IDE vs SCSI

[Prince]

Comparing SCSI with IDE, how much is the performence difference? (I know SCSI is best if everything is SCSI, wats "Everything"? HD and?)


P.S: Using the version of IDE / SCSI that currently main-stream in the market. Thanx in advnace

 

[CocaCola5]

It depends whats being done, for outright transfer rate the best IDE can be nearly as quick as SCSI. But not often do straight transfer speed matters, for everyday use, SCSI rules because of its 3-4 times faster seek/access time.

 

[Jerboy]

<< Comparing SCSI with IDE, how much is the performence difference? (I know SCSI is best if everything is SCSI, wats "Everything"? HD and?)


P.S: Using the version of IDE / SCSI that currently main-stream in the market. Thanx in advnace >>




SCSI supports bandwidth splitting and simultaneous transfer.

When you have two HDD's on one IDE channel, a master and a slave you can't access the drive simultaneously if I recall correctly.

On a U 160 SCSI with 15 drives, all the drives can be accessed at 10.5MB/s each at the same time. Not something a typical G /-\ /\/\ ! |\| G F /-\ |\| boys need.

 

[Peter]

not quite. IDE channels are blocked until the device completes the current command sequence -
meaning that the other drive on the same channel can't do anything.

SCSI allows asynchronous and out-of-order completion of commands, so that the time between command
issue and data return isn't wasted. These features are called Tagged Queuing and Disconnect/Reselect.

Whazitmean? On IDE, if you use a drive at 100 percent of its possible _media_ throughput, the channel is
blocked, regardless of channel bandwidth. On SCSI, each drive consumes only as much as it actually needs,
idle time is open for others to get some work done.

Example, say you have an UDMA33 capable IDE CD-ROM reading a large file at 4 MB/s (about 24x speed).
You'd think that your UDMA100 IDE channel would be more than 80 percent idle, but it's actually blocked
completely because the IDE system can't issue another command while one is being worked on.

Now move the stuff to SCSI, your same CD-ROM drive with an even worse Ultra-SCSI (20 MB/s) interface
will, at 5 MB/s media throughput, eat only 25 percent of your U160 channel bandwidth, leaving the rest to
others.


regards, Peter

 

[Wolfsraider]

thanks peter anything else you can say about scsi or ?

 

[jfunk]

You can put a lot more devices on SCSI.

SCSI is more expensive.

SCSI can also be used for external devices (scanner, etc..)

Thats about all I can think of. Not sure if SCSI RAID implementation is any better than IDE or not...until recently it was all there was.





j

 

[KGB1]

SCSI just rules. There is nothing better. I got my adaptec adapter (UW 160) last week, hooked up 2 cheetahs (18gig , 10K) I love it man, I have lots of videos and transferring them has never been faster. I got 1 old scsi scanner and it works like new. (who needs usb) They said firewire would replace scsi, NOT gonna HAPPEN!!! SCSI will be around even when EIDE is REPLACED by Serial ATA. Hehe so heres the facts.
1. EIDE will not be around after summer, being replaced by Serial ATA
2. SCSI will just get faster (upcoming UW360 or is it 320?)
3. IDE device is only 7200 rpm... pfft SCSI is 15K RPM on some Cheetah models.
4. ATA 100 can barely go at ATA 33. Please what a joke.
5. SCSI access faster and MUCH more QUALITY!!!!(last way longer than IDE)
6. WHY do MOBO makers have integrated IDE?? Because its cheaper to implement IDE on a mobo than SCSI? We all know scsi have been better for yeas.. why have computer makers implement IDE istead, HINDERING the speed of a computer??

 

[Peter]

One last point, SCSI HDDs. Two advantages here:

(1) There are no really high performance IDE drives - simply because the
lack of multi request capabilities on the interface defeats the purpose of having fast media.

(2) SCSI drives have internal features that drastically enhance reliability. They have factory
reserved spare areas, which they secretly and quietly
use when areas in the regular storage tracks get weak - so unlike IDE, where you lose data when writing to a bad spot and
subsequently get read errors there, SCSI HDDs notice the write weakness, internally replace that spot with a spare, re-write
the data into the latter, and all is fine. The host system will never know until the drive has gone really bad and run out of spares.

IDE drives do have internal factory defect lists, but they don't manage the so called grown defects. These do happen.
In fifteen years of using harddrives I have yet to see a bad sector from a SCSI drive, but I have seen piles and piles of lost data
and scandisk/reformat action on other interfaces, MFM, RLL, ESDI, IDE, you name them.

regards, Peter

 

[Peter]

btw, KGB, the original idea behind IDE was to build something cheap. Remember MFM or RLL? You'd have an ISA
controller card and one or two stupid drives. IDE moved this controller onto the drive electronics, and gave the
drive a stripped-down ISA interface. So back then, you'd put a couple of line driver chips onto your ISA bus, and
there's your IDE channel. The controller is in the drive now (if you want two drives on the cable, you disable one
controller by setting the respective drive to "slave" - then the "master"'s controller will take over).

Next, the IDE channel got separated from ISA - still functioning the same way, but with faster timings. In came PIO modes 1 and 2,
and shortly thereafter Enhanced PIO modes 3 and 4. Finally, a DMA engine was added on the mainboard side of things,
enter Bus Master IDE.

Still today, the controller is in the drive, and all the chipset does is shovel data and commands to and fro without caring
what's actually going on. (This is also why _ANY_ stupid old board can technically handle any size and flavor of IDE drive,
even those above 128 GBytes!)

SCSI requires some amount of grey matter on the host system side, thus it's more expensive to do ... which is why at some early point,
the market split up because those people who wanted cheap bought IDE, so the drive makers pushed IDE there and moved
SCSI to the high performance sector.

regards, Peter

 

[Trygve]

<< You can put a lot more devices on SCSI.

SCSI is more expensive.

SCSI can also be used for external devices (scanner, etc..)

Thats about all I can think of. Not sure if SCSI RAID implementation is any better than IDE or not...until recently it was all there was.
>>



Onboard IDE raid controllers and inexpensive add-in cards only offer mirroring (redundancy, but you need twice as many drives for the same effective storage space) or striping (better performance and it looks to the computer like it's a single big drive instead of multiple little ones--but if one drive goes, everything's gone).

With SCSI, I can get controllers like the AMD Enterprise 1500 (64-bit PCI, two Ultra2 LVD channels) or 1200 (32-bit PCI, three ultrawide channels) for $75-$150 on the surplus market, and they'll handle up to 30 or 45 drives, hardware RAID-5 (if you have seven drives, you get six drives worth of storage and speed and any one of them can fail without even shutting down the system; you swap out the bad drive and it regenerates it in the background).

With that kind of controller, it's got its own CPU, its own memory (typically up to 128Meg on the card), and in the case of the 1500, a little battery pack that preserves any to-be-written cache if the power fails.

If you want to get ambitious, it's also possible to implement load-balancing cluster servers with one or more shared SCSI busses, which also allows a fault-tolerant array of drives to be shared between multiple machines running in parallel, allowing fault-tolerance on the machine side as well. (I've only done this on test systems; I've never put this type of setup into service).

For a typical desktop system, IDE is lots cheaper; for a server or a system that's doing a lot of audio and video editing, SCSI does offer some significant advantages.