Jump to content


Photo

Solid State Drive (SSD) in SE/30


  • Please log in to reply
67 replies to this topic

#41 johnklos

johnklos
  • 68000
  • LocationEarth

Posted 12 January 2010 - 08:36 AM

The difference, if measurable, between CF and SATA on a polled I/O SCSI bus on a 16 MHz m68030 is going to be so insignificant as to be noise.

Pure speculation until I see actual benchmarks. Benchmarks are very easy too. Most of us have a copy of Norton Utilities, which includes a benchmarking program. There were also other popular benchmarking apps for 68k Macs.

No, not pure speculation - it's observation, although not observation specific to the SE/30. Take, for instance, the Amiga 3000. For quite a while it had the fastest SCSI bus of any non-Unix workstation and was used to benchmark drives. It's much faster than the SE/30's SCSI bus because it supports real DMA, has a FIFO to allow for 32 bit bus transfers and can even be persuaded to do synchronous mode. When I compared an 80 gig 7200 RPM 3.5" IDE drive on a SCSI to IDE adapter with a 250 gig 5400 RPM 2.5" SATA drive in an Acard SATA to SCSI adapter, there was no measurable difference. There was more of a variability between individual tests than there was between the different drives.

I'll let you know how things go when I compare the CF card with the SCSI-SATA.

#42 bbraun

bbraun
  • 6502

Posted 12 January 2010 - 08:54 PM

I have benchmarked a couple of storage solutions I had lying around, including:
Seagate Cheetah 10k RPM ST9102LW
SanDisk UltraII 4GB CompactFlash
Quantum ProDrive LPS connected externally
Kingston 8GB Standard CompactFlash (no manufacturer speed rating)

I had difficulty formatting a chart correctly in this post, so results are available at: http://synack.net/~bbraun/se30disks/
JDW indicated a preference for the Norton Utilities benchmark, so I used System Info from 4.0.3.
When interpreting the results, please read the notes I wrote at the bottom. There seemed to be pretty high variability in the results between multiple runs in the same configuration.

Sadly, I do not have any UDMA capable CF cards as that would provide an interesting comparison. As it stands, the low end CF card, which does not seem to have a speed rating from the manufacturer, performs the poorest. The UltraII card, which is pretty middle of the road, running in PIO mode, does as well as the 10k RPM Cheetah drive within the variability of the benchmark.

#43 JDW

JDW
  • 68000
  • LocationAichi-ken, Japan

Posted 12 January 2010 - 09:47 PM

bbraun, I certainly appreciate your time in testing CF solutions versus spinning platter drives with Norton "System Info."

It is interesting to see from your results that even the rather speedy SanDisk Ultra II did not beat the Cheetah in any of the tests except Random Read, and even then it wasn't an enormous difference. Overall the SanDisk Ultra II was a few points slower than the spinning platter drive (albeit a fast one).

The low end 8GB card you tested scored less than half of the SanDisk and Cheetah, although it did marginally better on the Random Read test. But clearly, that 8GB flash is a choice none of us would prefer in terms of performance.

It is interesting to note that the SanDisk Ultra II CF card is rated at 15MB/sec speeds for both READ and WRITE. You would think that should saturate the rather slow SCSI bus on the SE/30.

I wish to point out that while this test was only conducted on two different flash cards, it does confirm what I personally felt prior to the test. Namely, that given the specifications of the flash card, one would naturally think that the SanDisk Ultra II 4GB CF should saturate the SCSI bus of the SE/30 and be the winner over spinning platter drives in every test. But the fact that a spinning platter drive, the Cheetah, was in fact marginally faster than the SanDisk and winning all the drive tests except one, indicates that we cannot simply look at the flash card specifications alone. Benchmarks really do help us determine the performance of CF solutions for our vintage Macs.

Further benchmarks performed on spinning platter drives are available for viewing from this LEM SE/30 Page.

If you want to run your own tests but lack the software (or the knowledge where to download it freely), just PM me.

Thanks again, bbraun.

#44 bbraun

bbraun
  • 6502

Posted 12 January 2010 - 10:37 PM

It is interesting to see from your results that even the rather speedy SanDisk Ultra II did not beat the Cheetah in any of the tests except Random Read, and even then it wasn't an enormous difference. Overall the SanDisk Ultra II was a few points slower than the spinning platter drive (albeit a fast one).


I disagree with your interpretation of the benchmark results here. These are single data points with an observed variability of up to 12%. There is not sufficient data to be drawing the conclusions you are with any confidence at all. I'd really rather not dust off my old statistics book and do a proper work up on these benchmarks. That'd eat up the better part of a weekend to do properly.

It is interesting to note that the SanDisk Ultra II CF card is rated at 15MB/sec speeds for both READ and WRITE. You would think that should saturate the rather slow SCSI bus on the SE/30.


It is not entirely clear to me that both the Cheetah and UltraII card, and in some tests even the Quantum drive, aren't maxing out the machine. I lack sufficient knowledge of the system and protocol to make an educated guess in this area. There may be other bottle necks than the SCSI bus and the device. Clearly both are capable of much higher speed transfers. The standard for the protocol would indicate a 5MB/s bus, and both devices can exceed that in faster systems, yet we don't see anything close to that in these results. Clearly smaller transfers require a higher overhead of both bus throughput and CPU processing. Again, I do not believe there is sufficient data to support the conclusions you are drawing.

Further benchmarks performed on spinning platter drives are available for viewing from this LEM SE/30 Page.


After preparing these benchmarks, I am skeptical. For one thing, the article makes the assertion the SE/30 SCSI bus maximum is 4Mbps, yet provide data that it is operating at higher speeds (>14Mbps). Case is significant folks! Even if they meant the bus maxes out at 4MBps, their data doesn't get to 2MBps so I'm not sure where the 4 comes from. Additionally, they present data without specifying how it was gathered. That may simply be the result of the TimeDrive program they used to benchmark, in which case I argue the results are not very meaningful. I do not believe the results gathered there are comparable to the results gathered with Norton.

I personally use the lowest rated Kingston CF card and am very happy with it. I'm not out to run the fastest possible system, and I'm not out to convince you of anything at all. However, you have discounted informed reasoning as guesses, and misinterpreted gathered data to preserve your own preconceptions. If you really want to stick with a spinning platter or a SATA SSD or something else, I suggest you go for it. You don't need to justify or rationalize your decision to us.

#45 bbraun

bbraun
  • 6502

Posted 13 January 2010 - 01:11 AM

JFYI, I've updated the charts to include the SanDisk Ultra II and Cheetah drive on the G4 to get a feel for whether the SE/30 is maxing out the device or not.
The update includes the UltraII card both on the G4's PATA bus as a slave, and using the SCSI-IDE converter setup as in the SE/30.
http://synack.net/~bbraun/se30disks/

The UltraII card has a much higher top end than the Cheetah drive. The SCSI-IDE converter seems to max out at about the 5MB/s of narrow SCSI-1 even though it is supporting the SCSI-2 protocol. But all of the above exceed the maximum values we're getting out of the SE/30.

#46 JDW

JDW
  • 68000
  • LocationAichi-ken, Japan

Posted 13 January 2010 - 01:40 AM

I disagree with your interpretation... I'd really rather not dust off my old statistics book... I personally use the lowest rated Kingston CF card and am very happy with it. I'm not out to run the fastest possible system, and I'm not out to convince you of anything at all. However, you have discounted informed reasoning as guesses, and misinterpreted gathered data to preserve your own preconceptions. If you really want to stick with a spinning platter or a SATA SSD or something else, I suggest you go for it. You don't need to justify or rationalize your decision to us.

There is no need to get so flustered over this. I apologize for having clearly upset you, but my intention was not to anger anyone. Such has never been my intention through all the years I have posted in this excellent forum.

Nevertheless, I stand by what I have said about the need for comparative benchmarks. I also stand by my thanks which I extended to you for providing such benchmark data. You did not have to do that, but you kindly did so. Indeed, I am truly humbled by the amount of time you clearly spend on that dedicated web page for the results. Again, I wish to thank you for that. No doubt others appreciate your efforts as well, even though they have not yet offered you their thanks yet.

But just because one or two among us is "satisfied with the performance of a low end CF card" does not mean we all must follow suit and be satisfied with that. I don't say this to upset you at all. I state it merely as a fact. Some are satisfied with the 16MHz processor in a stock SE/30. I have a 50MHz DiiMO 030 in mine. It's still an SE/30 running 68k software, upgraded yet in a largely stock condition. I simply say this to point out that the SE/30 allows for us to add various things to it to make it faster, and no, I am not talking about transforming it into a G4 or Intel Mac Mini either. I am talking about taking a stock Mac and upgrading it such that it can continue to operate as it was originally intended (to run 68k software natively on the CPU), but at faster speeds. That is what this thread is all about, but of course the "flash drive" discussion extends to Mac other than the SE/30 as well.

Such that we can all make "better informed decisions" I suggested some benchmarks in this thread. Then for whatever reason, the harsh words began. Perhaps some of you are hurt that I trust in benchmarks more than personal experience or educated guesses (however good those guesses may be). But I've not been attacked like this in other threads, so I am rather confused by all the harshness. Again, if I have offended, I apologize.

But I must stand by my original call for benchmarks, because such data is helpful in determining performance. And quite frankly, the more benchmarking data we can compile, the more easy it will be to select an appropriate flash drive solution for our vintage Mac, in accordance with our individual needs on performance.

Some seek the lowest cost solution. I seek a low cost solution, yet with the best performance possible. I personally want to know, without a shadow of a doubt, that the flash drive solution I am going to pay money for, for use in my vintage Mac, is going to be FASTER (in most situations/tests) than the 7200rpm spinning platter drive I currently have installed. The only way to know which is the best is to have a robust number of comparative benchmarks.

If such benchmarks were utterly irrelevant, then it would not have been necessary for Macworld/Mac User magazine to offer them all these years, nor would Macworld today need to use their Speedmark testing. The fact is, there are many among us, myself included, who still like to see benchmarks. Right or wrong, benchmark data makes us feel better. And I personally think it helps me make a more informed purchasing decision. My main reason for even participating in this thread is because I have been in the market for a flash drive solution for my SE/30 for years, but I've not purchase yet because I've been holding out for comparative benchmarks.

From reviewing your excellent benchmark results web page, I see that the SanDisk Ultra II card is very close to the 10krpm Cheetah, which is a very fast drive to have in an SE/30, and faster than the 7200rpm IBM DGHS server drive I currently use in my SE/30. So if I dare to "guess" here (based upon what I read in your benchmarks anyway), it would probably be that even a faster card like the Extreme III or IV series may not yield much greater performance than what your benchmarks have shown. Of course, we would need to conduct a benchmark with a faster card to know for certain. But as you point it, it may be that the faster cards would not exceed the numbers shown in your Speed Info results.

So again, thank you for testing. You were very kind in doing so. I greatly appreciate it. And if others who own flash drive solutions would like to add to the benchmarks, then we all benefit as a result.

THANK YOU!

#47 tt

tt
  • 6502
  • LocationCalifornia

Posted 13 January 2010 - 06:42 AM

Thanks bbraun, these results are encouraging. I am in the process of procuring hardware for the conversion. I am a little put-off about getting one of the IDE-SCSI adapters that plug into the back of a CD drive since it seems it would be difficult to cleanly install it inside the case. Does anyone have some photos of their solution to installing one of these adapters?

Please backup and copy that floppy. Preserve softwaredeletion is irreversible.


#48 JDW

JDW
  • 68000
  • LocationAichi-ken, Japan

Posted 13 January 2010 - 07:01 AM

I also have a question along side tt. Do any of these flash drive solutions have an output for the LED?

As you know, some vintage Macs like the SE/30 have an LED on the front case that allows you to see drive access. I currently have that hooked up to my IBM DGHS drive and it works perfectly. It would be nice to have the LED also work with a flash drive solution.

#49 MacJunky

MacJunky
  • 68000
  • LocationBC, Canada

Posted 13 January 2010 - 08:29 AM

If you are using a SCSI -> IDE board then you can use an IDE -> CF board with an LED on it. Would have to remove the current LED and stick some wires on there but it is easy enough to do.
http://www.divshare....oad/7966012-6b0
LC 630, P580CD, LC 575, Mac Plus X2, PowerBook 160, PowerBook 150 X2, PowerBook 145

#50 Von

Von
  • 6502

Posted 14 January 2010 - 12:12 AM

My Acard AEC-7720U scsi to ide card has a connection that drives an LED. When I had this installed with a 3.5 inch IBM 7200 IDE drive in my SE/30, I had the drive mounted backwards in the drive sled so that the Acard and drive power and scsi connections were pointing towards the front of the case rather than the rear (stock orientation). This allowed me to use the stock LED (with short lead wires) built into the drive sled to connect to the Acard and this gave me the desired flashing of the drive light when it was in use. I don't think I have a photo of this rig as I too am working my way towards an SSD but in a slightly different route. My plan was to connect the Acard to a 4GB Transcend IDE flash drive that I picked up from Buy.com: http://www.buy.com/p.../207652979.html.

The snag with this approach is that the Acard and the Transcend drive have the same connection on them (like what is on an IDE ribbon cable) so I needed to track down some sort of gender changer/coupler. My first attempt was with this http://cgi.ebay.com/...#ht_1074wt_1051 however this straight pin approach doesn't line the pins up correctly. Next I tried this ribbon cable http://cgi.ebay.com/...1#ht_500wt_1037. Going this route initially didn't work in that the SE/30 (stock configuration with 8MB RAM) didn't see the drive when I booted to a 7.01 floppy. The Acard was in the same configuration as it was when it was successfully working with a hard drive so I don't think that was the issue. Next I pulled the Transcend drive and installed in a Mac with an IDE connection, formatted the drive and copied a working 7.01 install onto the Transcend. I then installed the Transcend back into the SE/30 and booted to the Sys 7 floppy and now the drive appeared. I used system picker to bless 7.01 system on the Transcend drive and then booted successfully to the Transcend drive. The issue that remains is that after the system boots, it will lock up when I attempt to open any programs. From there I need to power off... That is were I left the project a few months ago. I'll have some free time in the next couple of weeks to fiddle with this again. If anyone has thoughts on what went sideways, I'd love to hear them.

Assuming I can get this working, I was planning on using 3M double sided tape to secure the Acard and Transcend drive to the drive sled after I coat the area under the pieces with electrical tape. One more thing...when the system was booting with the Transcend drive, the Acard LED was flashing..

#51 bbraun

bbraun
  • 6502

Posted 14 January 2010 - 02:03 AM

If anyone has thoughts on what went sideways, I'd love to hear them.


I've found differences in behavior and performance depending on what does the partitioning and filesystem creation. The most reliable combination I found for 4GB and larger drives was to put them in an OSX machine, partition them using Disk Utility using an Apple Partition scheme, but only 1 partition of free space. Then take the drive and put it into the SE/30, and create a Mac Driver partition and Macintosh partition using a patched version of HD SC Setup 7.3.5 (patch mirrored here: http://synack.net/~b...p-735-patch.hqx) to allow partitioning of non-Apple disks. Just remember, only 2GB or smaller partitions for maximum compatibility.

The next best solution I've encountered is Lido: http://www.euronet.n...toud/lido7.html

#52 JDW

JDW
  • 68000
  • LocationAichi-ken, Japan

Posted 14 January 2010 - 02:27 AM

The most reliable combination I found for 4GB and larger drives was to put them in an OSX machine, partition them using Disk Utility using an Apple Partition scheme, but only 1 partition of free space. Then take the drive and put it into the SE/30, and create a Mac Driver partition and Macintosh partition using a patched version of HD SC Setup 7.3.5

Does this advice pertain to flash drive solutions? I ask because my IBM DGHS spinning platter drive is 4.5GB (i.e., "4GB and larger") and I formatted and partitioned it on my SE/30. I run System 6, 7 and 8 on it flawless. I have done so with the SE/30 in stock condition and upgraded. I too used the patched Apple HDSC app though.

In other words, I must assume the "reliability issue" being spoken of here is being directed merely at flash drive solutions for vintage Macs.

#53 bbraun

bbraun
  • 6502

Posted 14 January 2010 - 02:59 AM

Does this advice pertain to flash drive solutions? I ask because my IBM DGHS spinning platter drive is 4.5GB (i.e., "4GB and larger") and I formatted and partitioned it on my SE/30. I run System 6, 7 and 8 on it flawless. I have done so with the SE/30 in stock condition and upgraded. I too used the patched Apple HDSC app though.


I did not intend to propose this as the one and only guaranteed to work in all cases and nothing else could possibly work solution. Merely what has worked most consistently for me in similar situations.

#54 johnklos

johnklos
  • 68000
  • LocationEarth

Posted 03 February 2010 - 10:45 AM

Here are some actual numbers showing fast CF cards on a Quadra 605's SCSI bus. The differences in speed will be less noticeable on an SE/30. This Quadra is 40 MHz and it is running Mac OS 8.1. The cards are connected via an Addonics CF to IDE adapter which is then connected to an Acard AEC-7720U IDE to UltraSCSI adapter. The 16 gig CompactFlash card is a Transcend "133x" and the two 2 gig cards are both SanDisk Ultra II ("15 MB/s") cards. The base reference is a 5400 RPM 300 gig Hitachi HTS54323 2.5" SATA drive connected via an Acard ARS-2000SU SATA to UltraSCSI adapter / case.

Test System: Disk speeds
MacBench Version: MacBench® 4.0
Name: John Klos
Organization:
Test Date: Tuesday, February 2, 2010 7:25:59 AM

Quadra 605 (bbk) SATA-SCSI Quadra 605 (bbk) 16 gig CF Quadra 605 (bbk) 2 gig CF-SCSI Quadra 605 (bbk) 2nd 2 gig CF-SCSI

Disk Tests:
Disk: 295 250 208 218 score
Publishing Disk: 204 160 152 157 score
Sequential Read 512: 554.57 424.10 379.80 372.71 kilobytes/sec
Sequential Read 1K: 905.70 688.57 628.56 618.76 kilobytes/sec
Sequential Read 32K: 3733.92 2976.15 2932.26 2929.68 kilobytes/sec
Sequential Read 64K: 3965.87 3148.29 3126.13 3121.35 kilobytes/sec
Sequential Read 128K: 4042.00 3229.06 3218.32 3212.37 kilobytes/sec
Sequential Read 1024K: 4255.29 3372.93 3359.72 3354.13 kilobytes/sec

Random Read 512: 63.85 268.64 251.88 248.18 kilobytes/sec
Random Read 1K: 121.79 410.56 391.13 386.83 kilobytes/sec
Random Read 32K: 2129.38 2885.72 2846.58 2838.69 kilobytes/sec
Random Read 64K: 2849.16 3133.56 3112.95 3105.43 kilobytes/sec
Random Read 128K: 3500.34 3277.61 3267.98 3260.38 kilobytes/sec
Random Read 1024K: 4192.22 3398.23 3384.74 3377.70 kilobytes/sec

Sequential Write 512: 1093.36 972.48 956.25 993.21 kilobytes/sec
Sequential Write 1K: 1389.43 1208.63 1176.33 1243.98 kilobytes/sec
Sequential Write 32K: 2245.34 1826.06 1734.37 1904.70 kilobytes/sec
Sequential Write 64K: 2659.02 2167.95 2083.36 2218.39 kilobytes/sec
Sequential Write 128K: 2909.86 2361.84 2307.74 2408.79 kilobytes/sec
Sequential Write 1024K: 3257.64 2602.49 2532.98 2654.68 kilobytes/sec

Random Write 512: 249.00 102.68 32.15 36.21 kilobytes/sec
Random Write 1K: 467.74 278.42 63.92 72.11 kilobytes/sec
Random Write 32K: 2193.76 1693.77 243.38 282.05 kilobytes/sec
Random Write 64K: 2652.49 2079.39 473.03 527.24 kilobytes/sec
Random Write 128K: 2957.37 2258.68 970.47 980.85 kilobytes/sec
Random Write 1024K: 3291.23 2559.92 2032.18 2121.75 kilobytes/sec

#55 ojfd

ojfd
  • 6502

Posted 03 February 2010 - 05:03 PM

My SE/30's stock Apple drive was getting *very* noisy indeed, and a replacement 9gig IBM didn't improve things at all.

/Anders


Just FYI -

One of the quietest narrow 50-pin SCSI drives that I've ever tested (and I've tested > 50 different models in the past) was Seagate ST318418N - 18gig, 7200 rpm. The other one was ST318417N, also quiet, just slightly slower. Both models are (were) relatively recent - I think I've tested them in 2007, on loan from the dealer. I kept the ST318418N and now it resides inside the overclocked Q605 @40 MHz. In fact, Q605's fan is louder than the drive.

So, Anders, just google a bit and I'm sure You'll be able to find one somewhere. At the end it's much more elegant solution than all that IDE/ATA/adapters etc. hassle.

Just make sure the boot partition is 2 gig or less and away you go :-)

ojfd

P.S. That Q605 mentioned above is a real workhorse - it runs Cubase 3.0.1 (not XT and not VST, just plain MIDI) with two Opcode's Studio 4's under MROS (no OMS) and it's timing is almost on par with ST1040 :)
Looking for: WGS-9150 m'boards, Apple TwinTurbo 128MA Rev.3.7 (short), Asus TX-97 (Socket 7) m'boards

#56 jellyfin

jellyfin
  • 6502

Posted 02 March 2010 - 05:18 PM

Thanks for the tips on this thread. I installed a 40GB Kingston SSD using the ACARD ARS-2000SUP. I had a hell of a time getting it to format in my SE/30, but based on this info formatted it on my MacBook Pro and then things went fine in System 7. The box is SO much quieter now.

#57 jellyfin

jellyfin
  • 6502

Posted 02 March 2010 - 07:16 PM

Hm. I spoke too soon. I can't seem to get the SSD to auto mount or to be the boot disk.

Any thoughts? Partitioning worked fine with Apple SC HD Setup (patched), and I could copy things to the drive, run them off of it, etc., but when I restart it's not mounted and I can't get it to boot. I have to open Apple SC HD Setup to get it to mount (by clicking update) or I can hit mount in Lido.

#58 johnklos

johnklos
  • 68000
  • LocationEarth

Posted 02 March 2010 - 09:39 PM

The ARS-2000SUP has different firmware than the ARS-2000SU. I had a problem where my ARS-2000SU units would not auto boot on m68k Macs or on Amigas when they had the ARS-2000SUP firmware. I downloaded the latest ARS-2000SU (no P) firmware from the Acard site and flashed it (had to install Windows on a spare motherboard to do it... what a pain!), and everything auto booted just fine after that. I bet your issue is the same.

Separately, there are instances where the Mac SCSI driver which gets installed onto the drive isn't the right one. You may have both problems, but I'd recommend fixing the firmware problem before looking at the SCSI driver problem.

#59 jellyfin

jellyfin
  • 6502

Posted 03 March 2010 - 05:41 PM

Hmm. That means I'm going to need a SCSI card for my Windows box? Is that how the firmware update is done?

#60 johnklos

johnklos
  • 68000
  • LocationEarth

Posted 03 March 2010 - 07:55 PM

At least temporarily. Acard has flash update programs for Mac OS 9 (maybe 8.6, too), but not for OS X. However, the program to update the SATA-SCSI devices is apparently Windows only. I only upgraded mine because my drive kept spinning down way too often and the newer firmware fixed that.




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users