VOGONS


ET4000AX variants

Topic actions

First post, by Grzyb

User metadata
Rank Oldbie
Rank
Oldbie

The original ET4000 (probably the same as ET4000AX) datasheet only mentions up to 8-bpp modes - https://archive.org/download/bitsavers_tsengL … roller_1990.pdf
Early ET4000AX boards, eg. Orchid ProDesigner II, are equipped with RAMDACs like KDA0478, ie. 8-bpp.
Markings seen on the ET4000AX on such a board - "TC6058AF"

8bpp.jpg
Filename
8bpp.jpg
File size
718.87 KiB
Views
1059 views
File license
Fair use/fair dealing exception

Later ET4000AX boards, eg. Orchid ProDesigner IIs, support up to 15-bpp (but not 16-bpp), with RAMDACs like SC11483, ie. 15-bpp.
Markings seen on the ET4000AX on such a board - "TC6100AF"

15bpp.jpg
Filename
15bpp.jpg
File size
1.29 MiB
Views
1059 views
File license
Fair use/fair dealing exception

Even later ET4000AX boards, eg. Micro-Labs Ultimate VGA/TrueColor - http://www.microlabs.com/ut.html - support up to 24-bpp, using ATT20C490 RAMDAC.
Markings seen on the ET4000AX on such a board - "TC6100AF"

24bpp.jpg
Filename
24bpp.jpg
File size
1.23 MiB
Views
1059 views
File license
Fair use/fair dealing exception

Can't find any ET4000AX chips other than TC6058AF and TC6100AF.

So...
There are probably just two variants of ET4000AX:
* 8-bpp TC6058AF
* 24-bpp TC6100AF

And boards based on the latter are often limited due to a less-than-24-bpp RAMDAC.

Reply 2 of 23, by Grzyb

User metadata
Rank Oldbie
Rank
Oldbie
j^aws wrote:

I've seen more variants. I've got a 59AF as well.

Any idea about its capabilities?
What RAMDAC is it coupled with?

Anyway, with the help of http://vgamuseum.info/index.php/companies/ite … -tseng-et4000ax , I compiled a list of RAMDACs found with both variants of ET4000AX:

TC6058AF

JT82C176 - probably 8-bpp, like other 176 chips
KDA0476 - 8-bpp
KDA0478 - 8-bpp

TC6100AF

ATT20C490 - 24-bpp
CIC10476 - probably 8-bpp, like other 476 chips
KDA0476 - 8-bpp
MU9C9760 - 8-bpp
SC11483 - 15-bpp
TR9C1710 - 8-bpp

Reply 3 of 23, by BloodyCactus

User metadata
Rank Oldbie
Rank
Oldbie

these are just revisions of the chip, the ramdac is responsible for the BPP mode capabilities. The 6058AF is same pin count as 6100AF. The old 8bpp ramdacs are mostly designed to be dropins for the original inmos g171 ramdac (i have quite a bunch of ramdac chips here).

I'm aware of
TC6046AF (seen datecode 9025)
TC6051AF
TC6058AF (seen datecode 9035, 9050, 91xx)
TC6059AF (seen datecode 92xx)
TC6100AF (seen datecode 92xx, 9307, 94xx)

4 year span of revisions before we hit the W32 chipset.

--/\-[ Stu : Bloody Cactus :: http://kråketær.com :: http://mega-tokyo.com ]-/\--

Reply 4 of 23, by Grzyb

User metadata
Rank Oldbie
Rank
Oldbie
BloodyCactus wrote:

these are just revisions of the chip, the ramdac is responsible for the BPP mode capabilities

Do you want to say that even the earliest revision can provide 24-bpp, if combined with a proper RAMDAC?
I would expect the datasheet mentioning such a feature, really high-end by 1990 standards...

Edit:
After browsing through the datasheet more carefully, I've found a few mentions of 65536 color graphics, so I guess there was no 8-bpp ET4000 chips, they were 16-bpp from the very beginning.
But nothing about 16777216 colors, 24-bpp must have been added in some later revision, possibly TC6100AF.

Reply 5 of 23, by root42

User metadata
Rank Oldbie
Rank
Oldbie

I'm not sure that in linear mode a distinction between 15, 16 and 24bpp modes actually makes sense. After all the ET4000 is just a pretty dumb framebuffer in that mode. In contrast to plane based 4bpp modes for example. But.I don't know enough of the hardware. Either way, with only 1MiB of RAM, the possible resolutions in 24bpp would be rather limited. I think all the non-indexed color modes were a bit useless and slow on ISA cards without acceleration. My Trident cards can also do 15 and 16bpp in Windows, but at weirdly low resolutions and with lots of flicker on CRTs. Also those modes are slooooooow.

YouTube and Bonus
80486DX@33 MHz, 16 MiB RAM, Tseng ET4000 1 MiB, SnarkBarker & GUSar Lite, PC MIDI Card+X2+SC55+MT32, OSSC

Reply 6 of 23, by BloodyCactus

User metadata
Rank Oldbie
Rank
Oldbie
Grzyb wrote:
Do you want to say that even the earliest revision can provide 24-bpp, if combined with a proper RAMDAC? I would expect the data […]
Show full quote
BloodyCactus wrote:

these are just revisions of the chip, the ramdac is responsible for the BPP mode capabilities

Do you want to say that even the earliest revision can provide 24-bpp, if combined with a proper RAMDAC?
I would expect the datasheet mentioning such a feature, really high-end by 1990 standards...

Edit:
After browsing through the datasheet more carefully, I've found a few mentions of 65536 color graphics, so I guess there was no 8-bpp ET4000 chips, they were 16-bpp from the very beginning.
But nothing about 16777216 colors, 24-bpp must have been added in some later revision, possibly TC6100AF.

Well, they didnt have 24bit ramdacs back in the early 1990's. Everyone started by copying the inmos G171 ramdac which was just 8bpp from a 18bit lut internally, later ramdacs expanded on that adding 15bit + 16bit colour as a side channel then 24bpp.

early ramdacs were really slow, 35mhz etc. later ramdacs hit 110mhz, 160mhz which could do those high colours and higher resolutions.

the ramdac speed is what determines the resolutions are capable. The tr9c1710 can hit 110mhz pixel clock.
eg: 800x600x60hz needs 40mhz dot clock. 1280x720x60hz needs 75mhz... 1024x768x60 needs 65mhz

the early modes like 640x480x60 only needed 25mhz which I think all ramdacs had at least 33-35mhz

Everyones 15/16/24 bpp modes were not register compabitle so you couldnt swap ramdacs around and hope they worked. The bios or vesa driver had to know/detect your ramdac to know what modes to support.

--/\-[ Stu : Bloody Cactus :: http://kråketær.com :: http://mega-tokyo.com ]-/\--

Reply 7 of 23, by Grzyb

User metadata
Rank Oldbie
Rank
Oldbie
root42 wrote:

I'm not sure that in linear mode a distinction between 15, 16 and 24bpp modes actually makes sense.

OK, I've done some more research, and now I'm inclined to believe that for a graphics controller chip like ET4000AX it doesn't matter how many bits the RAMDAC uses to create a single pixel.

The graphics controller outputs 8 bits/cycle, and that SC11483 HiColor RAMDAC latches the 16 bits from 2 cycles.
Can't find the datasheet for ATT20C490 (can anybody help?), but I guess it latches 24 bits from 3 cycles.

So, for the graphics controller, 640x480x24bpp is identical to 1920x480x8bpp, and all it needs is a fast enough pixel clock.

The datasheet for WD90C31 also hardly mentions 16-bpp, and doesn't mention 24-bpp at all, but there are 24-bpp cards based on that chip - Windows 3.1 drivers for Diamond Speedstar 24x

Either way, with only 1MiB of RAM, the possible resolutions in 24bpp would be rather limited. I think all the non-indexed color modes were a bit useless and slow on ISA cards without acceleration.

1 MB is enough for 640x480x24bpp, and this was the standard in early TrueColor SVGA cards.
Indeed, it wasn't fast, and rather inconvenient for typical work with a GUI, but very useful eg. for displaying JPEG pictures - a slow graphics card was still faster than JPEG decoding on a 386.
And ray-tracing was even slower...

Reply 8 of 23, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie
Grzyb wrote:
j^aws wrote:

I've seen more variants. I've got a 59AF as well.

Any idea about its capabilities?
What RAMDAC is it coupled with?

I'm certain it's not capable of 24bpp.

I can't be sure what RAMDAC it has without digging it up. If you recall the warehouse filled with unlabelled boxes at the end of the movie Raiders of the Lost Arc, it's pretty much buried like that.

Budget RAMDACs were from many manufacturers back then, including Samsung, AMD?, INMOS and many others. The only tests I did was to compare output clarity, which varied quite a bit.

Reply 9 of 23, by BloodyCactus

User metadata
Rank Oldbie
Rank
Oldbie

outside of 8bpp, the ramdac just passes the bits straight through. the ramdac was originally a look up table to convert 256 palette colours into R/G/B values. once you leave 8bpp your leaving half the function of the chip behind. all your using is the dotclock to do all the sync and reads from your dram.

j^aws wrote:

Budget RAMDACs were from many manufacturers back then, including Samsung, AMD?, INMOS and many others. The only tests I did was to compare output clarity, which varied quite a bit.

Inmos wasnt budget, they created the first VGA ramdac for IBM. They were also heading into bankruptcy too 😀

even a modern 32bpp ramdac running at 180mhz like TI TLC34076 https://www.exdwh.com/TLC34076.pdf still has Inmos G171 and BT476 compatability in it.

--/\-[ Stu : Bloody Cactus :: http://kråketær.com :: http://mega-tokyo.com ]-/\--

Reply 10 of 23, by imi

User metadata
Rank l33t
Rank
l33t

my trashpicked Diamond SpeedSTAR24 has a Samsung KDA0476PL-80, idk about color depth, but all I can say is that the output of that card is clean af.

hardware32_07.jpg
Filename
hardware32_07.jpg
File size
238.59 KiB
Views
859 views
File license
Fair use/fair dealing exception

Reply 11 of 23, by Grzyb

User metadata
Rank Oldbie
Rank
Oldbie
BloodyCactus wrote:

the ramdac was originally a look up table to convert 256 palette colours into R/G/B values. once you leave 8bpp your leaving half the function of the chip behind.

In other words, RAMDAC in modes greater than 8-bpp doesn't use the RAM section, only the DAC.

imi wrote:

my trashpicked Diamond SpeedSTAR24 has a Samsung KDA0476PL-80, idk about color depth

Now I'm confused...
I thought that in "SpeedSTAR24", the "24" stands for TrueColor support.
But KDA0476 is definitely 8-bpp, the datasheet is easy to find.

Edit:
And this SpeedSTAR24 - http://old.vgamuseum.info/images/stories/palc … v.c1_top_hq.jpg - is equipped with SC11483, ie. 15-bpp.
Curiouser and curiouser!

Edit2:
And see what I've found in Windows 3.x drivers for that card...
"SpeedSTAR 24 (Small Font) -640x480x16.7M"

Reply 12 of 23, by BloodyCactus

User metadata
Rank Oldbie
Rank
Oldbie

both cards are REV C1 pcb layout.. both are 44pin PLCC chips. I bet you could swap them and they would work.

--/\-[ Stu : Bloody Cactus :: http://kråketær.com :: http://mega-tokyo.com ]-/\--

Reply 13 of 23, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie
BloodyCactus wrote:
j^aws wrote:

Budget RAMDACs were from many manufacturers back then, including Samsung, AMD?, INMOS and many others. The only tests I did was to compare output clarity, which varied quite a bit.

Inmos wasnt budget, they created the first VGA ramdac for IBM. They were also heading into bankruptcy too 😀

even a modern 32bpp ramdac running at 180mhz like TI TLC34076 https://www.exdwh.com/TLC34076.pdf still has Inmos G171 and BT476 compatability in it.

I've seen their gold top RAMDACSs which look fancy and highend, but I've also seen them in typical packaging in budget cards as well.

Reply 14 of 23, by Am386DX-40

User metadata
Rank Newbie
Rank
Newbie
Grzyb wrote on 2019-11-11, 21:07:

Can't find the datasheet for ATT20C490 (can anybody help?), but I guess it latches 24 bits from 3 cycles.

ATT20C490: https://datasheetspdf.com/pdf-file/530532/Age … ems/ATT20C490/1
ATT20C491: https://datasheetspdf.com/pdf-file/530534/Age … ems/ATT20C491/1

I have a really nice Genoa MultimediaVGA Model 7900 with ET4000AX (6100 version) and it has an ATT20C491-80 DAC, that's why I'm posting both datasheets.

s-l1600.jpg

Reply 15 of 23, by mkarcher

User metadata
Rank Member
Rank
Member
Grzyb wrote on 2019-11-12, 01:49:
BloodyCactus wrote:

the ramdac was originally a look up table to convert 256 palette colours into R/G/B values. once you leave 8bpp your leaving half the function of the chip behind.

In other words, RAMDAC in modes greater than 8-bpp doesn't use the RAM section, only the DAC.

Not necessarily. In 15/16bpp modes, you only have 32 or 64 levels of R/G/B. Especially true-color RAMDACs often allow you to map those levels arbitrarily to the 256 levels the hardware is capable of. Note that R, G and B is mapped individually. This feature is typically used for gamma correction.

Reply 16 of 23, by Eep386

User metadata
Rank Member
Rank
Member

Here's a really old, lesser known variant of ET4000AX for you: 24340F0004. Date code 9009EAI.
Seen on an extremely old Orchid ProDesigner II.
It's also really slow, though I suspect that's more to do with the low clock speeds it's configured with.

IMG_1546.JPG
Filename
IMG_1546.JPG
File size
1.51 MiB
Views
336 views
File license
Public domain

Life isn't long enough to re-enable every hidden option in every BIOS on every board... 🙁

Reply 17 of 23, by Grzyb

User metadata
Rank Oldbie
Rank
Oldbie
Am386DX-40 wrote on 2021-03-04, 22:36:
Grzyb wrote on 2019-11-11, 21:07:

Can't find the datasheet for ATT20C490 (can anybody help?), but I guess it latches 24 bits from 3 cycles.

ATT20C490: https://datasheetspdf.com/pdf-file/530532/Age … ems/ATT20C490/1

Thanks... so it indeed does latch 24 bits from 3 cycles.
With such a RAMDAC, it might be possible to add TrueColor even to plain VGA, though in this case the resolution would be useless - 320 x 200 x 8bpp -> 106 x 200 x 24bpp.

I would like to know if there's any difference between all those ET4000AX variants...

Reply 18 of 23, by Grzyb

User metadata
Rank Oldbie
Rank
Oldbie
mkarcher wrote on 2021-03-04, 23:06:
Grzyb wrote on 2019-11-12, 01:49:
BloodyCactus wrote:

the ramdac was originally a look up table to convert 256 palette colours into R/G/B values. once you leave 8bpp your leaving half the function of the chip behind.

In other words, RAMDAC in modes greater than 8-bpp doesn't use the RAM section, only the DAC.

Not necessarily. In 15/16bpp modes, you only have 32 or 64 levels of R/G/B. Especially true-color RAMDACs often allow you to map those levels arbitrarily to the 256 levels the hardware is capable of. Note that R, G and B is mapped individually. This feature is typically used for gamma correction.

Yes, and there's a difference between ATT20C491/492 and ATT20C490/493 - the former do support TrueColor gamma correction, the latter don't: "True-color bypass operation only", ie. the RAM section is bypassed, only the DAC is used.

Reply 19 of 23, by Am386DX-40

User metadata
Rank Newbie
Rank
Newbie
Grzyb wrote on 2021-03-06, 08:33:

I would like to know if there's any difference between all those ET4000AX variants...

Yeah, me too. Does the "plain" ET4000 even exist, or are all of them ET4000AX?