{110} Intel's 2113 RAM: Never Sell It, And Never Speak Of It Again

Рет қаралды 73,191

10 ай бұрын

In 1976, Intel couldn't get their production process for the 2114 4K SRAM working. Not only were chip shipments delayed, but board and system level products waiting for the 2114 RAM were also facing delays. As a stopgap measure, Intel used half-working 2114 chips to create a new memory device, the 2113, specifically for internal products. The 2113 is just a 2114 with bit failures on one half, but a fully functional on the other half. This chip was apparently never sold outside of intel and is used only in Intel's PROMPT 48 and Intel's iSBC 80/20 single board computer.
To be clear, near the end of the video when testing the memory chips, they are all function properly as 2113s. When I talk about bad bits, I am testing them as 2114s (i.e., using both sides). All the bad bits are all on the unused side of the 2113. So the RAM is fine and isn’t the source of my PROMPT 48 ‘s glitches.

Пікірлер: 116

@doogie812 9 ай бұрын

National Semiconductor had to have had the worst quality 2114 in the 70's. I had worked on 100's of terminals in the 80's and any time the complaint was garbage on the screen it was likely a a defective National 2114 in the video memory. If I found Nationals on the board I would replace them prophetically. At the time I found the NEC 2114's to be the most reliable. Memories...

@RonJohn63 9 ай бұрын

prophetically, or prophylactically?

@PainfulRenegade 8 ай бұрын

@@RonJohn63could it be both?

@RonJohn63 8 ай бұрын

@@PainfulRenegade _prophetic_ means "accurately describing what will happen in the future". _Prophylaxis_ means "protective or preventive treatment". So... no. But they're related.

@ShadowTronBlog 10 ай бұрын

From 1980 thru 1994 I worked as a test engineer at NatSemi focused on 2716-27C256 EPROMs. Your discussion of the 2704/2708 is a bit misleading in my opinion. While a 2704 may pass as a 2708 there are potential issues. My guess is you tested at nominal Vcc and the access speed of the programmer (typically much slower the speced speeds). A cell that doesn't fully program or fully erase per the spec can be an issue for the full expected life time data retention spec. This can often be revealed by Vcc margins testing (Vcc lower/higher then the speced Vcc range). It can show up as a full failure of the cell or as an issue with TACC or TCE timing. Memory testing is also guard-banded meaning it will work in a wider Vcc window then the speced range and will be tested to be faster the speced access times. Long term burn-in testing can be used to determine the window of acceptable voltages/timing (shmoo plots). These then set the required test specs to guarantee the part will hold its programmed/erased cell state within the life expectancy of the part. In my time at NatSemi I loaded hundreds (thousands?) of parts onto burn-in boards, baked them while powered up / being accessed and pulled them at various times for testing, then back in the boards for more burn-in, then out for testing. Rinse and repeat many times. We tested fully programmed / fully erased / checkerboard (55/AA) and topological checkboard (adjacent cell on the physical die have cells in the other state around them). I also wrote the test and analysis software for the parts. Point being that a 2704 with a 2708 die indication programmed as a 2708 doesn't guarantee the part will retain or present the programed values over the full speced life time.

@HelloWorldETX 10 ай бұрын

That is an extremely good point, I wasn't thinking about Vcc margins testing or other conditional issues that would reduce long term reliability, i was stuck on seeing simple go/no-go defects. It should have come to mind after testing that one 2113 that didn't show up until after the cell had been cycled a number of times before it finally stuck high. Since they were equivalent pin-out, I guess it wouldn't make sense to take a perfectly good '08 and mark it as an '04. I fully expect that Intel's "1976 4K production problems" were, in fact, not completely solved by the spring of '77 and maybe ran through mid-year. I am hoping that others with 2113s can give us a range of dates they were produced. thanks

@ShadowTronBlog 10 ай бұрын

@@HelloWorldETX I worked through a few yield crashes during my time at NatSemi. We also brought a new CMOS fab online and it took a while to the get the process dialed it. Those were fun days and feel like a couple of lifetimes ago.

@johnjakson444 10 ай бұрын

When I was a student I spent many happy months at TI in Bedford UK working on Teletext applications and the 9900 boards. I got to design a graphics card and populate it with maybe 16 or 32 of TI 2114 SRAM chips, these were quite valuable too. One day I went back into the lab and found my board had been depopulated, I suspect one of the technicians had filched the chips but who knows.

@scottlarson1548 9 ай бұрын

I'm surprised you didn't use dynamic RAM since video display circuits are usually constantly reading the RAM so you don't need a separate refresh circuit.

@johnjakson444 9 ай бұрын

In a commercial application that would have been the proper path, but I was just interning so it was a quick wire wrap hack that worked until the chips went away. There was only one of these systems, forever grateful that TI let me do what ever I wanted.@@scottlarson1548

@deang5622 8 ай бұрын

@@scottlarson1548 Still simpler when using 2114's. No latches needed for either address or data demultiplexing purposes. Just a 2114 and some address decide combinational logic, IIRC, but been 40 years since I used a 2114.

@scottlarson1548 8 ай бұрын

@@deang5622 ??? The 4116 had no multiplexing. The only annoying thing was that it needed -5V and 12V power but static RAM needed more +5V power.

@senilyDeluxe 10 ай бұрын

The 2114 was used in a ton of equipment, home computers, arcade machines, you can guess the pressure when one of the world's biggest memory manufacturer can't reliably make enough of one of the most popular memory ICs. (although TBH I don't know how popular the 2114 was this early on - but between '78 and '83 they were hugely popular)

@HelloWorldETX 10 ай бұрын

Yes, the 2114 became a huge success. To be the last manufacturer to get their 4K into production must have been humiliating.

@ClausB252 10 ай бұрын

Used in Exidy Sorcerer, Sinclair ZX80 and ZX81, to name a few.

@mrnmrn1 10 ай бұрын

And in the C64 as a color RAM! @@ClausB252

@scottlarson1548 9 ай бұрын

"2114" immediately brings back my memories of mowing lawns to save up enough money to buy sixteen of these expensive chips and soldering them onto a board so I could have another *8K* of RAM.

@nobodynoone2500 9 ай бұрын

@@scottlarson1548 Baller

@TomTRobot 9 ай бұрын

Used the Prompt48 along with an MDS system and ICE-48 at my first job in 1980 programming firmware for a prototype voice-controlled robot toy at Milton Bradley. I could tell ahead of time that there were going to be assembler errors from the the rhythm of "thunk-a-thunk' from the head load solenoid sounds from the 8" drives. We were paying about $60 apiece for 8748s. The 2114's were the goto RAM for those protos. One of the simplest 'cookbook' micro designs was to use the 8085 along with a couple of 2114's, XTAL, 2732 and a bit of glue logic. 1979 breadboarded such an 8085 system in my dorm room as a senior lab project using the guts of a broken TI-58 calculator for the keyboard and display and wrote a rudimentary monitor for it. After demo'ing it I had to take it apart and give all the chips back because they were computer lab inventory and didn't have the cash to buy for them for myself. Not unusual considering that designing things you can't afford for yourself is a common experience among engineers.

@HelloWorldETX 9 ай бұрын

That is interesting prompt first hand experience. I may do a pair of videos just for the two prompts. Ha! Yea, the 8” disk solenoids were great about telling you when the old MDS was finishing up.

@mhuffstutter1803 10 ай бұрын

Uh-Oh.... Now I need to dig out my prompt 48 and do a memory test! I haven't used it in quite a while now. Thank You for another education, Craig. i was completely unaware of the 2113/2114 story. Always learn something new in your videos. I started thinking about the prompt 48 again when I happened upon a pretty nice 8048 based clock that Matt Millman came up with. Of course, as He has done in the past, He designed his own MCS-48 programmer.... Man.

@HelloWorldETX 10 ай бұрын

Hi Mark, Matt”s also has the adapter for his programmer to program 8755s also. It will be interesting to hear about your prompt 48, if it has the piggyback board or not. I have seen a photo of a prompt that had four places for RAM but only two populated and two bare (no sockets). I couldn’t see if they were 2114s or 2113. But no piggyback.

@Erhannis 9 ай бұрын

I kinda feel for 'em - I've been on projects where we were like, "Well, this just isn't working. What do we do?"

@HelloWorldETX 9 ай бұрын

Yea, there were probably some sleepless nights.

@mk1cortinatony395 9 ай бұрын

I am fascinated to see chips' circuitry up close and although I dont understand the workings of it all your version on events is amazing! wish I could understand it all.

@GodmanchesterGoblin 10 ай бұрын

A fascinating piece of history, and the deep dive was much appreciated. Thank you. And I'd love to see something on the 1103 its features and its history. It was in use when I entered the computer industry back in 1976.

@HelloWorldETX 10 ай бұрын

Thanks. Yes, the 1103 was a pivotal device. I agree, it would be interesting to dig deep into the 1103.

@MartinPiper6502 10 ай бұрын

The 2114 was often used in old arcade machines around the 1980s. Mostly as register RAM or temporary work RAM for video display built with 74LS TTL.

@bskull3232 9 ай бұрын

24576 is total transistor count for memory array (4*1024*6), there must be additional IO and logic.

@WatchesTrainsAmdRockets 9 ай бұрын

Never dealt with the 2113 or even heard of it but the story brings back a lot of memories from thos early days of small computers.

@robertsneddon731 10 ай бұрын

A single-bit static-RAM cell is usually six transistors, IIRC. It's basically a flip-flop.

@HelloWorldETX 10 ай бұрын

Yes, two that are cross connected to deplete the gate of the inactive one. Then two to load either a 0 or 1, then two to trickle charge into the latched FET to keep it on. Chapter 6 of intel's memory design handbook is very good description. I should have included a figure of the storage cell, but that wasn't my main focus. My initial curiosity was why A0 was used as the good-half control. for anyone interested, DeRamp has a copy of Intel's '77 memory design handbook here: deramp.com/downloads/mfe_archive/050-Component%20Specifications/Intel/Memory%20Components/1977%20Intel%20Memory%20Design%20Handbook.pdf

@GodmanchesterGoblin 10 ай бұрын

Hence the estimated transistor count of 24576 that was shown briefly on screen, although that ignores the overhead of the decode circuits, buffers, etc. that were main subject of this video.

@HelloWorldETX 10 ай бұрын

@@GodmanchesterGoblin Wait.... wikipedia could be wrong? I looked at the references they had with this number of transistors, but neither reference had the actual number as far as I could see. I plan on removing the passivation layer and then the top metalization layer so we can get a good look at the storage cell. Rather than load transistors, they may have used load resistors which would free up 8K of the transistors to use for addressing.

@GodmanchesterGoblin 10 ай бұрын

@HelloWorldETX I have no idea regarding the correct number. I just saw that very round binary number (24576 = 6 x 4096) which immediately drew my attention (my brain works like that), and I also went with the six transistors per cell approach. Yes, I assume that resistive loads could have been used instead, but I don't know the history of the various topologies to judge further.

@vintageteardown5291 9 ай бұрын

Yes back then SRAM was always 4 transistor with high resistance polysilicon load resistors. @@HelloWorldETX

@KallePihlajasaari 9 ай бұрын

There was probably a option to use the most significant row select address as well to partition the RAM top and bottom. Interesting analysis of the failure modes. Harris might have been selling their faulty chips profitably while Intel was embarrassed. The Sinclair Spectrum also had options to select budget RAM, it had H/L and 1/0 options to accommodate cheap RAM from two vendors. It seems that for quite a while it was profitable to market the memory that was half broken to thrift customers.

@uzlonewolf 9 ай бұрын

That's still regularly done with things like CPUs where bad cores are deactivated and sold as a lower-core-count chips.

@bloeckmoep 10 ай бұрын

It's fascinating how history repeats, back then struggling with ram chips, today struggling with 14nm, 10nm, and now rebranded 10nm to 7nm finfet transistors. In between all of this, struggling with x86 (p1,2), struggling with new architecture (p4), struggling with even newer architecture (intanium), struggling with in house graphics solutions (i740), struggling with even more in house graphics (gmas and Larrabee), struggling with more ram (3d xpoint), struggling with even more ram and memory (optane). It reads like a disaster report.

@HelloWorldETX 10 ай бұрын

In fairness, that is the cost of pushing technology forward. Once Intel got the 2114 process into full production, it became the standard simply because it was superior. Quick, low power, and reasonable cost. while it was embarrassing at the time, in the end I think it more than made up for the delay in roll-out. Eight or ten years after this 2113 problem, I was at a conference where intel's CEO was the keynote speaker. He announced that fabrication R&D was just getting too expensive and Intel was no longer going to be on the cutting edge for chip technology; but they were going to focus on designing the chip and then outsource to offshore fabrication houses so that the fab shops could spend *their* R&D money perfecting the process. This was to a room full of semiconductor and fabrication researchers at a national conference !!! Of course, that is how we got to where we are today, i.e., in a situation where none of the domestic chip manufacturers can actually make their own chips that require the absolute latest process. The federal government can buy all the new fab lines they want for Intel and others, but when the bottom line says it is cheaper to send fabrication off shore, they will no doubt, once again, outsource the latest fabrication processes.

@bloeckmoep 10 ай бұрын

@@HelloWorldETX: That's all nice and dandy, but with intel, it always feels like they TRY to create the next best thing after bread but halfway through, they loose interest or goal. The NuCs are the next entry to this list, intel just lost interest three months ago, leaving vendors for their own, telling them to "please be so kind and keep developing and making them, because I got bored". Netburst... "failed" architectur, intel lost interest. Itanium... same game. It would have been the same for x86, if amd hadn't outdone and gone above and beyond them, fixing many architectural flaws and short comings. AGP... intels take on a highspeed high bandwidth bus with many fantastic features such as "AGP aperture ram" and "side band adressing", never fully utilized, never truly completed. I swear, the longer I think, the more dropped, bodged or sold projects come back from the nether realm of my brain where they should have stayed as corpses to rot forever. And now... we are at the realm of disabled, broken, obsolete cpu instruction sets. Disabled from THE very day it was put into silicon... (drum roll)... TSX 🎉🎉🎉. Next contender... AVX 512...🎉🎉🎉. I could swear, there was another one, intel silently did away, something to do with "randomized" memory mapping, inherently flawed and defeated by a precise timer like java or even javascript could provide and a whole load of ROPs. 😂 I'm not salty at intel, I just think and feel, for once they should pull something through, finish it, fix the bugs from a first revision whatever "that" is. But honestly, I can see the next butchered project already, their just recently born graphics card segment. 😮

@internetguy1260 9 ай бұрын

@@bloeckmoep this is i think actually a issue all across US corporate culture. Its a symptom of short term profit preference due tp shareholder primacy.

@johnclement5903 9 ай бұрын

@@internetguy1260This is THE reason why NOTHING is made in the USA anymore, and we are at the mercy of a bunch of Communist Chinamen who hate our guts. You would think that one of those Big Brains at RAND or MITRE or other branches of the Military-Industrial Complex would have had some serious reservations about sending ALL of our high tech secrets to a Communist Country! But, I guess all those dollars flowing into the Clinton Global Fund overrode the concerns of the Big Brains. I am simply disgusted

@ClausB252 10 ай бұрын

The same sort of thing happened with early 4116 16Kx1 DRAMs. 8Kx1 versions were sold cheaper and appeared in some computers such as Atari 800 and 400.

@adriendestugues4202 10 ай бұрын

And it happened as well with the 4416 16Kx4, where defective chips were sold as 4408NLT and 4408NLB. It is impossible to find a datasheet for these (as far as I know), but they were used in Matra Alice and Exelvision EXL100 computers, with the "T" and "B" indicating if the top or bottom of the address space was usable.

@theelmonk 10 ай бұрын

I did my final year undergrad project on a prompt-48 in 1980. I didn't have a teletype or an assembler so every byte was punched in on that keyboard, and I saved work-in-progress on an 8748 EPROM processor. I don't know what memory chips were fitted though !

@HelloWorldETX 10 ай бұрын

I have an 8048 coding form that I can scan and send you if you want to relive the good old days of hand assembling !!

@theelmonk 10 ай бұрын

@@HelloWorldETX Ha! I was a student - I used ruled paper :)

@bryede 10 ай бұрын

Maybe Harris thought the idea of selling defects as a 2K chip was a good idea and they could entice someone into incorporating into their design as a cheaper alternative, or as you said, they were just being snarky.

@davidconner-shover51 9 ай бұрын

I remember running stacks of the Harris version of this chip back in high school in the '80s

@0MoTheG 9 ай бұрын

Or their chips weren't 4k to begin with?

@warphammer 9 ай бұрын

@@0MoTheG Not a bad thought but if that were the case there wouldn't be a -H and -L variant.

@0MoTheG 9 ай бұрын

@@warphammer Why not?

@uzlonewolf 9 ай бұрын

@@0MoTheG Because if the chip was only 4k then A0 would just be "no connection" and a single chip would work in both -H and -L systems.

@dwightelvey645 15 күн бұрын

Intel did the same thing with 2716s. The even sold the single voltage 2708. They were mentioned. As the price for the 2716 dropped over time, the 2708 were still in the catalog but at close to the original price of over $30 each. The same catalog had 2716 for around $15 or so. I always thought it was funny that one could directly replace a 2708 H or L with a 2716 for a cheaper part.

@hattix6713 9 ай бұрын

The 24,576 count is made by simply multiplying the 6 transistors by the 4096 cells, naively. It ignores bus drivers, signal drivers, etc. The real number will be a little higher.

@HelloWorldETX 9 ай бұрын

I am not sure yet if the 2114 used active loads or if they were just resistors. As a low power device, i expect they were active, but if passive there may only be 4 transistors per cell. Hopefully before too long I get around to removing the passivation layer and then rephotograph and probe the address lines to confirm my A0 hunch, then start removing metal.

@HelloWorldETX 27 күн бұрын

Someone just pointed out to me that Intel’s ICE-85 control board has a jumper to use the 2113L, 2113H, or the 2114. So that is the third system I know of that was designed to use the 2113.

@mhuffstutter1803 10 ай бұрын

And of course by "Clock" I mean "Temperature Sensor"..... more brain cells MIA....

@waynesmith2287 8 ай бұрын

IBM had the same issues with its new 1 MB memory module for the 4331 in 1979. They knew they were going to have bad yields of their new 64K FET memory chip so they did as you have describe and had 4 chip types come off the wafers (1) A side ok and B side ok (2) A side bad and B side ok (3) A side bad and B side ok (4) Both side bad. The good chips were put into machines with 2 4 or 8 MB of memory. The chips with only half good were matched and joined on the memory module for the 1MB memory 4331. The bad chips were put in resin cubes and handed out as gifts. April 18 1986 IBM successfully release the first 1 Megabit memory chip.

@38911bytefree 10 ай бұрын

CRAZY tuff !!!!. Very interesting !!!.

@cameramaker 9 ай бұрын

Interesting naming convention vs physical layout - if the A0 jumps the largest portion on the die, one would think that it is the MSB, yet with classical logic design, we always assume it being the LSB. Even when you talk about is as even/odd address, but on the die its no longer repeating, thus its totally just left and right half of the whole array :)

@HelloWorldETX 9 ай бұрын

I agree, it is a bit disorienting. Most memories are in arrays where the address lines specify the rows and columns, so the idea of starting at one corner and counting up in binary works well in our brain, but maybe not so well on the chip. Since any one address enables or disabled half the cells, which *external* address becomes any particular *internal* address depends more on mundane issues like convenient carrier bonding.

@apsdev 10 ай бұрын

sinclair was cheap when they build the ZX spectrum computers and they bought half broken ram too. They already did that with their amplifiers where they bought rejected transistors, so they had to make something to test and pair transistors. So with the 2704 did they wire the address line internally or did you haveto do that externally ? .. I still have a vic-20 with 2114 ram chips. It actually has an odd amount of 2114's as for the character color ram you only needed 4 bits.

@HelloWorldETX 10 ай бұрын

Testing transistors to see if they are 'good enough' seems like the exact opposite of when we used to test to get matched pairs. I didn't need to do anything to the 2704s. pin 22 is either A9 on a 2708 or ground on a '04. i was only expecting the A0 half to program before it started throwing errors, but it programmed all the way to the top and matched the buffer.

@apsdev 10 ай бұрын

@@HelloWorldETX Well found this on a website of someone who worked for sinclair "The output transistors (mounted on the rear of the amplifier) were germanium and Clive Sinclair in those days used to buy reject devices: in the early days there was a lot of trouble using these, until I built a dynamic tester which plotted the gain at high currents, showing the graph on an oscilloscope." .. so i guess he bought transistor with current gain out of spec. which might be fine for an end stage, but they might need to match gains. I do actually have one of these sinclair amps, so that's why i googled it and found out about the use of rejected transistors.

@xlerb2286 9 ай бұрын

I remember a couple comp sci buddies in the old days talking about some machines that used half broken RAM but I couldn't remember who, if they ever said. Thank you.

@scottlarson1548 9 ай бұрын

I just realized something. In the early 80s it seemed like every fab on the planet was making 4164's to keep up with the huge demand mostly generated by the popular Commodore 64. Those computers always had a random collection of 64K bit chips from Toshiba, OKI., NEC, Motorola, Micron and others... but I don't recall ever seeing a 4164 made by Intel. Had they given up on the RAM market before it exploded?

@HelloWorldETX 9 ай бұрын

Intel’s was the 2164. It had to begin with a “2” because it was n-channel.

@scottlarson1548 9 ай бұрын

@@HelloWorldETX Did anyone use it? I don't think I've ever seen a 2164 in my life.

@HelloWorldETX 9 ай бұрын

@@scottlarson1548 i think you are on to something here…… you nailed it when you listed the manufacturers that made the 64kbit chips. In about ‘78, Mostek second sourced their memory to a number of Japanese firms. Within a few years those firms were making better chips than any domestic manufacturer, including intel. By the mid ‘80s, it was all over for US production. I looked at the adverts in the back some magazines from the early 80s. All the DRAM were Japanese. I think the answer is intel may have been used on some high end systems, but for the consumer market they were just too pricey.

@andygozzo72 10 ай бұрын

bit like part duff 64k x1 bit drams(4164 or equiv.) but these were definitely available, being sold as 4532s, with 4 combinations of 'dead' section, bad row or bad column and in each one of those could be higher or lower one dead,, the zx spectrum used these for their upper ram

@HelloWorldETX 10 ай бұрын

Did they bond the die differently depending on which quad worked properly? or same pinout like the 2704/2708 and 2113/2114?

@andygozzo72 10 ай бұрын

@@HelloWorldETX identical pinout , there were differences in row or column dropout depending on brand, OKI ones were one, TI the other cant remember which one had which dropout offhand

@KallePihlajasaari 9 ай бұрын

@@HelloWorldETX They had holes for links on the board to select the correct High or low. The selection was done of different signals from the different vendors so needed quite a few links.

@boristheengineer5160 9 ай бұрын

I read that at least one board variant of the Dragon 32 used those, there was one jumper link to flip the A7 line. All the boards I saw used two banks of 16k*1 parts, in one case the board was only designed to hold 8 DRAMS and it had a neat little daughterboard installed to hold the other 8. I don't know if the manufacturer was hit by shortages or just using whichever parts were cheapest. My friend converted his to 64k.

@andygozzo72 9 ай бұрын

@@boristheengineer5160 my current dragon 32 has 16 4116s with some on that daughterboard, a previous one had 8 32kx1bit chips which arent half duff 4164s but effectively have dual 16kx1 bit rams inside them, they have separate ras and cas pins for each 'half'

@Clancydaenlightened 8 ай бұрын

25:20 well why do they still make 6502 and z80 cpu when i can get an intel or amd 64 bit running at several ghz?

@AmauryJacquot 9 ай бұрын

suspect there was some classified military device using 2113 and needed a 2nd source...

@HelloWorldETX 9 ай бұрын

That may not be far from the truth. If the government were to use any iSBC 80/20s, they would have required a second source. But, as far as i can find, the harris 6513 doesn't show up in catalogs until after 1977, and by then (or at least sometime in 1977) the 2114 was in full production and would have been a direct replacement for the 2113 in any existing board using the 2113. There would soon be many second sources for the 2114.

@AmauryJacquot 9 ай бұрын

@@HelloWorldETX considering the time it takes to create military devices, they could have required the second source at the time the device was put in production. also they would have had access to pre-release 6513, which would be before 1977

@h7qvi 10 ай бұрын

Like looking at old retro gear

@HelloWorldETX 10 ай бұрын

Me too. But I especially like working and writing software for them.

@wiebel7569 9 ай бұрын

So is the 2113 from week 35/77 aptly named A0X?

@HelloWorldETX 9 ай бұрын

Good point, but it is a A0H like the others. I wonder if that means this chip has some sort of latent defect that would show up in extended testing? Or did intel really not think their customers could comprehend a 2114 working in a 2113 socket ?

@crystalsheep1434 9 ай бұрын

8:26 yes they do some cool stuff

@HelloWorldETX 9 ай бұрын

Indeed, a wealth of knowledge and information there

@pascalcoole2725 9 ай бұрын

When i was slightly younger i had a Sinclair ZX-Spectrum they used half of 4164 64kb Rams. the story was the same as you are explaining here... bad chips that where ok as long as you used only half of it. For the intel part, in the 90's on my job i used to work with an iSBC system running CP/M86. pffff even than i was not enthousiastic about it.

@HelloWorldETX 9 ай бұрын

I am presuming the iSBC system was considered archaic at the time- but one of those old control systems that was impossible to replace? The iSBC design was long in the tooth by the 90’s, but when they first came out they were a pretty sweet platform.

@pascalcoole2725 9 ай бұрын

@@HelloWorldETXIt was used in a pretty large Hardware Diagnostic System. Embedded into a large Desk/Cabinet where you could install PrintetCircuitsBoards to diagnose them (Aviation Industry) I started work to create a PC-Based system, and as a mather of fact later systems made use of a smaller unit in combibation with a PC running MS-DOS. (a that time I already only did development on Linux Systems) For your question on time. Wel In Aviation like in most industries, in general equipment will be used well over 30 years. At the time the technologie was about 8 years old, So delivered early 90's. There were a lot of different technologies from different era's used, Varying from TI980 (Harvard Architecture) alike computers, as well as small mainframes (today you could run that stuff on your Phone), Lots and Lots (i mean a really lot off) MicroControllers or Small Micro embedded Computers using 8086 CPU's, Weird never heard of computers and even i once had to deal with a transputer system. Most unique however would be me..... Yes I turned out to grow up as a UNIX-crack, but i never have had a beard !

@rinner2801 10 ай бұрын

You have defeated my colorblindness =)

@HelloWorldETX 10 ай бұрын

Nothing like a strong red/green/blue persistent image to let someone continue studying a diagram long after closing their eyes.

@grossteilfahrer 9 ай бұрын

I have a SBC 80/20 with 2113's on it. But I'm missing the 8080 for it. it has eight 2113s all A0H. Its PWB 1000818-01 rev 3, s/n 411 _ 3 or possibly 411 _ J

@HelloWorldETX 9 ай бұрын

Very nice. The 80/20 is a good single board computer and 8080s can be found for a reasonable price.

@grossteilfahrer 8 ай бұрын

@@HelloWorldETX As of yesterday there is a AMD 8080A from 1977 in my board. I also sourced three of the card edge connectors so I can start hooking up power and serial to it. I'm excited.

@leon545b 9 ай бұрын

Very interesting. Makes me wonder if Intel had a boatload of 2113's after the 4k crisis was over and they sold 'em to Harris at a bargain basement price...?

@HelloWorldETX 9 ай бұрын

Good idea, but i don't think so. In the harris spec sheet there is one diagram that indicates there are subtle differences between the 6513 and intel's 2113. They say: 2113 requires the address to remain valid throughout the cycle 6513 requires valid address for only a small portion of the cycle, but requires /E to fall to initiate each cycle. Part of what gave the 2113/4 its speed is that the addresses were not latched, they just rippled through. evidently the addresses in the 6113 are latched by a falling /E

@johndmcmaster 8 ай бұрын

What microscope did you use to capture this? Was it motorized?

@HelloWorldETX 8 ай бұрын

Amscope metallurgical. Yes, I have the stage motorized.

@johndmcmaster 8 ай бұрын

@@HelloWorldETX cool! do you have some info on your setup? I don't recall that being a factory option

@HelloWorldETX 8 ай бұрын

@@johndmcmaster I have added the motorized stage. The design is not as good as I would like, so I have been thinking of revisiting and making an improved version. It is a little out of the usual projects I share, but perhaps it is of general interest.

@0MoTheG 9 ай бұрын

Any one of the bits would have deselected half of the IC. They could have decided based on a statistic of two bit errors but that wouldn't have looked good to customers.

@cameramaker 9 ай бұрын

I think the issue was rather a contamination that spread multiple bit cells - like if you have a huge spot on your screen - picking even/odd is or 2 or other 2 is not really making go away the spot - but the MSB that drops the "physically largest region" is the key here. I am not good at statistics, but the common sense makes me agree with the commentary in this video.

@0MoTheG 9 ай бұрын

@@cameramaker You are likely right, the chance that two errors have a common cause is high, thus whatever pin disables adjacent cells will yield the most working chips.