• Updated 2023-07-12: Hello, Guest! Welcome back, and be sure to check out this follow-up post about our outage a week or so ago.

PowerTower Pro: ATA CF Slowness

Cory5412

Daring Pioneer of the Future
Staff member
Gotcha. ATA disks have an amount of write cache in RAM, so any contention could be resolved by the disk prioritizing the read operation and accepting the write operation into cache.

My guess is that your CF card has little or no of this kind of cache.

Looking at the CF card you got, my guess is that you could mitigate this getting a better card. Nobody seems to give exact specs here but a better card will overall be less bad in this scenario.

You've probably seen it before but for an overview on this kind of problem, but with SD card storage, there's some good general discussion in here: 





There are a couple higher-end CF cards you could look at, but I'll offer the caveat that I haven't used any very high end or new CF cards. Sandisk has some rated at 120 and 160 megabytes/second and I'd go right for that "extreme pro" level, even though you'll never actually get 160 megabytes/second out of it, if it's closer to being able to do, say, 40 for realsies, then you'll be in a way better position.

Unfortunately, I don't have a good way to test any of this, so I can't make any recommendation without the warning that: You might buy the highest-end available CF card and still have problems, even if it ends up doing better on benches. A high end CF card strikes me as an intriguing option for something like, a beige G3 or even a 630 or 5000/6000 series mac, but you're also sort of getting into the realm of this being enough money to justify posting a WTB for an SIL3112 and getting a SATA card and using like a used intel 520 or even a cheap modern SSD. 

Incidentally, if you still have those macbench results and have some time, would you mind dropping them somewhere in public on vtools?

 

Cory5412

Daring Pioneer of the Future
Staff member
Tangentially, I actually need to set up a few of my macs for some writing stuff and also just ot have more macs online in the near-mid future (before november) and one thing I have been wondering is how Mac OS 9 will handle on SMR hard disks, so one thing I might do is pull out my blue-and-white G3 and drop an SIL3112 and put 9.2.2 on the WD blue laptop disk. (These were going to be the disks I used in vtools, so I have one or two hanging around, one's in my 2011 mac mini as well.)

 

johnklos

Well-known member
Writing to CompactFlash will typically be pretty slow. Writing flash is a slow process, and CF cards really don't make for good general-purpose use drives. For cameras, they have a burst of a write, then a pause during which the CF's controller can actually write out the data. For video, it's a constant stream of writes that's typically slower than the maximum write speed rating, anyway.

If performance really matters, and you want something new (not older IDE), then get a SATA-IDE adapter and any new SATA drive.

 

jessenator

Well-known member
Here are some numbers from my StarMax as well as a Power Macintosh G3 (with SIL3112 SATA card) from MacBench v4.0. Fun fact: the results files are just Western (Mac) encoded tab-delimited(?) text files. I copied the disk test into a Google sheet and made a bar graph.
 

i6TeRHp.png.d9093c2cac98143dbda8da196d387b3b.png


bveu6hv.png.644027aaabf6d2c900fad39176de76b4.png


Sadly I don't have any spinner results handy to show in this way :/  

Here is a similar StarMax configuration (different OS version, CPU speed and L2 cache) in FWB HardDisk Toolkit v2.0.6 bench test (3 consecutive tests) for corroboration purposes:

1EuBwC4.jpg.0439871c2fd596b82003388253021ba1.jpg

 

 

Cory5412

Daring Pioneer of the Future
Staff member
Those CF results are very similar to what I get on my SCSI2SD v6, especially on the write side. Your reads are faster than my v6 was, but I think that's because the 8600's fast scsi bus is still only 10 megabytes/second. I've got it on reasonably good authority from a friend that the scsi2sd v6could do better on a better scsi bus.

IDE to SATA adapter would be the low-hanging way to get much better disk performance with what you already have without needing to buy that much new hardware, esp. a SIL3112, or have to go find an old IDE disk.

 

LaPorta

Well-known member
MacBench 5 proved to work for me. I got the results in an Excel file that it exported. I'll post them today for you to look at.

So...you think an adapter would be best?

 

jessenator

Well-known member
See what you can make of it by comparison.
Well, I want to try MacBench 5 on the Starmax now, because that gap is fairly wide:
sBMutQL.png.f207d89a11e2e60a16daa950a7077923.png

Not as many disk tests appeared in the MB5 file, so I limited it to what was common between them

Edit:
Yeah, I wonder if there's something else going on under the hood of MB5. Added the PMG3's SATA results

2MOKveY.png.3e0a6dbfee9c7c22e49324106ebbec42.png


 
Last edited by a moderator:

jessenator

Well-known member
Well now that's interesting...

r4sAxlE.png.3320163bae8ea6e5eae0642e167f386f.png


That board bus plus Sonnet has some secret sauce... I'm really surprised by the test against the SATA. I may just have to get out the PowerTower and run it again with MB5.

Also, correction to my supposition above, there really are just fewer disk tests in MB5 it seems. I ran "all disk tests" and that's all it spat out :/  

 

LaPorta

Well-known member
So let me get this straight: the thing is a speed demon, but it has crazy, unknown slowdowns?

 

jessenator

Well-known member
I guess ¯\_(ツ)_/¯

Ran the PowerTower 210 w/ SIL3112+samsung SSD: a bit more uniform with 1MB test files, but still hovering around that 20,000 mark, where the Sonnet climbs up almost to 33,000. Now I'm impressed by the 604e/210 with half the L2 cache outperforming the G3/300 (in disk)  :lol:  
LV2drxt.png.6b6237f6e60470846403e30cfd2b69b2.png


 
Last edited by a moderator:

Cory5412

Daring Pioneer of the Future
Staff member
I hadn't had to download the macbench results yet, thank you jessenator!

It seems like there's some random slowdown, and, come to think of it:

LaPorta: Did I give you an account on the new vtools? It might have been like on a share called "users3" or something? That new/testing vtools is offline because I had to clean my room, I wonder if the hang-up is because of that.  (At one point you mentioned using an automatic share mount and that might be causing delay if any of the servers are offline or otherwise misbehaving.)

 

LaPorta

Well-known member
I hadn't used vTools for this, I just downloaded MacBench from Macintosh Garden instead. I'm just trying to figure out what the results say, are longer bars better or are shorter ones?

 

jessenator

Well-known member
Longer is better, as far as I can tell, for every MacBench suite (disk, CPU, FPU, Graphics, etc). For the disk suite, it represents the kilobytes(kibibytes)/second rating for each disk test in said suite. At least that's what MacBench says over each of the results… But again, these are the detailed "Disk Inspection" tests: My PowerTower is peaking at ~20 MB/s in those tests, whereas your PowerTower Pro is hitting ~32 MB/s, skewing the graph a bit.

Unfortunately, there are no "Disk Inspection" test results for the benchmark files provided by ZD with the MacBench Program* (e.g. the Power Mac G3/300 I'm using in this specific post below—as well as a PowerMac 9600 for reference), which is somewhat lame, so we can only compare the "Disk" and "Publishing Disk" test numbers as follows:

A (ZD) Power Macintosh 9600/300 ranked an overall "Disk" score of 678 | a "Publishing Disk" score of 667
A (ZD) Power Macintosh G3/300 ranked an overall "Disk" score of 1000 | a "Publishing Disk" score of 1000
My StarMax 5000/225 (603ev) ranked an overall "Disk" score of 1450 | a "Publishing Disk" score of 661
My Power Tower 604e/210 ranked an overall "Disk" score of 2297 | a "Publishing Disk" score of 1165
Your Power Tower Pro (G3) ranked an overall "Disk" score of 1881 | a "Publishing Disk" score of 760

So there's just more detail we're not seeing on what the "Disk" and "Publishing Disk" tests are, as they're different(?) from the "Disk Inspection" tests ¯\_(ツ)_/¯ So on paper(?) your PowerTower Pro is running faster than a stock/standard Power Macintosh G3/300.

But back to the root issue, I can only assume there's something with the Sonnet card in general, or some fluke intrinsic to your specific card, that causes the slow downs.

Anecdotally, I do notice quite a delay, especially on shutdown and startup with Mac OS 9.x, especially 9.2.x, on my icebook, and on that Power Macintosh G3 you now have, running on a standard IDE drive.

I wonder if you were to create a new startup disk with system 7.6.1: what would the day-to-day "feel" operations, and everyday things like startup and shudown be like. On my 7.6.1 machines, shutdown is almost instantaneous, and startup is fairly quick as well, no real long gray screen delays, etc.

Sorry about the moire :/

*
rRyoOZs.jpg.a4b2d2ec7090620dbeefa1d7ace5d347.jpg


Also, I didn't have your file handy on this Mac to show alongside it :/

 

LaPorta

Well-known member
I have partitions on the card for 7.5.5, 8.6, and 9.1. Startup on 7.5.5 is quite speedy, and shut down is almost instantaneous.

 
Top