hcyandluodayou wrote on 2025-04-23, 13:25:

Does anyone have any details or knowledge about it?
This is a strange graphics card for more than one reason. RasterOps video cards are mostly used on Macintosh and adopt the NuBus interface. As far as I know, RasterOps did not launch ISA interface graphics cards nor use S3 display chips. Having three display chips but only one VGA output interface is not a traditional multi-screen display card. How do the three display chips work? Can they bring effects similar to Voodoo SLI to enhance color depth or increase resolution? There are resistors soldered next to the chips and manually soldered jumpers. Is this an engineering sample graphics card?

That's definitely a very intersting artifact. I suggest you open a new thread for this card, instead of posting pictures and questions about a historical ISA card in a thread about newly designed ISA cards. I don't suggest this because I want to be a rules enforcer, but posting your topic in a different thread with a specific title (like "RasterOps ISA video card with 3*S3 924 - what is it?") will likely get a lot more attention of the forum members that can actually help you with this card.

Nevertheless, I can do a quick assessment of what I see: The card uses a Bt473 RAMDAC. That chip can receive 24-bit true-color data on 24 dedicate input lines. IMHO, it's likely the card is made for True-Color operation with each S3 chip supplying one of the 3 components of R, G and B. This is in-line with the component numbering in the tripled sections. Take a look at the S3 chips being labeld UR1, UG1 and UB1, and similar labels for the RAM chips UB2..UB9, UG2..UG9 and UR1..UR9. I got the Bt473 datasheet, and I can clearly identify the traces that provide the red pixel data the RAMDAC, those are the traces below CB158 through RPR3 to the upper right corner of UR1. The corresponding green pixel values are sent from UG1 via RPG3 into some vias below the chip, but I cant see any traces on the front or back side, so this card obviously has signals on inner layers.

It's surprising the card manufacturer took this route, though. The S3 924 is perfectly capable of handling and accelerating TrueColor within a single graphics chips, the number "24" at the end most likely alludes to this fact. While the S3 86c924 only has an 8-bit output port, which is not well suited to output true-color at high resolution, the suggested way of operating the 911 series of S3 chips (best-known members: 911, 924, 928, 964, 966, 968) is to only use that 8-bit output port in VGA compatible mode, and use a direct connection from the VRAM (a special kind of DRAM that includes a row-sized shift register which is perfectly suited to output one pixel after the other) to the RAMDAC, bypassing the S3 chip, for advanced high-color modes (or even very high-res 256-color mode, in which multiple pixels can be passed in parallel to the RAMDAC). The "standard" RAM width of the 924 is 32 bits, and that's exactly what you see here: Each S3 chip has 8 VRAM chips that are four bits wide, resulting in 32 bits. Wiring these chips directly to the RAM-DAC can send 4 256-color pixels, 2 16-bit pixels or one true-color pixel to the RAMDAC per clock cycle. For some reason (maybe increased memory bandwidth?), your card runs three separate S3 graphics engines in 8-bit modes and combines their output into one true-color output.

mkarcher wrote on 2025-04-23, 15:14:

That's definitely a very intersting artifact. I suggest you open a new thread for this card, instead of posting pictures and que […]
Show full quote

hcyandluodayou wrote on 2025-04-23, 13:25:

Does anyone have any details or knowledge about it?
This is a strange graphics card for more than one reason. RasterOps video cards are mostly used on Macintosh and adopt the NuBus interface. As far as I know, RasterOps did not launch ISA interface graphics cards nor use S3 display chips. Having three display chips but only one VGA output interface is not a traditional multi-screen display card. How do the three display chips work? Can they bring effects similar to Voodoo SLI to enhance color depth or increase resolution? There are resistors soldered next to the chips and manually soldered jumpers. Is this an engineering sample graphics card?

That's definitely a very intersting artifact. I suggest you open a new thread for this card, instead of posting pictures and questions about a historical ISA card in a thread about newly designed ISA cards. I don't suggest this because I want to be a rules enforcer, but posting your topic in a different thread with a specific title (like "RasterOps ISA video card with 3*S3 924 - what is it?") will likely get a lot more attention of the forum members that can actually help you with this card.

Nevertheless, I can do a quick assessment of what I see: The card uses a Bt473 RAMDAC. That chip can receive 24-bit true-color data on 24 dedicate input lines. IMHO, it's likely the card is made for True-Color operation with each S3 chip supplying one of the 3 components of R, G and B. This is in-line with the component numbering in the tripled sections. Take a look at the S3 chips being labeld UR1, UG1 and UB1, and similar labels for the RAM chips UB2..UB9, UG2..UG9 and UR1..UR9. I got the Bt473 datasheet, and I can clearly identify the traces that provide the red pixel data the RAMDAC, those are the traces below CB158 through RPR3 to the upper right corner of UR1. The corresponding green pixel values are sent from UG1 via RPG3 into some vias below the chip, but I cant see any traces on the front or back side, so this card obviously has signals on inner layers.

It's surprising the card manufacturer took this route, though. The S3 924 is perfectly capable of handling and accelerating TrueColor within a single graphics chips, the number "24" at the end most likely alludes to this fact. While the S3 86c924 only has an 8-bit output port, which is not well suited to output true-color at high resolution, the suggested way of operating the 911 series of S3 chips (best-known members: 911, 924, 928, 964, 966, 968) is to only use that 8-bit output port in VGA compatible mode, and use a direct connection from the VRAM (a special kind of DRAM that includes a row-sized shift register which is perfectly suited to output one pixel after the other) to the RAMDAC, bypassing the S3 chip, for advanced high-color modes (or even very high-res 256-color mode, in which multiple pixels can be passed in parallel to the RAMDAC). The "standard" RAM width of the 924 is 32 bits, and that's exactly what you see here: Each S3 chip has 8 VRAM chips that are four bits wide, resulting in 32 bits. Wiring these chips directly to the RAM-DAC can send 4 256-color pixels, 2 16-bit pixels or one true-color pixel to the RAMDAC per clock cycle. For some reason (maybe increased memory bandwidth?), your card runs three separate S3 graphics engines in 8-bit modes and combines their output into one true-color output.

Thank you for your detailed analysis and suggestions! As a new member of the vogons forum, I don't have the permission to start new threads yet.
The images I provided are public, so anyone can repost and share them. I find this discussion very interesting.
Regarding your analysis, I find it very professional and well-supported. You mentioned that the manufacturer's choice of design is surprising. Are there other graphics cards that use a similar design?
Additionally, Did the acquisition of TrueVision in 1992 influence RasterOps to not only continue developing NuBus interface graphics cards but also start developing ISA interface graphics cards? Information on this topic is quite scarce, and I haven't been able to verify it.
Thanks again for your help!

hcyandluodayou wrote on 2025-04-23, 13:25:

Seeking Help: Information on RasterOps Graphics Card with S3 P86C924 Chip Hi everyone, […]
Show full quote

Seeking Help: Information on RasterOps Graphics Card with S3 P86C924 Chip
Hi everyone,

I recently acquired a vintage graphics card with an S3 P86C924 chip, and the boot screen displays the RasterOps logo. I'm looking for more information about the origin and specific model of this card. Does anyone have any details or knowledge about it?
This is a strange graphics card for more than one reason. RasterOps video cards are mostly used on Macintosh and adopt the NuBus interface. As far as I know, RasterOps did not launch ISA interface graphics cards nor use S3 display chips. Having three display chips but only one VGA output interface is not a traditional multi-screen display card. How do the three display chips work? Can they bring effects similar to Voodoo SLI to enhance color depth or increase resolution? There are resistors soldered next to the chips and manually soldered jumpers. Is this an engineering sample graphics card?

Super cool graphics card. Exactly the kind of thing I am interested in.

Quite off topic for this thread though, this was about designing new manufacture ISA cards.

hcyandluodayou wrote on 2025-04-23, 17:04:

You mentioned that the manufacturer's choice of design is surprising. Are there other graphics cards that use a similar design?

I checked for other RasterOps cards, and it seems the "paste an 8-bit graphics engine three times and connect a TrueColor DAC" theme had already been tried on NuBus cards, the ColorBoard 364 for example, or the PaintBoard 24. And that's why...

hcyandluodayou wrote on 2025-04-23, 17:04:

Additionally, Did the acquisition of TrueVision in 1992 influence RasterOps to not only continue developing NuBus interface graphics cards but also start developing ISA interface graphics cards?

... I think the RasterOps card you showed likely has more "RasterOps" than "TrueVision" in it, although some know-how in interfacing to the ISA bus might have come from TrueVision.

hcyandluodayou wrote on 2025-04-23, 17:04:

Thank you for your detailed analysis and suggestions! As a new member of the vogons forum, I don't have the permission to start […]
Show full quote

mkarcher wrote on 2025-04-23, 15:14:

That's definitely a very intersting artifact. I suggest you open a new thread for this card, instead of posting pictures and que […]
Show full quote

hcyandluodayou wrote on 2025-04-23, 13:25:

Does anyone have any details or knowledge about it?
This is a strange graphics card for more than one reason. RasterOps video cards are mostly used on Macintosh and adopt the NuBus interface. As far as I know, RasterOps did not launch ISA interface graphics cards nor use S3 display chips. Having three display chips but only one VGA output interface is not a traditional multi-screen display card. How do the three display chips work? Can they bring effects similar to Voodoo SLI to enhance color depth or increase resolution? There are resistors soldered next to the chips and manually soldered jumpers. Is this an engineering sample graphics card?

That's definitely a very intersting artifact. I suggest you open a new thread for this card, instead of posting pictures and questions about a historical ISA card in a thread about newly designed ISA cards. I don't suggest this because I want to be a rules enforcer, but posting your topic in a different thread with a specific title (like "RasterOps ISA video card with 3*S3 924 - what is it?") will likely get a lot more attention of the forum members that can actually help you with this card.

Nevertheless, I can do a quick assessment of what I see: The card uses a Bt473 RAMDAC. That chip can receive 24-bit true-color data on 24 dedicate input lines. IMHO, it's likely the card is made for True-Color operation with each S3 chip supplying one of the 3 components of R, G and B. This is in-line with the component numbering in the tripled sections. Take a look at the S3 chips being labeld UR1, UG1 and UB1, and similar labels for the RAM chips UB2..UB9, UG2..UG9 and UR1..UR9. I got the Bt473 datasheet, and I can clearly identify the traces that provide the red pixel data the RAMDAC, those are the traces below CB158 through RPR3 to the upper right corner of UR1. The corresponding green pixel values are sent from UG1 via RPG3 into some vias below the chip, but I cant see any traces on the front or back side, so this card obviously has signals on inner layers.

It's surprising the card manufacturer took this route, though. The S3 924 is perfectly capable of handling and accelerating TrueColor within a single graphics chips, the number "24" at the end most likely alludes to this fact. While the S3 86c924 only has an 8-bit output port, which is not well suited to output true-color at high resolution, the suggested way of operating the 911 series of S3 chips (best-known members: 911, 924, 928, 964, 966, 968) is to only use that 8-bit output port in VGA compatible mode, and use a direct connection from the VRAM (a special kind of DRAM that includes a row-sized shift register which is perfectly suited to output one pixel after the other) to the RAMDAC, bypassing the S3 chip, for advanced high-color modes (or even very high-res 256-color mode, in which multiple pixels can be passed in parallel to the RAMDAC). The "standard" RAM width of the 924 is 32 bits, and that's exactly what you see here: Each S3 chip has 8 VRAM chips that are four bits wide, resulting in 32 bits. Wiring these chips directly to the RAM-DAC can send 4 256-color pixels, 2 16-bit pixels or one true-color pixel to the RAMDAC per clock cycle. For some reason (maybe increased memory bandwidth?), your card runs three separate S3 graphics engines in 8-bit modes and combines their output into one true-color output.

Thank you for your detailed analysis and suggestions! As a new member of the vogons forum, I don't have the permission to start new threads yet.
The images I provided are public, so anyone can repost and share them. I find this discussion very interesting.
Regarding your analysis, I find it very professional and well-supported. You mentioned that the manufacturer's choice of design is surprising. Are there other graphics cards that use a similar design?
Additionally, Did the acquisition of TrueVision in 1992 influence RasterOps to not only continue developing NuBus interface graphics cards but also start developing ISA interface graphics cards? Information on this topic is quite scarce, and I haven't been able to verify it.
Thanks again for your help!

Short article on the cards introduction...

PC Hoarder Patrol wrote on 2025-04-23, 18:40:

hcyandluodayou wrote on 2025-04-23, 17:04:

Thank you for your detailed analysis and suggestions! As a new member of the vogons forum, I don't have the permission to start […]
Show full quote

mkarcher wrote on 2025-04-23, 15:14:

That's definitely a very intersting artifact. I suggest you open a new thread for this card, instead of posting pictures and questions about a historical ISA card in a thread about newly designed ISA cards. I don't suggest this because I want to be a rules enforcer, but posting your topic in a different thread with a specific title (like "RasterOps ISA video card with 3*S3 924 - what is it?") will likely get a lot more attention of the forum members that can actually help you with this card.

Nevertheless, I can do a quick assessment of what I see: The card uses a Bt473 RAMDAC. That chip can receive 24-bit true-color data on 24 dedicate input lines. IMHO, it's likely the card is made for True-Color operation with each S3 chip supplying one of the 3 components of R, G and B. This is in-line with the component numbering in the tripled sections. Take a look at the S3 chips being labeld UR1, UG1 and UB1, and similar labels for the RAM chips UB2..UB9, UG2..UG9 and UR1..UR9. I got the Bt473 datasheet, and I can clearly identify the traces that provide the red pixel data the RAMDAC, those are the traces below CB158 through RPR3 to the upper right corner of UR1. The corresponding green pixel values are sent from UG1 via RPG3 into some vias below the chip, but I cant see any traces on the front or back side, so this card obviously has signals on inner layers.

It's surprising the card manufacturer took this route, though. The S3 924 is perfectly capable of handling and accelerating TrueColor within a single graphics chips, the number "24" at the end most likely alludes to this fact. While the S3 86c924 only has an 8-bit output port, which is not well suited to output true-color at high resolution, the suggested way of operating the 911 series of S3 chips (best-known members: 911, 924, 928, 964, 966, 968) is to only use that 8-bit output port in VGA compatible mode, and use a direct connection from the VRAM (a special kind of DRAM that includes a row-sized shift register which is perfectly suited to output one pixel after the other) to the RAMDAC, bypassing the S3 chip, for advanced high-color modes (or even very high-res 256-color mode, in which multiple pixels can be passed in parallel to the RAMDAC). The "standard" RAM width of the 924 is 32 bits, and that's exactly what you see here: Each S3 chip has 8 VRAM chips that are four bits wide, resulting in 32 bits. Wiring these chips directly to the RAM-DAC can send 4 256-color pixels, 2 16-bit pixels or one true-color pixel to the RAMDAC per clock cycle. For some reason (maybe increased memory bandwidth?), your card runs three separate S3 graphics engines in 8-bit modes and combines their output into one true-color output.

Thank you for your detailed analysis and suggestions! As a new member of the vogons forum, I don't have the permission to start new threads yet.
The images I provided are public, so anyone can repost and share them. I find this discussion very interesting.
Regarding your analysis, I find it very professional and well-supported. You mentioned that the manufacturer's choice of design is surprising. Are there other graphics cards that use a similar design?
Additionally, Did the acquisition of TrueVision in 1992 influence RasterOps to not only continue developing NuBus interface graphics cards but also start developing ISA interface graphics cards? Information on this topic is quite scarce, and I haven't been able to verify it.
Thanks again for your help!

Short article on the cards introduction...

The attachment RasterOps PaintBoard PC.jpg is no longer available

It's amazing that you found the news screenshot from back then! This gives me a deeper understanding of the history of this graphics card.

mkarcher wrote on 2025-04-23, 17:28:

I checked for other RasterOps cards, and it seems the "paste an 8-bit graphics engine three times and connect a TrueColor DAC" t […]
Show full quote

hcyandluodayou wrote on 2025-04-23, 17:04:

You mentioned that the manufacturer's choice of design is surprising. Are there other graphics cards that use a similar design?

I checked for other RasterOps cards, and it seems the "paste an 8-bit graphics engine three times and connect a TrueColor DAC" theme had already been tried on NuBus cards, the ColorBoard 364 for example, or the PaintBoard 24. And that's why...

hcyandluodayou wrote on 2025-04-23, 17:04:

Additionally, Did the acquisition of TrueVision in 1992 influence RasterOps to not only continue developing NuBus interface graphics cards but also start developing ISA interface graphics cards?

... I think the RasterOps card you showed likely has more "RasterOps" than "TrueVision" in it, although some know-how in interfacing to the ISA bus might have come from TrueVision.

This information makes me cherish my collection even more. The history of graphics cards is truly fascinating. Every time I see these antique graphics cards, I think of their glorious moments.

PC Hoarder Patrol wrote on 2025-04-23, 18:40:

hcyandluodayou wrote on 2025-04-23, 17:04:

Thank you for your detailed analysis and suggestions! As a new member of the vogons forum, I don't have the permission to start […]
Show full quote

mkarcher wrote on 2025-04-23, 15:14:

That's definitely a very intersting artifact. I suggest you open a new thread for this card, instead of posting pictures and questions about a historical ISA card in a thread about newly designed ISA cards. I don't suggest this because I want to be a rules enforcer, but posting your topic in a different thread with a specific title (like "RasterOps ISA video card with 3*S3 924 - what is it?") will likely get a lot more attention of the forum members that can actually help you with this card.

Nevertheless, I can do a quick assessment of what I see: The card uses a Bt473 RAMDAC. That chip can receive 24-bit true-color data on 24 dedicate input lines. IMHO, it's likely the card is made for True-Color operation with each S3 chip supplying one of the 3 components of R, G and B. This is in-line with the component numbering in the tripled sections. Take a look at the S3 chips being labeld UR1, UG1 and UB1, and similar labels for the RAM chips UB2..UB9, UG2..UG9 and UR1..UR9. I got the Bt473 datasheet, and I can clearly identify the traces that provide the red pixel data the RAMDAC, those are the traces below CB158 through RPR3 to the upper right corner of UR1. The corresponding green pixel values are sent from UG1 via RPG3 into some vias below the chip, but I cant see any traces on the front or back side, so this card obviously has signals on inner layers.

It's surprising the card manufacturer took this route, though. The S3 924 is perfectly capable of handling and accelerating TrueColor within a single graphics chips, the number "24" at the end most likely alludes to this fact. While the S3 86c924 only has an 8-bit output port, which is not well suited to output true-color at high resolution, the suggested way of operating the 911 series of S3 chips (best-known members: 911, 924, 928, 964, 966, 968) is to only use that 8-bit output port in VGA compatible mode, and use a direct connection from the VRAM (a special kind of DRAM that includes a row-sized shift register which is perfectly suited to output one pixel after the other) to the RAMDAC, bypassing the S3 chip, for advanced high-color modes (or even very high-res 256-color mode, in which multiple pixels can be passed in parallel to the RAMDAC). The "standard" RAM width of the 924 is 32 bits, and that's exactly what you see here: Each S3 chip has 8 VRAM chips that are four bits wide, resulting in 32 bits. Wiring these chips directly to the RAM-DAC can send 4 256-color pixels, 2 16-bit pixels or one true-color pixel to the RAMDAC per clock cycle. For some reason (maybe increased memory bandwidth?), your card runs three separate S3 graphics engines in 8-bit modes and combines their output into one true-color output.

Thank you for your detailed analysis and suggestions! As a new member of the vogons forum, I don't have the permission to start new threads yet.
The images I provided are public, so anyone can repost and share them. I find this discussion very interesting.
Regarding your analysis, I find it very professional and well-supported. You mentioned that the manufacturer's choice of design is surprising. Are there other graphics cards that use a similar design?
Additionally, Did the acquisition of TrueVision in 1992 influence RasterOps to not only continue developing NuBus interface graphics cards but also start developing ISA interface graphics cards? Information on this topic is quite scarce, and I haven't been able to verify it.
Thanks again for your help!

Short article on the cards introduction...

The attachment RasterOps PaintBoard PC.jpg is no longer available

Lots of empty pcb there.

What that card does with 3 chips became a consumer standard for normal video cards in 3-4 years.

High color (15/16 bit) was extremely popular during that time so you have to wonder if support for that mode was just dropped.

My guess is that it must have a lot of 2d windows acceleration features and if it didn’t have support via bios for fixed frequency screens I would be surprised.

A very strange, very expensive, very short lived solution for 1992 (PCI was a year away)

Reminds me of a non-proprietary version of a high end cornerstone video card.

Had a quick search for RasterOps own drivers ( Windows 3.1 & AutoCAD according to this graphics card group test https://www.worldradiohistory.com/Archive-Byt … yte-1994-02.pdf )but no luck so far 🙁

PC Hoarder Patrol wrote on 2025-04-24, 05:24:

Attached: benchmark table

While I don't know how representative the "speed" benchmark used in this table is, and whether some drivers were, ahem, optimized for certain benchmarks, you clearly see an advantage of the 3-chip approach to the single-chip approach, with the RasterOps PaintBoard at around 2000s, and the #9GXe at 2700s, i.e. more than 30% slower.

It's interesting to see two signigicantly different cards in https://dosdays.co.uk/topics/Manufacturers/nu … bernine/gxe.php as ISA cards claimed to #9GXe cards. According to that DOS Days report, the #9GXe only hits 800*600 at True Color, while the PaintBoard goes up to 1024x768. I'm surprised by the #9GXe being that as the benchmark quoted by PC Hoarder Patrol indicated "4MB", but possibly it's an architectural limit of their RAM interface. In the version with the ZIP (zig-zag inline package) RAM chips, the RAM chips are likely 256 x 4, so each block of 8 (there are two such blocks) make up for a 32-bit bank of 1MB. With two banks of 1MB and 32bpp (yes, I do know that "True Color" only requires 24bpp, but unless you pull strange stunts like the MiroCrystal 24S did, you need to provide RAM for 32bpp), you are indeed limited to 512KiPixels, topping out at 800x600. There are two extra RAM chips on the #9GXe with ZIP chips, obviously optional, but those are just EDO/FPM without VRAM capabilities. These chips are 256K x 16, so they provide a third bank of memory, yielding a total of 3MB (not 4MB!), but as they are not VRAM chips, the 911 series is not able to directly display the contents of those memory chips on the monitor.

Instead, these chips are likely used for caching font shapes and icons, so they can be painted faster into the VRAM chips than if the fonts/icons would be repeatedly transferred over the ISA bus. Another use of off-screen memory is filling polygons. With the 8514 architecture the S3 chips are based on, you can't just instruct the chip to fill a triangle, a trapezoid or something like that. Instead, the vendor recommended algorithm is to draw the outline of the polygon to fill into off-screen memory using the hardware accelerated line-drawing feature. Having done that, the chip can in a second pass read the rasterized outline back line-by-line. When reading the outline, the chip is able to toggle between "draw pixel" and "skip pixel" everytime it crosses the outline, so the outline then can be used as mask for a constant-color fill or drawing an image from another part of (possibly off-screen) memory or the ISA bus. If you operate the #9GXe at 1152x864 with 16bpp, the VRAM is fully used with screen pixels, so there is no significant off-screen memory left, which severely limits which acceleration features are available. So adding another megabyte of off-screen memory actually makes a lot of sense.

Looking at the other kind of ISA #9GXE with 16 SMD RAM chips, those chips most likely are 256K x 8 VRAM chips, which is actually 4MB of VRAM, so unless the card uses a creative design (which was quite common back in the daym, though), it should be able to handle 1024x768 TrueColor.

hcyandluodayou wrote on 2025-04-23, 13:25:

Seeking Help: Information on RasterOps Graphics Card with S3 P86C924 Chip Hi everyone, […]
Show full quote

Seeking Help: Information on RasterOps Graphics Card with S3 P86C924 Chip
Hi everyone,

I recently acquired a vintage graphics card with an S3 P86C924 chip, and the boot screen displays the RasterOps logo. I'm looking for more information about the origin and specific model of this card. Does anyone have any details or knowledge about it?
This is a strange graphics card for more than one reason. RasterOps video cards are mostly used on Macintosh and adopt the NuBus interface. As far as I know, RasterOps did not launch ISA interface graphics cards nor use S3 display chips. Having three display chips but only one VGA output interface is not a traditional multi-screen display card. How do the three display chips work? Can they bring effects similar to Voodoo SLI to enhance color depth or increase resolution? There are resistors soldered next to the chips and manually soldered jumpers. Is this an engineering sample graphics card?

I love the clean layout of this card. You can really tell that whatever engineer designed it put a lot of love into it. The little flying resistors that are floating above the board are also adorable, whether a bodge or intentionally added. There is just so much attention to detail here.

I did a similar floating bodge recently with electronic capacitors, so I can especially appreciate the time it takes to get it right, both technically and artistically.

Kahenraz wrote on 2025-04-24, 17:08:

I love the clean layout of this card. You can really tell that whatever engineer designed it put a lot of love into it. The little flying resistors that are floating above the board are also adorable, whether a bodge or intentionally added. There is just so much attention to detail here.

I did a similar floating bodge recently with electronic capacitors, so I can especially appreciate the time it takes to get it right, both technically and artistically.

The attachment 20250401_222720.jpg is no longer available

is that a SNES mini ?
I am waiting for that voultar edge enhancer to go back in stock, thinking to install it in my JP snes - though no idea if he actually ships in EU

Kahenraz wrote on 2025-04-24, 17:08:

I love the clean layout of this card. You can really tell that whatever engineer designed it put a lot of love into it. The little flying resistors that are floating above the board are also adorable, whether a bodge or intentionally added. There is just so much attention to detail here.

On the other hand, you can also see that the core part of the card is just (lazily?) copy/pasted three times. Perhaps that makes it even more aestethic, though. The flying resistors are a 330/220 ohm termination, just like a passive SCSI terminator would do. As I don't have a '924 datasheet, I can't look up what the pins are, but I have a very strong suspicion what's going on here: The memory clock and the pixel clock are generated by the ICS2494 chip in the lower right corner (between the ISA edge contact and the slot bracket), and then routed to the three S3 chips. To prevent reflections of the clock signal causing false edges at the intermediate S3 chips, the clock lines are terminated at the third S3 chip.

I'd guess the designers originally assumed that there is no need to terminate the clock line at the far end - I can't make out whether there is series termination of the clock lines near the clock generator. You'll notice that the other bodge on the card is also quite involving clock stuff. I can't figure out what's going on in detail with that bodge, because the sharpness of that (quite excellent!) photo in that corner isn't sufficient, so I can't even read what kind of chip U12 is.

Well, I got an idea what U12 is: It's used as clock buffer: It's some kind of quad gate (likely non-inverting like quad AND (74F08) or quad OR (74F32). The two input pins of the first gate are connected to one bodge wire from one end of R12, and the output of the same gate is connected to the other end of R12. The original design intention was likely to use R12 as series termination resistor, but in the end the clock signal turned out to not be strong enough to drive all the chips it is supposed to drive, so the series termination resistor is now "overridden" by a clock buffer ("buffer" is an electronics term for an amplifier that provides more current for a signal without increasing the voltage level, i.e. turning a weak source into a strong source). Especially as this clock line (only one of the two lines, obviously) is no longer series terminated at the point of feeding it, the termination at the far end is likely essential. I.e. the two bodges are related and all about clock signal integrity.

Clock integrity is a huge thing compared to data line integrity, because for data lines you have specific points in time in which they need to be valid, but they may have strange levels in-between, like ringing or slow rise times. On the other hand, the clock signal is used as reference to when the data signals need to be valid (typical specifications are something like "the data needs to be at the receiver 5ns before the rising edge of the clock, and it needs to be kept valid for at least another 5ns"), and unclear signals on the clock line make the whole idea of a synchronous system break down. Especially bad are ringing and reflections that may cause wave forms that look like extra clock transitions that cause effects in chips (like sampling the data lines) at points in time when this is not expected to happen.

keropi wrote on 2025-04-24, 17:21:

Kahenraz wrote on 2025-04-24, 17:08:

I love the clean layout of this card. You can really tell that whatever engineer designed it put a lot of love into it. The little flying resistors that are floating above the board are also adorable, whether a bodge or intentionally added. There is just so much attention to detail here.

I did a similar floating bodge recently with electronic capacitors, so I can especially appreciate the time it takes to get it right, both technically and artistically.

The attachment 20250401_222720.jpg is no longer available

is that a SNES mini ?
I am waiting for that voultar edge enhancer to go back in stock, thinking to install it in my JP snes - though no idea if he actually ships in EU

It's a Nintendo 64. I thought that adding these capacitors would improve the RGB signal, but I later realized that they need to be on the cable end, after the output out of the RGB mod, not before.

Kahenraz wrote on 2025-04-24, 18:17:

It's a Nintendo 64. I thought that adding these capacitors would improve the RGB signal, but I later realized that they need to be on the cable end, after the output out of the RGB mod, not before.

On the SNES, this did depend on the SNES revision and/or the region type (PAL vs NTSC).
Some SNES units required the cable to have capacitors, some units had them on mainboard PCB already.

This confused the heck out of me when looking for an RGB cable for my sister's SNES.
Because she had an LCD TV with RGB scart input and using normal composite looked so badly on that TV.
The whole 288p vs 576i thing was less of a problem with the RGB input somehow, not sure why.

Edit: That being said, I think that Composite or RF might still be good for those Nintendos.

My Super Nintendo (PAL) looked much better on my old Commodore 1702 via composite rather than S-Video.

I mean okay, the S-Video to Luma/Chroma cable I had made wasn't ideal.
I didn't take the voltage differences betwen Commodore Luma/Chroma (those RCA inputs) and regular S-Video (Hosiden input) into account.

Then there's the situation with RF.. Modern TVs have an RF input meant for cable TV.
They do expect a rock solid signal that's on frequency.

By contrast, old tube TVs do have an RF tuner meant for terrestrial reception.
Such tuners were very sensitive and had an AFC, automatic frequency control.

Thing is, that RF modulators in old home computers aren’t the greatest.
So they may not be good enough for modern TVs.

Using an VCR as a receiver used to be agood workaround.
The TV tuners inside VCRs were among the best there is.

The Yaesu FRG-9600 communications receiver of the 80s did use an VCR TV tuner, for example.
The internal computer did talk to the TV tuner essentially, to set frequency, modulation type and so on!

Moved to new topic.

@hcyandluodayou don't ever hijack a thread again! That's really really bad netiquette