@Paul_v
Hi, I am using VORTEX86DX processor, and I can't find a way to write BIOS online through JTAG port. Fortunately, I saw your post. Can you tell me whether to flash through JFLASH.exe? What other devices are needed on the hardware? Can you send me the Jflash.exe file? Thanks!

lesocreates wrote on 2025-07-24, 22:47:

@Paul_v
Hi, I am using VORTEX86DX processor, and I can't find a way to write BIOS online through JTAG port. Fortunately, I saw your post. Can you tell me whether to flash through JFLASH.exe? What other devices are needed on the hardware? Can you send me the Jflash.exe file? Thanks!

I'm sorry, but I've since moved on to other projects and don't have the files at hand. You'll have to do some legwork.
This link seems to be up to date.
Re: Vortex86DX - A beastie of it's own (General info and discussion)

Archive jflash127.zip.asc is AES256 encrypted binary. Should be easy to decrypt using "gpg" (or similar) utility with the password provided.
The cable is XILINX Parallel Cable III or any clone. Match output TDI\TDO\TCK etc to jtag port pins on the target device.
I used HP thin client with LPT port and WinXP as a host.

P.S. Make a full backup of your current ROM.
Or at least write down network MAC. It is stored on the same IC. Can be reprogrammed using BIOSMP tool.
Flashing through jtag may reset it to FF's

@Paul_v
Thank you very much！, I will try your method.

Hi there, first of all, thanks for this thread, this helped me gather the information I needed for a little project of my own related to one of these boxes 😁

I have a couple of DMP eBox 3310A-L2, which happen to have a mini pci slot occupied by a second network card ...

I don't need a second network card, but I always thought that mini pci slot could come handy.

And then recently I stumbled across a MSI MS-9513 which apparently was meant for this server board https://theretroweb.com/motherboards/s/msi-e7 … ter-ls2-ms-9121 for a good price

It's basically a mini PCI ATi Rage XL card. Nothing revolutionary, but I actually think this card might be exactly what the DMP boxes needed. The chip itself is readily available (thanks to a massive supply of aliexpress cards), and unlikely to be scalped due to its very cheap nature and the fact it got mass produced, and while its performance isn't revolutionary at all, I think it's actually just right for the Vortex86 chips ! Furthermore it has a very good OS support (MS-DOS vesa modes, Win3.1, win 9x, win NT4, OS2, win2k ...)

So first of all, I had to test the card to make sure it was working so I got myself a mini pci to pci adapter and used a bog standard MSI board to test.

However if you pay close attention you'll notice that the card has no bios onboard ...
And yep, it needs the BIOS to be integrated to the main bios, hence why I'm commenting here ! So anyways I first checked the BIOS from the original motherboard it was meant for and extracted the VBIOS, then I injected it into my "sacrificial" motherboard and ... I got video !

Now that I made sure the litle potato worked, I had to mod the DMP bios ... First I extracted it using uniflash, then I modded it with mmtools to inject the ragexl vbios and remove the volari vbios:

And then I inserted the card into the DMP box

And it worked !

So then I proceeded to install win98 on my CF card, install the drivers, directx, etc ...

And then I made couple of benchmarks:

First of all 3dmark99 in 640x480x16bpp

Then 3dmark99 in 800x600x16bpp

Then I tried a game of UT99 in 400x300x16bpp in low details on the first map

After the game I had ~24FPS on average, not bad !

Just to compare, I checked again in software rendering mode (same res, same settings) and here's how it looked like

After the game I had ~11FPS on average. So the RageXL definitely accelerates the rendering, even if it's not a rocket, it's turning something totally unplayable into something """acceptable""". Also it looks a LOT better in hardware mode (the two screenshots don't quite make justice for the hardware rendered mode, but just look at the way the weapon looks, the software rendered mode looks blocky and even has some "green dots" that shouldn't be there, whereas the hardware rendered one looks "okay")

Now this specific card and my setup have some drawbacks that may impact severely the performance of the tests:
- I'm using USB HID because DMP placed the second ethernet jack in place of the PS/2 connector. I'll probably remove it and replace it with a proper PS/2 connector if I want to make this mod "definitive"
- This specific Rage XL card uses only one chip of 32 bit SDRAM, but the chip is actually capable of 64bit SGRAM (or 64bit SDRAM). I'm not exactly sure how much exactly this impacts the performance, but yeah I'm pretty sure as potato as it is, even this chip is starving for bandwidth !
- I didn't have sound installed

Also fun fact, despite not having a VBIOS anymore, the original volari graphic chip is still detected by windows 98, so technically I can also do a dual monitor setup 🤣 (I might desolder it entirely and remove the original VGA connector to replace it with a connector that goes to the rage card). For those who want to attempt this mod and aren't familiar with soldering, you can also just remove one of the two serial ports if you don't need two of them.

Now that doesn't fix the issue with the sound chip, but hey now 50% of the problems this machine has are fixed 😁

Deksor wrote on 2025-09-22, 02:27:

Now that doesn't fix the issue with the sound chip, but hey now 50% of the problems this machine has are fixed 😁

I'd love to find a soundblaster-compatible solution for those eboxes. I'm not sure if the best approach would be SB-LINK/PC-PCI, LPC, or something else (TDMA?)

I believe SBEMU works on them though

Well mine has the CMI119, aka usb sound ... So literally anything pci would be better than this ...

That's a challenge to say the least.
EBOX machines can have same model number with differrent SoC.
If I recall, 3310 with DX has PCI slot and USB based CM119 soundchip (like on the pic above). SBEMU won't work with it.
MX version has integrated audio which works with SBEMU, but only has LPC bus.

The boards I have do not seem to have any LPC/ISA pins exposed, all LPC devices are integrated.
IDK, maybe there's a software way to do this, there should be at leat some registers to decode LPC over Serial I think.
Or a super hardcore BGA-to-SoC adapter with ISA.

I was thinking of something different ... Now that the volari chip is basically useless, how about taking it off and stealing it's PCI signal for a highly integrated sound chip (Yamaha ymf724 ?) on a flex cable?
And then also remove the cmedia chip and connect the other end of the flex cable to the amplifier circuit of the cmedia ?

Maybe I'm not considering some other issues, but so far the main issue have is that I cannot find the pinout of the volari chip ...
If anybody has it 🙏

PCI is a parallel connection bus, it would be easier to make miniPCI splitter.
Would probably also require modifying IRQ routing table.

Well in that case how about making a specific mini PCI card ?
Because I don't think a mini PCI splitter + two cards would fit in the case anymore.

I dunno if the mini PCI slot here is fully wired but from the pinout I have I can see that it even supports audio passthrough so it may not even need any hardware modification to the PCB then.

I don't want to tease too much, but I'm currently working on making a mini PCIe rage XL with the idea of using it with @eivind's tinyllama for the same reasons I explained here : the chip is super simple, capable enough for what the vortex86 offers in terms of performance, have a very good retro compatibility and they're readily available for very cheap if you accept to do bga desoldering (I may also have found a source for nos chips ...). Just grab any shitty Aliexpress rage XL for ~10 bucks and off you go 🤣

I basically see it like the Crystal CS4237B is for sound on DMP boxes. Not a "good" chip, but for this niche I think it's probably the best pick.

It even has an integrated tdms output, aka DVI/hdmi (although it's limited to 1024x768, compared to the maximum of 1600x1200 for the VGA output according to the specs)

I can't promise anything yet because I'm not even finished doing the tracing and I don't know it I will be able to make a mini PCIe card that will *fit* below the tinyllama yet (the main issue is that I need to add a PCIe to PCI bridge on the same PCB which of course requires more space).
But yeah that's definitely doable by someone patient and skilled enough, especially for mini PCI like here (and it also allows you to improve the vram bus speed issue I talked about earlier)

Whether I abandon it or not, I'll eventually share the kicad project

Also according to ati's datasheet the rage xl is almost pin-compatible with the ATI rage 128VR chip so if the graphics speed isn't satisfactory enough to some, there's also room for improvement there (although the rage 128vr had to ditch the tdms output down in the process...)