At 12:23 PM -0500 02/03/2003, Mike wrote: >According to Adaptec,SCSI or at least the latestSCSI, is still the least >processor intensive I/O capable of simultaneous data processing of >multiple devices.
Yes, that's true, to a point. It takes less processor time to set up the DMA command blocks necessary to do a SCSI operation than it does a Firewire op. But we're talking a diff of only a few instructions & bytes here. A few nanoseconds at that point is nothing significant, IMO. The real issue is what happens (and how fast) after the bus has the commands. Firewire has the potential to leave SCSI in the dust. (Keep in mind also that companies have had YEARS to optimize their SCSI drivers; leaving that investment behind isn't going to happen overnight). [ObUnSimplification] These busses are pretty smart. The bus and processor talk by leaving each other "mail" in the form of encoded commands in RAM, then they kick each other with DMA Requests (interrupts). The target reads the "mail", does what it says, then kicks the other guy when its done. In the old days, the busses were dumb - the processor had to talk to the peripherals directly and tell them to do each individual task (position head, read sector x, send data, etc). Very time consuming. >Despite the claims that many, many devices can be >connected to a firewire bus, the bus becomes degraded as more devices >are added. This is because each device grabs a portion of that bus >when it is connected whether it is actually being used or not. The real degredation is primary due to power availability. If you want to run a lot of power hungry devices, you just have to get self-powered ones. The same goes for USB. The alternative would be for Apple to have designed Firewire to carry something like 110v AC. Then we'd have cables bigger than SCSI-1 to drap all over the room! The secondary degredation is due to token passing. It's trivial, nothing compared to power issues. >So if I understand this correctly, it certainly would explain why Apple or >other third parties haven't pushed peripheral devices for firewire, >especially if they would take up excessive bandwidth on the bus and >degrade performance for the higher throughput devices that Apple >intended. Exactly. Mixing high and low speed devices is Bad. Very Bad. A bus full of large packets produces very very high throughput (maximizing your bandwidth usage). Add the itty bitty packets produced by slow devices into that - and it's like parking a bicycle across a major highway in rush hour. This issue will become more profound as busses get faster. This problem, btw, has been studied and simulated at length, resulting in a number of published papers and dissertations. Insert terms like token, cdma, etc. IOW, slow peripherals need to be on slow, lower priority busses, and fast peripherals should be on faster, high priority busses. Now... USB 1 vs USB 2 vs Firewire 1 vs Firewire 2. USB is a great protocol for slower peripherals. Firewire (IEEE-1394) is a great protocol for faster peripherals. Computers got faster, but Firewire 1 was only 400 Mbps! :( So Apple et al upped the anty with Firewire 2 at 800 Mbps. They also laid the groundwork for "Gigawire" - firewire at 1 Gbps and faster. MS blinked. IDE is too slow. But let's face it - hell would freeze before they paid Apple royalties! So they pushed and got Intel et al to come up with USB 2, a cheesy wedge onto 1 that competes with Firewire 1. Good enuf for the lemmings to use... All in the name of Competition, er a, Not Invented Here. So we're in the exactly the same boat as we were x years ago when IDE came out. The world was moving toward SCSI when suddenly MS pushed this slower cheaper interface into the market. The MS mentality proclaimed "cheap is great!" and that was that. The server market is SCSI and moving toward Firewire 2. The peecee market is IDE and moving toward USB 2. Personally, hell will have to freeze before I buy fast peripherals and put them on a USB 2 bus. I couldn't care less if Apple ever "supports" USB 2. IMO, their energies are better spent moving us to firewire 2. *shrug* *duck* - Dan. -- PCI-PowerMacs is sponsored by <http://lowendmac.com/> and... Small Dog Electronics http://www.smalldog.com | Refurbished Drives | -- Sonnet & PowerLogix Upgrades - start at $169 | & CDRWs on Sale! | Support Low End Mac <http://lowendmac.com/lists/support.html> PCI-PowerMacs list info: <http://lowendmac.com/lists/pci-powermacs.shtml> --> AOL users, remove "mailto:"; Send list messages to: <mailto:[EMAIL PROTECTED]> To unsubscribe, email: <mailto:[EMAIL PROTECTED]> For digest mode, email: <mailto:[EMAIL PROTECTED]> Subscription questions: <mailto:[EMAIL PROTECTED]> Archive:<http://www.mail-archive.com/pci-powermacs%40mail.maclaunch.com/> Using a Mac? Free email & more at Applelinks! http://www.applelinks.com
