What is the future of I/O?
And will we see it in 2002 macs?
Some of you know that I think Apple ought to make a headless micro-tower iMac. 1 PCI slot and 1 AGP slot for sound and video cards is a must for a consumer machine that wishes to woo windows customers. OR, so I thought, but what if Firewire got a whole lot faster?
Not 400Mbps but 800-1600Mbps, or even 3.2Gbps? Would this be fast enough to enable the firwire bus to effectively replace the PCI bus for most pro-sumer applications? I'm talking multiple devices connected to a processor independent bus that can easily handle real-time (and accelerated) audio-video capture, encoding, surround sound steering, mixing... whatever is needed.
I'm pretty sure this can be done even with the current spec, but the bandwidth overhead of a bus 2 to 4 (or maybe 8) times faster would pretty much make it as fast as PCI? I don't claim to know, but I think there's a slight difference. -- doesn't the current PCI spec top out around 215MB/sec? A 1600Mbps (200MB/sec) firewire bus might finally make us forget the PCI bus altogether. Currently most people use the firewire bus as a fast portable storage bus. They claim the extra speed (above the current spec) is unneccessary for that purpose, and unless you want a really fast firwire RAID set-up, they're probably right. However, I thought that the original mission statement for Firewire was broader than just storage. I think some people already use firewire based sound-cards.
Apple should go for the extra bandwidth, whether storage uses it or not, it could finally answer the expansion niggles of consumer macs, and give a laptop solution beyond anything available to any other platform.
[ 12-03-2001: Message edited by: Matsu ]</p>
Some of you know that I think Apple ought to make a headless micro-tower iMac. 1 PCI slot and 1 AGP slot for sound and video cards is a must for a consumer machine that wishes to woo windows customers. OR, so I thought, but what if Firewire got a whole lot faster?
Not 400Mbps but 800-1600Mbps, or even 3.2Gbps? Would this be fast enough to enable the firwire bus to effectively replace the PCI bus for most pro-sumer applications? I'm talking multiple devices connected to a processor independent bus that can easily handle real-time (and accelerated) audio-video capture, encoding, surround sound steering, mixing... whatever is needed.
I'm pretty sure this can be done even with the current spec, but the bandwidth overhead of a bus 2 to 4 (or maybe 8) times faster would pretty much make it as fast as PCI? I don't claim to know, but I think there's a slight difference. -- doesn't the current PCI spec top out around 215MB/sec? A 1600Mbps (200MB/sec) firewire bus might finally make us forget the PCI bus altogether. Currently most people use the firewire bus as a fast portable storage bus. They claim the extra speed (above the current spec) is unneccessary for that purpose, and unless you want a really fast firwire RAID set-up, they're probably right. However, I thought that the original mission statement for Firewire was broader than just storage. I think some people already use firewire based sound-cards.
Apple should go for the extra bandwidth, whether storage uses it or not, it could finally answer the expansion niggles of consumer macs, and give a laptop solution beyond anything available to any other platform.
[ 12-03-2001: Message edited by: Matsu ]</p>
Comments
Hmmm.... I just re-read your post. Are you talking about internal or external devices? AFAIK, HyperTransport and such are practically limited to internal devices, so my reply may not be particularly relevent. Oh well, I think I'll post it anyway
I think that to effectively implement your idea, there would have to be places set aside in the tower for industry-standard sized cards... hm... like PCI cards are done now.
I don't think that Firewire is the medium we need to switch our upgrade cards to. Probably something dedicated to that purpose like what Whisper talked about.
-Ender
If a company put a soundblaster Audigy in a firewire enclosure and it could run at the same speed as a PCI version, would you buy it? Given that there really isn't any other way to get this on your iMac?
I suppose that this could work on the FireWire bus, given enough bandwith. What controls the FireWire bus though, isn't it the processor? Is FireWire similar to SCSI, which handles all of it's own stuff, or ATA, which relies on the processor?
If it is the latter, I believe we would need some significantly faster processors before we go and stick a bunch of high bandwith stuff on an iMac. 40 MB/s is one thing, but > 200 MB/s is entirely different. I still remember the early USB systems where the mouse would become jumpy when copying files to a USB Zip drive... we can't have that happen all over again.
-Ender
<strong>it seems to me that everyone is coming up with very exotic ideas about this firewire bus thing... how does this idea hold? If apple were to make firewire bandwidth comparable to pci, why doesn't someone make a firewire pci box that would be very similar to the boxes used to mount internal ide drives in an external box on firewire? What about the magma boxes? I would think they would be the first to make their boxes connect via firewire considering scsi doesn't exist on many consumer machines yet firewire exists on just about every machine. This way developers of cards don't have to make special firewire cards and can stick with pci for everyone.</strong><hr></blockquote>
I like that idea. I'm not sure what the technical hurdles might be, but it sounds good to me
<strong>Wouldn't all the external stuff become unmanageable quickly? I have two PCI graphics cards running 2 monitors (and one on the AGP bus), and a SCSI card which my burner and an internal hard drive are connected to.</strong><hr></blockquote>
What if Apple (or another company) built the video cards directly into their monitors then had them connect through FireWire?
<strong> [SNIP] I still remember the early USB systems where the mouse would become jumpy when copying files to a USB Zip drive... we can't have that happen all over again.
-Ender</strong><hr></blockquote>
Well, that issue had nothing to do with the bandwith of USB, according to an Apple note. They attributed the behavior to a bug in the driver. (I can't remember the title of the specific Apple document, but I know that's the official explanation. I'm pretty sure it's online at Apple's Knowledge Base.)