A Look At the Fastest IDE Drive Yet

muks writes "Here's an article on Tom's Hardware about IBM's Deskstar 75GXP. It has some good points on why we still need UltraATA/66 and faster IDE interfaces while hard drive transfer rates don't keep up. "

Comments

back in my day

[acomj]

we didn't have these fancy namby pampy hard drives. We had cassete tape recorders to load and save programs.. And we liked it..

Now I feel old.

But at what price?

[technomancerX]

Very nice... now how about some price comparisons:

IBM 20GB 75 GXP $136
IBM 60GB 75 GXP $417
Seagate 18.4GB X15 $482

(figures from pricewatch)

So that's over 3x the capacity for LESS THAN the
same price... SCSI rocks if money is no object,
but it'll still be a while before I build a PC
with it...

.technomancer

Re:WHY the absurd price differential?

[Anonymous Coward]

I used to write SCSI firmware and the reason SCSI drives cost more is this, they have more features and functionality, fewer SCSI drives are sold each year causing parts to cost more. Also SCSI drives usually have lower areal densities so they need more disks and heads.

For example most SCSI hard drives support the following features:

Command Queuing and reordering, support restricted and unrestricted reordering

Mode pages which allow you to adjust the following:
Read-write recovery
Verify recovery
Caching
Logging
XOR controls
Send receive diagnostics
Skip read/write
And many more

EIDE has very few parameters that you can adjust so you have less code complexity. The sad thing is most operating systems don't take advantage of some of the features of SCSI. For example tag queuing is not taken advantage of in most operating systems. Windows 95 never sends a queued command and NT never really uses much queuing (less than 4 outstanding commands at most that I have seen in a bus trace) and if you think Linux does, think again. I looked at one device driver that only supported one outstanding command per initiator and some other UN*X's are not much better. We had to develop our own tools to get 64 or more outstanding commands to a disk drive to measure re-ordering performance gains.


I wish people would research the differences before making an opinion that has no credibility. Yes, Joe Blow user running Windows 95 won't see much of a difference running Word and Excel or playing UT, but a smart OS that understands SCSI will have significantly better IO and data integrity from disk to controller.

Just my $0.02

1394 not fast enough by a long strech...

[maynard]

It's definitly fast enough, 200 Mbits/sec IIRC.

No, 1394 supports 100mbit, 200mbit, and 400mbit transfers, not just 200. However, even given 400mbit Ultra2LVD SCSI at a theoretical burst of 160MB/sec will beat the pants off of a serial 400mbit connection. Think about it, that's 40MB/sec without protocol overhead, and it's shared (though so is SCSI). Why do you think most external firewire drives on the market are just IDE disks with a IDE to 1394 interface? It's not because of price.

I'm about ready to buy a new system and recently looked into moving from SCSI to 1394... forget it. I still want 1394 for it's multimedia potential -- yes, those SONY Digital handycams with 1394 look real promising. But fiberchannel, 1394 ain't.

Frankly, why don't people use a two tiered approach with their systems. Folks here always degenerate the argument into an either/or debate. Either Linux or Windows, either GNOME or KDE, either IDE or SCSI; what bullshit. I use SCSI for boot and swap, and IDE for cheap storage. Do I need gigs of mp3's on a fast boot disk? No. Do I want my OS and swap on a set of chained master/slave IDE drives, with all the known contention this involves? No.

Use both to your advantage. You'll love the system speed of SCSI for what really counts, and the savings of IDE for what's a little less important.

Fastest IDE hard drive, not that fast

[Jeffrey Baker]

Compare the fastest IDE hard drive (IBM 75 GXP), with the fastest SCSI hard drive (Seagate X15), using the IOMeter benchmark's workstation access pattern:

• Total I/O per second: X15 117% faster
• Total throughput: X15 116% faster
• Average I/O response: X15 53% faster

Thanks to Storage Review for the benchmark figures.

And remember kids, maximum length on those UltraATA cables is 18 inches(!), versus a luxurious 12 meters for Ultra160 SCSI, which transfers data much more quickly anyway.

Bah!

[GoRK]

The old (E)IDE/ATA vs. SCSI flamewar again. I can't believe it! All I have to say is BAH

Each have their own applications.

As far as new equipment goes, SCSI is FAR more expensive. Yes I know you can buy old 9GB SCSI drives for peanuts and hook them up in a RAID and beat the pants off of any IDE out there. What about new systems? What if you dont have space or cooling for 8 cheap scsi drives?

What about devices such as the TiVo that demand 30 gigs of storage in a single drive (not to mention the other electronics) to make a profit at $500? Try doing that, SCSI!

After this, it may sound like I'm pro-IDE but really I prefer SCSI. But I'm not a bigot either and like to give IDE its fair due.

I am quite impressed with ATA/100. I set up a RAID-0 with the HPT370+Raid ATA/100 controller which came built into my new Abit mobo and two Maxtor 20.5GB ATA/100 Drives with 4M cache each. There is one drive per chain per channel. Ive got 41GB of storage now and I'm getting burst transfers from cache of 170-180MB/s! (In other words, my max transfer is better than Ultra160 SCSI) Also, I will never need to add extra drives etc in this particular system, so SCSI doesnt have it on me there. To top it all off, all the files on this system that are getting shuffled around are in the neighborhood of 100K. It's all run through the cache. SCSI would have cost me an extra $1000 to get the same performance. That's all there is to it.

~GoRK