Sonnet Tempo ATA133 and ATA100 PCI - PROMISE cards flashing quest

[Ataman Honcho Honchev]

This is why at first I would like to have a modern driver-set design.
Than we can see, which ASIC is the optimal solution.

There is no such thing as "best" solution of course.
But based on the feature set + performance the 3114 (and even the 3112) is far from being "optimal".

When we can see is of course few more months to wait.

In addition to two SATA ports the VIA6421A has one PATA port.
But do people need any PATA card at all?

I see that the Sonnet Tempo133 cards fetching an insane amount of money.
I remember that with a custom flash utility written in macOS 9 (and based on what Sonnet provided to me) I was able to flash any corresponding Promise card back and forth.

I still have that utility (of course) - and just before the COVID I tried to flash a given card.
Without success. Than I remember, all flashing was done on an "old world" Mac, maybe 9600.

So I am going to pick up an "old world" G3 Mac in Erlenbach (close to Heilbronn, Germany) some time this or next week. I don't live in Germany, just (once in a while) close enough.

Won't be around here tomorrow - and now have to back up and pack.

 

[reddrag0n]

This is why at first I would like to have a modern driver-set design.
Than we can see, which ASIC is the optimal solution.

There is no such thing as "best" solution of course.
But based on the feature set + performance the 3114 (and even the 3112) is far from being "optimal".

When we can see is of course few more months to wait.

In addition to two SATA ports the VIA6421A has one PATA port.
But do people need any PATA card at all?

I see that the Sonnet Tempo133 cards fetching an insane amount of money.
I remember that with a custom flash utility written in macOS 9 (and based on what Sonnet provided to me) I was able to flash any corresponding Promise card back and forth.

I still have that utility (of course) - and just before the COVID I tried to flash a given card.
Without success. Than I remember, all flashing was done on an "old world" Mac, maybe 9600.

So I am going to pick up an "old world" G3 Mac in Erlenbach (close to Heilbronn, Germany) some time this or next week. I don't live in Germany, just (once in a while) close enough.

Won't be around here tomorrow - and now have to back up and pack.

i have a 7500 mac with pci slots. and i still have that blasted ata 100 card. would anything like that work?

 

[DearthnVader]

Same about SAS controllers - but without a proper PCI-X bridge we are stuck at PCIe G5 or later.

Well I did test a SAS card in a PCI-E to PCI bridge and got way better speeds in benchmarking than the 3112. Tho it only worked well with the G5, it has issues with the builtin USB on the G4 systems.

That is likely an issue you could fix.

 

[Ataman Honcho Honchev]

Maybe - can you provide some more details? I am curious what kind of bridge adaptor it was and of course what kind of SAS adaptor. I know how to program LSI SAS 9200 (and later) cards, soon probably the Microchip / PMC Sierra as well.

A PCIe <--> PCI-X bridge board would solve many problems. The performance of 3112 (and similar) cards sux horribly, the 3114 is the worst offender.

OK, for the "beige" "old world" machines the 3112 or 3114 is just fine.

But starting with "Yosemite" these chips are underachievers and shouldn't be used.

The Vitesse / Intel used in SeriTek/1Vxx series is far better (but it's very rare). The Silicon Image 3124 (or better) would be a good and fairly commen choice.

 

[Ataman Honcho Honchev]

Starting the "Beta 6" today. Found that there is a problem with interrupts if the Silicon Image chip is "behind" an extra PCI bridge. Also want to eliminate as much global variables as possible. Than, thinking that maybe having each (S)ATA channel each own queue (as it is done in "X") could be better. Than, using "tags" as it is done in my AHCI driver and also the driver for Marvell 88SX7042 would make sense. At first, let me see these elusive interrupts.

Tomorrow hopefully picking up a "beige" G3 in Erlenbach (in the hope to deal with Promise cards later on), today looking for the interrupts.

Hope will have some time to walk around in Erlenbach and Heilbronn (with my Sony Alpha 7IV and some "hacked" lens).

Looks nice...

Château Berwartstein Dans Dahn Rockland Image stock - Image du grès, château: 122639511

Photo à propos Château Berwartstein dans Dahn Rockland, Allemagne. Image du grès, château, rouge - 122639511

fr.dreamstime.com

 

[joevt]

A PCIe <--> PCI-X bridge board would solve many problems. The performance of 3112 (and similar) cards sux horribly, the 3114 is the worst offender.

I think we need someone to make a PCIe to PCI-X board that can satisfy all the various PCI buses that a Power Mac can have. The only easily obtainable bridges appear to be for only 32-bit 33 MHz which is only optimal for Old World Macs.

There's some notes in the PCI Adapters and PCI Speeds sections at:
https://forums.macrumors.com/thread...eige-power-macintosh-g3.2303689/post-30082400

 

[DearthnVader]

Maybe - can you provide some more details? I am curious what kind of bridge adaptor it was and of course what kind of SAS adaptor. I know how to program LSI SAS 9200 (and later) cards, soon probably the Microchip / PMC Sierra as well.

A PCIe <--> PCI-X bridge board would solve many problems. The performance of 3112 (and similar) cards sux horribly, the 3114 is the worst offender.

OK, for the "beige" "old world" machines the 3112 or 3114 is just fine.

But starting with "Yosemite" these chips are underachievers and shouldn't be used.

The Vitesse / Intel used in SeriTek/1Vxx series is far better (but it's very rare). The Silicon Image 3124 (or better) would be a good and fairly commen choice.

#170

Through 174 in that thread.

This SAS card:

LSI Logic SAS3041E PCIe x4 4-Port 3Gbs SAS/SATA Host Bus Adapter | eBay

Find many great new & used options and get the best deals for LSI Logic SAS3041E PCIe x4 4-Port 3Gbs SAS/SATA Host Bus Adapter at the best online prices at eBay! Free shipping for many products!

www.ebay.com

And this PCI Bridge:

Pardon Our Interruption...

**Note, not my auctions.

 

[Ataman Honcho Honchev]

I do have both the card and the bridge; The bridge card does not work in G4. I have an other one which works.
I do know few things about the programming interface for that card, even for much later ones.
Given the proper bridge the LSI cards would be a clear winner.

 

[DearthnVader]

I do have both the card and the bridge; The bridge card does not work in G4. I have an other one which works.
I do know few things about the programming interface for that card, even for much later ones.
Given the proper bridge the LSI cards would be a clear winner.

Also, I think this card may work as 64bit PCI in a PCI G5 and PCI-X in those G5's. Not tested tho....

I went ahead and ordered one to test with my PCI( not PCI-X )G5 so we will see....

LSI Logic Sas3041x-r Pci-x-133 SAS RAID SATA Controller Card 4-port for sale online | eBay

Find many great new & used options and get the best deals for LSI Logic Sas3041x-r Pci-x-133 SAS RAID SATA Controller Card 4-port at the best online prices at eBay! Free shipping for many products!

www.ebay.com

Sadly it won't work in the 64bit PCI slots of the G4 systems do to how it is keyed.

 

[Ataman Honcho Honchev]

Also, I think this card may work as 64bit PCI in a PCI G5 and PCI-X in those G5's. Not tested tho....

I went ahead and ordered one to test with my PCI( not PCI-X )G5 so we will see....

LSI Logic Sas3041x-r Pci-x-133 SAS RAID SATA Controller Card 4-port for sale online | eBay

Find many great new & used options and get the best deals for LSI Logic Sas3041x-r Pci-x-133 SAS RAID SATA Controller Card 4-port at the best online prices at eBay! Free shipping for many products!

www.ebay.com

Sadly it won't work in the 64bit PCI slots of the G4 systems do to how it is keyed.

100% correct. I too, do have that card because I am familiar with the LSI API-s. That card is a no-go for G4 and not only due the keying - it's a voltage question as well. A mere PCI-X rise-card is not enough, one needs to convert the strictly 3.3 Volt 64-bit PCI card with the LSI chip to be able to deal with the 5 Volt 64-bit PCI slot of G4.

The keying of G4 is such, that a strictly 3.3 Volt PCI-(X) cards von't fit. So far I wasn't able to find a strictly 3.3 Volt 32-bit PCI card in my collection (maybe they do exist tho) - but the LSI card is a strictly 3.3 Volt 64-bit PCI card, so it is a no-go.

The G4 is smart enough to switch off instantly before anything is damaged if you force such card into it (via a "dumb" riser card). Unfortunately I don't have a "smart" riser card which would absorb the extra voltage.

Maybe my riser card is not very dumb (to let everything burn), just a little bit dumb (like me, who experimented with such dumb things). As far as I see, the G4 wasn't burned at all - and the card is hopefully OK. That was a while ago, way before the COVID. I did not play with that card since. Hopefully it is still alive, not killed by my dumbness.

 

[Ataman Honcho Honchev]

Just tested the "beta5" with the tri-interface (3112 + FireWire400 + USB 2.0) card in 9.2.2
That card features some *** CHEAP *** PCI - PCI bridge.

Accordingly, the performance is very-very CHEAP: 70-something on reads, 30-something on writes.
Using a (kinda) modern Kingston SHFS37A 120GB SSD drive.

Switched to the tri-3112 RAID PCI-X controller, featuring six(!) SATA ports (three 3112 chips, right?).
That board features *** EXPENSIVE *** Intel 960-derived PCI-X - PCI bridge.

I jumpered the board that the Intel bridge became transparent.

The result: reads and writes very *** EXPENSIVE *** : between 100 and 90 MB/Sec. Pretty much the max. I can achieve with (still) UDMA-100 and the SImage 3112.

Conclusion: with *** CHEAP *** bridge we get *** CHEAP *** performance. With *** EXPENSIVE *** bridge, if set up correctly, the performance is all right ("*** EXPENSIVE ***")

I don't think, this came as a major revelation

The problem were the cables: the G4 case does not close and if I use external box I need "I-to-L" SATA cables.
And because of the G4 case, the "L" end of the cable has to be "L" shaped in outside (not only inside).
Such cables do exist of course, but not a common household item.

So for now the G4 case is half-closed. The Adaptec version of the same tri-chip board has smaller 3512 chips, so it is narrower and it is easier to mount the cables. I did not build the 3512-flavor of the driver yet, so for a day or two this is how it will be.

Ultimately I will buy few extra proper SATA cables.

But first this one:

apple power macintosh g3 | eBay

Find many great new & used options and get the best deals for apple power macintosh g3 at the best online prices at eBay! Free shipping for many products!

www.ebay.com

Yes, I know, 100 Euro is a lot.
But hey, a free MacMini comes with it...