IDE versus SCSI

[Scotty35]

Hi all,

I thought this was interesting enough for me to post these results.

<b>Using HDTach 2.52 for XPpro I tested the following hard disk drives for read speed only:</b>

1 x IBM Ultrastar 73LZX 18.3GB U160 10k SCSI hard drive (Boot/OS)
Max: 584444 Kps
Min: 7523 Kps
Average: 42824 Kps
Random access time: 7.9 ms

1 x IBM Ultrastar 36LZX 18.3GB U160 10k SCSI hard drive
Max: 36100 Kps
Min: 21200 Kps
Average: 28500 Kps
Random access time: 8.5 ms

4 x Western Digital WD12000JB in raid 0+1 array (240GB)
Max: 60387 Kps
Min: 26129 Kps
Average: 45783.5 Kps
Random access time: 13.9 ms


<b>Related hardware:</b>

Epox 8k5a3+ M/B with onboard 4 ch IDE raid.

Adaptec 29160 SCSI controller card.


<b>Ok lads start disecting that lot </b>

Now keep in mind these IBM drives are certainly not top of list of best drives and are a bit dated, and I am sure other brands would score better, also and the PCI bus would possibly alter the true reading of the SCSI drives due to the limitation of its bandwidth.
But still, I am impressed with the Raid 0+1 array performance in the test and has changed my way of thinking in regard to the SCSI versus IDE arguement.



<A HREF="http://forums.btvillarin.com/viewtopic.php?t=326" target="_new"> My networked systems</A>

 

[lhgpoobaa]

Yes... bout what i would expect.
2 times the write speed of one 30mb/sec ish drive.

Nice setup. Lovely and fast and REDUNDANT 😀


<b><font color=purple>[Rik_]</font color=purple> I wonder how many people have made their own phasechange system?
<font color=blue>[LHGPooBaa]</font color=blue> I get phasechange whenever i eat a hot chillie :lol: </b>

 

[Ncogneto]

Now keep in mind these IBM drives are certainly not top of list of best drives and are a bit dated, and I am sure other brands would score better, also and the PCI bus would possibly alter the true reading of the SCSI drives due to the limitation of its bandwidth.
But still, I am impressed with the Raid 0+1 array performance in the test and has changed my way of thinking in regard to the SCSI versus IDE arguement.

Not quite sure I understand the comparison. You are comparing essentailly 2 of the newest IDE drives in a striped set to one dated SCSI drive by itself? Hardly a fare comparison. Even still the fastest SCSI drive you listed more than keeps up and its access times are quite a bit faster, myself I would rather my OS reside on this drive than the array.

It's not what they tell you, its what they don't tell you!

 

[Scotty35]

The comparrison was to prove and justify to myself the continueing use of the SCSI's, also by announcing the results it may give an idea of the SCSI v IDE performances differences if anyone was interested.

No you are right, it is not a fair comparrison as the SCSI would have to have been in a simular raid config', but I found it interesting enough to point out that even an older U160 drive can keep up with some of the best IDE drives in a Raid situation and this is why they remain as my boot, program and pagefile drives.

Also as I mentioned above that the PCI bus is a limiting factor for the SCSI devices at 32/33 where a pci bus of 64/66 like those only found on a dual cpu boards (afaik) would allow my SCSI controller to work at its full potential and increasing the bench scores.
The final reason to post the scores is to inform the ill-informed that SCSI is still faster than IDE (even in raid)in a practicle sense in my mind with all things being fair and equal.

* If anyone can show me a single CPU board with a 64/66 PCI slot/bus I would like to see it.

<A HREF="http://forums.btvillarin.com/viewtopic.php?t=326" target="_new"> Scotty's Toys</A>

 

[Ncogneto]

Ahhh I see well I took this comment:

But still, I am impressed with the Raid 0+1 array performance in the test and has changed my way of thinking in regard to the SCSI versus IDE arguement.

To mean exactly the opposite of what yor were trying to get across.


It's not what they tell you, its what they don't tell you!

 

[Scotty35]

Well yes, it is kind of baited and is meant it to go either way to encourage debate.

But still the raid did impress me and I also feel better about the SCSI's I invested in, what I learned is that even fast IDE raid arrays are still no match for todays SCSI HDD's, contrary to some peoples beliefs, so I had to find out for my self.

So my answer to the IDE V SCSI question of which is better would be; It depends on the application and the budget requirements as SCSI will cost twice as much per GB completely setup with controller(non-raid)IMO.

Anyway, these are my own thoughts and opinions so take it with as much salt as you like :lol:



<A HREF="http://forums.btvillarin.com/viewtopic.php?t=326" target="_new"> Scotty's Toys</A>

 

[UsHeR_564]

I too have wondered about these setups and the info is appreciated. One thing that keeps me from doing anything with it is the fact that once a game is loaded (the primary use of my computers), doesn't it run from RAM memory anyway? And that being the case, both RAID and SCSI are moot point for me. I spent the money on 512Meg CAS-2 2700DDR instead...🙂 UsHeR_564

"You can run, but your punk ass will only die tired!"

 

[MarkN]

I kinda found the IDE verse SCSI comparison meaningless, as all it did was show that an IDE drive gives the same performance through an IDE controller as through a SCSI controller. Which just says that the IDE drive is the limiting factor.

I have no never-dying love of SCSI (hell, it ain't like BASIC to Billg) but what I always want to know is this: Is the current generation of SCSI worth the extra cost over the current generation of IDE? What I would like to see:

Compare:

SCSI 160 10K 32bit PCI
SCSI 160 10K 64bit PCI
IDE 133
IDE Serial
SCSI 320 15K 32bit PCI
SCSI 320 15K 64bit PCI

And compare the costs as well.

 

[Crashman]

If you're running RAID in multiple channels, U160 is no faster than Ultra 80. Why? The PCI slot is only 133! Most Ultra 80 cards will run both channels at full speed, 2x80=160=133, because you're limitted to 133MB/s by the PCI bus!

Now the higher RPM's are helpfull of course. But I don't think the 15k drives are worth the extra money.

If you want to get into SCSI RAID cheaply, I have a server card here that I can't afford drives for. It supports up to 45 hard drives on 3 channels in any RAID configuration you can think of. I installed 64MB of cache also. You should buy it from me and then get yourself a few 10k used drives to get started.

<font color=blue>You're posting in a forum with class. It may be third class, but it's still class!</font color=blue>

 

[theoldwizard]

The numbers posted are "interesting", but they overlook the, IMHO, big advantage of SCSI. Or to say it another way, why do "big" servers still use SCSI when IDE is Soooooo much cheaper ?

It's not the raw numbers. I'm sure Seagate, IBM and others use the same housing, motor, actuator, arm, and read/write amplifiers on both their SCSI and IDE products. Obviously the elecritcal interface is different, but we haven't gotten to the "meat" of it yet !

SCSI drives have 2 big advantages.

First, they can "queue up" multiple I/O requests from the host adapter. IDE can only handle 1 request at a time (the next gen serial IDE addresses this).

Second, it can do "out of order" execution (not the correct term, but, hey it's late). In other words, if the host send 5 commands, say from 5 different users, then command 3 might be returned first (because the heads were nearest to that hunk of data), then 2, then 5, then 4, then 1. The processor (yes, there is a CPU on the disk) re-orders to requests for "optimal" perfomance.

These benefits don't show up in most disk performance tests. At least none the disk performance tests that I know of.

Why do people buy IDE, if SCSI is better for multiple I/O's ?

$$$

And your Aunt Millie probably isn't running Linux or trying to listen to her MP3's while surfing the web and re-compiling the latest set of GNU tools (i.e. simulating multiple users)

Now my beef is, why are the disk manufactures continueing to r_pe the *nix world with such enormous price differences between SCSI and IDE ? Don't tell me volume. SCSI is a cash cow for them ! "Hey we change a few component for an additional $10-20 dollars and we can charge a extra $100-200."

What a SCAM !!!

 

[MarkN]

This assumes that the SCSI controller is always saturating the PCI bus. 160 is faster between the controller and the drive, and it is also not to say that the controller and drive controller for 160 devices aren't "smarter" than the ones for 80s - the 160 is a newer "generation".

 

[Crashman]

80-160 will only make a difference on a server PCI slot (either 64-bit or 66MHz) or on a single channel. The drives (even fast ones) are not faster than the controller.

<font color=blue>There are no stupid questions, only stupid people doling out faulty information based upon rumors, myths, and poor logic!</font color=blue>