noshutdown wrote on 2025-03-04, 03:26:

windows operation should be smoother than plain svga cards without acceleration

This is only relevant for GDI, and not WinG or DirectX. I seriously question the effectiveness of GDI benchmark scores when it comes to actual real-world usage, specifically what CPU is required for there to be any significantly perceived difference in real-world usage. I feel like that it stops becoming relevant past a 386 outside of one specific shareware game which really utilises GDI to its limits. Granted, outside of industrial SBC+backplane setups there's only three known Socket 5/7 motherboards that are ISA-only so there's a practical hard limit on what to investigate.

I was hoping that the Windows 3.1 version of Spectre VR was going to be an incredible example of GDI usage, but it seems like it calculates everything internally and spits everything out via blitter operations - which makes sense given its DOS/Mac/Win3.1 cross-port nature, and the fact that mach32 cards screw up massively and shifts the entire output inside the window roughly a quarter/third to the left.

cirrus: 5426/28/29/34.

When it comes to synthetic benchmarks like WindSock GDI Detailed and WinBench's full benchmark (and not the useless Graphics WINMARK that everyone uses), there is significantly less difference between the GD5424 and those you listed when it comes to ISA when using the latest Windows 3.1 GD5424 drivers, version 1.50. The same is most likely true for the GD5402 and GD5420 as well, but I haven't tested those outside of a 286.

headland: little info but ht216 said to be an accelerator.

This is correct - synthetic benchmarks are relatively poor compared to ATi/S3 cards, but like I said above I have good reason to believe that it's far less significant than what is perceived. Also, HT216 cards and WD90C31 ISA cards are very rare. If you're going to talk about HT216, at least talk about the Avance Logic ALG2101 and Weitek W5186/W5286 while you're at it given they all existed in the same time period of late 1992/early 1993.

oak: 087.

The 087 and 087X have differences - in synthetic benchmarks the 087X is better overall in GDI functions, but the 087 has better blitter performance. I have a lot of Win 3.1 footage recorded from an 087X (along with ET4000, GD5424 and S3 928) captured via OSSC on a 386DX40 and 486DLC-40 on the same motherboard that I've been meaning to edit into a video for Youtube, I'll need to focus more on this when time is on my side and get more footage from other cards especially the 087, along with trying to find some more good examples of GDI usage. At least Doom map editors (WadAuthor, WinDEU etc) fill the CAD criteria easily, and I can find plenty of custom maps that would be perfect for testing.

realtek, trident and umc didn't make any isa accelerator cards.

Not true, but they're unicorns that exist in only one domestic market in the same way that WD90c33 ISA and HT216 cards are unicorns in general.

cirrus: 5422/5424 and some 5420 are truecolor. note that the early version of 5420 is 256colors only, and with 16bit dram width. but according to the 542x tech ref manual, 5420rev.b and newer are updated to truecolor and 32bit dram width, just like the 5422 except slightly lower clock.

Do you have a photo of a 5420 chipset that has this high/true colour mode support? My 5420 card doesn't support it at all.

headland: no info, but some ht216 cards were seen with hicolor ramdac.

My HT216 card only has 8-bit output - trying to use the two HT216 drivers that I know of which are only found on Packard Bell OEM CDs fails to get to the desktop with 640x480x32K and 800x600x32K and the driver files for 640x480x65K and 800x600x65K modes are missing, but considering that only a single ISA card is known (the Video 7 Win.VGA) and that the only known usage of VLB HT216 is integrated in Packard Bell motherboards from late 1993, I'd like to know what you actually mean by "some ht216 cards".

oak: few oti077 cards was shipped with hicolor ramdac, and most if not all 087.

I have four 087X ISA cards and none of them have high colour RAMDACs, yet my single 087 card does.

realtek: probably none, but the uncommon 3106(with 1mb dram max) is rumored to have bios support for hicolor modes, however i have never seen one shipped with hicolor ramdac, not sure if ramdac swapping is possible.

There are Windows 3.1 drivers for 3105/6 that have high colour modes, but I haven't really spent much time with my 1MB 3106. It does have a socketed RAMDAC, so at some point I'll need to experiment.

s3: all but a few 911 cards shipped with only 256colors ramdacs. 924 and later support truecolor but some cards may have hicolor ramdac only.

From what I can tell this is purely a VBIOS thing - my Orchid Fahrenheit is a 911 with high colour support, probably true colour support given the RAMDAC it has but I haven't tested the true colour Win 3.1 driver yet. My Diamond Stealth VRAM is a 924 yet lacks 65K colour support in its BIOS though 32K colour works fine - if I swap the BIOS chips from my Orchid card, it suddenly gains 65K support.

western digital: some but not all wd90c11/30/31 cards have hicolor ramdac, few wd90c31 even has truecolor. all wd90c33 are believed to be truecolor. note that the wd90c11 supports 512kb of ram at max so it would only go up to 640*400 in hicolor, just like the trident9000c except being much faster.

I have four WD90c30 cards, and not a single one has high colour support. The datasheet for the WD90c30 doesn't even mention it, though curiously there are a couple of Win 3.1 drivers aimed specifically at the WD90c30 (which amusingly only matters for 16-c0lour modes, everything else uses the same files as earlier WD90c31 drivers and there are clear performance differences between both cards using those same files) which do list high colour modes in their OEMSETUP.INF.

DEAT wrote on 2025-03-05, 06:05:

Do you have a photo of a 5420 chipset that has this high/true colour mode support? My 5420 card doesn't support it at all.

no i don't have the card, but the cirrus tech ref manual appendix b5 says so. could it be restrained by bios to make some difference from the 5422?

My HT216 card only has 8-bit output - trying to use the two HT216 drivers that I know of which are only found on Packard Bell OEM CDs fails to get to the desktop with 640x480x32K and 800x600x32K and the driver files for 640x480x65K and 800x600x65K modes are missing, but considering that only a single ISA card is known (the Video 7 Win.VGA) and that the only known usage of VLB HT216 is integrated in Packard Bell motherboards from late 1993, I'd like to know what you actually mean by "some ht216 cards".

this, from yjfy a chinese collector site, the same card can also be found on "salted fish", the chinese junk market app. the card from octek has hicolor ramdac but strangely only 512kb of ram, no idea what octek was thinking about. actually i was wondering "why spend a hicolor ramdac on a 512kb card" to find out that its a headland.

From what I can tell this is purely a VBIOS thing - my Orchid Fahrenheit is a 911 with high colour support, probably true colour support given the RAMDAC it has but I haven't tested the true colour Win 3.1 driver yet. My Diamond Stealth VRAM is a 924 yet lacks 65K colour support in its BIOS though 32K colour works fine - if I swap the BIOS chips from my Orchid card, it suddenly gains 65K support.

this is a grammatical mistake, what i intended to mean is that all s3 cards, other than a few 911s that shipped with 256color ramdac, have hicolor or truecolor.

noshutdown wrote on 2025-03-05, 08:18:

this, from yjfy a chinese collector site, the same card can also be found on "salted fish", the chinese junk market app. the card from octek has hicolor ramdac but strangely only 512kb of ram, no idea what octek was thinking about. actually i was wondering "why spend a hicolor ramdac on a 512kb card" to find out that its a headland.

Interesting, thanks for the link - the 1994 copyright date on the PCB is interesting seeing as Headland/Video 7 went out of business in 1993. The only explanation I can think of for that card existing is making a quick buck out of liquidated assets.