Disk Speed Upgrades (aka The Bootable PCI SATA & SSD thread)

[V.Yakob]

Now don't get me wrong a boost of 11.65% is awesome, but we should also note that SATA card eats up PCI bandwidth to achieve that extra 11.65%, so I'm just saying for some people who fill their G4s with ProTools cards, or PCI audio interface cards with DSP cards (Powercore, UAD-1) then the IDE to SATA adapter may still be the best option since it's on a separate bus... just saying

Any productivity growth from 5% can already be considered significant.
As our FBz testing showed, my SATA-PATA bridges on the JM20330 controller are slower, so at first it seemed to me that the performance gain was greater.
But I agree with you, if the user needs to use many or all of the PCI slots on PM, then the best practice is to use the bridge, even on the JMicron controller.
Again, if the user needs to connect large partitions in Mac OS 9, the only possible option is to use a PCI card. As you probably remember, as an experiment, I connected an HGST disk with one 1 TB partition.

In any case, dealing with this issue has been interesting over the past few months. 

[IIO]

2TB volumes work fine with the cheap green adapters, but what you dont have without PCI solutions is multiple ATA controllers.
in the case of MDDs you can double and in the case of quicksilvers you can triple your internal storage with 1 additional 2-port IDE or 4-port SATA card, which is why a proper (and available) 4-port SATA card would be awesome to have.

to be honest, i bought a sonnet in 2005 mostly because i did not know about cheap adapters. but then later i was happy for the additional storage it allowed.

[Borgmac]

As our FBz testing showed, my SATA-PATA bridges on the JM20330 controller are slower, so at first it seemed to me that the performance gain was greater.

I made some tests with SIL 3114 and I was quite surprised to see that SIL 3114 is slower than SIL 3112 AM28F010 replaced by AM29F010 flashed with new flashrom from dosdude1. I thought I read somewhere that SIL3114 based cards were supposed to be faster than SIL3112.



It looks like the Sinloon adapter has still some good days to live...

In any case, dealing with this issue has been interesting over the past few months. 

Same for me, it gave me the opportunity to lear Linux on PowerMac and to solder my first chip!

[chrisNova777]

its the year 2032...
diehard.. still refuses to die...
still emphasisizing for you to not over-use your PCI BUS!! ehehehe
MACOS9! FOREVERRRRRR

whos gonna do the benchmark of two ssd's on two seperate PCI cards in RAID0 stripe
PCI BUS is trembling in fear

in all seriousness tho, if the sil3114 wont boot, couldnt there still be a way to force it to boot with an OF command?

[DieHard]

its the year 2032...
diehard.. still refuses to die...
still emphasisizing for you to not over-use your PCI BUS!! ehehehe
MACOS9! FOREVERRRRRR

OK, Chris, you got me there

Still reminds us of the old debates... hehe

I am sure, as stated an 11% increase is very beneficial, and that for the majority of users, I admit, SATA PCI card is the way to go; I am only mentioning this for the users that might not have a chassis and are using the slots for core/farm cards or the users that have (2) or more DSP cards and an interface PCI card.  DAW users are the ones that need to weight it all, even you, if you decide to add (2) or more cards with your MotU PCI card will see the light, audio pops and clicks are a real thing when you max out the PCI bus and then use it to write data.  I had to personally test and load up G4s back it the day, and it was a factor, even though, I know you like to ignore that, but hey, everyone can experiment and make their own choices, I am just putting it out there for those who google their way here and are going for an old DAW.

[chrisNova777]

my SIL3112 card has the winbond w29C011AP-15 thats 128k right?

for me personally -- re ssd its not even about the benchmarks its about the actual real world feel of feeling that the system is more responsive
that is what i experienced the first time i got the sonnet tempo sata card (purple one)

it definately was a huge thing, back in the late 90s + early 2000s .. but the disk performance was total shit compared to now
mechanical vs ssd theres no comparison in seek/access/wait times

[n8blz]

I FINALLY got it figured out, well, sort of... I couldn't figure out what in my implementation was causing the issue, so I decided to make a new implementation, based off a different C implementation along with a (broken) Forth implementation I came across. With this decompression routine implemented, the USB probing issue stopped happening on my MDD! So now, finally, we have a fully working ROM that fits onto a 128K EEPROM! I have attached this new ROM below, along with my patched copy of the OS 9 SeriTek flasher tool with said ROM embedded, as well as the Forth source of the new decompression implementation.

Please tell me you made a video of you doing this!!

[IIO]

did you notice? huge pricedrop for consumer SSDs.

2 TB SSDs are now below €100, about where 1TB models were 4 weeks ago.

for massive data writing jobs or for archival puposes (i.e. in the cupboard) you still need HDDs, but for everyhting else you have no longer any excuses.

[chrisNova777]

it feels alot more than 11.65% .. like anyone going from a normal hdd to a ssd the change is instantly felt.
every read + write is faster... it adds up

[Yrouel]

As it was discussed here: https://forums.macrumors.com/threads/pci-sata-cards-what-is-the-state-of-compatibility.1650568/post-31775847 the issue with the QuickSilver seems related to the slow 3.3V regulator which makes the 1.8V rail that feeds off of it lag behind just enough to make the Mac hang.

I didn't have a FS8860 33C (I tried another random one but didn't work) so I lifted the input of the 1.8V regulator and fed 5V directly to it (so both regulators feed off of the 5V rail instead of being chained together).

It works! However the 1.8V regulator (AMS1117 1.8 ) gets quite toasty so I don't think this is a sustainable mod in the long run and I wouldn't recommend it

*Some images cropped and re-sized for display purposes. -FBz

[ssp3]

Put a few 1A or so diodes in series to drop the voltage. Chinese did it on some JM20330 adapters.

[Yrouel]

Put a few 1A or so diodes in series to drop the voltage. Chinese did it on some JM20330 adapters.

Yeah that's also how some of my CF to IDE adapters are able to be dual voltage. I didn't have SMD diodes at hand so I just used a wire.
This was just to prove the card could work and it was meant to be temporary anyway.

I just got the Adaptec card and it's working perfectly pretty much as is (minus having disabled the EEPROM).
Strangely however it wasn't recognized by flashrom using the "satasii" programmer even after disabling the EEPROM, I was able to flash it from Mac OS 9 with the patched dosdude1 flasher though so all is well

[dosdude1]

I FINALLY got it figured out, well, sort of... I couldn't figure out what in my implementation was causing the issue, so I decided to make a new implementation, based off a different C implementation along with a (broken) Forth implementation I came across. With this decompression routine implemented, the USB probing issue stopped happening on my MDD! So now, finally, we have a fully working ROM that fits onto a 128K EEPROM! I have attached this new ROM below, along with my patched copy of the OS 9 SeriTek flasher tool with said ROM embedded, as well as the Forth source of the new decompression implementation.

Please tell me you made a video of you doing this!!

I mean... Not sure what I'd really show? It'd just be me sitting there writing a bit of Forth code haha. Not exactly the most exciting of videos.

[DieHard]

Yeah, that would be kinda boring...

But, I watched this video 2X for it's entertainment value...

[youtube]apEKAY11NQs[/youtube]

I must say, you are my hero, I am kinda feeling a little gay saying that... I better cancel my Netflix subscription... whatever it's 2023.

All jokes aside, you are one tenacious MF, have you ever given up on anything ?