On 6/18/07, Cameron Kaiser <spectre at floodgap.com> wrote: Sure. I think the 6502 was heavily favored by Atari (for obvious reasons) and not many other vendors. The 6809 (Defender, et al.) was, I think, more popular; but I agree, the Z-80 was exceedingly popular. It was powerful, inexpensive, and there were lots of folks with lots of experience to make code that performed well. There were plenty of 8080-based games in the black-and-white era, as well - those were the vendors that moved up to the Z-80 later. -ethan

Atari actually had a good mixture of CPU's, though there was a sort period (very early 80's) where they moved to 6502 reliance (some of it coming out of their PCS development in the late 70's, which was initially intended to be simultaniously designed for arcade use but never panned out because of different display issue needs. The only thing that made the leap out of that was the 6502 and their custom POKEY chip. The closest hardware to the PCS's was Missile Command and Centipede). For a sampling: Quiz Show (1976) used an S2650 CPU. Tank 8 (1976 and the first color game), Destroyer (1977), Drag Race (1977), Pool Shark (1977), Sprint 8 (1977), Super Bug (1977), Triple Hunt (1977), Sky Diver (1978), Fire Truck (1978), Orbit (1978), Smokey Joe (1978), and Monte Carlo (1980) all used the 6800. Arabian (1983) and Kangaroo (1982) used Z80's. I, Robot (1983 - the first 3D/Polygonal game) used a 6809 with supporting 2901 bit slice cpu's for graphics support. The color vector Star Wars also used the 6809. System 1 games (starting in 1984 with Marble Madness) ushered in the 68000 series era, though the System 2 (also in 1984 with 720 and Paperboy) used T11's. Marty

On 6/18/07, Jim Leonard <trixter at oldskool.org> wrote: As mentioned further up in this thread, the Atari vector games had, in effect, a vector math co-processor, the Analogue Vector Generator (AVG) or "math box" built from 2901s to do the heavy lifting. The 8-bit processor fed shapes to the AVG and sounds to the Pokey and handled the yoke/joystick I/O. -ethan

On 6/18/07, Ensor <classiccmp at memory-alpha.org.uk> wrote: Yes, four. -ethan

Jim Leonard wrote: It's made of bit-slice logic chips. I don't think you can necessarily call them microprocessors. I believe the most popular chips were of the AMD Am2900 family. I've used them myself. They're righteous. http://en.wikipedia.org/wiki/Bit-slice Peace... Sridhar

On Jun 18, 2007, at 11:39 PM, Jim Leonard wrote: "Bit slice" chips are building blocks that can be used to construct a processor. They typically contain things like an ALU and stuff to help with the decoding of microcode. The neat thing about them is that they can be cascaded to make a processor of any data bus width that you want. The ubiquitous AMD Am2901, for example, is four bits wide. Many 3rd-party disk controllers for DEC machines used two or four of them to implement an 8- or 16-bit processor to handle disk control operations. They've also been used for much more ambitious things: The VAX-11/730 and DECsystem-2020, for example, are both built around Am2901 bit-slice chips, running microcode from fast RAM which implements the VAX and PDP-10 instruction sets, respectively. That microcode is loaded at boot time, from TU58 in the case of the VAX-11/730, or from a TU45 9-track magtape on the DECsystem-2020. Before that microcode is loaded, the machines aren't really VAXen or PDP-10s...as they have no idea how to implement the architecture or execute the instructions, until the microcode is loaded. With a good bit-slice chip like an Am2901, one can build a processor to implement any architecture one wants. Many, many machines were built in this way...I only use the VAX-11/730 and DECsystem-2020 as examples because I have them sitting here. Further, the Am2901 isn't the only bit-slice CPU around...there were many others. Intel made at least one, I think it was the 3002. -Dave -- Dave McGuire Port Charlotte, FL

On 6/19/07, Robert Borsuk <rborsuk at colourfull.com> wrote: Yes (see the thread where it appears in a few video games). DEC used it as a PDP-11-instruction-set-compatible microcontroller. It appears on a few peripheral boards (the RQDX3 comes to mind), and a few other places. For some real boots-on-the-ground history, we turn to Bob Supnik... http://simh.trailing-edge.com/semi/t11.html (I did not know about his T-11-based RT-11 box - I should ask him about it). In what? In video games? the 6100 _might_ have been a competetive architecture when it was new in the mid-to-late 1970s, but as competition against an early 8-bit micro, not a 16-bit micro, which the T-11 is. From looking at the Atari line at the time, it seems that they were positioning the T-11 against the Motorola 68000. Presumably there was some engineering or marketing or production reason to go with the T-11 over the 68000, but, as much as I like the PDP-11, I can't imagine what that would be. Perhaps the $10/unit cost that Bob Supnik cites was favorable compared to, say, trying to go over 8MHz on a 68000, but that's mere speculation. I know that at the low-point in the 68000 timeline, it was going for about $3 each in reasonable quantities, but I don't know where that curve compares to the $10 each for the T-11. I'm not saying you _couldn't_ make a video game based on the IM6100, but it didn't happen to have been done, and would probably just end up as a demonstration of engineering prowess, not something that would have made sense from a business standpoint in 1976. -ethan

On Tue, 2007-06-19 at 10:01 -0400, Robert Borsuk wrote: Another thought that occurs is, surely the FPU-less T11 would be no better (and in some cases not as good as) a 68000? I can see where the F11 or J11, although expensive, might be able to handle maths-intensive tasks rather better. Gordon

On 6/20/07, e.stiebler <emu at e-bbes.com> wrote: Fixed point enhanced with fast (but sometimes large) lookup tables instead of MUL and DIV. Yep. Even without 32-bit registers, there are tricks one may employ. Twenty years ago, I had to implement a high-frequency filter in 286 assembler as part of an auto-focus routine for a vision system. It had to be fast, and the company would not pay for a 287 co-pro. The normal technique for a simple 3x3 kernel is to sum up all the values in a 3x3 area and divide by 9 to get the average value, then iterate over the entire frame. Dividing by 9 is expensive for a 286. Dividing by 8 isn't, and produced acceptable results. There are frequently tricks to avoid expensive math operations. Video games use most of them. -ethan

On 6/20/07, Ensor <classiccmp at memory-alpha.org.uk> wrote: I saw the preliminary data sheet around 1979, after I'd been banging around the 6502 for a bit, and totally fell in love with the 68K. I finally got my hands on one in 1984 - much fun ensued. -ethan

On 6/19/07, Dave McGuire <mcguire at neurotica.com> wrote: Every once in a while, a fit overcomes me and I contemplate what it would take to re-implement the KS20 with a stack of 2901s and a pile of fast SRAM. Fortunately, the madness passes before I can do much more than think about it. It'd be cool, but I'd probably be better off trying to follow in the footsteps of one of the 10-in-an-FPGA projects. For the forseeable future, though, I'll probably stick to klh10 (and should get around to ordering the last parts to finish my "Panda Display" blinkenlights). It would still be a thrill to boot TOPS-20 on real hardware that first time, though. -ethan

Dave McGuire wrote: So how much $$ do you expect parts for such a beast would be? So build a RISC flavor of a 9/18/36/72 bit machine. :) Perhaps that would be a better project to emulate. Spare Time Gizmos makes TU-58 Emulator. You might ask them to emulate a tape drive too.

Dave McGuire wrote: Providing the word length is a multiable of 4 bits that is. Did going bit slice make the machines more profitable for DEC rather than lower the price for the end user? That was a 2 bit bit slice. TI had a weird 4 bit slice. That is all the ones I can think of.

woodelf wrote: ..flipping through IC Master '81: - The 74S481 4-bit slice. - Fairchild CMOS 470x series (4705 4-bit ALU/register slice). - Motorola ECL MC10800 series (4-bit ALU slice) - Siemens ECL SH100B480 4-bit slice - Mustn't forget the 74181 & 74281 4 bit ALUs in the standard TTL series, although I suppose some may argue because they don't contain the register portions. Before bit-slice chips, the HP 2116 was constructed in 'bit-slice form' from SSI: 4 boards plugged into the backplane are identical, each board containing 4 bits of the 16-bit ALU and main registers. I expect other SSI-era minis were done this way. Were any of the early PDP-11's constructed this way, or did the first PDP-11 use the 74181? A weird chip I ran across recently in an early 80's protocol analyser is the AMD 29116. The overview in IC Master calls it a "16-bit Bipolar Microprocessor" but it only contains the ALU, registers, and instruction register/decoding. It appears the program counter/sequencing have to be provided externally, and it doesn't appear to be cascadeable so you can't really call it a bit-slice.

Philips made a 4-bit wide slice ASIC called 'SPALU' (Scratch Pad + Arithmetic Logic Unit -- 'scratch pad' meaning general purpose registers). AFAIK it could only be used to make a 16 bit processor (I think all 16 bits went ito each slice, but each slice only outputted 4 result bits). and was used only (?) in the P851 machine. It may not really count, but it's a curiousity. AFAIk you can't casecade them (things like the carry signals are not available externally in the right form, etc). The PERQ 3a (AGW3300) has a pair of them on the CPU board for the graphics processor, according to the schemaitcs one does the actual data updating, the other calculates the addresse to be updatate. There an external control store (I forget how big) with a simple counter as a sequencer (it can be loaded for jumps, but there's no subroutine stack or anything like that). Obviously you could use a more advanced sequencer, like a 2910 or something -tony

Hi Bit-slice refers to the fact that the ALU is designed to be stacked for whatever word size you'd like. Most come with chips that are 1, 2,4 or 16 bits wide. Say you wanted a 20 bit computer, 5 ea of a 4 bit wide would do. Also, these are truly microcoded. These usually have there own microcode ROMs and can have any number of bits wide to run them. Often when used for hardware control ( such as disk drive controllers ) many of the bits in the micro code directly control I/O functions so that the processor part could be doing something with data at the same time the I/O was doing something else. The ROM was usually controlled by a chip called an address sequencer. This part usually handled things like conditional jumps, subroutine calls and just sequential addressing. The AMD 2900 series was relative straight forward and seemed like an address sequencer used in any number of todays uP's. The 3000 series sequencer by Intel was a strange beast. It had all kinds of row and column jumps. Every instruction had bits that defned where the next instruction was. It was quite complicated to program and most likely why it was not that popular. I know that there is someplace a design that used 2900 parts to emulate a 8080 processor. Remember, the micro code was a separate structure Bit-slice machines were the way to go for speed until more recent times. Now, the time lost in driving external I/O lines between chips means that uP's can actually run much faster. Still, for hardware control, knowing when an instruction executes relative to external hardware can be important. This means that there are still places that bit-slice methods are still needed. Today it is often handled with a FPGA. Although, it looked like one could build most any computer model with these bit-slice chips, most were register oriented and did poorly at creating other computer models, such as stack processors. Dwight _________________________________________________________________ Picture this ? share your photos and you could win big! http://www.GETREALPhotoContest.com?ocid=TXT_TAGHM&loc=us

A bit slice CPU is noe made from bit slice chips. OK, that's no help at all. It's probalby easiest to explain using a concrete example, the AM2900 series (which are, AFAIK the most populer bit-slice chips). The AM2901 is a 4 bit ALU + regitster 'slice'. It contains 16 4 bit registers, a 4 bit ALU, and the clrcuitry to link them together. It's the data path for a 4 bit CPU. What makes it 'different' is that you can link several togther. 2 will give you an 8 bit data path. 4 would give you a 16 bit data path. 8 would give you a 32 bit data path, and so on. For example, the VAX11/730 has 8 2901s on one of the CPU cards for this reason. Of course the data path is not the complete processor, you need the control logic as well. There were microcode sequencer chips like the 2909 and 2911, which producde 4 bits of microcode address, and which could easily be cascaerd to make as wide a microcode address as you wanted, but which needed some external logic, and the 2910 which was a 12 bit-wide adress seqencer, incldued the 'extra' logic, but which couldn't easily be exteded (cue for the PERQ 2-bit-kludge story, I guess :-)) There;'s a useful book 'Bit Slice Microprocessor Design' by Mick and Brick (that is not a joke!), whcih I am told is one the web soemwhere. And of course the AM2900 series data sheets. -tony

On Monday 18 June 2007 10:52, Ethan Dicks wrote: I actually probably have more familiarity with the z80 than any other part, and was surprised about 10-12 years ago to be talking to a guy about getting paid for that bit of expertise. The job offer was contingent on some sort of a merger deal and who would or wouldn't be willing to relocate to Houston afterwards so it eventually fell through, but it was rather interesting to see some demand for what I'd thought were obsolete skills at that point in time. -- Member of the toughest, meanest, deadliest, most unrelenting -- and ablest -- form of life in this section of space, ?a critter that can be killed but can't be tamed. ?--Robert A. Heinlein, "The Puppet Masters" - Information is more dangerous than cannon to a society ruled by lies. --James M Dakin

Z80 is just the sound processor. The main processor is a 68000. Not only dreamcast, but almost all videogames derived arcade machines. - Namco has a boardset based on Playstation 1 - Konami has a boardset based on Playstation 2 (and some bemani machines uses the entire PS2 inside! - The Chiriro hardware is xbox based (take a look at outrun 2006 on arcade and xbox) - Some machines uses N64/Gamecube based hardware - Saturn had a boardset called ST-V, but it was cartridge based www.system16.com has tons of info on all kinds of boards. You'll loose 2 or 3 hours of your life, but very pleasant ones :o)

Zane H. Healy wrote: And this is why some games are *still* being released (in Japan) for Dreamcast, because they keep making NAOMI arcade machines (shooters, mostly). The release schedule is once every 18 months or so, but that's not bad for a system that was declared dead in 2003. -- Jim Leonard (trixter at oldskool.org) http://www.oldskool.org/ Help our electronic games project: http://www.mobygames.com/ Or check out some trippy MindCandy at http://www.mindcandydvd.com/ A child borne of the home computer wars: http://trixter.wordpress.com/

Jim Leonard <trixter at oldskool.org> says: There's still smaller publishing companies releasing games for it in the US as well.

Crazy Taxi should run exactly like the arcade version (NAOMI) when on the Dreamcast, as both are almost identical. The Dreamcast has 10MB less RAM than the NAOMI board. Alot of the games ported from the arcade to the Dreamcast were because they ran on the NAOMI board. That is why Dreamcast games, in Japan, have continued to appear as they are largely arcade ports. I'm not sure what changes were made in the NAOMI 2 board, and I know nothing about the Lindbergh arcade board. Regards, Andrew B aliensrcooluk at yahoo.co.uk Cameron Kaiser <spectre at floodgap.com> wrote: The Dreamcast remains my favourite modern-era console, and really epitomizes "being ahead of its time." I think it merits a "cool" exception to the 10-year rule. As you say, things like Crazy Taxi run just like the arcade, because the code (with minor changes or home-play modes at most) is just about the same! Still need to put up my DC server farm :) -- ------------------------------------ personal: http://www.cameronkaiser.com/ -- Cameron Kaiser * Floodgap Systems * www.floodgap.com * ckaiser at floodgap.com -- LOAD"STANDARD DISCLAIMER",8,1 ----------------------------------------------

Ahh, yes. I forgot about the extra modes. Crazy Taxi 2 was great with using the Taxi to hit a giant golf ball into the air!!! The giant jump ramp was great too. Bringin this back on topic, and diving into my childhood, what chips were used in Speak & Spell? The bright red/orange childrens speaking/spelling tool from the 80's. Our one was destroyed years ago by my brother who loved taking apart and destroying toys :( Regards, Andrew B aliensrcooluk at yahoo.co.uk

Pete Turnbull wrote: I think they're talking about arcade games, specifically. Anyway, the newer ones are all on POWER-architecture chips. Peace... Sridhar

On Monday 18 June 2007 11:23, Pete Turnbull wrote: The PS3 is a IBM Cell proc (PowerPC), the XBOX360 uses a triad of PPC cores, and the Nintendo Gamecube (possibly Wii also) use a PPC chip. Pat -- Purdue University ITAP/RCAC --- http://www.rcac.purdue.edu/ The Computer Refuge --- http://computer-refuge.org

On Mon, Jun 18, 2007 at 04:23:40PM +0100, Pete Turnbull wrote: The PSP isn't ARM architecture, it has two MIPS CPUs - both R4000s or similar AFAIK. --

On 25/06/2007 16:49, Pete Turnbull wrote: Aha, I've found out where the ARM info came from. According to ARM's website (dated late 2005), the PSP has an ARM 9 core and the Playstation 3 has an ARM9E. http://www.arm.com/markets/home_solutions/armpp/8732.html http://www.arm.com/markets/home_solutions/armpp/15672.html -- Pete Peter Turnbull Network Manager University of York