IDE/SCSI vs. USB 2.0/IEE 1394(b)

[Elledan]

It's probably because I'm overlooking something, but to me it seems like IDE and SCSI are just about obsolete technologies.

One must admit that an USB 2.0-like interface for HD's would make far more sense than even Ultra160 SCSI, for the simple fact that the former is both hot-swappable, provides a higher transfer-rate, can provide all connected devices with power without the need for an additional connector and doesn't have to be terminated. It would also allow many more devices to be connected (127 for USB 2.0; 63 for IEEE 1394(b)), which would save IRQ's.

Now, I know that the transition from IDE/SCSI to a new standard would be difficult, but it would make sense as far as I can see.

Anyone here cares to point out the flaws in the above idea?

 

[CTho9305]

USB 2.0 is 480 megaBITS per second. that works out to 60 megaBYTES per second.
Ultra160 SCSI is, IIRC, 160 megaBYTES per second, or almost 3x the speed.

I also thought you could get SCSI hot-swap trays. you just can't get scsi scanners, video cameras, etc, and scsi needs to be terminated.

edit: the difference, of course, being in the "b" of USB's 480 Mbps vs the "B" of SCSI's 160MBps

 

[Elledan]

Okay, so I neglected to see whether it was a 'b' or a 'B'

But still, SCSI doesn't seem to have that many advantages over IEEE 1394b/USB 2.0.

Hmm... that makes me wonder: why are SCSI peripherals (HD's, CD-drives etc.) so much more expensive than comparable IDE peripherals?

 

[Peter]

Because they aren't comparable. Cache sizes typically are much larger on SCSI drives, they have much more advanced firmware, spare capacity to auto-replace bad sectors, higher performance and much higher reliability (see MTBF values).

On other peripherals like CDRW drives and such (that tend to be more similar in IDE and SCSI flavors), it's a matter of production volumes, and price differences there are much smaller.

regards, Peter

 

[Elledan]

Thank you, Peter

 

[Nefrodite]

how about cpu overhead?

 

[Elledan]

<< how about cpu overhead? >>

As you might know, with IEEE 1394(b), the whole system can in principle operate without any PC, i.e., a digital camera can send its data to a printer, without the need for a PC.

This means that there would be zero CPU overhead.

 

[Peter]

CPU overhead on IDE is always the same, since the command handling (done by the drive) and the bus master engine setup (in the IDE bridge chipset) are always the same.

On SCSI however there are more and less intelligent host adapters. Brainiacs like all the LSI Symbios (formerly NCR) chips and the top-of-the-line Adaptecs and QLogic chips have a RISC engine and RAM of their own, and can handle complex command sequences without ever bothering the system CPUs ... while stupid low-end stuff needs much more assistance, like Tekram's own SCSI chip, Adaptec's low end devices and many more.

With FireWire it's basically the same, how much CPU load you get is a function of the controller chip's capabilities.

regards, Peter

 

[DHL]

One thing I noticed - both USB and IEEE1394 are not bootable, at least at this moment. So both of them can be used as interface of storage devices, but cannot replace IDE at this moment

Also USB (and may be 1394) can only supply 5V @ only 500mA and no 12V, therefore an external pwr source might still be necessary

Serial ATA, on the other hand, might answer all the problem

 

[Elledan]

DHL, that's why I was talking about an "USB 2.0-like" interface, i.e., able to provide the required 12 V and the ability to boot from connected devices.

I agree, though, that IEEE 1394b is an excellent choice for (external) HD's and other storage devices which do not have to be bootable (just storage).

 

[Daovonnaex]

<< One thing I noticed - both USB and IEEE1394 are not bootable, at least at this moment. So both of them can be used as interface of storage devices, but cannot replace IDE at this moment

Also USB (and may be 1394) can only supply 5V @ only 500mA and no 12V, therefore an external pwr source might still be necessary

Serial ATA, on the other hand, might answer all the problem >>

Nah, SATA is only running at 600 megabits/second, which doesn't come close to catching SCSI or IEEE 1394 2.0 (1,344 megabits/second).

 

[Nefrodite]

article on firewire and usb 2.0

anyone have one with benchmarks and stuff?

 

[Peter]

Serial ATA will be 150 MBytes/s, folks. Hardly a match for SCSI's 320 MByte/s that we'll have until then, and not a replacement for FireWire's peer-to-peer and hotplug capabilities either. Each attachment methods has its market and reason, in no way is one of them going to replace the others.

regards, Peter

 

[Mday]

serial ata is slated to go into the giga range with further revisions.

the concept of irq itself is outdated.

 

[Elledan]

<< the concept of irq itself is outdated. >>

Please elaborate

 

[Uuplaku]

One more advantage SCSI has over the external drives are its silly-fast spin rates. It'll be a cold day in hell when Firewire/USB2.0 drives can spin up at 15,000 rpm!

 

[Elledan]

<< One more advantage SCSI has over the external drives are its silly-fast spin rates. It'll be a cold day in hell when Firewire/USB2.0 drives can spin up at 15,000 rpm! >>

Why would this not be possible?

 

[RegisPhilbin]

<< One more advantage SCSI has over the external drives are its silly-fast spin rates. It'll be a cold day in hell when Firewire/USB2.0 drives can spin up at 15,000 rpm! >>

current Firewire / USB 2.0 Drives are just IDE drives encased in a firewire /usb 2.0 enclosure....if you stick in a faster IDE drive, it will be a faster firewire/usb 2.0 drive....

Does anyone have remarks on burst speeds / latencies due to interface ?....i'm curious if those are variables.....