Serial ATA vs. SCSI - Will it Compete?

fazzumar asks: "I've been checking out serial ATA (SATA) and it seems like it's got a lot of potential. The first generation spec was finalized August 2001 and members of the SATA group anticipate a 12-18 month acceptance period. They've planned for a cut-over phase and adapters that allow connecting SATA devices to ATA adapters and vice versa. The cables alone are a worthwhile advantage (4 pins, up to 1 meter in length), and the 150MB/sec bandwidth is a (minor) improvement over current ATA drives & adapters. Infoworld has a story on SATA that provides a few tidbits of information. What I really want to know is, will manufacturers of the new host adapters be able to integrate many of the advantages that SCSI provides or will the cost of adding these features push the retail cost too high for the anticipated market?" I just picked up a new WD Hard drive just yesterday for the planned MP3 jukebox I hope to be building near the end of the summer. I really wanted to go SCSI, but couldn't. While the poster claims a near ~7x in price difference, I saw about a ~5x difference in my local store. Is SCSI in danger of falling behind IDE drives (especially serial IDE drives) in popularity?

"I love SCSI, and I can bring myself to accept the additional cost of the controller, but with IDE hard drive prices dropping, I frequently wonder if SCSI drive prices are artificially inflated. Just a few years ago, SCSI drives were ~10-20% more than IDE and now they're ~7X more than an IDE drive. (Seagate 10k RPM SCSI - ~18 gig for ~175. Western Digital 7200 RPM IDE - ~120 gig for ~175) If the option comes out to get SCSI performance from an IDE drive I'm going to take it."

Comments

Damn hard drive mannufacurers.....

[reaper20]

I frequently wonder if SCSI drive prices are artificially inflated.

Hell yes - There's no way they aren't. I'm sick of this price barrier myself.

Look at a SCSI drive and an IDE drive. Sure, there are some differences, MTBF, blah blah ... but generally speaking, they're the same thing. There are certainly not enough differences to justify the price. Is there some magic spell they cast on SCSI drives that quadruples the price? SCSI's "enterprise capabilities" make using SCSI on the desktop really expensive, so they continue to gouge us.

What do we get in return, Technology that Should Not Exist(tm) - Things like IDE RAID.

Sucks being a SCSI zealot. :(

Re:Damn hard drive mannufacurers.....

[mosch]

How many of you drive a car with a Wankel engine? This is the sorry fate of humanity.

If it wasn't for slashdot, I'd never get to see the failure of the rotary engine described as a bellweather signaling the collapse of humanity. Thank you!

Re:Damn hard drive mannufacurers.....

[mosch]

I mean you no disrespect, but you're an idiot.

Hard drives prices are set by a complicated formula where somebody sits down, looks at the costs involved in developing the drive, the costs involved in testing the drive, the costs involved with creating the equipment to manufacturer the drive, the number of units they expect to sell, the expected return rate and the cost per unit of manufacture (and a lot of other things). They then attempt to set a price which will result in a reasonable profit.

Why attempt to set a price which results in reasonable profit? Because if one aims for an unreasonable per-unit profit, a competitor (and the hard-drive market is hardly a monopoly) will decide that they can sell enough extra units by lowering their price that they'll undercut the first vendor.

So SCSI drives happen to be the most expensive in nearly every fashion. They're the most likely to implement cutting edge technologies, thereby funding the refinement of these technologies for later use in cheap IDE drives. 15K RPM SCSI drives have been around for quite some time now, but I'm not aware of any IDE drives that beat 10K RPM yet.

SCSI sells fewer units, thus the fixed costs are distributed over a much, much smaller number of units, driving up the price dramatically.

SCSI drives are more likely to be returned as defective. Not because of a higher failure rate, but because SCSI drives are likely to be put places where they get used heavily, 24 hours a day. IDE drives tend to get used in applications which involve idling 98% of the time, so even cheap shit will have low return rates.

Additionally, the SCSI command set is larger than the ATAPI command set. In fact, ATAPI is just a standard way to communicate a subset of SCSI over an IDE connection. This means that SCSI drives really can do more things than an IDE drive... for instance, they can accept multiple commands, then execute them in a more intelligent manner than an IDE drive, which only knows how to do one thing at a time (IBM DTLA and DPTA not withstanding)... So this SCSI drive is also more complicated than the IDE counterpart, thus requiring more testing, and more design, raising the costs further.

Saying that SCSI is artificially overpriced is like claiming a WRC rally car is artificially overpriced, because you managed to get 500hp out of your Dodge Neon by bolting on a turbo, intercooler, straight pipes and a cold air intake.

Somehow I think that my explanation for the price disparity between IDE and SCSI is a lot more plausible than one which requires every single hard-drive manufacturer on the planet to be organized in a cabal, whose sole purpose is to drive up the price of server-class hardware.

Re:Damn hard drive mannufacurers.....

[HalfFlat]

Somehow I think that my explanation for the price disparity between IDE and SCSI is a lot more plausible than one which requires every single hard-drive manufacturer on the planet to be organized in a cabal, whose sole purpose is to drive up the price of server-class hardware.

Have you looked at the tape-drive/media market recently? My money's on the cabal.

Re:Damn hard drive mannufacurers.....

[scotch]

When the new Maxda RX-8 comes out next year, you can bet I'll be driving the rotary engine (again).

Re:Damn hard drive mannufacurers.....

[HalfFlat]

That's one hell of a rejoinder to a throw-away comment.

More specifically: there are much larger margins on server-targeted hardware, such as tape drives, than there is in the desktop arena. While 5-10% margins are common at the retail level for consumer hardware, tape systems seem to start at 20% and go up from there.

Even acknowledging the fact that these are mechanical devices and so will naturally not benefit so greatly from the price drops in other computer hardware, tape backup prices have barely fallen at all in the last 4 years. As higher capacities and speeds are introduced, they just come out at higher and higher pricepoints.

There doesn't need to be an organised cabal - when you have a market which is willing to pay these prices, and where there are often established long-term relationships between the buyers and vendors, it's unsurprising that there hasn't been the same downward pressure on prices. Unlike in consumer hardware, where there is typically little brand loyalty and price-competition is fierce.

Final concrete example: getting SCSI hardware in an external enclosure typically costs about 12% more than when it is bare. At the cheap end, it's about what you'd expect to pay for a decent enclosure. At the expensive end, it's about 3 or 4 times the retail cost of a similar enclosure. Where's the competition driving prices down here?

Oh, and don't be so quick to call people retards. It's rude on a number of levels.

Apple seems to think ATA is better...

[knabar]

Apple uses ATA in its new Xserve rackmount server. (See here and here)

Quote: The ATA drive subsystem has a high-bandwidth I/O bus that minimizes bottlenecks, even when all four drives are engaged at once. That's how Xserve can achieve a theoretical peak performance of up to 266 megabytes per second, compared to a 160MB/s theoretical performance with SCSI Ultra160 disk drives -- at a significantly lower cost, and while generating less heat than SCSI drives.

Re:Apple seems to think ATA is better...

[DietFluffy]

Apple uses ATA in its new Xserve rackmount server.

Jobs also said Apple is "humble" as it enters the [server] market. "For everything we know, there are 10 things we don't know," he said.

Perhaps he was referring to the use of ATA instead of SCSI for a "serious" server? *grin*

ATA becomes more popular than scsi

[DietFluffy]

Is SCSI in danger of falling behind IDE drives (especially serial IDE drives) in popularity?

In other news, Microsoft Windows takes a majority of the desktop operating system market.

Re:ATA becomes more popular than scsi

[tenman]

The major difference is Queuing and Parallel Processing. Even today's IDE RAID controllers can't get more than one instruction down the pipe at the same time, and queuing is non-existent. Recently I have seen IDE cards that support extra RAM, but specs on the boxes don't read like they support queues. SATA will introduce the concept of a queue, and that will speed I/O greatly, but the biggest speed barrier is the speed at which the instructions are sent to the disk.
I run a small video editing shop. Real time video editing taxes disk I/O on PC's more than anything else I know of. I HAVE to run SCSI, and still, even with SCSI, I have to wait. Most of the time I am waiting on the I/O.
That being said, after reading the SATA standards, I will feel safe replacing more than half of my high-speed I/O channels with it. I am willing to have it manage everything except the most extensive reads and writes. I will soon be dedicating a portion of my web site to the explanation of SCSI vs. SATA. Bookmark it now.

scsi performance in an ide drive

[DietFluffy]

If the option comes out to get SCSI performance from an IDE drive I'm going to take it."

Review of the Western Digital 1200JB (The 8MB Cache Special Edition.)

With desktop performance and capacity vastly superior to the competition as well as a surprisingly low operating temperature, the Caviar WD1200JB reaffirms Western Digital's preeminence in the IDE desktop performance segment. In fact, for desktop usage, the JB bests all 10k RPM drives save only Maxtor's Atlas 10k III.

Once again we're obligated to point out an interesting fact. The hardware enthusiast market, comprising a significant portion of StorageReview.com's readership, has always pledged it would respond enthusiastically to the world's first 10,000 RPM drive. These folks want the performance of a 10k RPM SCSI drive without the SCSI premium. The WD1200JB, like the WD1000BB-SE, delivers the desktop performance of a good 10k RPM drive according to tests constructed from real-world, high-level applications. If you want SCSI's performance without its price or capacity limits, the WD1200JB is the drive for you.

Re:scsi performance in an ide drive

[nosferatu-man]

I have two. They're hellaciously fast, to say nothing of MUCH quieter and cooler running than a 10k SCSI, let alone a 15k.

'jfb

Re:scsi performance in an ide drive

[mosch]

Interesting portion of the article to quote. Odd that you didn't mention that the Maxtor Atlas 10k III (a U160 SCSI drive) beats the WD1200JB by:

• 88% in the SR File Server DriveMark 2002
• 85% in the SR Web Server DriveMark 2002
• 43% lower average read service time
• 39% lower average write service time
• 20% in the Business Disk WinMark 99
• 16% in the SR Gaming DriveMark 2002
• 13% higher transfer rate beginning
• 10% in the SR Office DriveMark 2002
• 8% higher transfer rate ending
• 2% In the High-End Disk WinMark 99
• 1% In the SR High-End DriveMark 2002
• 0% (a tie) in the SR Bootup DriveMark 2002

In fact, there's no performance test where the IDE drive in question beat the SCSI drive in question at all. Next time you make an argument, you should really cite a source that supports your argument, on the off chance that somebody reads it.

Just bought Qlogic 2200 Fibre Channel adapter, $50

[NoMoreNicksLeft]

Sometimes, you have to love ebay. I still need a few drives, for which the smaller ones go about $1 per gig. 256 drives before I run out of room. The only thing that will hurt me price wise, is a decent rackmount drive case. Software raid, at very nearly the max theoretical throughput.

SCSI was nice, but it is dead. IDE is alive, but it sucks. Fibre channel is seen as an enterprise-only tech, and they no longer have use for the small stuff on their SANs. I'm not so proud that I won't humbly accept their table scraps.

Oh, and before anyone else mentions it... I don't have to use expensive fiber. That's only for ridiculously long cable runs. 20 meters max for stp cat5. More than enough.

Adaptec's Opinion

[xrayspx]

I remember being at Adaptec Live SPAM Session at MS Technet a few years back. The Adaptec rep told of his many trips to drive manufacturers, particularly Seagate.

He said that SCSI drives come off the same lines as IDEs, if the batch can handle running at 10000 RPM, they're Cheetahs, if they can do 7200, they're Barracuda SCSI's, if they have a lower MTBF, they're IDE 'cudas.

I never really believed him, since the physical drive cases are different, but his contention was that they were all the same drives, with different electronics slapped on depending on MTBF.

Take it for what you will, BIG Grains of Salt...

I'll...I'll put strychnine in the guacamole.

Re:Adaptec's Opinion

[xrayspx]

The 'who cares' factor is decided by the 'how fast can I run this without it melting' factor.

I agree that the MTBF must be the result of more stringent engineering, maybe, who knows. I myself think the guy is full of shit, but what do I know? I was jaded by the fact that logically, of friggin COURSE he's going to say that, he wants to sell more Ultra160 SCSI adaptors, duh. But maybe he wasn't full of shit, I leave that decision as an excersise for the reader.

Re:Just bought Qlogic 2200 Fibre Channel adapter,

[NoMoreNicksLeft]

Am I bad, you're right on the 127 figure. Not that I'll get close to that either.

Re:Just bought Qlogic 2200 Fibre Channel adapter,

[NoMoreNicksLeft]

The card itself is on the way. Need to get drives yet. I really want a nice looking rackmount box... but I don't even have a rack yet. Would love to be able to get matching 4U ATX and disk enclosure cases. My current home server is a crappy old dual p100 tower, but I'm getting ready to build my rackmount dual athlon server. Heck, I may even build more than one, eventually, cluster the things. The fibre channel raid is primarily going to be used for my soon to be rackmounted DirecTivo's (only have 2 so far, gonna up it to 4 directivos soon).

Re:What ever happened to fay ray...

[larien]

SSA is still alive and well, although it tends to only get used with IBM systems. The IBM marketing says it will work with other vendors (including Windows), but I haven't seen it used elsewhere.

When motherboards support it

[jhines]

When the next generation of motherboards comes out with support for 4-8 serial disk ports, as well as the USB, Firewire, LAN, and other ports, then ATA can be left for cdroms and the like.

Re:Taged Comand Queueing

[Chaostrophy]

Is what seperates SCSI from IDE, but it's in the IDE spec, and people are working on it for the 2.5 linux kernel, there have been two recent patches implementing it.

Right now, you can bet that most drive and chipset suport for it is bugy as hell, but if people start to use it, it will improve.