Comparison of SCSI2SD v5.0b and BlueSCSI v2 on Centris 650

[Stan Johnson]

Hello,

Below is a MacBench 4.0 omparison of the SCSI2SD v5.0b and BlueSCSI v2 on a Centris 650 with 136 MiB memory, running Mac OS 8.1. (The screenshots were reformatted by Finn Thain for better readability.)

Methodology:
1) Write zeros to all sectors of a 16 GB SanDisk Industrial microSD card.
2) Install the card in a SCSI2SD v5.0b and assign the first 1 GiB of the card as SCSI ID 0. Enable SCSI2 and SCSI Disconnect.
3) Copy a Mac OS 8.1 partition to the SCSI2SD.
4) Run tests on a Centris 650:
SCSI2SD v5.0b <--> CD-ROM <--> Centris 650
(terminated)
5) Move the SD card to a BlueSCSI. Without an MBR partition or ExFAT filesystem, BlueSCSI interprets the SD card as raw sectors (1 GiB), and assigns a SCSI ID of 1.
6) Run the Centris 650 tests again:
BlueSCSI v2 <--> CD-ROM <--> Centris 650
(terminated)

SCSI2SD did better on some tests; BlueSCSI did better on many others. I wasn't able to find anything from Apple regarding the maximum speed of the Centris 650 SCSI bus.. According to MacBench, the BlueSCSI reached 4200 kBytes/sec for sequential and random 1024k reads. The SCSI2SD did better than the BlueSCSI for 512 and 1024 byte reads and writes, except for sequential writes.

Notes:
1) I chose the SCSI2SD v5.0b instead of the v5.1 because the 5.0b has termination resistors while the v5.1 is terminated via a firmware setting, and the termination doesn't seem to work well on some systems, including the 650 and Lombard PowerBook.
2) 128 GB SD cards seem to be unstable on BlueSCSI. The ExFAT partition becomes corrupted and all data is lost). 64 GB (and smaller) cards work without any problems.

 

[cheesestraws]

where are your zuluscsi and macsd benchmarks?

 

[Stan Johnson]

I didn't do those. ZuluSCSI is too expenive ($150 USD compared to $50 for BlueSCSI). And, as I understand it, MacSD uses FAT32 filesystems instead of ExFAT, so it only supports file sizes (raw partitions) up to 4 GiB, which is big enough for Mac OS, but not big enough if you also want GNU/Linux and NetBSD. However, if anyone has ZuluSCSI benchmarks, I would like to see them.

 

[cheesestraws]

($150 USD compared to $50 for BlueSCSI).

what

https://store.rabbitholecomputing.com/ZuluSCSI-Compact-RP2040-p/zuluscsi-compact-rp2040.htm - $49.00

are you in marketing, by any chance?

 

[saybur]

I was also about to ask where on earth that number came from. The price difference between BlueSCSI and ZuluSCSI is IMHO close enough to be negligible.

big enough for Mac OS, but not big enough if you also want GNU/Linux and NetBSD

This is awfully arbitrary reason to completely discard MacSD. For many (most?) use cases people here are likely to have a 4GB partition limit is not a big deal and the device offers some neat features that others do not have.

While I appreciate performance metrics I agree with @cheesestraws that this feels like an attempt at marketing BlueSCSI. If that's what you want to do, that's fine, but its disingenuous to trash the competition without giving them a fair shake.

 

[cheesestraws]

If that's what you want to do, that's fine

it's only fine to do marketing if you are clear that that is what you are doing. This is just something I like to call "dishonesty".

 

[joshc]

What is the point of a comparison between a product that is no longer sold, and one that is made by a bunch of people who refuse to use this forum?

The Homebrew ZuluSCSI kit is only $42: https://store.rabbitholecomputing.com/ZuluSCSI-Compact-Homebrew-p/zuluscsi-compact-homebrew-kit.htm and IMO is a far better product.

 

[robin-fo]

Maybe I missed some of the backstory, but I can't see any issue with someone randomly posting disk benchmarks. Even if it might be of little interest to most, it can still be an interesting insight for some as long as the data shown is not biased or manipulated. We are all using obsolete devices, so why not benchmark or compare them? And since all we do here is only our hobby and mostly 'just for fun', why should we expect completeness of the data shared with us?

 

[Stan Johnson]

No, I'm not in imarketing. ZuluSCSI v6.4 (the one I would get that is comparable to BlueSCSI v2) is only $120, but that's still a bit more than $50. https://store.rabbitholecomputing.com/ZuluSCSI-V6-4-p/zuluscsi-v6.4-rev2024a.htm

I would enjoy seeing helpful comments and other benchmarks, including tests of ZuluSCSI v6.4 and SCSI2SD v6 (the predecessor to ZuluSCSI). And MacSD too. Its 4 GiB limit is too much for me, and I get to have that opinion. I'd also like to hear more about RASCSI / PI SCSI. As it turns out, I just summarized what I happen to have. If the information is not helpful to you, then please just ignore it rather than make this forum appear to be hostile.

 

[saybur]

The best comparison on the ZuluSCSI side is one of the RP2040 models. BlueSCSI v2 is a clone of those, they will perform identically.

Benchmarks for PiSCSI are here if you're interested: https://github.com/piscsi/piscsi/wiki/Benchmarks

So you are aware, there is a history of animosity between the above projects and BlueSCSI's marketing is quite extensive: both have primed people to be suspicious of benchmarks showing only one device.

 

[Cory5412]

Moderator note // as a word of caution:

It's fine to prefer whatever you want.

But there's good reason people are suspicious here. Historically, often in any comparison thread or in any thread where legitimate problems (especially with previous BlueSCSI designs) were mentioned, we'd be joined by people whose express purpose for being here was to cause problems.

If that's not happening and we're all on our good behavior, then, this is fine, just, be aware that there's a reason people are hesitant about this type of thread. Especially when it's so obviously stacked in a specific direction.

 

[Cory5412]

In terms of SCSI replacers:

In my experience, SCSI2SD v5 is Good Enough, basically for anything most people would want to do on any 68k Mac, or any compatible hardware of a similar era. (~1994 and older) especially in microcomputers and single-user systems. I've got a V5 in my Quadra 840av and it runs great. Feels at least as good as the stock disk did.

The SCSI2SD v5 (and really almost all the other options that were hovering at roughly that performance level) are all Good Enough because they benefit significantly from near-zero seek/access times, even if they're slower in actual r/w operations.

Classic Macs aren't fast

For a long time the SCSI2SD v6 was the performance champion in the hobbyist SCSI replacer scene. (Obviously there's other SCSI replacers for if you're maintaining a $million-dollar money-counter class computer, but nobody's shopping for that product for their 68k Mac.) My SCSI2SD v6 is good enough that it matches or beats AV disks for machines like the Mach5 8600/9600, and may even outperform early stock G3 disks.

One thing I did find is that at least in the SCSI2SD v6, getting a good SD card matters a lot. We had a thread here where we were able to get someone's benches up almost literally 10x, I believe on their either Jackhammer or SEIV. I believe this is probably true for now all of the modern SCSI replacers, because most of them are getting pretty close to that 10MB/second point. (ZuluSCSI RP2040 Compact is quoting 9.5MB/sec read speeds, which is above and beyond Good Enough because most 68k Macs have fairly slow SCSI, I believe SCSI-1 at 5MB/sec. (This honeslty persists in most PowerPC Macs as well.)

In terms of the ZuluSCSI, the one you linked, "v6.4" is named that way because it's being marketed as a SCSI2SD v6 replacement, for anybody who was using them in specifically approved situations, and uses a similar (or possibly even direct descendant of?) configuration utility.

The ZuluSCSI and ZuluSCSI Compact RP2040 should have most or all of the same performance but with the EXFAT filesystem flexibility, for $50. To reapeat it: ZuluSCSI reports it should do 9.5MB/sec reads.


I'd say that for anybody who has a SCSI2SD v5 installed, it's probably doing good enough.

If you want to continue this thought process, to make the data useful what i'd say is consider including the numbers for the stock disk, or a good-health period disk, and consider including more of the options.

I don't have a ZuluSCSI yet, but for what it's worth, I trust the team at Rabbit Hole a lot, I've got 3 old Macs with SCSI2SDs in 'em and they're all great. So for me, it'd be worth the extra couple bucks for a Zulu over one of the other options, especially since under no circumstance will I ever (this is just a personal thing) buy a DIY SCSI replacer. It's just not something I have the time, skill, or patience for. (Again, that's just a personal preference! and I love that the DIY kits are available for people who do want that!)

There's clearly room in the SCSI replacer market for several options, and it's been so neat to see the evolution in terms of making things easier and more convenient. At this point I think most of these things now do more, faster, than most people need, which is great.

(There's kind of a crossover once you get to PCI where you can just toss a SATA card in, and once you get to Macs that have IDE there's some IDE<>SATA options, but who knows, the meta could change again as we get faster SCSI options.)

But now I'm just chuckling to myself imagining a FiberChannel disk replacer for like a Sun V880.

 

[ymk]

And MacSD too. Its 4 GiB limit is too much for me, and I get to have that opinion.

That isn't accurate though. Here's MacSD running a 206GB HFS+ image.

 

[Stan Johnson]

That isn't accurate though. Here's MacSD running a 206GB HFS+ image.

ok, thanks

I was referring to the MacSD web site, which says:
"... A single image file may not exceed 4GB due to FAT32 limitations."

It sounds like if image files aren't being used, then MacSD can support volumes of any size that can be defined as a partition in the MBR partition table, is that right?

 

[ymk]

It sounds like if image files aren't being used, then MacSD can support volumes of any size that can be defined as a partition in the MBR partition table, is that right?

That is correct. It can use images inside and outside of the FAT32 partition.

 

[Tarantulas]

I'm just a simple country Mac user, but I'm just curious who fromagepaille thinks they are to demand that other people spend money on anything just so that their favorite brands get equal airtime?

If fromagepaille feels so strongly about it, they should put their money where their mouth is - and send Stan the tracking info.

 

[Daniël]

Historically, often in any comparison thread or in any thread where legitimate problems (especially with previous BlueSCSI designs) were mentioned, we'd be joined by people whose express purpose for being here was to cause problems.

I'm just a simple country Mac user, but I'm just curious who fromagepaille thinks they are to demand that other people spend money on anything just so that their favorite brands get equal airtime?

If fromagepaille feels so strongly about it, they should put their money where their mouth is - and send Stan the tracking info.

Case in point.

 

[cheesestraws]

It's alright: I wouldn't expect a dubious youtuber to understand things like good methodology.

 

[finkmac]

I'm just a simple country Mac user, but I'm just curious who fromagepaille thinks they are to demand that other people spend money on anything just so that their favorite brands get equal airtime?

If fromagepaille feels so strongly about it, they should put their money where their mouth is - and send Stan the tracking info.

who the heck is fromagepaille

are you trying to be a "funny guy" and not directly mention @cheesestraws ?

don't you have better things to do? like getting scammed or whatever it was

 

[Tarantulas]

Case in point.

I think you mean:



Is the person that started this thread not allowed to post their comparison without stan accounts trying to tell them they OWE MONEY to their favorite brand?