For ultimate compatibility I hear the 32 bit versions are best because the 64 bit versions are bugged for proper legacy filesystem handlig.

creepingnet wrote on 2022-07-26, 07:12:

Pulled out the Tandy to start recording some footage for #SEPTandy this year.....christ....all sorts of issues... […]
Show full quote

Pulled out the Tandy to start recording some footage for #SEPTandy this year.....christ....all sorts of issues...

- Most of the bigger games are corrupted and refuse to run
- having issues copying over files using mTCP (file corruption)
- Now looking around at working 8088 compatible pkunzip tool so I can copy ZIPs over and unzip them on the PC (more reliable in general)

If you think mTCP is causing the corruption I'd like to hear about it. I think it's unlikely and it usually turns out to be a hardware problem, but you never know.

The data is checksummed before it is transmitted, and the checksum is confirmed when it is received. Ethernet also protects the packets with a 32 bit CRC. If something is corrupted it is happening either before the data checksum is computed at the source, or after it is received and verified by the receiver.

I would start running memory testers on both machines.

SteveC wrote on 2022-07-26, 09:56:

I had trouble copying to my PS/1 286 using the mTCP FTP Server (copying from the 286 was fine) - by setting WinSCP to its slowest possible setting it helped a bit but was still really unreliable to the point of being useless. Ended up using Windows 3.0's terminal tool and a serial->usb adapter and copied files that way! One file at a time was a bit painful though 🙁

That sounds like a flow control problem. TCP/IP has a send and receive window, and the local programs have buffers. Most machines have fairly large buffers, but that still should work when running against a smaller machine.

Are you running with the latest mTCP? (I made some flow control changes that would help in this situation.) If you are how about sending me a bug report with a trace?

mbbrutman wrote on 2022-07-26, 16:14:

SteveC wrote on 2022-07-26, 09:56:

I had trouble copying to my PS/1 286 using the mTCP FTP Server (copying from the 286 was fine) - by setting WinSCP to its slowest possible setting it helped a bit but was still really unreliable to the point of being useless. Ended up using Windows 3.0's terminal tool and a serial->usb adapter and copied files that way! One file at a time was a bit painful though 🙁

That sounds like a flow control problem. TCP/IP has a send and receive window, and the local programs have buffers. Most machines have fairly large buffers, but that still should work when running against a smaller machine.

Are you running with the latest mTCP? (I made some flow control changes that would help in this situation.) If you are how about sending me a bug report with a trace?

I'm running the old version on the Tandy, I still need to update. The newer versions do work a lot better. I'm always hammering my 486s with files with no problem, and a 286 and a 386SX.

I think I got the Tandy sorted this morning. Probably will update with the new version tonight. I got LSL and Space Quest on there up and running perfectly. I traced it down to not having speed-limits set in Filezilla and trying to push too much at once. Over night I pushed LSL, BeQuest, and bunch of other Sierra games and all but LSL had issues, also the Tandy hung over night, I think I overwhelmed the poor thing (I got too used to being able to just drop a ton of stuff on my super-fast 486 machines and let them run all night). I think one-at-a-time with the old version is the trick. I also tweaked the MTU.

This morning xfers for several games went through no problem and were quite fast. Probably also helps my wife was not streaming TV shows at the time 🤣. Ran Space Quest for a few minutes before work to test. I think I'll upgrade the mTCP client tonight.

If one good thing came out of this outside of file xfer stuff, I now have a working archive unzipper on the Tandy I can use. So if anyone wants an Intel 8088 compatible unzipper for zip files, PkUnZip 2.04 works well. I was a tad surprised how fast Space Quest and Leisure Suit Larry extracted on an 8088.

I remember back in the day, older machines couldn't cope with a full MTU size, only when 486s were common could "everyone" set to 1500.

The newest version (2022-07-01) should be better on slow/crappy connections but on slow machines some care and feeding is still needed:

• A modern network is very noisy; there is broadcast traffic everywhere. Run pkttool in the network packet sniffer mode to get an idea of how noisy your network is. On my network there is lot of ARP and broadcast traffic, and those extra packets take up valuable space in the Ethernet hardware and cause extra overhead in the TCP/IP code. If your network is busy and mTCP is reporting a high number of retransmitted packets or SEQ/ACK errors then you should consider isolating the machine onto a quieter subnet. (I develop and test under these conditions so they should not be a problem, but extra packets coming in are overhead.)
• FileZilla assumes you are connecting to a modern server. It really shouldn't be an issue as the TCP flow control should work, but the combination of the app opening multiple connections and the OS having large buffers is kind of like a laser beam of death onto older/slower machines. I can't remember what I did to FileZilla way back when, but I think I restricted the number of concurrent connections and also set reasonable speed limits. Reasonable speed limits for FTP on XT class machines are kind of low; even if the machine can do 100KB/sec over a socket, file I/O makes things almost 1/3rd of that speed.
• For best file transfer performance over Ethernet the MTU should be set to 1500. 1500 is generally safe, unless you have some weird networking setup. I have the MTU set to a default of 576 because that is supposed to be completely safe. If I implement MTU path discovery things will get even better.

BitWrangler wrote on 2022-07-26, 14:40:

For ultimate compatibility I hear the 32 bit versions are best because the 64 bit versions are bugged for proper legacy filesystem handlig.

Debian based 32bit systems reads fat16 fine for me, haven't tried on a 64bit machine

sirotkaslo wrote on 2022-07-25, 05:59:

Started working on my Compaq Proliant 1500, also have the scsi disk tower, but that will have to wait, because it seems the PSU died after I dusted it with the air compressor.

Didn't die, I just didn't put the side panel on which has a metal thingy that pushes into the PSU, without this, you can't turn the machine on. 😁

There is no metal object that goes into PSU. I have the Proliant 1500 correct power supply and it does not feature one. Show us photo of this one on yours?
Cheers,

I am wondering if he means a sprung punchout that grounds the case.

Reason I said this since I do not have chassis case and have all the necessary parts to make Proliant 1500 running. All I used is motherboard, processor card, PATA hard drive and floppy and original power supply nothing shorting the pins except for fan to not say error was jumpered. I'll recheck my information and.

Second, there is no springs or something since I knew compaq's designs well. For motherboard mounts stand offs they're permanently riveted to the chassis. And only thing that has grounding springs is limited to the rear panel i/o ports shield.

Once again, there is no metal bit that *goes into* power supply, second, there is no security switch at all except for one detail. There is a power supply disable built into the power switch assembly. This is poked into from front through metal chassis (shaped hole) via plastic pin on the front bezel to close the second switch. For testing computer with covers removed, there is a key need to be inserted and turned 90 degrees to keep this switch closed to make power switch usable. What I did was open up the power assembly, this is black clam shell and reroute the one spade connector to the power switch instead bypassing the power disable switch.

Here's this:

download/file.php?id=95845&mode=view

Cheers,

There's not a jumper sitting on a chassis intrusion switch header? That jumper gets installed 100% of the time I come across one 🤣

Tandy's all setup, going to start recording tonight or within the next few days if not. Thinking a "Let's Play" of Leisure Suit Larry (AGI version) is in order. Might also do some humor with Ultima VI as the day-night cycles are literally almost synched with day and night. Also a candid Space Quest run.

Next up might be ordering a V20, 8087, and I'm toying with getting one of those SmartWatch+ modules as well. I'm not doing overclocking to the Tandy, at least not yet. May eventually do EEVBLOG's mod (he has the same model I do) but that'll be for next year or the year after that when I have some time to engineer my own PCB for that (since I want it switchable like his).

Still some memory tuning left. Brought it up from 445K to 490K clean boot, so I think some stacks/buffers/files tweakery might be in order too at the very least.

sirotkaslo wrote on 2022-07-27, 07:49:

sirotkaslo wrote on 2022-07-25, 05:59:

Started working on my Compaq Proliant 1500, also have the scsi disk tower, but that will have to wait, because it seems the PSU died after I dusted it with the air compressor.

Didn't die, I just didn't put the side panel on which has a metal thingy that pushes into the PSU, without this, you can't turn the machine on. 😁

Does it kill the power when opening or just scream an alarm?

bofh.fromhell wrote on 2022-07-27, 20:22:

sirotkaslo wrote on 2022-07-27, 07:49:

sirotkaslo wrote on 2022-07-25, 05:59:

Started working on my Compaq Proliant 1500, also have the scsi disk tower, but that will have to wait, because it seems the PSU died after I dusted it with the air compressor.

Didn't die, I just didn't put the side panel on which has a metal thingy that pushes into the PSU, without this, you can't turn the machine on. 😁

Does it kill the power when opening or just scream an alarm?

Just simply dead if cover is off. The power disable is a open switch in series with main switch toggle, need a plastic "key" that you put in a specific hole at front panel and twist 90 degrees to keep key in place to push the disable power switch closed if you want to work on this Proliant 1500 without the cover.

Cheers,

Rummaging through my stack of old mobos I came across a 286 mobo from days gone by. Not sure why it was there as I've tossed out a lot of stuff like this over the years. The barrel cmoss battery had been removed. It has 1mb in dimms on the mobo. Put an isa video card in it , hooked up an AT psu using a hdd to provide some sort of load. Well bugger me it fired up with the usual cmos errors you'd expect.

I've soldered a coin cell battery holder and inserted a new battery. Fitted isa card with fdd and ide hdd headers. Did a bit more testing. Looking good. Fitted an 8-bit IDE controller, then an EIDE Promise card. Yip the 286 mobo seems to be doing exactly what it should.

I scrounged around a bit more and found an Adaptic scsi controller card with fdd support , cable and a scsi hdd.Booted into a IBM PC Dos 6.3 installation boot disk, removed a partitions on the hdd. Created one active primary partition and saved the settings. Next rebooted back in to PC Dos, then formatted and installed PC Dos on to the hdd and rebooted. Up came the Dos prompt-happy me.

Did some more scrounging amongst my parts and pulled out the isa ram expansion card it kept off a system I wrecked years ago. Turned off the psu to the mobo, fitted the isa ram expansion card and then fire the system up again. All two megs of ram on the expansion card was counted along with the available conventional memory. Tho bios complained about the added memory, which was good and expected. I was presented with a text prompt saying press F1 or CTL-ALT-ESC to enter the bios. I pressed F1 and the system presented the C:>. Type mem then pressed enter and all the memory on the system was displayed.

Now I put this mobo in storage because I'd had a problem with it that I can't recall at present so more testing required.

So far I'm very please with the out come. 😀 I'll post a separate thread in due course.

Just finished lapping a new heatsink for one of my super 7 builds I'm working on. It certainly needed it, had a noticeable concave curvature to it when mounted on the chip, as the contact was mostly on the corners. I never really gave much thought to lapping heatsinks before, but now I think I'll have to going forward.

Possibly an indication of a place that just extruded them and chopped them, rather than properly tensioning them and aging them in an oven. Extrusion die is usually specced a tad oversize, it's pushed/pulled out onto a rack, should cool in air, then get stretched until dimension conforms, reduced to portable lengths, packed in a rack, aged in an oven for 4-8 hours depending on alloy, then be cut to finished size and have additional finishing operations, like cut for the clip slot, and maybe also crosscut through the fins. Anyhoooooo, if no stretched and aged, it coulda curled up with ageing in real time, rather than oven accelerated.

Started getting video footage of Leisure Suit Larry off the Tandy 1000A and discovered that it runs surprisingly well from clean boot at stock 4.77 MHz.

Going to download and move some vids from my other retrocomputing channel to my main variety channel for throwback thursdays for september and december as well.

BitWrangler wrote on 2022-07-28, 03:13:

Possibly an indication of a place that just extruded them and chopped them, rather than properly tensioning them and aging them in an oven. Extrusion die is usually specced a tad oversize, it's pushed/pulled out onto a rack, should cool in air, then get stretched until dimension conforms, reduced to portable lengths, packed in a rack, aged in an oven for 4-8 hours depending on alloy, then be cut to finished size and have additional finishing operations, like cut for the clip slot, and maybe also crosscut through the fins. Anyhoooooo, if no stretched and aged, it coulda curled up with ageing in real time, rather than oven accelerated.

The base was pretty rough regardless of the shape, so it needed it anyway. To be fair they are some of the cheaper heatsinks you can buy, so I'm not expecting a flat mirror finish. Checking the flatness of heatsinks should be standard practice anyway if you care about optimal cooling performance.