I have read through this entire thread and, I have to admit, I am not sure if this device is worth having or not. Seems there's quite a few issues with it. Are people, in general, happy with the purchase?

I have Pocket 8086 and I am very happy with that. I bought it to try out old XT-class games and to learn 8086/8088 Assembly language on.

I don't have a 386 in my collection. Though, I am building a pseudo 386 out of a Pentium MMX at 100Mhz with cache disabled, and ISA video card, and EDO RAM, with SB Pro 2.0.

I was considering the Pocket 386, as potentially useful for running games that won't run on my Pocket 8088 or Pseudo 386. But, that's theoretical though, as I haven't found any games that I can't run yet. Lol.

RetroPCCupboard wrote on 2025-07-11, 06:43:

Are people, in general, happy with the purchase?

As a gaming device, I have a ton of issues mainly related to its size. I was aware this would be an issue before I purchased. If I was going to use it for an extended gaming session, I would defiantly hook external mouse, external keyboard, and use an external monitor. Stuff like the RTC not working reliably doesn’t bother me at all, sometimes it loses the bios settings which is annoying. A 386 class machine wasn’t necessarily something I was nostalgic for, I would have preferred a fast 486 or slow pentium.

I find it to be a convenient device for me to test things on mainly due to the integrated usb storage support- I can do some programming in dosbox on my laptop and then transfer my work to usb and instantly test on the Pocket386 right next to my laptop.

So it’s not perfect, but I like having a portable dos machine for testing.

BinaryDemon wrote on 2025-07-11, 11:15:

As a gaming device, I have a ton of issues mainly related to its size. I was aware this would be an issue before I purchased. If I was going to use it for an extended gaming session, I would defiantly hook external mouse, external keyboard, and use an external monitor. Stuff like the RTC not working reliably doesn’t bother me at all, sometimes it loses the bios settings which is annoying. A 386 class machine wasn’t necessarily something I was nostalgic for, I would have preferred a fast 486 or slow pentium.

I find it to be a convenient device for me to test things on mainly due to the integrated usb storage support- I can do some programming in dosbox on my laptop and then transfer my work to usb and instantly test on the Pocket386 right next to my laptop.

So it’s not perfect, but I like having a portable dos machine for testing.

Thanks for the reply. Yes, I wasn't planning to game with the built-in keyboard / mouse functionality. I have a PS/2 & VGA KVM. So I would just connect the device to one of the KVM ports.

I'm happy with mine - despite all the long posts Ive done on the faults/shortcomings, most arent that bad or that much of an issue - and if you really do want to fix them & youve some soldering skills many are easily fixed/overcome.
Mine has had loads of use Im certainly getting my money's worth out of it. I like the small form factor, yeah the keyboard is cramped, and I get lots of typos... and sometimes hit the wrong key when playing a game and use an energy bomb or and escape capsule when I dont want to in Elite+.... but Im still OK with the compromise, it's a natty little thing to have in the living room or take on hols to play games on. Still wish it had at least 8 bit digitised sound capability though.

Outside of that I like having something from that hardware era that I dont feel is so precious that I cant get into minor hardware mods etc... back in the day I wouldnt have dared to take a soldering iron to the motherboard of my (expensive!) main, everyday PC!!
Ive even thought of buying a second one!
Having said all that, if your planning on using a KVM and using it sat at a desk then I suspect the machine you are building will serve you better - better sound capabilities, if you use an S3 graphics card as good or better DOS game compatibility etc, a built in CDROM & floppy drive would make life easier....
Far easier to add or change bits if you dont like something or want to try more things - like adding more memory for Linux or whatever...
I have 2 such similar machines, a 233mmx with S3 graphics with OPl3+16 bit audio on the motherboard and a 1.3GHz Athlon with TNT2 M64 and an ESS soundcard; with these on my KVM for DOS/Win98 games (and alegedly some old ISA based electronics test equipment 😉 ) I cant see me using the Pocket 386 at that desk... tempting to hook it up, but in reality I doubt Ill ever use it like that...
If you want something smaller than a desktop machine (and are up for a soldering project) to plug ito your KVM maybe look at something along the lines of the Kharon 386?

I'm willing to help with any new mods to the Pocket386. I can do schematic/pcb design on Altium or Kicad and I know how to use a soldering iron. My pocket386 works fine apart from the usual clock resets.
I mainly use it for a bit of C programming practice using turbo C++. I'm still looking around for uses for a DOS database app like Foxbase which is installed. Expansion of the parallel port functionality would be good for simple robotics or industrial control type applications. A parallel and serial port is sadly missed on the rest of my modern PC's. Also learning about how to read/write from an ATA based hard drive can be useful. Also the same with any VGA or sound card improvements. I bought mine so I could use it to learn something useful and maybe have some fun at the same time.

The mystery of the unused footprint on the V1.2 serial PCB I mentioned on a previous post has been partially solved:
https://forum.vcfed.org/index.php?threads/poc … 40/post-1446285
Says that it is a match for ESP32-C3F devices - a module aimed at adding wifi & bluetooth functions to simple devices, very much the kind of thing used in external serial wifi modems others have mentioned on here. It seems the vendor was looking at doing something like an internal version...
The ESP32 are 3v3 and there is a regulator on the PCB, but I cant see any official info that says the inputs on this module are 5V tolerant (the 16c550 on the PCB are on 5V supplies) - in fact the datasheet definately says not, in its max voltages for the IO (VDD+0.3V), some seem to say that they get away with it, or even that these modules were designed with potection...

The though hole connector footprint on the V1.2 serial PCB is mostly connected to the pads on the ESP footprint & the signal names on the silkcreen seem to tally with the ESP32-c3f module to some extent, perhaps minor differences could be on the specific vendor module I looked at, or just how they decided to configure the module....
I think the footprint pin out does seem a better match (ADC, IO numbers, EN) for the older, cheaper (fits with this vendor 😉 ), ESP-12F
https://ar2uino.wordpress.com/tag/esp-12/

If you are prepared to gamble on the IO being 5V tolerant (add series resistors?), an ESP-12f module could prob be made to work with some programming & are very, very cheap!

Interesting news, thank you. I am hoping I get a "boring" winter as regards other events this year, so I can play with this and other retro projects, feel like I've only logged about 6 hours retro playtime all year so far.

TBH this is a new area for me - Ive never used any of these modules, & Im not a programmer....
I cant find any schematics for the V1.2 PCB, so Ive done a partial reverse engineer of the v1.2 Serial PCB, the following are the connections from the ESP-12F (or E) module:
ESP-12F RXD0 > R5 100R > pin 8 16C550 SOUT
ESP-12F TXD0 > R4 100R > pin 7 16C550 SIN
ESP-12F EN > R3 4k7 > 3v3
ESP-12F IO0 > R6 4k7 >3v3
ESP-12F IO2 > R7 4k7 > 3v3
ESP-12F IO15 > R8 4k7 > GND

So it seems there are 100Ohm resistors between the ESP-12F RX & TX and the UART, this will a least offer some protection from the 3v3 5v difference, though the value seems low & wouldnt keep current within the 16c550 limits of 1.8mA, more like 1kOhm should...
but 1k might cause issue with high data rates... hardware manual for the ESP8266 (used on the ESP-12E/F) explicitly states a level shifter should be employed between it and a 5V UART & gives the attached reccomended circuit.
I cant see any typical serial flow control signals connected.
Im not sure why IO0 & IO2 have pull ups fitted - of course with reverse engineeering its possible I have these connections wrong 😀
That the vendor hasnt offered this for sale does make me wonder if they ever got this working!

Ive ordered a few ESP-12E to have a play...

IIRC correctly "Node MCU" variants of ESP dev boards were using 5V serial somehow.

Edit: 🤣 or correctly even. IIRC already got a correctly in it, but leaving it for humor.

There seems to be quite a few forum posts from people saying that they have used these with 5V io with no issue.....
but the hardware manual says you shouldnt do it ¯\_(ツ)_/¯
My shrug emoticon didnt work, I'll leave it in for humour 😉

it seems this has already been done:
OS/2 Warp and 115200bps SLIP Connections

Ive put an ESP-12E on the V1.2 serial card with a little mod & it works, my pocket386 is now acessing my WiFi network!
OK, so the resistors on IO0, IO2 & IO15 set the ESP-12 boot mode and should be left alone.
I changed R5 to 1k8 & put a 3k3 resistor from the ESP12 side of it to ground (scraped off a little resist on the ground plane to solder to) as in the pic (honest the soldering looks worse in the pic than it did under my mircoscope!) , this forms a potential divider on the serial tx line from the 16c550 to drop the 5V signal down to about 3v3. I was wrong in my post above, with a 5V supply the 16C550 can supply 4mA from this pin, we are way under that with these resistors. A quick guestimate says the rise fall times should be OK with these resistors for the speeds we are looking at, but I havent put a 'scope probe on it yet to check... You might get away without doing something like this, but I wouldnt risk it myself.

R4 can be left @100R, though the hardware guide for the ESP8266 does suggest a 500Ohm resistor here to reduce RF emmisions - there should be one in the module really, close to the screening can, but the datasheet on mine says there isnt one.... So you change it 500Ohms or so if you like (might help with speaker clicks below)
A quick look at the 3v3 signal levels from the esp to the input on the 16c550 looks there is a decent enough noise margin without level translation in that direction.

Talking of RF inteference, I did notice some clicks from the speakers, in particular the right one, though oddly that seems to have dropped in level now (perhaps now its connected its dropped its RF output level??), this should be no great suprise - Ive just put and RF module & aerial right over the audio circuits!! If it gets annoying I may look at shielding the audio ccts somehow, but the level its dropped to doesnt seem tooo bad, given we arent talking quality audio anyway..
I programmed the modue with the binaries straight from:
https://github.com/martin-ger/esp_slip_router

soldered it in place...
I used EtherSL, & mTCP library

All works a treat so far!
but ive only got as far as pinging computers on my home network so far....

Good going, great job!

Ive hooked a scope probe up to the RX & TX on the ESP-12. results attached..
I dont see any issue, the rise time on the ouput of the resistor potential divider I added to the output of the 16C550 isnt fast, the scope prob probably adds 25nS to it because of the high impedance... whilst slow for logic, take tha off & its still below 0.5% of a 115200 baud pulse width. The voltage levels here give several hundred mV of noise margin o the ESP-12 input. On the output of the ESP-12, its fine driving the 16C550 on its 5V rails, high output is waaaay above the max threshold of 2V & low logic level is a similar margin below the minimum low level threshold.... I wouldnt trust the rise time measurement too much - the scope used doesnt have fast enough sampling for good measurements of one shot events, but its prob in the ball park.
So whilst just using a potential divider in one direction to do the level transation is a bit 'bodgy', It shouldnt give anyone any trouble if they wish to use it.

Now lets see if i can get it to play with a modern windows file/folder sharing.... bet I spend more time on that than I have getting the ESP fitted!

A word of warning to those that are thinking of trying an ESP-12 on the serial card...
Im glad Ive done it, but so far its very, very slow - at 115200 baud (which, with no overhead I thought would imply a max of 14kB/s ish) I expected it to be slow... but its slower still!
In wfw 3.11 using trumpet winsock and using explorer 3 Im seeing transfer rates of around 1Kbyte/s often a bit less. Ive tried other browsers and opera, netscape 2 etc seem pretty similar, opera maybe a little faster.
TBH this looks like the poor little 386 running flat out trying to render webpages and images (simple webpage, html and a few .gif s or .jpg s ) - it seems there isnt enough processing time left to get the most out of TCP/IP over the serial link - if you save the webpage files to hard drive, or just have them cached in the browser, yes of course its faster, but not as much as you might expect, so max possible transfer data rates arent really all thats limiting browsing speed...
I went on to win95 and ie3 seems to run a bit better! Im using OSR2, and I didnt see a null modem driver in the standard install, I found a .inf file that seems to have come from cisco, this lets you select a TCP connection to a unix system that works OK
Browsing on ie3 on Win95 is actually faster, but not much! system monitor does report the processor is at 100% doing this kind of thing..
when running Win95, Ive got windows network file/folder shares to other windows 95/98 machines working, but I couldnt get WINS to work over the connection so had to resort to an LMHOSTS file (not such bad thing for me no a little home network...)
Im seeing about 6KByte/ second transfering large files over the windows shares like this... so I assume there is quite a large overhead on the connection!!
So, yeah of course its slow over a serial port, but may be a little slower than you might expect (well certainly I expected a bit more! I was hoping for more like 10KByte/s) .

Peopl on here seem to report wildly different speed for win95, Ive stuck to osr2 because its meant to have a smaller memory footprint. But also I dont load a CH375 driver for the USB, at boot, unless I need it; I ran a benchmark to see how much slower for CF is when its forced to run in 16 bit.
SiSoft SAndrareports
with no CH375 driver loaded:
unbuffered read 5MB/S
sequntial read 1203KB/S
random read 1256 KB/S
Buffered write 1596KB/S
sequential Write 773 KB/S
random write 628KB/s
estimated access time 2mS (ive seen this as 1mS on previous tests)

if you load the dirver for the usb these drop to transfer rates in the range of something like 330KB/S to 400 KB/s ish (cant rember exactly...) and an 'estimated' access time of 10mS
So the real mode disk access from using a driver for the USB really does slow down Win95...

You might want to play with MTU settings in TCP/IP config, think modem type connections prefer something like 576 vs network card doing 1500. Gotta bring it down until the buffers aren't spilling over and you're getting lots of retries.

However, yes, gif and jpg are gonna straight up murder a 386 with no 387, and beat it unconscious WITH one. Even DX2s could struggle with "in era" sized gifs, which were typically actually optimised for web load and display, whereas nobody has bothered since 1998 or so. If you put a magazine cover CD on the network from another machine, that has cached webpages on it from 94-96ish then you may find performance merely sluggish looking at those. Some older browsers have better settings for improving performance on limited hardware/connections like images off, or low quality image rendering.

Ta, I was hoping someone would reply with sugestions!
I remember at uni ( along time ago!) the electronics dept where i was studying had a whole lab of 386 (sx 25MHz I think) machines with 4MByte of RAM & no hard drives -everything over 10Mbit Ethernet - I gave up trying to browse the web (pretty early days for the www...) on those!
It does feel like something isnt right, I was expecting the transfer rates to be a bit higher. I had a quick play with MTU etc in Trumpet winsock, but I'll have another go now you've said that. At one point I did notice one picture stoped after only about 90% of it was loaded, it did feel like it had stopped after so many re-tries somewhere along the line, or something similar.
The website Ive been using is very undemanding, 50KByte of html & a 55KB jpeg, the next page I use is more demanding, 145KB or so of html & 51 jpegs, mostly between 9KB and 20KB a couple over that.
Even though Im moaning about the speed, I did still spend a while browsing that site, it wasnt sooo bad...

But even just accessing windows shares over the network seems a little slower than I thought it should be, I mean even a 386 can handle TCP/IP over a network adapter at decent rates, so surely it cant be the processor struggling with TCP/IP.... so 6KB/S seems a bit poor for a 115200 baud rate link, unless the overhead is much higher than I thought?

Another ref to save others trying this some time: I tried a Win95 update, an windows unoficial update pack (ok, maybe asking for trouble!) to update to newer versions of the dial up modem & Winsock. Id be wary of this! The bloat from the updates slowed win 95 to a crawl (much higher use of swap file, even just after boot), but the modem/dial up adapter updates broke the dial up that I was using: the null modem definitions Im using still go through the pantomime of dialing up, even though there is no modem, but it seems to get through this OK, perhaps just ignoring what happens with phoone dialing & trying to connect anyway.... With the updates I had to re-install the modem and then it just complains that the modem doesnt respond...
I suppose I could try sending the esp-12 some AT commands via hyperterminal to see how it handles them, but I am using the slip version of the software for it, not the modem software.... But its a very long time since Ive used AT commands!!! Id have to brush up on those!

Fortunatley I took an image of the compact flash before updating & Ive reverted to that!

Jules_nerd wrote on Yesterday, 14:14:

But even just accessing windows shares over the network seems a little slower than I thought it should be, I mean even a 386 can handle TCP/IP over a network adapter at decent rates, so surely it cant be the processor struggling with TCP/IP.... so 6KB/S seems a bit poor for a 115200 baud rate link, unless the overhead is much higher than I thought?

Hi there! Please have a quick look at this thread over here: how do i get MS-DOS 6.22 online on PCem?
There are a few programs with their own network stack.
I got them working on as low as an PC/XT class computer with 4,77 MHz..
Maybe it's useful for testing purposes. There are also solutions for serial connection.

Good luck! 🙂🤞