ok thanks!
Incidentally, talking of things that are listed for the GD5426/8, the display memory overclocking utilities in utilities.exe

https://vogonsdrivers.com/getfile.php?fileid= … 4&menustate=9,1

50mclkn.com (normal 50MHz speed)
&
57mclkn.com

do seem to run without reporting any errors, Ive only tried the vendor supplied 3D bench so far and there wasnt any difference there...
It does seem likely that video memory bandwidth isn't the bottleneck in this test, perhaps a better benchmark is needed?
anyone got any ideas for other tests?

Ive also not looked at the speed of the memory fitted, it might be rated for it, even if the DG5420 isnt!

Yoghoo wrote on 2026-08-30, 16:08:

I just used the 1.41 version for Windows 3.11 which is included on the CF card. Don't forget the patch (=overwrite a DLL if I remember correctly). Without patch it doesn't work correctly. Max is also 800x600 and 256 colors with this driver btw for my GD5240/512KB.

Yes, that was it, I was missing the patch. Even on the stock CF that I got from vendor (v1.41 drivers too), trying 800x600x256 would result in black screen with only white text appearing, but after applying the patch seems to work fine.
So I guess v1.50 must have that patch applied or no longer need it.

Jules_nerd wrote on 2026-08-30, 17:48:

50mclkn.com (normal 50MHz speed) & 57mclkn.com […]
Show full quote

50mclkn.com (normal 50MHz speed)
&
57mclkn.com

do seem to run without reporting any errors, Ive only tried the vendor supplied 3D bench so far and there wasnt any difference there...
It does seem likely that video memory bandwidth isn't the bottleneck in this test, perhaps a better benchmark is needed?
anyone got any ideas for other tests?

OK I tried some tests in Phils Dos benchmark collection.

Landmark shows a significant increase in video speed, from 3321 to 4703 chars/mS, however, back in a more real world...
both the Doom benchmarks (one tiny screen at low res,the other fullscreen) showed no discernible change in fps.
I guess this is no great surprise, the bottleneck in most real games/applications is going to be the processor and/or the ISA bus...
so it seems the utilities do work, but I cant think of a real instance where its going to help much!!

EDIT, increasing the video memory clock to 57MHz seems to work (at least it gives different results in landmark), but the utilty to set it back to 50MHz doesnt!
a reboot sets it back though...

has anyone got a Crystal CX4235-XQ3 or similar soundcard working with theirs?
Ive been struggling, Ive tested the card in another PC (though its more modern and has plug and play BIOS), it works OK..
In the BOOK386, using the latest crystal 'drivers' from the review in DOSdays, the card initailization program seems to find the card and set it up, but then the crystalware volume utility says it cant find the card?
Diagnostics programs find that there is a soundblaster at address 220, but tell me nothing is on IRQ5? Edit - red herring, there wasnt anything at 220

Unisound cant find the card when it tries to initialize it, but does find a CS4236 when you do UNISOUND /CS
Oddly at one point it found the network chip, but ive put that on address280 IRQ11 now, and ive not seen it find it since...

I got an Aztech based card working no problem...

OK I know people will be thinking why bother with the CZ(s)4235, but my plan was t use the internal OPL in the book, disabling it and the joystick in the crystal chip..
since we all know the opl emulation is poor in the CZ4235, they are one of the cheaper cards out there...

the only thing I havent tried is taking the internal soundcard out to see if a conflict there is the problem, but I think its should work with it in?
The OPL side at least shouldn't be an issue....

I suppose i could have soldered up my ISA adapter wrong! 😁, but it works OK with the aztech card??? or at least it does set the jumpers to use jumper selected setup, if I dont it doesn't boot, but the setup utilities seem to let me change the settings...

Looked in my stash but don't have a similar card like the Crystal CX4235-XQ3. Tried an ALS100 card and that worked correctly.

Have some unused OPTI, Sound Blaster and other ALS cards which I didn't test yet. But none are like the Crystal card unfortunately.

Thanks for looking..
So are you using the internal OPL & joystick & disabling them in the external soundcard?
all im after really is a simple one chip solution to add PCM sound, my idea was to track a PCB and maybe do a 3d printed case that plugs on the back to make it rather more practical than the ISA adapter +ISA card....
I think the ALS100 are reasonably cheap as well, so if they work like in tandem with what's already in the book.....

To be honest I didn't test it extensively. Just did a quick test with multiple ISA cards. I didn't need to adjust/disable anything. Unisound worked correctly and I observed that OPL3 sound came from the internal speaker and games recognized a SB16 compatible card and worked correctly.

Need to test more cards to see if there are any quirks with those but don't have a lot of free time. Recapping a motherboard (not the BOOK386) in the little free time I have atm. Maybe coming weekend I have some time to check it out.

I've come across an issue that may be relevant, it hasn't solved my issue with the CX(s)4235(yet), but I thought ID mention it as it may help others....
The more experienced on here may well correct me and indeed may well be aware of the issue ! apologies if I'm teaching my grandmother to suck eggs...
But it looks like in parts of the book386 only partial io address decoding is done, so you do get aliases, worth keeping in mind when you set IO addresses for sound cards etc...

TLDNR, keep io address assignments below 3FFh to be 100% safe. edit now ive done the maths AFAIK anything under 600h should be OK

The internal sound card, so that's the COVOX, OPL & Joystick ports have addresses fixed in hardware, but they are only decoding address lines A0 to A9, so will alias above 3FFh - apparently this was par for the course for ISA cards in the AT era...
The covox and the OPL only read from their addresses, so the only danger there is that if there is a clash they will just produce noise. The joystick port has data in both directions, I believe one address to tell it to 'fire' the timers and the results go on another address...
The M5113 multi IO chip does better, it seems to have A0 to A10 connected, so prob not an issue, it only aliases above 7FFh, i cant ever recall setting an address above this!
The RTL8019AS has 20 address lines, so better still...
Note that even though the M5113 & RTL8019 do better address decoding , that doesn't mean its 100% safe to use addresses over 3FFh for them, the other devices can alias into them...
So the easy option is to avoid using addresses above 3FFh (default on the CX4235 for WSS & CTRL are 534 & 538), I don't think the odds of a clash above 3FFh are that high, but if you want to use addresses above this and be 100% sure you'd have to do the maths...

Edit Ive done some maths, 530h, 530h & 538h should be safe from the internal soundcard alias, youd need to have something with 10bit partial address decoding @ address of 130h, 134h, and 138h to be a problem.

For ref, the aliases for the internal soundcard are:
Joystick:
201h aliases to 601h, A01h, E01h, 1201h, etc
202h aliases to 602h, A02h, E02h, 1202h, etc

covox:
278h aliases to 678h, A78h, E78h, 1278h, etc
OPL:
378h aliases to 788h, B88h, F88h, 1388h, etc.

Which are all above addresses Id expect to see, so, after all that long post, maybe not such a big issue!...

So, homework assignment for anyone with a Book 386 😀

Go and get yourself a Vonets RJ45 to WiFi adapter. I have the vap11n: https://x.com/RetroTechChris/status/2089079817685909894

Based on how the routing works for wireless, I actually used this little device at my local library recently, using a modern PC to get past the "I agree" page.: https://x.com/RetroTechChris/status/2093781956181852672

Other fun I have had with this:
* First impressions: https://threadreaderapp.com/thread/2087000458715877436.html
* I did add a math coprocessor: https://x.com/RetroTechChris/status/2091563963095929338
* This was my Win3.11 For Workgroups setup thread: https://threadreaderapp.com/thread/2087215009881285094.html
* My mouse fix was to use errorlevels and continue to reboot until detected. I like the patched mouse driver better, which I will use! https://x.com/RetroTechChris/status/2089091138917466340
* For some reasons I had issues with Netscape Navigator 2.0 and lockups, I think related to the Cirrus Logic video drivers I used. This was my fix: https://threadreaderapp.com/thread/2091695235009065049.html
* Floppy drive and ISA expansion testing: https://threadreaderapp.com/thread/2091202973376413798.html
* General exploring with video CH375, etc: https://threadreaderapp.com/thread/2087282849099506026.html

Worth noting that the ADLIBWV driver will give you WAV support in WFW 3.11, I think someone already covered that perhaps.

Oh, and of course for MIDI in Win3.1x, you can use Jamie O’Connell’s FM MIDI Synth drivers!!

Good stuff! i like the look of the Vonets adapter!
Some nice pics of inside, we can see that R2 is indeed still fitted (many removed a similar part on the clock circuits of the pocket386)
Shame I cant make out the value of R152, the NTC pull up resistor - do you have a high res original?
Unfortunately it seems that R1, C1 & U14 are on the bottom bottom side of the PCB 🙁 I'll have to take it in bits myself at some point!
javispedro1 worked out the trackpad issue & patched the CTmouse to get around the issue, neat use of error levels though! EDIT, download & discussion is further back in this thread
Ive been using the Yamaha (the driver seems to actually be from Yamaha themselves) OPL3 drivers in win311 & the covox driver (in bios set theLPt1 address to 278 &tell the driver that the covox is on LPT1), not sure where i found them, more than likely the vogons driver section.
The covox works much better than I thought it would, though there is a little bit of noise when it accesses the CF card (a little like hard drive head noise 😁), this is presumably power supply noise making its way on to the R2R ladder - but I havent measured anything to confirm that. if anyone wants to fix it filtering the supply should do the trick, but its not very loud TBH
Great Idea to use the video input to check out the panel res! I'd seen the feature, but got nervous about the idea of potentially connecting two VGA outputs together!!
I may have a bash at this, whatever the 1024x768 drivers that i tried were Im not sure, I presumed they were non-interlaced, the GD5420 datasheet says it can do both, it can do non interlaced @60 &70Hz I think, though the 75MHz max pixel clock on this chip limits it, so it cant manage the 72z that chips higher in the range can do (you're v lucky on the pocket have that graphics IC!)
I may have a play, see if its possible to find a 1024x768 setting that the LCD controller will "sync" at!

Ive tried using the book386 as a display for an EEPC701, it synced at all sorts of high resolutions, unfortunately the eeepc topped out at 1600x900, it doesn't do 1600x1200...
BUt at any rate, it looked pretty blurry at anything over 1024x768, which it would sync at at both 60Hz & 70Hz.
so im guessing its native res is 1024x768. hard to be 100% sure again...

But those 1024x768 resolutions/refresh rates are a basic match for the non interlaced modes of the GD5420 so either it is interlaced modes we've all been trying, or there is some timing detail that the WFW311 drivers configure it for that the LCD controller doesn't like...
It certainly seems worth investigating the drivers, the basic capabilities of the hardware seem to say 1024x768 should be possible...
At the very least Im going to see if i can somehow try both 60 & 70Hz

Thanks for the interesting observations re the panel's native resolution!

Jules_nerd wrote on 2026-09-07, 14:02:

But those 1024x768 resolutions/refresh rates are a basic match for the non interlaced modes of the GD5420 so either it is interlaced modes we've all been trying, or there is some timing detail that the WFW311 drivers configure it for that the LCD controller doesn't like...
It certainly seems worth investigating the drivers, the basic capabilities of the hardware seem to say 1024x768 should be possible...
At the very least Im going to see if i can somehow try both 60 & 70Hz

One option could be to try Linux & XF86, it allows you to set basically everything, including all mode timings by hand (not something I'd want to do 😀 ) . Back in the p386, this is how I first realized the "LCD" scaler would actually support 800x600 input (and downscale).

Re DOS, there was a TSR if I remember, CLMODE, it comes at least with the 1.50 driver I was using but likely the older ones as well. Seemed to allow setting the refresh rate, but not sure if it did anything better than the windows SETRES.
I can't play with it at the moment. I remeber that at least from DOS side the 1024x768 VESA mode also results in "No support" from the OSD, so it's not just the windows driver.

The Dram chips used for the main memory in the book386 appear to be GM71C17400CJ, 60nS, fast page mode, standard refresh
whilst mine seems to have been stable with the refresh set to 60uS in the BIOS settings, unfortunately if you've got those chips fitted, really I think it should be set to 15uS...
As I mentioned before with the pocket, the refresh timing needed can be v temp dependent etc, so trying settings and seeing if they are OK may be rather error prone....YMMV

The memory on the graphics card is 4x M514256C-60J, which are 4bit wide chips, so the graphics chip is working with 16 bit wide memory interface. I think this has adequate bandwidth for non interlaced 1024x768 4bit...

RetroTechChris wrote on 2026-09-06, 20:03:

* For some reasons I had issues with Netscape Navigator 2.0 and lockups, I think related to the Cirrus Logic video drivers I used. This was my fix: https://threadreaderapp.com/thread/2091695235009065049.html

If you really want to run Netscape Navigator try version 4.08. Works for me. But it's extremely slow.

Nice alternative if you need it would be Microweb which is a much much faster but runs under DOS and not under Windows 3.11.

javispedro1 wrote on 2026-09-07, 14:38:

Re DOS, there was a TSR if I remember, CLMODE, it comes at least with the 1.50 driver I was using but likely the older ones as well. Seemed to allow setting the refresh rate, but not sure if it did anything better than the windows SETRES.
I can't play with it at the moment. I remember that at least from DOS side the 1024x768 VESA mode also results in "No support" from the OSD, so it's not just the windows driver.

Ive tried the CLMODE that was installed with the V1.5 divers, it doesn't seem to work - it doesn't recognize the graphics IC, I presume thats why it only allows 640c480 @60Hz...
in its 'GUI' regardless of the monitor selected that's all it gives you, same from the command line, it just rejects settings for anything other than 640x480...
Shame, it looked promising!
I guess its possible that this version of CLMODE is only meant for later chips in the range - the bundle does say its for later chips....
may be worth seeing if earlier versions support this IC, it doesnt seem to be bundled with the V1.2 i was using before...

Further random idea on mouse and keyboard problems... now this only applies because maker previously did Book8088 then used a very similar keyboard, so did the keyboard routines get reused??

So apparently there is a gotcha in XT keyboard handling on an AT class, where the address or resources used by XT keyboard conflicts with PS/2 mouse address. This is just something I saw in passing over at DOSDays as a brief thing about using XT keyboards, so didn't really cite chapter and verse. Anyway, probably wouldn't really apply at all apart from it seems like some things are repurposed between the Book models, so might be an outside chance.

When playing some games Ive seen the keyboard suddenly stop working as it should in one case it behaved as if the right arrow is pressed all the time...
Only just had it happen the first time on the Book a few days ago, had similar things happen with thepocket386 as well.
Resorting to pressing every key in frustration sometimes brings it back to normal operation...

IN the Book386 it looks as if the external keyboard data and clock on the M6117 (KBCLOCK, KBDATA on the schematic) go to connector P1, which is the rear panel connector for an external keyboard.
The built in keyboard seems to send /recieve data from the motherboard 8051 (KB_TxD, KB_RxD, presumably for the control of the FN+key functions you get displayed on the trackpad), and it seems to send data to the M6117 KBCLOCK, KBDATA lines via a couple of 510R resistors R19 &R20. Presumably the idea is that if you plug in an external keyboard it has a lower output Z and overcomes any drive from the internal keyboard via the resistors. Might be worth poking around to see if the signal is still clean/sharp edges after going through these resistor - its prob OK, that's a pretty low resistance & its low datarate stuff...

From the PCB pics on RetroTechChris's links It looks as if the keyboard has its own 8051 doing the matrix scanning etc... which sends data to the motherboard 8051 and directly (via the resistors) to the M6117.
Im not saying it would be easy, or indeed if it this is where the problem lies, but if that micro is just dedicated to keyboard scanning, then it might be possible to look at the data its sending to the motherboard 8051, work out what its doing and write code for the keyboard 8051... presumably the data its sending to the M6117 must be standard ps2 keyboard data - though I dont understand how that fits in with the scan code errors that javispedro1 mentions.

Ive been playing around trying to speed up the Book386 as mush as possible.. learning to use debug etc to interrogate/program the M6117 registers..
so far I can say:

1. The BIOS is shadowed - read only.
2. The Video bios is shadowed, but both read and writes are allowed to the shadowed area...
3. Whilst the pocket uses 50nS EDO & the Book uses 60nS FPM ram(with interleaving for the FPM), i think both are set to 1 wait state for a miss & 0 wait states for a hit, so there will be very little difference in performance - both are very good for a 386 system.

When setting the ISA clock to higher rates, 8MHz seems stable, even for windows (311, not tried 95). Anything higher crashes windows.
The PCLK/6 setting (13.33Mhz, pclk2 is 2x40MHz) seems stable for DOS (its a slight overclock for the GD5420, rated for 12.5MHz), though I haven't tried using the parallel or serial ports, nor the Ethernet. My suspicion here is that the multi IO and ethernet may well struggle at that speed, Id guess 10MHz may be ok for them, but.. if your not using them in games... you can prob use the 13MHz bios setting for DOS games.
Im only seeing something like a 10% increase in frame rates on things like Doom, I suspect that its just freeing up a bit more processor time because writing to the graphics IC is taking less time.
Ive totally removed any additional wait states etc for the ISA bus in the BIOS , seems OK, though again Ive not tested Ethernet or IO chip.

Ive been educating myself on what the 8254 PIT is used for in a PC, apologies to those that know this, its just generic PC hardware stuff... Its a programmable interval timer (the 'I' doesn't stand for interrupt, though it is connected to an interrupt). It is usually running from a 1.19318 MHz clock derived from the 14.31818 MHz- the 7.15909MHz default ISA clock is half this same clock), it has 3 timers in it:

1. Prob the most important to us, its connected to IRQ0 and programmed by the BIOS to give an 18.2Hz 'tick' used by DOS to keep time (it only reads the RTC on boot and when you run 'TIME'). I believe some programs like games may also use this to ensure they run at consistent speeds on different systems. Other operating systems like windows or linux re-program this to run faster (like 100Hz or 1kHz) & use it for lots more, time-slicing etc etc..
2. DRAM refresh, Im not sure if it is used for this on the M6117.
3. PC speaker Beep timing.

Now when I run the ISA clock high, it speeds up graphics, but if you use a utility like checkit2 to verify the coherence of the DOS clock and the RTC it all looks OK, so I suspect the PIT clock is being left at alone, regardless of the isa clock, so I dont think the issue on the M6117 is that changing the ISA clock changes the PIT clock...
If it was an old external IC id suspect that speeding up the ISA clock means writes to configure the 8254 are just to fast to work correctly, but inside the M6117? who knows!!

Ive done some batch/text files using debug that allow you to increase the ISA clock & use the feature where the M6117 can use dual clocks, with the VGA address range set to a higher clock...
Ill test them later and post them on here if they seem to work/offer anything over just setting a fixed clock in the BIOS!!!

OK well Ive given these a quick test and they seem to work.
They use the DOS debug to manipulate the M6117 setup registers to give higher ISA clocks also another to set the video shadow to read only (in case it bothers you that default allows writing).
put them all in the same folder and execute the .bat s from there, that way they will find the.txt s that the debug input is redirected to.
I'm sure someone will recommend a tidier way of doing it!!

isa10_13 sets a high ISA clock rate of 13.33MHz for the VGA address range, 10Mhz for every other address
isa8_10 sets a high ISA clock rate of 10MHz for the VGA address range, 8Mhz for every other address
isa8 just sets the ISA clock to 8Mhz
ISA7 sets it back to the default that works with windows etc so you dont have to re-boot

ISA8 seems to work with windows, but isa10_8 does not.....
Im not sure why.. nor can i guarantee 8 Mhz is OK for windows, i only gave it a quick go, and i still dont understand the nature of the problem with high isa clocks and windows...

TBH dont expect much!!! its quite hard to tell that the ISA clock is higher, it doesn't make that much difference, if your looking for a benchmark then speedsys 's VGA bandwidth seems to be a good way to see it. I think it prob illustrates that the 386 really is the bottleneck in these.... still you can expect 5-10% at least

incidentally , i notice bypassing my startup files on my machine gave as much improvement to doom benchmarks!! 😁
I haven't investigated that yet though...

But even if these aren't much use, put them together with the m6117 datasheet & it serves as an example of manipulating the registers....
I accept no responsibility for what may happen!!!

Edit, i can see no reason why these wouldn't work on the pocket as well as the book, though I not sure you get much advantage over using BIOS settings on the pocket, with there being so little else on the ISA bus...
Edit 2 ive corrected an error in one of the files and added a file to change the control register index 32 "ISA high speed switching" and set this to 80, I think this may give an additional benefit, not sure...yet again teh datasheet seems rather terse! its default value is 00