• Please review our updated Terms and Rules here

Which XT-IDE/CF? Blue Lava or XT-CF Lite or Lo-Tech?

1200XL M.U.L.E.

Experienced Member
Joined
May 9, 2021
Messages
84
Now that my 5150 has 640kB installed, I am setting my sights on installing a XT-IDE/CF card to use as my storage device. It seems like there are a couple of versions available which integrate a slot bracket.

I see one from Blue Lava on eBay that comes with a 64GB CF card.

https://www.ebay.com/itm/124118342613?hash=item1ce6079bd5:g:vLQAAOSwAkZebZd B

I see another on eBay called XT-CF Lite.

https://www.ebay.com/itm/224546590178?hash=item3448050de2:g:yLcAAOSwlTJhIdEm

There is also the Lo Tech design sold by Texelec.

https://www.ebay.com/itm/284431651859?hash=item4239729413:g:tG0AAOSw7w1hLWtd

I'm leaning toward either the Blue Lava or the XT-CF Lite card since it would only use one slot whereas the Lo Tech solution would use up two slots. But, maybe that is the wrong reason to favor the first two?

Is any one of these "better" than the others if I all I am looking for is a simple storage solution? Any recommendations?

Thanks!
 
If you feel comfortable doing kit assembly, the XT-CF Lite is a good deal and Sergei generally does good designs.

I've got both the original XT-CF; just attached a right-angle header and plugged in a CF adapter to it. It's all internal, but only occupies a single slot. It's the one I worked out the "Chuck mod".
I also assembled and programmed the CPLD on James' original XT-CF. It's stil working, but it was obviously a bit more involved assembly with SMD.
 
XT-CF is a simpler design, but in my experience it's slower. I get about 250KB/s with XT-IDE on 8088 @ 4.77, and only 150KB/s with XT-CF.

Interestingly enough, a 1993 card called sIDE/8 gives me 1100KB/s on the same 8088 @ 4.77. And it recognizes and works fine with 2GB CF and 4GB microdrive.

Bluelava guy is highly recommended. I got many items from him, they are all very well made, competitively priced and he ships fast and does excellent packaging.
 
XT-CF is a simpler design, but in my experience it's slower. I get about 250KB/s with XT-IDE on 8088 @ 4.77, and only 150KB/s with XT-CF.

There are reasons not to, of course, but if you're willing to swap in a V20 that can give the XT-CF a significant boost. At 4.77mhz I get around 380KB/s in my Tandy 1000 HX. (~500KB/s at 7.16mhz.)

(Although, honestly... when it comes to stuff like this it doesn't hurt to be realistic. The original Xebec controller with the standard 7:1 interleave in the IBM 5160 only pushed around 80KB/s, while typical late-XT-era MFM/RLL controllers were still only around the 200-300KB/s ballpark. And, probably more importantly for most applications, the track and average seek time of flash devices blow away those of any realistic mechanical XT-class hard drive. Even the most basic XT-CF solution is going to give you perfectly acceptable, IE, faster than almost any period solution, performance for just about anything you'd do with an XT today, so unless you specifically want to break old benchmarks I'd say it's not worth worrying too much about picking the econobox option.)

In terms of performance I'd guess the OP's options are functionally identical. A possible downside of the Lo-Tech design is it assigns 32KB of UMB for its flash memory. Likely not a big deal unless you're planning to stick it in a machine with a bunch of other option ROMs and *also* going to try to set up some kind of upper memory RAM solution, but it is worth calling out. One possible advantage of it is since it uses a standard 40 pin IDE plug if you wanted to substitute an IDE to SD adapter for the CF backplane socket you could do that. (Although with the others if you'd prefer using SD cards, I do, CF to SD adapters are cheap and available as well.)
 
I have XT-IDE Deluxe from Blue Lava and it's working well with my Acer/Sakhr AX-1000 (8088 clone with VGA card). I only had to flash it with a newer XTIDE BIOS and change the I/O address from 300h to 320h in order to make it working. I still haven't done any benchmark for it to see how fast it is.
 
(Although, honestly... when it comes to stuff like this it doesn't hurt to be realistic. The original Xebec controller with the standard 7:1 interleave in the IBM 5160 only pushed around 80KB/s, while typical late-XT-era MFM/RLL controllers were still only around the 200-300KB/s ballpark. And, probably more importantly for most applications, the track and average seek time of flash devices blow away those of any realistic mechanical XT-class hard drive. Even the most basic XT-CF solution is going to give you perfectly acceptable, IE, faster than almost any period solution, performance for just about anything you'd do with an XT today, so unless you specifically want to break old benchmarks I'd say it's not worth worrying too much about picking the econobox option.)

My sentiments exactly. A fair number of 5150 users ran without a hard disk for quite some time. As far as speed goes, it mostly depends on how ports are mapped, which was the basis for my mod--to transfer 16 bits with a single instruction using the 8088's BIU to do the consecutive accesses.

For the kind of stuff I do, however, speed doesn't matter, so long as it's faster than a floppy drive.

Has anyone worked up a SD-card version? Seems that with today's high-speed SDHCs, the speed might be comparable, provided that 4-bit SDIO was being used.
 
One more note about performance: I have the equivalent of a Lo-Tech EMS card built into my Tandy 1000 on the same board as the XT-CF, and for laughs I actually tried running the same disk performance benchmark against an EMS RAM disk as the XT-CF. And the end result was essentially a tie. Someone correct me if I'm wrong, but doesn't the 8088 take something like 4 cycles to access a byte of memory? Given that it would seem to follow that the maximum performance you could theoretically get with any kind of CPU-driven memory transfer (read a byte, write a byte) somewhere around 600K/second @4.77mhz?

(If there does exist an IDE controller that can to 1MB/s in such a system I can only assume it achieves this by doing DMA direct from the HD buffer to system RAM and completely owning the bus while doing it. I guess that would check out.)

In any case, I 'd figure even if your hard disk "only" streams data 1/4 as fast as a memory-memory copy, as is the case for a PIO XT-CF with an 8088, that's still pretty ridiculously fast, all things considered.
 
Has anyone worked up a SD-card version? Seems that with today's high-speed SDHCs, the speed might be comparable, provided that 4-bit SDIO was being used.

I actually wonder if just plain 1-bit SPI would be adequate for an XT-class storage device. Pulling up a datasheet for a random SD card says a 25mhz clock is acceptable for data transfer mode, that's 3MB/s. If there's a suitable 8-bit to SPI bridge chip out there that would relieve the CPU from having to do too much bit-banging it seems like it should be very doable to just talk to an SD card directly.

That said, I love the FC1307A IDE/CF to SD bridge chip-based adapters muchly, so as long as they're easily available my solution is to use the standard XT-CF design and put a 2mm 44 pin right-angle header on the board to hang these guys off of. Nice compact solution as long as you don't care about accessing the card from the outside. (I don't, that's what a network card is for.)
 
Just about any MCU with enough memory can do SPI. I do most of my SD work using SDIO and DMA and it's very fast--at least faster than I can feed it.
 
Just about any MCU with enough memory can do SPI. I do most of my SD work using SDIO and DMA and it's very fast--at least faster than I can feed it.

Mostly I'm thinking if you had some suitable SPI glue, which really essentially boils down to a UART by another name, you could ditch the MCU and just let the 8088 drive it directly. I know people have cobbled together solutions using shift registers. (Also looks like someone kind of managed to pull it off with a 6522 VIA.)

That's again why I really like the FC1307. It supports 8-bit mode so, yeah, considering you can do an XT-CF pretty much with one GAL or a '688 and a couple inverters (and you really should have a '245 as well) that's probably less glue than needed to hang an MCU off the bus.
 
I am not about to rain on anyone's parade. Lord knows some of my hobbies over the years have earned me some funny looks. But some of the above gave me a chuckle. In the days of my misspent youth, I campaigned a D/MP 1968 Chevy II "Nova" in NHRA's Division 2 points wars in the Southeast USA. I went fast, made a lot of noise, broke a lot of expensive parts, and had one heck of a good time. I now own a 1996 Catalina 30 Mk III (sailboat LOL). People ask me if I race it. The idea of trying to make a sailboat go fast brings a chuckle. If I want to go fast it isn't gonna be in a sailboat.

To me, asking about speed on a 5150 is about like asking these folks about their fuel mileage.

https://www.youtube.com/watch?v=QVXLGE1LHDU

I own two IBM 5150's. One is bone stock, the other has two CF card slots in the front. The cards I put in them pretend to be the C: and D: hard drives. But when I want to compute fast, I don't go to my 5150's.
 
Last edited:
Yeah, really, it's silly. I mean, it's great and all if you have a benchmark tell you your storage device can get 1MB/s from a flash disk into your 8088's memory, but is that actually good for anything? The 8088 at 5mhz is optimistically rated at something around .3 MIPS (millions of instructions per second), there is literally nothing you can do with that data at a rate that even approaches the speed of the transfer. Once a storage device gets even into the 200KB/s-ish ballpark in terms of data transfer it might as well be considered "instantaneous". Bragging about your 5150 having a 1MB/s hard disk transfer rate is kind of like bragging about having a water fountain equipped with a firehose nozzle. You're still only going to be able to actually drink so fast.

(And again, and probably more importantly, the flash disk in my XT-CF bencharks at around 1.5ms for *both* track-to-track and average seek times. The figures for real mechanical hard drives from the 1980s are around one order of magnitude slower for the first figure and a good fraction of two slower for the second. Response times like this are really what makes any of these things essentially RAM-fast compared to the originals.)
 
Lots of interesting notes here!

I'm not too worried or concerned about speed. As Chuck(G) said, as long as it's faster than a floppy drive then I am happy.

It seems like there are no bad options here. I am going to order the fully built and tested Sergey Kiselev card from the seller numberf55 on eBay tonight after dinner unless someone points out a big "gotcha!" that I am missing.

Thanks!
 
Mostly I'm thinking if you had some suitable SPI glue, which really essentially boils down to a UART by another name, you could ditch the MCU and just let the 8088 drive it directly. I know people have cobbled together solutions using shift registers. (Also looks like someone kind of managed to pull it off with a 6522 VIA.).

I suppose you could do it fairly easily with, say, a CPLD, but SD cards are 3V devices anyway. MCUs are nearly as cheap as jellybeans. Seem that everything has an MCU in it nowadays. The ironic thing is that a lot of these MCUs are more powerful than an 8088 running at 4.77 MHz.
 
I received my XF-CF Lite V4 card today and got it working. As a reminder, here's the card I got from eBay.

https://www.ebay.com/itm/224546590178

Here you can see the seller name too, for reference.

The card arrived in a thin, blue plastic bubble envelope. I was expecting a small cardboard box for more protection. The card bracket was not attached to the card and I didn't seem to get any screws. Thankfully, I have a stash of little screws that I gathered over the years and I just happened to have two that fit. I count this as a minus. If it weren't for my stash then I would not have been able to securely mount the card in my 5150.

The instruction guide says if I don't see the BIOS boot screen for the card then I must have an address conflict. Sure enough, I did have a conflict! About a week ago I added a Lo-Tech 1MB card and enabled all the addresses. I disabled segment 0xC000 on my Lo-Tech 1MB card and set the EEPROM address of the XT-CF Lite card to 0xC000. Viola! I saw the boot screen!

IMG-0990.jpg

If you don't press "A" to boot from a floppy within a time window then the XF-CF Lite card will push on through to try booting from the CF card. My card wasn't initialized and being so caused the computer to hard crash. I had to power cycle my 5150 here. After power cycling my 5150, I pressed "A" here to boot from my MS-DOS 5.0 floppy. When I got to the DOS prompt I ran fdisk to clear the default partition. Then I created a 150 MB primary partition that became my C: drive. Two things I missed/forgot.

1) Make sure your boot partition is active
2) Run "fdisk /mbr" to set the master boot record.

After doing those two things, my 5150 booted up from the CF card and blasted to a DOS prompt! Hooraay! :)

I remember reading somewhere that big drives take a while to create a directory listing using DIR. Yep, my 150 MB partition takes 5 to 10 seconds to show the free space available on the drive before completing the operation.

I'm thinking of deleting this partition and starting over with a smaller C: drive but I would like to start a new thread on that. :) This thread was about the hardware.

Overall, I am initially satisfied with the card and hope that it works reliabily.

Thanks!
 
Back
Top