Debunking the SCSI (in a Mac) is faster IDE (in the same Mac) myth?

[Trash80toHP_Mini]

There was a big hoopla back in the day about Apple switching to "cheaper" IDE HDDs back in the day. In other threads we've discussed just how awful Apple's implementation of SCSI was in terms of performance. Now I'm wondering if Apple's awful IDE implementation in the 630 and 580 might actually have been slower yet or if it was a tad faster than the SCSI bus implementation?

Did a very little bit of research to get the ball rolling and discovered that Apple's Disk Tools initialization of  SCSI drives tweaked SCSI interleaving to dumb the drive's inherent setup to better performance on a slow computer, something I'll assume Workstation SCSI implementations didn't require.

https://en.wikipedia.org/wiki/Hard_disk_drive_interface#Performance_and_communication_channels

Theory:

Given the terrible performance of the Mac's SCSI implementation, moving to IDE imposed no  verifiable speed penalty. Uninformed Mac fanatics used the "SCSI is faster" argument against the PC fanatic crowd, but that was a myth based upon the speeds achieved using SCSI implementations in Workstations of the day. Without a Fast Narrow SCSI2 controller card or a Fast Wide Jackhammer card, no Mac could come close to utilizing the benefits of any SCSI drive faster than a run-o-the-mill IDE drive, much less high end SCSI drives. The only reasons IDE was "cheaper" than SCSI (only the control boards were different) would be economies of scale regarding controller board manufacture and considerable cost of the required SCSI controller for that version.

More info?

Thoughts?

More importantly:

Benchmarking of SCSI control board version of identical HDD mechanism against the IDE controller board versionj? Test performance of the 630/580's moronic single HDD (and ONLY HDD) bus against its brain dead SCSI bus implementation?

edit: mods, please insert "than" into title

 

[Schmoburger]

I'm glad somebody addressed the elephant in the room finally... I shall watch this thread with interest and await some meaningful data to prove or disprove the suspicion.

Assuming I can find some identical drives in both SCSI and IDE configuration, and some proper benchmarking tools, I may be in a position to do some test myself at some stage upon one of my Performa 580's.

 

[uniserver]

Well this is interesting.

I know with norton system info,  if you just choose the HD section,  you can clearly see as you go from the SE to the LC to the Quadra  The performance increases greatly! 

not when it comes to the 630 and the 580,  VS the LC 575

maybe someone with these machines can do some benchmarks?

http://macintoshgarden.org/apps/norton-utilities-353

 

[uniserver]

i had a powerbook 145 on my bench last week too... would have been great to run http://macintoshgarden.org/apps/norton-utilities-353 <----- norton system info on both powerbook 170 and powerbook 145 as well.   but i don't have the powerbook 145 anymore it went back the customer with a nice re-capped screen.

 

[Cory5412]

Solid state media that you can connect to these things has been (for several years) loads faster the buses in these machines.

All you need is to find somebody with two pretty similar machines -- perhaps a 475/605 and a 630, and a 575/580. (To be honest, you probably only need one pair of those machine.) Run disk benchmarks on a CF card adapted directly to the 580's bus, and bench the same CF card connected to the 575 via an AztecMonster.

I suspect what you'll find is that in those particular machines, the performance is going to be similar.

Another question is how the AztecMonster and the cheaper SCSI2SD compare. For the purposes of a Mac, it probably really doesn't matter. At the time, CPU speed was at least as big a factor, and today, if you're applying modern usage habits to the old machines, RAM becomes a big factor, etc.

This all compares to today where storage is so important in part because Mac OS X doesn't know how to handle RAM and because it's the most significant bottleneck in computers these days, to the point where the actual interface to any spinning disk hasn't been the slow part for years, and few computers are picking up on PCIe based solid state disks.

But that's neither here nor there. I don't have the relevant machines on hand, but for somebody who does, and who has an afternoon for some benchmarking, I'll throw in that I'm also exceedingly interested in seeing the results.

 

[Trash80toHP_Mini]

I'm glad there's some interest! I dug up one interesting overview while putzing around on googling during some downtime at work:

http://www.milosoftware.com/mike/scsi_ide.html

I was originally thinking in terms of some Western Digital(?) drives with stamped aluminum covers I'd been wanting to replace with clear vacuum formed copies for the heck of it in a clear cased hack. ISTR having them in IDE and SCSI and I might even manage to find them  .  .  .  or not. :

In reading a few articles about the differences and wondering about what to benchmark, I came up with the notion that trying to match the drives, even with adapters still wouldn't be a true test. No two are created equal in either flavor. Even if you use the same IDE drive on both buses with and without the likes of an ACARD adapter, there must be some overhead involved.

Since we have faster, more modern drives available that will swamp both types of controllers in every category, wouldn't that be a better test? I'm not big on benchmarking, but so long as the cache on each drive were to be equal, wouldn't such drives be better at beating the snot outta either type of controller? Rotation speeds would naturally need to be the same, would any other factors be relevant?

Would that be a fairer comparison of the interface capabilities on the IDE and SCSI buses in a 630 or 580? I'm not sure that a 475/580 comparison would mean anything at all really. The only case where such might matter would be if Apple chose a lower grade SCSI controller for its first rev IDE machines. Chip ID would tell that tale.

ISTR there being Blackbirds shipped with SCSI Adapted IDE drives from IBM, unbeknownst to buyers. Did anyone ever even notice a difference?

 

[trag]

What exactly are you trying to determine? Back in the day, the pathetic disk interface performance (SCSI or IDE) really didn't matter in real life, because the real world performance of spinning disks was terrible.

I had a bunch of the first Seagate Barracuda drives, back when 7200RPM could elicit an oooh and an ahhh. They delivered 6 MB/s. Get the ST32550W, put it on a Fast & Wide Jackhammer or SEIV, theoretically capable of 20 MB/s and you'll still get 6 MB/s; that's just how fast the drive can drag the data off the spinning platters.

Create an array of four of them. You might get 10MB/s. The ancient CPU gets bogged down managing the array.

My point is that Apple's slow SCSI implementation just didn't matter that much because few disk drives could outrun it. If you look at these interfaces with historical perspective, they were good enough. There were faster interfaces available, but no drives that could saturate them -- until the PCI Mac era. The 5 & 10MB/s SCSI built into the X500 and x600 machines (PowerSurge family) did get passed up by the drives of the day. But the PowerSurge architecture was around for a long stretch. At the beginning of PowerSurge's run it was speedy compared to real drive performance.

Once the PCI age started, drive performance, what IBM used to list as Media Data Rate, really started to take off, not because of PCI, drive tech just got better at about that time.

So, sure, you put modern, fast drives on the old busses, their limitations really show, but in the day, they were fast enough for the typical hard drive.

I think the Q630's IDE performance was pretty decent at the time. I remember there was a notcicable improvement in performance when we replaced Apple's slow stock drive with a newer 500MB WD drive, so the interface had some head room.

Same story with the Beige G3. It sports a 16MB/s IDE interface, but again, the stock hard drive didn't even challenge that. It wasn't until faster drives became available, that the interface was realistically challenged.

 

[Paralel]

That's the problem with using faster solid state media with older systems, you'll just never get the performance you'd like to have out of them because the interfaces were just never meant for anything that fast.

 

[Unknown_K]

At the time SCSI drivers were faster then the same size IDE drives. Apple could have had a better SCSI bus but they figured nobody was going to have drives fast enough to matter, and if they did they would get a fast Nubus card to use them.

 

[Paralel]

Their thinking only falls apart with regard to the compacts that have either limited or no expansion.

 

[Unknown_K]

Who purchased a compact for video editing? Most people did basic office stuff.

 

[Paralel]

Yeah, but 1.25 or 1.5 MB/sec is damn pokey.

 

[Unknown_K]

Considering the original compact had barely enough RAM to boot and had a serial port connection for external drives consider yourself lucky.

 

[techknight]

1.5MB/Sec back in those days would have been BLAZING and a welcome sight! I remember on 56K dialup it would take a few minutes to download a meg. I remember it taking 15 to 30 minutes to download an MP3...