Re: SCSI drive replacement using SD?

From: Justin <shadow_at_darksideresearch.com>
Date: Fri, 9 Jan 2015 16:21:10 -0500
Message-Id: <7D24403B-94D6-46F4-91D6-6BB0123FAF82@darksideresearch.com>
I wanted to follow up on this and let everyone know how this thing worked out.  Out of the box it was maybe not the best user experience as the utilities for configuring it are not exactly mature yet, but once it works it is nice to have something dead silent to work with.  You have to use a utility to set it to pretend to be an Apple drive, which is a bit of a problem as his newer firmware and GUI for configuring it has a bunch of cool features (such as having it carve up the SD card into multiple virtual SCSI devices) but that utility doesn’t have a —apple switch that makes classic Mac OS’s Disk Utility initialize the drive.  So I picked the last of his firmware revisions that did allow the older command line utility to work since I can live without multiple emulated SCSI devices for the time being and I needed the —apple switch for the Mac to be happy (third party formatting tools would just hang).

I used the FWB toolkit benchmarking suite to bench it relative to my twin 7.2k barracuda on an Adaptec 2940 (weirdly one drive shows up as fast narrow and one as fast wide despite being the same model and firmware).  Attached to the internal SCSI bus on the PowerMac 9600, the SCSI2SD has pretty much dead on read speeds of 2MB/s (1.9xx).  Random writes however are so incredibly slow on this benchmark tool that I thought the thing had crashed.  By comparison, the striped Barracudas pretty much bury this thing in performance. I’ll have to try a less crappy SD card to see if that makes a difference, as the one I am using was a cheap $7 one I bought from a display next to a cash register at a store.

For a more real world test, I moved a system folder and some applications onto the drive and then booted from it.  It has acceptable boot speed - not quite as fast as those two striped drives but one heck of a lot quieter.  Application launch speed is also acceptable.  I did some light browsing in Classilla since I don’t have many applications to play with, and it seems to run acceptably.  I think if I was still using the original internal hard drive and did not have the stripe setup, I would be completely happy with it.  The consistent low latency on reads makes up for its much lower maximum sustained read speed relative to the striped drives.  I’ll be interested to try it in some other hosts that expect much slower drives anyway - or to be able to run the newer firmware on it.  I am using the SD card holder on a long ribbon cable plugged into the micros slot on the SCS2SD itself, which will be great if I permanently mount it in a machine as I can just swap cards to swap SCSI hard drives.

I’ll try to find time to shove it in an Amiga soon, but I’m not anticipating any drama now that I understand how to configure it etc.

Overall, this is a good product.

Justin

> On Nov 3, 2014, at 6:17 PM, Justin <shadow@darksideresearch.com> wrote:
> 
> I think it’s important to maintain perspective on the difference between the nominal performance limits of the bus and the real world performance of the system the bus is included in.  SCSI should’ve done 5MB/s but most of those 68k macs couldn’t do that even if the drive was just a box of RAM because of the realities of their overall bus architecture, clocks, driver quality, very weak multitasking in the kernel, who knows what else, etc. This thing seems to saturate the narrowest part of the pipe on the ones people have posted numbers for.  I have a PowerMac 9600 here because it was the last machine Apple built that could read their 400k floppy format, and is therefore the easiest to get onto a modern ethernet network and talk to something else with. That will probably be the machine that I stick this thing in first.
> 
> What I have not seen is a post regarding its latency.  I would be very interested in knowing what its random read latency is across the emulated drive, particularly as a function of card size and speed rating. A lot of the posts talk about using pretty small SD cards but they only talk about class 10 or unlabeled.  I have a lot of the 32GB SanDisk Extreme Pro cards that are rated for 95MB/s so I’ll probably be using those at least at first. I’ll be interested to see how that behaves. I think that latency has much more impact on the normal user experience than pure sequential read or write speed across the bus.  It’s not as if anyone is going to use this thing in a media server of some kind.
> 
> In that 9600 I currently use two 7.2k RPM Barracuda drives in software RAID on an Adaptec 2940UW card (a PC one that I bought for nothing and flashed the PowerDomain firmware into for a tiny fraction of the price). I’m always worried that these or the drives in my Amiga are going to crap the bed, and they are really annoyingly loud to leave on all the time anyway, so I do not leave these machines hot with the drives spinning. Anyway, I’ll be very interested to see how the system “feels” in terms of boot time and latency booting from this thing. Maybe I’m officially old, I used to have a ton of fans crammed into my PC’s and they all sounded like jet engines, and now I have a Mac Pro sitting here effectively dead silent even with all cores pegged.
> 
> FWIW for those who got errors, the site told me the thing was out of stock but let me place an order anyway.
> 
>> On Nov 3, 2014, at 11:14 AM, silverdr@wfmh.org.pl wrote:
>> 
>> 
>> On 2014-11-03 16:50, Per Olofsson wrote:
>> 
>>> At half the price of a 7720U it looks like a pretty good deal, but it
>>> looks like it's currently limited to about 2.5 MB/s which is a bit too
>>> slow for my taste:
>>> http://www.codesrc.com/mediawiki/index.php?title=SCSI2SD#Performance
>> 
>> That's what I was wondering - if that's the limit of the device or the Mac they tested it with. If that's the limit of the device/firmware then you are not alone in your taste.
>> 
>> P.S. Maybe we should look at it this way: less than half the price = less than half the performance ;-)
>> 
>> -- 
>> SD!
>> 
>>     Message was sent through the cbm-hackers mailing list
> 



       Message was sent through the cbm-hackers mailing list

Received on 2015-01-09 22:00:03

Archive generated by hypermail 2.2.0.