PC/XT-compatible-ish "AT" SCSI spotted

[SomeGuy]

Some people here might be interested in these, I stumbled across a seller selling a bunch of low-end AT ISA SCSI cards for a decent price. But what caught my eye is that unlike most AT ISA SCSI cards, this actually has 8-bit electronics. After obtaining and testing one of these, I confirmed my suspicion: This will work on an 8-bit PC/XT or compatibles equipped with an NEC V20 CPU.



At present the seller has them listed for US $17.50 each with free shipping, and the page says there are "more than 10".

https://www.ebay.com/itm/STORAGE-PL...A-SCSi-HOST-CONTROLLER-BOARD-FCC/232620890451

There is not a lot of information about this particular card, so I am posting my test results here:

This card is a Sumo Systems / Storage Plus, Inc "SCSI-AT". Reportedly this model was later renamed to "SPI 200".

This appears to be a fairly low-end Japanese SCSI card.

It uses an 8-bit NCR53C80 SCSI chip, and an NCL2020 Floppy Disk Controller.

Jumper settings can be found here: http://arvutimuuseum.ee/th99/c/S-T/20203.htm . For PC/XT operation you must set the IRQ to 3, 4, or 5.

I tested this on several XT clone motherboards as well as an IBM 5150 motherboard. If the motherboard is equipped with an 8088, the system probe for SCSI devices, but will freeze up as it tries to boot the hard drive.

If the system is equipped with an NEC V20, the system will boot from the hard drive, and appears to work normally. This indicates that the BIOS uses some 186 or V20 specific code. (Perhaps someone could hack the BIOS to enable 8088 operation?)

The caveat is that the high density floppy controller appears not to work at all in a PC/XT class system. However, there is a jumper to disable the FDC letting you use another card. I would have expected it to operate in low-density mode like some IDE/FDC cards, but it acts like the hardware can not be found at all.

Unsurprisingly, there appears to be no floppy support in the card's BIOS, and no way to select drive type without AT CMOS settings. I tried ImageDisk, that bypasses BIOS, and it could not talk to the FDC either. 2M-XBIOS also did not work.

There are SCSI DOS drivers for this card, but those are not needed to operate a normal hard drive. I did try attaching a SCSI ZIP drive, however it did not work with it. (It did show up in the probe list at boot)

I tested the speed with Coretest. It is no speed demon, but it is not bad.

IBM 5150 motherboard with NEC v20 and a Quantum Lightning 540S, Coretest 2.4 reports 385KB/Sec

For comparison, a Lo-tech ISA CompactFlash Adapter in a 4.77mhz 8088 machine (CDP 1600) reports: 175KB/Sec

Attaching drivers, and a dump of the BIOS chip.
View attachment Storage Plus SCSI-AT R1 BIOS.zip
View attachment scsi-at.zip

 

[ibmapc]

Hmm... Very interesting. I wonder if the floppy part could be made to work with something like Sergy's Floppy BIOS?

 

[gslick]

I have one of those Storage Plus, Inc "SCSI-AT" cards. I don't think it ever occurred to me to try one in a PC/XT system. Maybe I'll have to get around to getting an NEC V20 processor and give it a try. The HardCard in my XT died. This could be a good way to get the system going again.

That Sony CXK5816 chip is a 2KB SRAM. When I was looking at this a while I ago I didn't think that it was used directly by the other hardware on the card, and maybe it is just some scratch RAM for use by the BIOS without using any low system RAM.

 

[gslick]

I just checked my notes. I bought 2 of those from the same seller almost exactly 5 years ago. He must have a bunch of them if he is still selling them now. At the time when I bought 2 of them they were listed at $9.95 each with free shipping. I got them at a best offer of $8 each with free shipping. I found one of them right away when I went looking. Might have to dig a while to find the other. Too much stuff I've bought over the years and sometimes forget about.

 

[SomeGuy]

Well, considering eBay and shipping inflation, $17 is still not too bad. I'm wishing I had grabbed a few spare 50-pin SCSI hard drives back when everyone was just dumping them on eBay.

I'll have to do some more testing, but I think I may have figured out the issue with the floppy controller. The FDC chip looks like it is getting ground and +5v ONLY from the 16-bit portion of the ISA connector! The ground/+5v on that side is not even connected to the ground/+5v on the 8-bit side. W, T, and several Fs?

Dropping the card in an 8-bit slot on a 286 board causes the FDC portion to appear dead in the same way as the 5150 board.

Anyway, if I'm right, then it should act as a low density controller unmodified, in which case, yes it should work with an external BIOS, 2M-XBIOS, and ImageDisk.

 

[ibmapc]

...I think I may have figured out the issue with the floppy controller. The FDC chip looks like it is getting ground and +5v ONLY from the 16-bit portion of the ISA connector!

If that's the case, it should be an easy mod to run a wire from B31 on the 8 bit connector to D18 on the 16 bit connector for ground and another one from B3 to D16 for the 5v.

 

[SomeGuy]

Ok, a couple of ugly wires later with B29 (also +5v) connected to D16 and B31 (also ground) to D18 the FDC does indeed come to life on the 5150 motherboard.

Without any additional BIOS, however, it gets confused trying to boot or read from a disk. It looks like the chip is defaulting to high density mode at power up. It would not boot from a 720k disk. It would start to load MS-DOS from a 1.44mb disk, but falls down and goes boom. Interestingly, it does boot DR-DOS from a 1.44mb disk (DR DOS is known to override BIOS for certain things).

Once 2M-XBIOS is loaded on the hard drive, DOS will read, write, and format 720k and 1.44m disks fine.

Imagedisk also then worked with the FDC, although it didn't seem too happy about it.

An external BIOS should work. I tried an old trick, where I plugged in a Magitronic B215 8-bit HD floppy card with the FDC chip removed and the jumpers set to 1.4mb. That enabled it to boot from 1.44mb and 720k disks. (I used to use that trick on my 286 to boot from disks with more than 18 sectors per track and not require a TSR to read them)

I did confirm that even with the extra power connections in place, the DIP switch will still let me disable the FDC and use a separate one.

I'm just guessing the thinking was that since the FDC would not be supported on a PC/XT it should not be powered.

 

[VERAULT]

Hmm... Very interesting. I wonder if the floppy part could be made to work with something like Sergy's Floppy BIOS?

Id be very interested in the floppy part working. Already have 8 bit scsi cards. 8 bit floppy cards are another story.

 

[Lutiana]

Ok, a couple of ugly wires later with B29 (also +5v) connected to D16 and B31 (also ground) to D18 the FDC does indeed come to life on the 5150 motherboard.

An external BIOS should work. I tried an old trick, where I plugged in a Magitronic B215 8-bit HD floppy card with the FDC chip removed and the jumpers set to 1.4mb. That enabled it to boot from 1.44mb and 720k disks. (I used to use that trick on my 286 to boot from disks with more than 18 sectors per track and not require a TSR to read them).

So what you are saying is that with a couple of bodge wires and Sergie's BIOS thrown onto something like a 3Com NIC you'd end up with both a SCSI HDD and HD floppy support?

 

[SomeGuy]

So what you are saying is that with a couple of bodge wires and Sergie's BIOS thrown onto something like a 3Com NIC you'd end up with both a SCSI HDD and HD floppy support?

As long as your machine has at least a V20 CPU, then yes.

BTW, another interesting discovery. I noticed that the ROM dump has two different versions of the BIOS in it. The jumper settings linked above say the second jumper is "Factory configured - do not alter". On the card it is labeled as "CODE". Removing this jumper switches from BIOS version 6.3U (in the first half of the ROM) to BIOS 5.55 (in the second half of the ROM).

Off hand, that does not seem to do anything useful. Perhaps change compatiblity with some SCSI devices? (For a moment I was hoping that 5.55 might have 8088 compatiblity, but I tested that and it does not)

Id be very interested in the floppy part working. Already have 8 bit scsi cards. 8 bit floppy cards are another story.

Well, there shouldn't be a shortage of 16-bit ISA IDE or multi I/O cards with FDCs that can be rigged similarly.

 

[Krille]

This card is a Sumo Systems / Storage Plus, Inc "SCSI-AT". Reportedly this model was later renamed to "SPI 200".

It's a "SCSI-AT R1". The "R1" is significant because apparently there's another card called "SCSI-AT". See this. I wonder what the differences are? Maybe they are compatible on the software level (ie can use the same BIOS)?

If the system is equipped with an NEC V20, the system will boot from the hard drive, and appears to work normally. This indicates that the BIOS uses some 186 or V20 specific code.

It's all 808x code with a few exceptions; In the 6.3U-version the data is transferred to/from the drive with rep outsb/rep insb instructions respectively. The same is true for the 5.55-version but that also contains a single pusha/popa.

(Perhaps someone could hack the BIOS to enable 8088 operation?)

I've attached a zip-file containing the following;
6.3U-186.BIN <--- The first half of the original BIOS.
5.55-186.BIN <--- The second half of the original BIOS.
6.3U-086.BIN <--- 808x compatible version of the 6.3U-version.

I didn't bother doing an 808x compatible version of v5.55. The 6.3U-version had plenty of free space so I could unroll the transfer loops to a degree where the performance hopefully won't suck donkey balls. (I'd love to see the result from a benchmark.) I also did some optimizations to improve speed but there's a risk that this may have broken things (I don't know how fast you can do I/O to/from the NCR chip). If it doesn't work (corrupts data for example) then this is most likely the reason. There's still plenty of room for improvement - it should probably be done properly from source created from a disassembly using IDA. I used DEBUG which is also why I added a routine at the end to quickly calculate and update the option ROM checksum.

Anyway, just combine two of the files in whatever combination you want using copy /b, then select the version using the jumper.

View attachment SPI SCSI-AT R1 BIOS.zip

That Sony CXK5816 chip is a 2KB SRAM. When I was looking at this a while I ago I didn't think that it was used directly by the other hardware on the card, and maybe it is just some scratch RAM for use by the BIOS without using any low system RAM.

Spot on. That's exactly what it is.

Perhaps change compatiblity with some SCSI devices?

That would be my guess too.

 

[SomeGuy]

Thanks for pointing that out. I had not seen the non-R1 out there. It looks then like the R1 is mainly a board redesign with SMT components. There is a sticker over the NCR chip, so can't be sure if it is exactly the same. But more importantly, looking at the ISA connector, it probably could also be used in an 8-bit slot.

[Off topic: It looks like someone unsoldered a chip from that board, yet the seller wants $125 for that? Anyway...]

I gave the modified BIOS a quick test on the 5150 board with its 4.77mhz 8088 back in. The card works and boots. Copying files around seems to work. Running Coretest 2.4 against the same Quantum drive reports a transfer rate of 189kb/sec.

 

[gslick]

I gave the modified BIOS a quick test on the 5150 board with its 4.77mhz 8088 back in. The card works and boots. Copying files around seems to work. Running Coretest 2.4 against the same Quantum drive reports a transfer rate of 189kb/sec.

Previously:

IBM 5150 motherboard with NEC v20 and a Quantum Lightning 540S, Coretest 2.4 reports 385KB/Sec

That's a large change. Less than half the transfer speed.

How about with the modified BIOS with the NEC V20? I wonder how much of the speed change is from using the 8088 vs the NEC V20, and how much of the speed change is from the modified BIOS.

 

[1ST1]

What if the XT mainboard does have floppy controller onbard, and not disable? Is it possible to disable the floppy controller on that card?

 

[SomeGuy]

As I mentioned earlier, yes this card has a jumper that lets you disable its floppy controller. This disables it even if extra power wiring is present.

Here are some more tests: Drive used is Quantum 540mb SCSI, Coretest 2.4 and Disktest from https://www.lo-tech.co.uk/wiki/DOS_Disk_Tester , 5150 motherboard, Norton SI 4.5 reports CPU SI of 1.8 with a V20 installed.

CPU: 8088 4.77mhz, SCSI BIOS: 8088
Coretest 2.4: 189kb/sec
Disktest
Write: 135.42 KB/S
Read: 178.86 KB/S
8K Random, 70% Read: 7.5 IOPS
Sector Random Read: 12.3 IOPS
Access Time: 81ms

CPU: V20 4.77mhz, SCSI BIOS: 8088
Coretest 2.4: 204.6 kb/sec
Disktest
Write: 141.24 KB/S
Read: 196.26 KB/S
8K Random, 70% Read: 7.8 IOPS
Sector Random Read: 12.3 IOPS
Access Time: 82ms

CPU: V20 4.77mhz, SCSI BIOS: 186/V20
Coretest 2.4: 450.5kb/sec
Disktest
Write: 276.20 KB/S
Read: 406.14 KB/S
8K Random, 70% Read: 10.6 IOPS
Sector Random Read: 13.2 IOPS
Access Time: 74ms

Media test mode passed OK.

So yea, those v20 instructions really speed things up.

 

[gslick]

So switching from an 8088 to a V20 with the modified 8088 compatible BIOS didn't yield much of an improvement, but switching to a V20 with the original BIOS made a big improvement.

Looks like it is definitely worth picking up a V20 to run one of these.

Thanks for checking.

 

[SomeGuy]

V20s are great way to get a little speed boost from an 8088 based system. If, of course, that is what one is looking for. Thanks to Krille's modification this can now run on a a stock 8088, even if it is a little slower. Interestingly, because of this card's ability to switch between two different BIOSs, you can have it both ways without additional EPROM swapping.

It is my understanding that the XT-IDE can also greatly benefit from a V20, but it has to be flashed with a V20-specific version of the XT-IDE BIOS. Haven't had a chance to mess with that myself.

 

[keenerb]

V20s are great way to get a little speed boost from an 8088 based system. If, of course, that is what one is looking for. Thanks to Krille's modification this can now run on a a stock 8088, even if it is a little slower. Interestingly, because of this card's ability to switch between two different BIOSs, you can have it both ways without additional EPROM swapping.

It is my understanding that the XT-IDE can also greatly benefit from a V20, but it has to be flashed with a V20-specific version of the XT-IDE BIOS. Haven't had a chance to mess with that myself.

Is the eprom split between two 4k banks? I see the one in the photo is a 27C64, I think. If it could support Sergey's floppy bios (which is 8k binary size, not sure about code) it would be a decent choice as a bios-enabled floppy controller.

 

[SomeGuy]

The card only maps 4k of the EPROM in to memory at one time depending on the jumper. That means you can't fit any additional accessible content in there.

Oddly, it maps some scratchpad ram immediately after the ROM so even modifying it to map more ROM would be problematic.

 

[oleg_mishin]

Hi Guys,
I'm currently working on a clone of this card as a hobby project. My goal is to make a card with a boot capability on an PC XT clone computers.
The card I have as a reference is exactly the same as SomeGuy's.
First, I tested it with x286 PC - it works with 5.55 BIOS only. SCSI HDD detected and bootable.
With 6.3U - no go. SCSI HDD was not detected. Tried with different SCSI devices(two types of HDD, blueSCSI, SCSI2SD) - no difference.
With XT PC 6.3U-086 - SCSI devices not detected. With 5.55/5.53 - drops some garbage on screen when detecting devices. Which is expected as the BIOS is designed for higher CPUs.

Does anybody have an idea what could be wrong with my card? I wonder why 6.xU BIOS doesn't detect any SCSI device.

This is my PC XT mainboard: 8088-TXM-8-10

Krille, could you please if possible share the 5.55 BIOS source code. It would save me a lot of time digging in the current one. Or compile it for 8086 CPU, I would expect it will work on my XT machine.

The project's location for reference: SCSI-ISA-5380

Regards
Oleg