I'm looking for "off the shelf" graphic IC's that supported hardware sprites. By "off the shelf" I mean chips that were designed to be used in various machines and not designed specifically for one machine. So the VIC chips from MOS/Commodore wouldn't count. Even though the VIC-I was originally designed to be off the shelf it never materialized that way since it was used mostly in the VIC-20.

Also, I'm only interested in chips that were made any time during the 70's and 80's (1989 being a cutoff).

I've been doing some research on this and I've only been able to find one group of graphic chips that supported hardware sprites.

That was the TMS9918. The TMS9918 had a few variants such as the 9928, 9938 and 9958.

So, other than the TMS IC's from Texas Instruments, were there any others that supported hardware sprites?



I neglected to mention that in my research, I discovered a few "Pong" chips like the AY-3-8500 series. I'm not sure if they count because while I assume they had sprite support, they were not general purpose graphics chips that I could tell. They were designed to host a few "Pong/shooting"-like games and that's it.


I want to make sure I'm clear. I'm looking for single chip solutions that required very minimal external components. RAM and clock would be OK but not much else (other than passives, of course). I don't believe EGA would count (regardless if it had hardware sprites) because it required external components for memory decoding, DAC's, etc. The TMS9918 is a good example of what I'm looking for.


It looks like there was a "single" chip solution to EGA. However, I'm really pushing the boundaries of the term "single chip" when we add RAM/ROM and a clock. But, "single chip" isn't the primary focus of my question. I'm more interested in the off-the-shelf graphics chips that included hardware sprites and I'd prefer close to single chip if possible. I still, however, cannot confirm that the EGA cards (chips) supported any hardware assisted sprites. But if it did, I would certainly add it to my research.

  • 2
    I'm not looking for a recommendation or "what should I buy". I'm looking to document graphics chips from that time period that supported hardware sprites. In my research, I only found the TMS versions which surprised me. Hence, the question. I've tried to be very specific to avoid asking a list question and I clearly did not ask for recommendations. For a "single chip solution" I'm willing to let RAM/clock slide because onboard RAM in the 70's would have been too expensive. But needing decode logic, DACs, etc. would be too much.
    – cbmeeks
    Commented Sep 5, 2017 at 15:34
  • 3
    @Raffzahn go back and re-read what I wrote. I've edited my question three times. The ANTIC/GTIA wouldn't count because they were made specifically for one brand of computers. They were not "off the shelf" components that you could buy and put into a competing computer. Which I clearly mentioned in my original question. Also, the Tandy's chips wouldn't apply either because they didn't support hardware sprites. I think you need to re-read my question 2-3 times and maybe you will get it.
    – cbmeeks
    Commented Sep 5, 2017 at 18:19
  • 1
    @Raffzahn and, BTW, no...you do NOT have to have decode logic for a video chip to render video. The TMS will render video just fine all by itself. Now, if you want to be practical, then you need to have decode logic between the MPU and the video chip. But that's not the point of my question. You ALSO need electricity. But I didn't specify that either. You also need matter, plastic and metal. But I assumed the reader would know those are implied.
    – cbmeeks
    Commented Sep 5, 2017 at 18:20
  • 3
    @Raffzahn finally...I didn't ask for the BEST or the CHEAPEST. I just need to know what was available. My original question was very clear...at least I thought.
    – cbmeeks
    Commented Sep 5, 2017 at 18:22
  • 1
    @JeremyP the VIC-1 doesn't count because it doesn't support hardware sprites. The reason I am asking this question is that I'm doing some research on 80's (and some 70's) technology and what was available. Hardware sprites are very important to my research.
    – cbmeeks
    Commented Sep 6, 2017 at 12:33

7 Answers 7


The General Instruments AY-3-8900-1 supports eight 8x8 sprites with 1 color each. It was used in the Mattel Intellivision.

The successors of the TMS9918 you mentioned are the Yamaha V9938 (32 32x32 sprites with 16 colors), the V9958 (same as V9938), and the V9990 (125 16x16 sprites with 16 colors). The V9938 was used in various manufacturer's MSX2 systems. The V9958 was used in various manufacturer's MSX2+ systems. The V9990 was used in cartridges for the MSX systems.

  • Interesting find...but was the AY-3-8900-1 sold for more than just the Intellivision? Or was it made exclusively for the Intellivision?
    – cbmeeks
    Commented Sep 5, 2017 at 14:15
  • The GI chips where complete consols in a single chip. there's no way to use them for anything else, especially not as graphics controller.
    – Raffzahn
    Commented Sep 5, 2017 at 14:43
  • 1
    While no one else used the AY-3-8900-1 video chip, General Instruments also produced a CPU and sound chip, both of which were also used in the Mattel Intellivision, but the CPU was also used by Honeywell and the sound chips were used in a lot of computers and game consoles such as the Mockingboard for the Apple II, the Color Genie, the MSX systems, the Oric 1, the Vectrex, the Amstrad CPC systems, ZX Spectrum 128, Timex/Sinclair 2048, and the Atari ST. I think this strongly suggests their video chip was also available for other companies to use even if no others used it.
    – Tim Locke
    Commented Sep 5, 2017 at 14:50
  • @TimLocke you're right, I was thinking of all the pong chipsets. But wouldn't be the 8900 be a bit below specs? AFAIR it only supported 160x96 positions for the 8 sprites moving over a 20x12 blocky background. It also did need an external colour generator chip (8915?), basicly an external CLUT outputing NTSC encoding. THinking in that timeframe, RCAs 1861/62 didn't have sprites, did it?
    – Raffzahn
    Commented Sep 5, 2017 at 15:03
  • It came out around 1978/79 so it is commensurate with other offerings of that time. Atari's ANTIC/GTIA was at the end of 1979. The VIC didn't have sprites and had similar resolution 176x184. The video of the 1977 Trinity was higher resolution but less color, the PET and TRS-80 being monochrome and the Apple II having 16 colors at 40x48 or 6 at 280x192, although the way the pixels had to be arranged to produce the colors, the horizontal resolution was actually more like 140x192. The GI video chip could do 16 colors at full res, plus sprites!
    – Tim Locke
    Commented Sep 5, 2017 at 17:15

The TI TMS 34010 and 34020 (released in 1986 and 1988 respectively) were pretty much the first fully programmable graphics processing units. They were 32-bit processors with a fair amount of dedicated graphics capability, including PixBlt operations, which were basically sprites--i.e., instructions that worked with/on a 2D array of pixels (though they also supported sizes quite a bit larger than you'd normally think of as sprites).

Intel apparently decided the dedicated GPU business looked interesting as well. In 1988 they released the 82786. Somewhat like the TI processors, it doesn't really treat sprites as a separate sort of thing--but it supports sprite-like operations on bitmaps and textures of (almost) arbitrary size, including the relatively small ones you'd normally think of as a sprite (and if you wanted a 32767x32767 pixel sprite, it'd theoretically do that too).

Both were intended to be pretty much single-chip solutions. You'd need clock, power-supply, and external RAM, but both included things like the CRT controller, so you didn't need to add much there. They also included DRAM/VRAM controllers, so you didn't need much glue logic to talk to the memory either.

Most typical graphics cards using these probably did include a fair number of other chips though. On their own, neither provided anything similar to a EGA or VGA that the motherboard's BIOS would know how to talk to, so many (most?) cards using these also included some sort of EGA/VGA chip set--which (despite providing much more limited abilities) often added a fair number of other chips. Some just supported VGA pass-through through, so you needed a VGA card along with the high-end graphics card.


If this is about general available chips, not custom ones, Texas did prety much set the standards and own the market when it came to hardware sprites. Motorola beeing the only other major player/initiaor did focus on chips enabling bitmap displays with variable designs, but not higher functionality. And as TI licensees did improve on the 9918 design, Thomson and other Motorola fellows did stay within their linage.

Another, maybe lesser known chip was the Signetics 2637 Universal Video Interface. The best description would be as a single chip video console solution, as it offered a

  • video circuitry,
  • colour abilitiy (8colours),
  • four sprites,
  • collision detection,
  • 48x52 block graphics
  • 13x16 or 26x16 text,
  • character ROM (40),
  • 64 graphic characters,
  • 16 backgroudn characters (frames etc.),
  • 8 definable characters,
  • a 4 channel ADC (4 paddles or 2 joysticks)

Everything else with sprites, from Atari VCS (1977) and 8 Bit Atari (1979) over Commodore 64 (1982) and Famicom/NES (1983/85) to the Amiga (1985), did use proprietary, custom chipsets. Others didn't have hardware sprites at all. Well unless you considere lodable character sets as such.

Oh, and looking close, the PC got sprites also beginning in 1985 with some third party EGA and later SVGA cards - after all, the hardware mouse cursor is a single 16x16 2bpp sprite. Isn't it? :))

  • 1
    I wouldn't count the EGA/SVGA cards because they are not packaged in a single chip...that I know of.
    – cbmeeks
    Commented Sep 5, 2017 at 14:09
  • 1
    I don't believe EGA would count. I've updated my question to be more clear. Also, I wouldn't TECHNICALLY consider a TMS9918 to be a single chip solution because it required external RAM. But, I would say RAM and a clock would be about it.
    – cbmeeks
    Commented Sep 5, 2017 at 14:31
  • 1
    There where several singel chip EGA implementations that do fit your requirement of no external (logic) components beside RAMs. But serious, I realy have a hard time why you circle all arround the EGA remark, or do you realy considere a chip capable of handling a SINGLE 16x16 sprite?
    – Raffzahn
    Commented Sep 5, 2017 at 14:42
  • 3
    You mean the PC got a sprite, not sprites.
    – TonyM
    Commented Sep 5, 2017 at 20:32
  • 1
    @Jules Of course it's not an VGA, it's an EGA - it wuld be hard to premier a VGA chip in 1985, wouldn't it?. The board shown is a combination of an EGA (ET1000) and a Hercules compatible MGA (6845 based) with its paralell port (TTL components, lower connector) and glue logic to combine them, including dualporting the memory (unlike CGA, EGA was dual port). None of the additional components would be needed for an ET1000 only design. The IBM PC does need the EPROM as he board EPROM doesn't know how to handle an EGA.
    – Raffzahn
    Commented Sep 6, 2017 at 7:17

The 6845 supports one hardware "sprite", which is typically used for a text cursor but could be used for other purposes if desired. Horizontal positioning accuracy is limited to one character cell, and there is no support for the sprite to straddle two cells vertically, but it would not be overly difficult for a system to offer a crude mouse pointer which was a "dot" that was half a character horizontally and vertically, and which could be moved on half-character boundaries.

  • Interesting... although most systems that used a 6845 (including the EGA systems that @Raffzahn describes in his answer) used that capability for the text mode input cursor, so would probably not be usable for a mouse cursor at the same time.
    – Jules
    Commented Sep 6, 2017 at 0:46
  • 1
    I considered making the same 6845 observation, but was worried the original author both (i) won't accept a 6845 anyway because it can't generate pixels on its own and he wants a single chip; and (ii) won't count a solid rectangle as a sprite. But otherwise, yeah, it's a thing that is overlaid by many 6845 systems onto the video stream based on an independent comparison of position. Trivia! The Camputers Lynx did something even more interesting with the cursor active output: it uses it as its interrupt trigger. So you can place your interrupt anywhere on the visible display.
    – Tommy
    Commented Sep 6, 2017 at 1:11
  • The 6845 is certainly interesting. I was familiar (somewhat) of this chip during my research but I honestly didn't consider the mouse cursor sprite as a "true sprite". Yet, there it is. I believe what we call a "sprite" they called a "hardware text cursor". But it's close enough in scope that I will dig deeper into it. Although, the sprite abilities it has won't be adequate for what I'm looking for but that wasn't the focus of my question. +1
    – cbmeeks
    Commented Sep 6, 2017 at 12:47
  • Now, I wouldn't considere the 6845s cursor a sprite, as it can't be positioned on pixel coordinates. It's just a text cursor as found in many other CRT controllers. The (third party) EGA mouse cursor was different, as it worked on pixel coordinats and had a pixel definable shape, qualifying as a sprite.
    – Raffzahn
    Commented Sep 16, 2017 at 9:11
  • @Raffzahn: The 6845 doesn't do anything with a horizontal granularity finer than one character. It does, however, allow single-pixel positioning of the cursor in the vertical direction, and the positioning and triggering of the cursor is independent of the video memory system.
    – supercat
    Commented Sep 16, 2017 at 11:47

The Signetics 2636 Video Controller may qualify -- it supports up to 4 8x10 sprites (resizable to 64x80!) and was used in the Interton VC 4000 console released in 1978: http://www.old-computers.com/museum/computer.asp?st=2&c=726

If so, it may slightly predate the TMS9918 which came out in 1979, but I can't find exact dates on either chip.

  • The TMS9918 was of course used in the TI-99/4, which was released in October 1979, so that at least puts a limit on it in one direction...
    – Jules
    Commented Feb 2, 2018 at 23:20

The graphics chip developed by Nutting for Bally was used in many of their arcade games as well as the Astrocade console (ostensibly from Bally, but not really) and the Datamax UV-1.

It did not have hardware sprites in the typical sense, but instead used a clever bit-blitting system so that writes to memory over the 32k boundary were mirrored into memory below 32k while applying a selected bitwise function, normally XOR. In this way you could have any number of 2D objects that could be moved with simple commands.

I'm not sure it exactly meets the definition of either "off the shelf" or "hardware sprites", but I think its close enough.


Hmm... quite the puzzler! Other than what others have already written, I can't think of any - the TMS series were about all I was aware of.

There were some more-or-less single-chip, and very definitely intended as "off the shelf" PC options intended for use by graphics card OEMs that pretty much anyone building a system - PC or otherwise - could have bought, but I dunno if they'd count for your purposes. The EGA-clones had a variety of interesting accelerated writing and graphics manipulation functions and modes, including multiple software-definable character sets (usable in just the same way as console tilemaps) and smooth hardware scrolling even of text displays, but nothing that would so much count as a "sprite" engine. The VGA-clones added at least a single hardware sprite - for the mouse cursor - but it wasn't exactly very sophisticated or useful (limited to 32x32 pixels, 2-bit colour, and a fixed location and pattern that could only be updated at Vsync, IIRC) compared to pretty much any other sprite engine all the way back to the VCS. In a lot of cases it wasn't even used for the mouse, and programmers just blitted the cursor onto the display instead, making use of the chip's manifold other accelerated drawing functions to speed up everything else. You start getting Super-VGAs and XGA-clones just before the turn of the 90s, which may include a few such functions, but they're already moving more into the realm of Window Accelerators, and accelerating general memory transfers and blits as much as possible, with an emphasis on CPU-driven and GPU-assisted general drawing functions, with simple copying of a small, limited colour chunk of bitmap on top of it all being considered a fairly trivial thing for a target system to pull off without any particular assistance at all.

Apart from the TMS derivatives used in a huge array of consoles and console-ish computers, and the PC series (and exotic, expensive stuff like the 34010), most manufacturers of the era were pretty insular and relied on home-rolled designs for their graphics hardware, as far as I knew. The fact that MOS didn't bother marketing the VIC outside of their own machinery maybe indicates that they didn't even think there'd be a market for it (...that or they figured it and the SID were so far in advance of everything else, they didn't want to lose the advantage), and/or that they figured said market was already pretty well sewn up by Texas Instruments and whoever it was made the Teletext/Prestel character generator...

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .