t Eric,
[ 03-06-2002: Message edited by: Eric D.V.H ]</p>
Thus, unless Motorola decides to drop RapidIO in future chips in favor of HT, were not going to see a solution like this, but rather like the one I painted above.
As for HT being slottable: I said that it's out of question whether it can be done or not, but the questiom is, whether it would make sense or not.
Apart from a CPU/GPU, I can't think of any part that could possibly exploit 12.8GB bandwidth. And then again, as I said in the beginning, these cards would have to be manufactured and engineered first.
I'd find it likely that Apple would allow ofr one "personality slot" using HT, like they made a Personality slot in the Beige G3s, using a PCI variant.
From what I read, general consensus about HT is that it's not going to replace PCI (&co) as a peripheral interconnect for cards, but that it's going to replace PCI for mainboard applications, something the macs doesnt do anymore since the Beige G3, afaik. (FireWire, Gigabit Ethernet etc are all handled on their own busses, not over an integrated PCI bus, as seen on every PC today.
(Basically, as soon as a PC motherboard has something on board today, that you can have as a PCI card separately, that something is connected to the FSB using a PCI bus, just that it's integrated into the logicboard and not slotted. This creates a 133MB/sec bottleneck for all these PC peripherals, and it's only obvious that AMD and co want to remove that bottleneck, using a new standard, ie HT.
as for the Kyro and it's "render only what's needed" stuff. You may be aware that but nVidia and ATI also have certain features that reduce rendering overhead. In ATI's case it's called Hyper-Z and is aparently highly efficient, in nVidias case, it's write combining technologies and some driver tweaks.
As good as the idea may sound, atm the Kyro still lags way behind the ATI and nVidia offerings, overhead reduced or not. And with the upcoming AGP8x, performance is getting some headroom to go once again.
Last but not least, what remains to be seen is how HT performs REALLY. 12.8GB/sec sounds awfully nice, but it's pure theory and is certainly doing do depend on implementations etc. Maybe the Hammer will provide the first platform to experiment with.
G-News
<strong>That looks unlikely to ever happen, as this graph shows that the CPU has to support LDT (an older name variant of HT) on chip, while my newer graph shows that the CPUs can run a seperate bus to the HT controller.</strong><hr></blockquote>
Careful with that vocabulary fella. "name variant" makes it sound too much like LDT is a dead branch-off HyperTransport. for those folks in the audience. LDT was just the developmental codename for HyperTransport.
[quote]Originally posted by G-News:
<strong>Thus, unless Motorola decides to drop RapidIO in future chips in favor of HT, were not going to see a solution like this, but rather like the one I painted above.
As for HT being slottable: I said that it's out of question whether it can be done or not, but the questiom is, whether it would make sense or not.</strong><hr></blockquote>
I think the illustrator just drew it as being on-chip for simplicity's sake. but even if HyperTransport does require the I/O port to be on-chip whenever daisy chaining is used(Which makes absolutely no sense whatsoever). Apple could just preach the gospel of HyperTransport to Motorola. or promise them a RapidIO bridge would make it onto each motherboard.
[quote]Originally posted by G-News:
<strong>Apart from a CPU/GPU, I can't think of any part that could possibly exploit 12.8GB bandwidth. And then again, as I said in the beginning, these cards would have to be manufactured and engineered first.</strong><hr></blockquote>
The newer varients of nearly everything you can think of. 1Gb/10Gb Ethernet, Ultra 640 SCSI, even faster ATM and FibreChannel and many more. as well as new external busses like InfiniBand.
Also. lets not forget that Apple wouldn't be going it _entirely_ alone for all of the future. HyperTransport is an industry wide standard. and Apple would just be leading the way. Apple would. of course. cajole the rest of the computer industry to adopt it. Apple would be alone for a while. but they would be seen by the rest of the industry as an early adopter. instead of a futurist radical like usual. this strategy would temporalily combine the safety of when they started using USB with the tremendous advantage of when the got NuBus.
[quote]Originally posted by G-News:
<strong>I'd find it likely that Apple would allow ofr one "personality slot" using HT, like they made a Personality slot in the Beige G3s, using a PCI variant.
[QUOTE]Originally posted by G-News:
[QB]From what I read, general consensus about HT is that it's not going to replace PCI (&co) as a peripheral interconnect for cards, but that it's going to replace PCI for mainboard applications,</strong><hr></blockquote>
That's because most people don't know what I told you. so they never even consider the possibility.
[quote]Originally posted by G-News:
<strong>as for the Kyro and it's "render only what's needed" stuff. You may be aware that but nVidia and ATI also have certain features that reduce rendering overhead. In ATI's case it's called Hyper-Z and is aparently highly efficient, in nVidias case, it's write combining technologies and some driver tweaks.
As good as the idea may sound, atm the Kyro still lags way behind the ATI and nVidia offerings, overhead reduced or not. And with the upcoming AGP8x, performance is getting some headroom to go once again.</strong><hr></blockquote>
ATI and nVidia's techniques are cheap tricks compared to tile based rendering. once Imagine Tech gets in gear and raises the speed of the Kyro's core to higher levels. there'll be no stopping them. and although nVidia <a href="http://www.theregister.co.uk/content/archive/19962.html"; target="_blank">owns something like that called Gigapixel</a> that it got in it's purchase of (God bless their souls)3DFX. it's been about a year since nVidia <a href="http://www.nvidia.com/view.asp?IO=IO_20010530_5354"; target="_blank">acquired them</a>. and nVidia hasn't said word one about using it. that. combined with ATI and nVidia's legendary lazyness when it comes to radically altering the fundamental design of their chips. prompts me to think that they would probably choose shifting to HyperTransport over increasing the intelligence of their chips.
As for AGP 8X. the thing to remember is that HyperTransport is here _right now_. and has been for a over two years. as well. ATI and nVidia(Especially them) heve had very extensive HyperTransport experience for a quite a while. I'll bet some engineer in nVidia already has a design he/she thought of for a GeForce HyperTransport card taped on their wall(And even if they don't. a soldered on GPU chipset al´a iMac would still totally rock). AGP 8X(And PCI-X) standards are [I]just[I] exiting the _experimental stage_ as we speak. no one on earth has experience with a shipping version. and nearly no one has experience with prototypes. with the fact of that HyperTransport supports slots are perfectly easy. there's no good reason for them to choose anything else.
[quote]Originally posted by G-News:
<strong>Last but not least, what remains to be seen is how HT performs REALLY. 12.8GB/sec sounds awfully nice, but it's pure theory and is certainly doing do depend on implementations etc. Maybe the Hammer will provide the first platform to experiment with.</strong><hr></blockquote>
AMD has kept their promises so far. lets just sit back and hope for the best(As well as harassing Apple to use HT as soon as possible. and AMD to endorse some HT PCB
Eric,
[ 03-06-2002: Message edited by: Eric D.V.H ]</p>
As for HT vs AGP 8x: Both have been in development for a long time and both are not ready for the market aparently. even if HT is taped out and all, there's no product a mortal person can buy today. As we know from things like FireWire, sometimes it takes a long road from taping out to rolling out.
And last but not least: If you give me the choice between a slotted AGP 8x port around 2GB/sec and a soldered HT graphics chip with 12.8GB/sec, I'll take the AGP ANY day. non-upgradeable graphics is one of the biggest sins you can make designing a computer with a long lifespan. Do that for mobiles if you absolutely have to, but even the iMacs are really bad in that case.
I happen to own a Beige G3 with soldered on Rage2+ and I'd give an empire for an extra PCI slot instead of this now totally useless chip.
(especially because the chip is connected via PCI bus anyway)
If HT is ever going to be used for Graphics, CPU's and other quickly ageing devices, then it will absolutely have to be slotted, or it's going to fail. But I'm sure the consortium knows that better than Apple.
HT is going to get us fast motherboards without bottlenecks. With a little luc it's also going to give us extra slots for ultra fast expansion, but it's not going to kill PCI, PCI-X or AGP anytime soon.
G-News
[ 03-06-2002: Message edited by: G-News ]</p>