Fastest hard drive setup... USB 2.0, SCSI, Firewire?

[kabir2]

Just wanted to know what setup would give me the fastest hard drive.

 

[AkumaBao]

Uh, well. SCSI of course. That is if you feel like shelling out the extra $$$ to get it. As for the other two, I would say the firewire would be you best bet. (and probably more affordable)

 

[kabir2]

I was confused about that... I thought scsi was... but I read that USB 2.0 transfers at 460 mb/sec. Fastest SCSI setup I could find transfered at 400 MB/sec. Someone correct me if I'm wrong, just what I read though

 

[Tripleshot]

USB2 is faster than firewire and SCCI. It soon will be the norm on mobos, instead of the udma dinasaur. You can get external HDDS with USB2 hoookup at Staples.

 

[joeryu]

<< USB2 is faster than firewire and SCCI. It soon will be the norm on mobos, instead of the udma dinasaur. You can get external HDDS with USB2 hoookup at Staples. >>



uh, no. scsi is by far the fastest, and USB 2.0 will NOT take over ata. next in line is Serial ATA. technically usb 2.0 should be faster than firewire, cause usb 2.0 is rated at 480 mbps, while firewire is rated at 400 mbps. I doubt u will see a difference though.

 

[Goosemaster]

<< USB2 is faster than firewire and SCCI >>



I'm sorry, but you are very wrong. SCSI is still the fastest internal/external medium.

Fire wire: 400Mbits/s(Megabits) or 50MB/s(Megabytes)

USB 2.0: 480MBits/s or 60MB/s

SCSI: 2560Mbits/s or 320MB/s

You tell me which one is the fastest? see what I mean? These are the capacity of the bus, and as with other mediums other than SCSI, the amount of bandwidth quickly dwindles to nothing.


As for misc. devices such as webcams and DV camaras, they are promising on firewire which caters to these devices....SCSI does not. But never compare two different types of mediums.


Just clearring things up.

By the way, Ultra640 SCSI is coming out too. iSCSI is on the way too. ALlows for SCSI to transfer over ip...tell me that isn't cool...

 

[FishTankX]

Allright, well, Goosemaster is right. But his figures are a bit out of proportion.

He gave you the numbers for a SCSI 320 controller which is very pricey now. I believe that you could just go with SCSI2 and be fine if you're only using one device.

But there is one *very* important reason why SCSI is faster than USB2 and Firewire

here are more reasonable figures

Firewire=50MB/S
USB2=60MB/S
Ultra160=160MB/S
Firewireeer to Peer
USB2=slave based system
Ultra160=Peer to Peer

Firewire:Medium expensive
USB2=Elcheapo
Ultra160=Very expensive

Benefits of Peer2Peer:Suppose you were homeschooling your children. In a Peer2Peer network it'd be the equivelent of them homeschooling themselves.
With USB2's system, you would not only have to homeschool them yourself, you would have to check every single problem they ever did.

USB2 can consume 10% of the cycles of an Athlon800 just for moving a mouse! Think about how a 50MB/S harddrive would do!

And thus lies the true speed advantage. DMA like properties.

SCSI and Firewire are both independent of the CPU.

USB2 relies on the CPU for everything.

Just Firewire/SCSI would feel faster than a USB2 setup. Espically when transfering stuff between drives in the USB2 chain.

And just for your refrence UDMA will *not* be replaced for USB2. That would be a step back. DMA is impossible in a USB2 enviornment. And Ultra100 has 50% more bandwidth than USB2. It's just not gonna happen, Trippleshot. Remember the days where DMA didn't exist and Harddrives relied on the host CPU?

*shudders*

 

[Patenter]

Seems to me that you ought to consider the disk drive itself too, rather than just focusing on the interfaces (which seem to be fast enough??). As you probably know, SCSI drives are currently the fastest platter spinners -- spinning at 10,000 or even 15,000 RPM as compared to 5,400 0R 7,200 RPM as in the case of the current group of ATA/EIDE and external Firewire drives (and I assume USB2 drives also). The extra revolutions usually translate to faster access/reads/writes and unfortunately more expensive drives with less storage space.

For what it's worth:

I have a new 2 GHz P4 workstation that is running both Ultra160 SCSI and Ultra ATA 100 drives. According to my Matrox Disk Benchmark program, my two new 10,00 RPM Ultra 160 SCSI drives (18 GB and 73 GB) each have a minimum/average/maximum write capability of about 23/27/59 MB/sec and a read capability of about 23/27/39 MB/sec. My new ($90 backup) 80 GB 5,400 RPM Ultra ATA 100 drive is just a little farther behind -- it has a minimum/average/maximum write capability of about 13/27/39 MB/sec and a read capability of about 14/21/29 MB/sec.

Note, that according to my humble benchmark (which I know nothing about) both the Ultra 160 SCSI and Ultra ATA 100 interfaces are fast enough for my disk drives.

Big price difference however, my 73 GB SCSI drive was $800 -- my 80 GB ATA drive was only $90! I'll let you do the math, as I'm affraid too.

 

[LethalWolfe]

My 2 cents... Firewire 2 is in the final stages of coming to market and I bet we'll see Firewire 2 cards before the end of the year. And Firewire 2 is, IIRC, going to be 2xFirewire speed so 800Megabits/100Megabytes.


Lethal

 

[Wolfsraider]

<< According to my Matrox Disk Benchmark program, my two new 10,00 RPM Ultra 160 SCSI drives (18 GB and 73 GB) each have a minimum/average/maximum write capability of about 23/27/59 MB/sec and a read capability of about 23/27/39 MB/sec. >>






<< There also seems to be a big bug in XP related to slow SCSI performance and WinXP. >>



AND



<< HDtach isn't that good of a utility for SCSI drive benchmaking... use ATTO or IOmeter >>



so i wouldn't trust those benchmarks just yet

this came from 2cpu.com in their storage area

Patenter
you can read it here

also
storage review has more to base your judgements on harddrives

hope this helps

 

[Patenter]

Thanks for the info Wolfsraider.

I don't really trust too many benchmarks anyway -- especially if I don't know how they work, which is most of them.

That said, I must admit that reading the suggested links reminded to the fact that my reported results are skewed against the SCSI because write caching was disabled for the SCSI drives and enabled for the ATA drive. Sorry about that. Also, the Matrox benchmark doesn't measure access times, as might the other suspect HDD benchmark programs.