They fooled so many: Fake cache chips!

[@vincentpremel4817]

Back at the time, a french magazine took a suspicious motheboard to the nearest dentist, to make a X-ray pic of the fake chips : thoses were empty.

[@gblargg]

These days a thermal camera would easily show whether there is a chip inside, by the slightly warmer rectangle in the center.

[@cebruthius]

@@gblargg Or probe a data line with an oscilloscope? Would that work?

[@gblargg]

@@cebruthius Yeah, check an output line for voltage.

[@intron9]

@@cebruthius you'd need to know which lines are supposed to change only if a chip is connected. Watching a level changing isn't enough, since it could be another chip that is changing it.

[@l3xforever]

“None” and “256K” are both exactly 4 characters long. The simplest hex edit in the world.

[@Lilithe]

I had the exact same thought. They probably took the stock BIOS, searched for "None" and replaced it with "256K" and put the empty chips in the board.

[@robertcasey2490]

Maybe do a hex edit to the edited bois and have it say "ZERO". And if it turns out it had two places that say "256K" change just one and see if you see "ZERO" with no cache, or maybe it says "ZERO" when there really is cache memory present. In which case change "ZERO" to 256K" and "256K" to "ZERO"

[@SidneyCritic]

Isn't there a program that displays both bios side by side and highlights the differences in red, ie, seen it somewhere.

[@matthiastilly5480]

@@SidneyCritic They are different versions - so there will be a LOT of differences

[@AureliusR]

@@Lilithe Actually, the string '256K' is not hardcded anywhere in the BIOS from what I can tell. Not even the digits "256" so they are doing something a bit sneakier and likely just changed the return value from the function that checks for cache to always be 256k.

[@therealjammit]

A quick way to tell if there's any silicon inside a chip is to set your DMM to diode check mode and measure between Vcc and Gnd. With the black meter lead on the Vcc pin and the red meter lead on the Gnd pin you should be able to measure around 0.6 volts on an actual chip. Every chip has a reverse biased diode from Vcc to Gnd that's unintentionally made during doping.

[@diegolastra]

Didn’t know that. That’s quite interesting and useful to know.

[@bitsundbolts]

I need to remember diode mode! It comes up very often and seems to be a super power :) Thanks for sharing! I will try that right away! OK, I checked! You are correct! Real chips show something between 0.6 to 0.7 volts. The fake chips show nothing (OL).

[@serhio4275]

"Every chip" like every-every? Simple logic, memory, etc ? Cool info

[@adonisengineering5508]

It's not a bug, it's a feature

[@thiloreichelt4199]

At least every digital chip with CMOS or even NMOS. This includes any memory and processor chip made probably after 1990. I wanted to check on some 74 chips, but the ones I have in reach are all 74HC, so they are CMOS anyhow.

[@ascender14]

I actually purchased a MB with these fake cache chips on it back in 1995. At the time, I was really obsessed with measuring performance, so when I upgraded from a 486 DX2-66 to a DX4-100 and saw no increase in performance, I started investigating. This was difficult at the time, but I finally read something about this and was able to confirm the scam. When I returned the MB to the retailer and told them the story, the guy smirked and said, "Hm...well, you're smarter than the average bear." :-D He all but admitted to the scam and replaced the board with a genuine board. I was able to confirm that the cache was good and I saw an appropriate perf boost after that. This was an ordeal, because this was not widely known at the time. The "internet" in 1995 was of course not what it is today. I had to smile when I saw this video in my recommendations. Yep...been there, done that. :) Thanks for the video!

[@bitsundbolts]

Good that you figured it out! The majority didn't. As you said, it was difficult to test and find this information. Thanks for watching!

[@GamIngDoge.]

Interesting story!

[@richjageman3976]

Was the DX4-100 the Blue Lightning or Cobalt?

[@kc5402]

So did the retailer KNOW that he was selling fake boards? If yes, did you punch him in the mouth? Or if no, what action did he take against the people who had supplied the boards to him?

[@ascender14]

@kc5402 I was not able to determine whether he knew he sold fake cache, or simply knew about the scam.

[@DarkZenith]

Top of the article is dated 95 not 2000. The way back image date is 2000. Thanks for another video!

[@dereinem]

yeah i have the physical copy of the magazine right here, it´s from march 1995

[@bitsundbolts]

Yes, you're right, my bad.

[@DarkZenith]

@@bitsundbolts is there a private area or discord?

[@bitsundbolts]

I just recently set up a discord server. I will publish a link soon on my website.

[@mirek190]

who was even interested 486 in 2000 :)

[@VladoT]

The pinnacle of fake cache chips is a series of Socket 7 motherboards that we got here and had 4 black rectangle (QFP) chips with embossed "Write Back" on their plastic housing and the housing was just plain PVC with pins. Hilarious.

[@sierraboney1394]

Yep, I had a 486 board (might've been a PCChips board, can't remember) with those "WriteBack" chips on it (normal dip package ones). Ami WinBios said something like '256kb Writeback Cache Enabled' on bootup, you could even switch the cache on and off in the bios, and the bootup reflected the change. Just a shame it was making no difference whatsoever!

[@GodmanchesterGoblin]

​@sierraboney1394 PC Chips M919. There were a few versions, I think. Some had the fake QFP packages soldered down. I have one that I bought new, back in the late 90s. It doesn't have those parts mounted, but has a proprietary Cache On A STick module (256kB) which is functional.

[@RWBHere]

Yes. I've seen a few of them. Soldered directly to the board.

[@louistournas120]

@@sierraboney1394 When did you buy it?

[@RJRC_105]

PC Chips M919 I think. They also claimed to be whatever chipset the distributor paid then to sticker on the anonymous chipset.

[@stamasd8500]

I have a 386 motherboard that came with fake cache chips. Not only that but they weren't socketed - they were soldered on the board. And had a doctored BIOS of course that showed "cache" on POST. I unsoldered those chips, installed sockets and real SRAM in them and it does have real cache now. I was lucky enough to find on ebay a number of years ago a large lot of SRAM cache chips that were NOS in tubes from a surplus lot so I have plenty of those. They are not the faster 15ns variety but rather 20ns, and still they work fine as cache in all motherboards I used them in.

[@bitsundbolts]

Wow, that's a commitment to fake cache! They saved even on the sockets!

[@stamasd8500]

@@bitsundbolts soldered fake cache is actually a lot more common than socketed. In fact it's one easy way to tell if a motherboard's cache is fake or not: if the chips are soldered, it's almost always fake. This is valid for DIP chips of course. While fake cache does exist on some Pentium motherboards, most of those use the QFP-package burst cache chips which are always soldered. For those, a way to tell if they have fake cache is different: if there are almost no traces leading from the "cache" chips to other components on the motherboard. :) They saved on copper to fake the chip connections.

[@ЁбрагимИпатенкоибнАдхарма]

@@stamasd8500 So disgusting. Even MB manufacturers are responsible for this scam. Which brand it was?

[@StrikerX3]

I have disassembled both BIOSes and essentially what they're doing is making the function that checks for the cache size to always return a value that says it has 256K of cache. On the v1.2 BIOS, the routine that prints the cache size is at address 2406h. On address 2420h there's a call to a subroutine at A8A0h which does some magic with the CMOS registers and returns some value in AL (presumably reading the cache size from the motherboard). On the modified v2.01r BIOS this routine is located at 24E3h and the call is at 24FDh. Note that this time around, this particular call jumps to an intermediate routine that calls the original one at 9E23h, which does the same exact magic with the CMOS registers and returns, then it checks if AL is 5, and if it is not, sets AL to 5. So presumably AL=5 means 256K of cache. (A bit inefficient tbh, you could just set AL to 5 and return.) The easiest way to fix this (assuming there's no checksum validation) is to patch the bytes at 24FEh and 24FFh from 00 BA to 23 79, which will bypass the sneaky modified call that forces the cache size to 256K. You could also play around with the modified BIOS and make it return different values for AL by patching bytes DF04h and DF08h to some other value to see if it reports other cache sizes. They must both match and the other patch must be undone.

[@marcogenovesi8570]

is this the only change between the two versions?

[@StrikerX3]

@@marcogenovesi8570 Given that they're different versions, there are plently of other differences. It'd take too long to figure out what else was changed. Ideally we'd have a proper dump of an unmodified 2.00 BIOS to compare the two.

[@bitsundbolts]

Thank you for your thorough analysis of the modified BIOS. Very interesting how they implemented it. I guess the call to the original routine to get the cache size triggers a necessary side effect? Maybe real cache wouldn't work if this method wouldn't be called. Regarding checking AL first before changing it... Maybe it is better to avoid overwriting if it has already the correct value. I don't know for what reason, just guessing.

[@marcogenovesi8570]

@@StrikerX3 ok so they are actually different versions, it's not the same BIOS with the hack and a different version string

[@StrikerX3]

​@@bitsundbolts There might be side-effects, though we could do some science and try skipping that routine to see if that is really true. To do that, you can patch the bytes at DF00h-DF02h from E8 20 BF to 90 90 90, turning the call into NOPs. The routine reads a byte from CMOS at index 3Ah and also does a couple of writes to port E1h, which seems to be a motherboard-specific port that's written all over the place. You could also test the idea that checking AL before changing it is unnecessary by patching those three bytes to B0 05 C3 instead, which replaces the sneaky routine with a simple "set AL to 5 and return". Or, if you want to keep the original routine call, change the three bytes immediately after that, so the code at DF00h-DF05h reads E8 20 BF B0 05 C3 (call the original function, set AL=5, return). You can change the 05 in both of these to any value from 00 to 07 to make it report different cache sizes: None (AL=0), or from 16K (AL=1) to 1024K (AL=7).

[@argoneum]

BIOS with double characters comes from 286 era, when there were two 8b chips (EVEN and ODD, or HI and LO) for a total 16b-wide ROM function. There was copyright repeated on both chips, hence the double characters. 486 supports bus sizing, so it can read 8b ROM (almost) directly, yet apparently the characters remained as they were before. There is an option in setup to turn on ROM's shadow (copy it into RAM), which speeds things up (32b access). When it's set ROM chip can be removed after boot with no effect, but don't try it 😉 Later BIOSes were compressed (more advanced), and they decompressed into RAM after power on, so those could be hot-swapped for flashing. Did that when computers were expensive, and programmers weren't actually cheap to get either.

[@bitsundbolts]

That makes a lot of sense! Thanks for sharing. I am working on a 286 right now - which indeed has two BIOS chips (Hi/LO or EVEN/ODD). Thanks for clearing that up. Also, I did hot-swap a BIOS chip on an ASUS P3B in one of my videos! I never did that back then, but I wanted to try.

[@chris-tal]

@@bitsundbolts I reflashed and saved someone else's corrupted BIOS (also made backups) with my P2B back when, using this hot-swap method. I didn't have an EEPROM programmer and buying one wasn't worth it. I thought It wouldn't work with the network card boot ROM trick, because the chips were way bigger in capacity or different type. This SRAM quackery stuff is pretty sneaky. I've heard about it before and if I remember correctly I checked my 486 using these tools, but never really took a look on the hw summary table or removed the ICs. This got me real paranoid to check it again and maybe even upgrade the amount I have. 🕵

[@TonyLing]

I remember a board which had soldered in 'cache chips' but if you were to follow the PCB tracks, they went nowhere.

[@rtechlab6254]

Would that be the M919 I mentioned in my comment?

[@TonyLing]

@@rtechlab6254 I don't remember. That was many decades ago.

[@rojovision]

That's a pretty wild discovery. I've heard of fake cache chips, but a BIOS apparently designed to lie to you about it is something else entirely.

[@bitsundbolts]

I love finding these things at the scrapyard - each item seems to have some sort of story linked to it. I also didn't know they went to that extend to fake cache chips. But I am happy to hear that so many viewers are amazed how far those people went to make money.

[@LarixusSnydes]

It still happens, but now with modern Non-Volatile memory products. If you see a ridiculously high capacity medium sold at a ridiculously low price, they are fake. Your OS will show you that they have the capacity as shown by the label, but if you try to copy a (few) large file(s) to it, it will give you a write error. What's problematic is that this is occasionally also the case with media that are sold at regular prices for their stated capacity.

[@stuartmcconnachie]

My guess would be unscrupulous systems integrator patched the BIOS after finding themselves in receipt of a load of fake chips. Seems obviously patched, and so therefore not an official release by the BIOS vendor themselves.

[@fallingwater]

One follows the other. There couldn't be fake chips without a fake BIOS, or you'd twig to the scam immediately.

[@ssokolow]

@@LarixusSnydes That's why it's so important to use something like F3 to test what you buy AND buy from a reputable source, if for no other reason than to ensure a good return policy.

[@M_McFly]

This brings back memories. Also of the time when they used to put parity generators on the RAM PCBs instead of real RAM parity chips.

[@louistournas120]

I remember reading about Pentium chips being shaved. The package was made of a dark grey ceramic, which was normal. Some company used a machine to shave off 0.5 mm from the top and print a higher speed or something on top. The CPUs would work but it would be overclocked. In some cases, it would not run stable.

[@SianaGearz]

Haha Georg Schnurer from c't is an absolute legend, always worth a read! Still active, still core team, almost 30 years after this article. Also generally seen in c't videos, he's the cheerful guy. How much could they have saved by shipping fake chips? A couple dollars? 256k SRAM shouldn't be so expensive by mid 90s!

[@bitsundbolts]

There was definitely criminal energy behind all this - I wonder too if this was lucrative.

[@T3hBeowulf]

I would err on the side of preservation. Submit it and see if it can be tagged as modified to show invalid cache levels. I'd love to see a hex diff of v1.2 and your v2.01 BIOS.

[@bitsundbolts]

The Retro Web Team has accepted the BIOS and it is available for the board. Unfortunately, the two BIOSes are very different and I think a DIFF between them will most likely mark the entire file. Btw, I think there are some performance issues with version 2.01r. I have to look into this. You can see a huge difference in the moving benchmark between v1.2 and V2.01r.

[@MM-gd3be]

@@bitsundbolts I'm pretty sure they just changed the None string to 256k. Back in the day in the 90's i had a 486 motherboard with fake chips that was reporting 256k. I've bougth myself 128k of cache and the BIOS was reporting 128k with real chips.

[@phipli]

Yeah, that's what I'd expect. If I was them I'd search for "None" and overwrite with "256k". Perhaps you can do the reverse @bitsundbolts? Find 256k as a string, if you find two, change one to "None" and test, if it doesn't work, change the other instead.

[@myne00]

That's what I was thinking. Search the Rom for 256kb. Replace with none.

[@myne00]

That's what I was thinking. Search the Rom for 256kb. Replace with none.

[@advil000]

Just out of high school back in the 90s when I was working for my first employer, before I started my own business, I built so many of these 486 and early Pentium systems. You memorize the jumper configuration for the boards pretty quickly. I wonder if a bunch of them had fake cache and none of us youngsters on the build crew ever realized it? We didn't have the means of verifying it. And frankly, back in those days, there wasn't much reason to suspect. I mean, as you show, the BIOS even happily displays cache that isn't there. It was so long ago now I can't remember the specifics anymore but we did do some stability tests and benchmarks and we were all PC performance junkies. You'd think we would have picked up on the cache being totally non-functional. But it was a different time, and we were all pretty green.

[@danielktdoranie]

Old guy here again, I vaguely recall a scandal regarding fake cache chips back in the 90s

[@bitsundbolts]

How about "experienced" 😊? Even though I'm aware of fake cache, I don't think I knew it back then.

[@_LM_]

@@bitsundbolts I remember it quite well. It's one of the reasons we used to cal PC-Chips motherboards "PC-Shit".

[@MinceWalsh]

Hahaha I see what they did with that BIOS. They didn't really want to deal with anything but the checksum so 2.0ra became 2.01r so they didn't have to shift any bytes and refactor.

[@rigues]

Nice video. One of the first things I remember about starting university (Computer Science) here in Brazil in 96 was a warning on the student newsletter about motherboards with fake cache chips.

[@bitsundbolts]

Oh wow - interesting :) I should put one of those stickers on this board 😂

[@vswitchzero]

Excellent video! Amazing they went to such lengths to mod the BIOS to hide it too. I have an example of the infamous PC Chips M919 with the fake “write back” chips on it. I think it would still be good to upload this bios to the RetroWeb page for the board just for historical interest. I’m pretty sure the mods can add a custom message to warn people about its behaviour. Cheers! 🙂👍

[@bitsundbolts]

Thanks Mike! I uploaded the BIOS to The Retro Web and it has a note about the cache display. People are already at work to undo the changes. Let's see if we get an even better BIOS once the experts are done 👍

[@tony359]

I came to know about motherboards with fake cache when I repaired a PC-Chips board. Mine had genuine cache but I wasn't aware that "fake cache" was an option! It amazes me, the picture you showed shows an IC coming from manufacturing, they just didn't add the actual die which is probably the expensive bit! Fantastic! And someone edited the BIOS from "None" to "256 KB", so simple and so hilarious! My first PC, a 386/DX 25 had no cache. I couldn't afford it. Imagine I paid big money (for a teenager in 1992) for cache and not even realise that the long awaited cache was just a piece of LEGO. I'd share the BIOS anyways, with a note explaining what it does :) For historic purposes :)

[@bitsundbolts]

That is the sad part! You spend money because you want to have a better system - and then you realize that there is no difference. You would always be skeptical and claim that cache has no benefit: "I have seen it with my own eyes!" - Sad :( The BIOS is shared on The Retro Web for this board with a note that this BIOS always shows 256kb L2 cache.

[@codingwithculp]

OMG. A few years ago I put together a Tomato 4DPS 486 system with a Am5x86-P100 that I stably overclocked to 200Mhz. If I remember correctly, the board could hold a whopping 512 K of cache. So I went and searched for cache chips and bought a bunch from few different sellers and sources. All of them were fake. I finally gave up on it.

[@bitsundbolts]

Wow, a stable overclock to 200 MHz is not that common! Congrats! Sorry on all the fake cache chips you collected :(

[@GigAHerZ64]

All 386 and 486 bioses (and to certain extent pentium 1) are similar enough that you can easily go to retroweb and get latest bios based on your chipset. You don't have to pick bioses specific to your board. ;) (There are, of course, exceptions, though they are rare.) NB! I suggest checking out how to enable write-back on the cache, too. Another measurable performance jump. ;)

[@bitsundbolts]

I want to try that some day. Using BIOSes from different boards. Good to know, I'll keep it in mind. Thanks!

[@M0UAW_IO83]

Definitely, as long as the chipset is the same it's easy, used to do that back when these boards were new.

[@insanelydigitalvids]

Fascinating (and heart breaking). I very much liked your use of the Programmer to test the chips.

[@CosmoRiderDE]

PCChips M912 and M919 - notorious fake cache boards, but superb performers

[@GodmanchesterGoblin]

Yes! And I have an M919, but with the functional cache module fitted and the fake IC positions left empty. Currently it's running with AMD 5x86-133 and 64MB RAM.

[@bitsundbolts]

I have one of those boards that needs to be repaired. I don't know yet what is wrong with it - no sign of life. But I will try my luck again soon.

[@GodmanchesterGoblin]

@@bitsundbolts Cool. I look forward to the video if you do one.

[@CosmoRiderDE]

@@GodmanchesterGoblin Some dude managed 180-200mhz on the M919, it supports even 60mhz bus clock afaik.

[@CosmoRiderDE]

@@bitsundbolts the ones i have work perfectly well, the M919 with the cache module, and the 912 with real cache installed instead of fake placebos. Excellent stable performers. Hope you get it working. Gruss aus Deutschland.

[@michaelkreitzer1369]

It’s likely possible to undo the modification to the bios. It’s a pretty safe bet the bare minimum was done to achieve the false reporting. There were probably only a few instructions modified, and with some disassembly it should be fairly simple to locate and undo it.

[@myne00]

The absolute bare minimum is to change the string if zero.

[@DxDeksor]

@@myne00 probably needs confirmation but yeah my first assumption is that they've replaced the "none" string with "256kb"

[@bitsundbolts]

Yes, probably that is exactly what they did. I didn't try to reverse the change, but the BIOS is on The Retro Web if anyone would like to try. It also has a note about always displaying 256k of cache.

[@stamasd8500]

There is a complication to simply hex editing strings in the BIOS: you have to recalculate the checksum of the new file and write it in the right spot.

[@UltimatePerfection]

Not a huge deal though.​@@stamasd8500

[@GadgetUK164]

Another very interesting video =D I would upload that BIOS to the retro web anyway, with a readme that notes that the it will always report cache when not installed. Just based on the extra features that BIOS provides. That "mod" could probably be easily un-modded by someone if they can get a copy of the BIOS.

[@TosterCx]

It is on the wiki

[@ScottyBrockway]

Fake cache was a real problem, and ruined PCChips reputation as their boards started shipping that way too. Kind of sad.

[@bitsundbolts]

I do have one of those boards. It has solder pads, but the traces route from one cache chip to another - no traces go anywhere else.

[@ScottyBrockway]

@@bitsundbolts Begs the question, what were they thinking...

[@M_McFly]

Ah PCChips - the Maxtor of motherboards.

[@indask8]

I think I remember seeing a video where the cache chips were hollow... EDIT: Speaking of the devil 25:53

[@M0UAW_IO83]

They were crap from start to finish, the fake cache chips looked fake and were made of some gharbage soft palstic you could dig a fingernail into. Shame as the chipset was pretty OK but the quality of the PCBs was just awful.

[@envoycdx]

Excellent stuff, I should look at my board may have a similar issue or I have rose tinted glasses around how well Doom used to play on a 486! :)

[@bitsundbolts]

I guess it's good practice to test the cache on any board. Based on what I saw with this board, you can't trust any values displayed.

[@logipilot]

maybe this is the cause for lgr s woodgrain being so slow 😂

[@Stratotank3r]

Yeah, this is even one step ahead of the famous PCChips Boards with fake "write back" cache. I have two of this M919 Baords. One green PCB with the plastic fake chips and traces to a roundabout and a later version without the chips but with the right cache modul and real 256kB. There are also fake chips with marking and legs but they are hollow on the underside.

[@bitsundbolts]

I have one of those M919 boards that needs to be repaired. The CPU was inserted in a wrong orientation and blew up the socket. Crazy to hear all those stories about fake cache chips and the extend!

[@eformance]

I have 2 of these MV035 boards, one I purchased when they were new. This is interesting.

[@wtmayhew]

In the 1990s, counterfeit capacitors on motherboards were a problem. An exploding electrolytic capacitor can cause lost data.

[@AJMansfield1]

Fascinating, you've earned my subscription! Though, I'm surprised you didn't include a DOSbench run in the video showing the Fake BIOS + Fake Chips configuration. It doesn't affect your conclusion -- you proved your thesis regardless, and I'm sure that run would look absolutely identical to the one you showed for the Fake BIOS + No Chips configuration. But it would've improved the rhetoric at least slightly to show or mention that, as it's probably the strongest argument that could've been made (in that, it rules out even far-fetched fantastical scenarios, like 'special woowoo BIOS code somehow enables plastic blobs to cache data').

[@bitsundbolts]

Thanks! Yes, I did not check the fake BIOS with the fake chips - something I can do and use to start the follow-up video. People much more knowledgeable than I are working on reversing the modifications that were done to that BIOS. It would also be interesting to look into one of those cache chips - although I'm pretty sure there is nothing in them.

[@saimoncerise4870]

I had the exact same issues with a PCChips 80486 VIP board, as I couln't find any "safe" BIOS I just tried swapping fake with real cache chips, tested with Phils DOS Tools and then... IT WORKED :) What I like is that 30 years later I'm still learning computer things, I only knew about cache on a stick from Pentium era ! Too young to check the cache chips on my DX4-100, only wanted to play DOOOOOooooooom ?! :D

[@scytob]

we built and sold PCs in this era, had no clue there were fake cache chips, we speed tested every board we made so that would have turned up any fake cache chips we might have bought (we populated cache chips by hand) we were always amazed the number of machines from other vendors that came in for repair and just had no cache installed at all (looking at you AST and Compaq in the UK)

[@danielcarter305]

You know you're old when a 486 board is more familiar than a modern-day board! 😂😂😂

[@UpLateGeek]

I'd heard about fake cache chips back in the day, but I never came across them until the late 90s when I was working at a computer store. One day I was working on an old 486 machine when the older tech came in, pointed to the cache chips and said "Those are fake cache chips!". I was surprised and doubtful, so he told me to remove them and see what happens, so I did and it still reported 256K just like you saw. I asked what to do and he said just put them back, nobody ever asks about the cache, so they'll never know the difference. I did, but I wasn't happy about it. Looking back, there's not much you can do in those situations, other than offer a replacement machine if someone complains about buying one with fake cache. The boss was very dodgy, he definitely wouldn't go out of his way to buy new cache chips and replace them, especially considering it would probably cost more than the rest of the machine cost him!

[@bitsundbolts]

Wow... Funny story - that experienced tech pointing out the fake cache is funny. I guess all those boards were sold and probably nobody ever came back and claimed a replacement for the fake cache chips.

[@FalconFour]

There's a "notes" field in TRW's ROM section. A perfectly fine place to put a note that "this BIOS reports fake 256k cache", and it can still be put there for all to gawk at the nasty shenanigans. Regardless of whether it was an evil thing to have done by the manufacturer, it is still part of history.

[@bitsundbolts]

That is exactly right! The BIOS is already on TRW with a note similar what you mentioned.

[@PhilErringtonSK7]

I saw these boards when I first started my company back in 1994. The Deep Green BIOS rings a bell. If I remember correctly these boards were in the cheapest machines possible and often supported other non-intel 486s which made things even cheaper. The dominance of good board manufacturers supplying to the builder market wasn't yet established and there was a lot of cheap stuff cutting corners to sell boards to an industry that was very price sensitive. I believe these boards were around £5 cheaper which was around $8 cheaper at the time.

[@CharlesVanNoland]

The hacked v2.01 bios that came with the board tested with lower overall read/move speeds than the v1.2 bios you tested on it.

[@bitsundbolts]

I did notice that! 1.2 is quite a bit faster, but it has less features. Something for a follow-up video.

[@AureliusR]

Can you please upload your dump of the original BIOS somewhere? I would like to process it through Ghidra

[@bitsundbolts]

It is already on The Retro Web for this board. It has a note to always show 256kb of cache

[@ms2649]

Mind keeping us updated with what you find?

[@AureliusR]

@@ms2649 Well, they didn't do the obvious thing and hard-code the string "256K" into the BIOS anywhere, so it must be hard-coded as a value somewhere. Unfortunately as a word value, it's 00 01 which shows up a billion times so it's hard to track down. I'm also not that experienced with Ghidra and it's doing some odd things that I can't figure out. I assumed the BIOS is either at FFFF:0000h or 000F:0000h but the code seems to have references to labels in the segment 0000:0000h so I'm not sure what the deal is there.

[@ms2649]

@@AureliusR things like cache checks are probably going to be the same across versions. what about using the older bios to search for where "NONE" is used, and then search for the bytes around that in the new bio. have you tried a good ol' diff on the versions?

[@AureliusR]

@@ms2649 Haven't done a diff, I really want to get a somewhat clean decompile in Ghidra but I'm having issues. I need to dig out my book on Ghidra and jog my memory on some things.

[@Rasterizing]

Great vid, thanks! It makes me wonder how many systems I've used in the past with fake cache now :D

[@Ray_of_Light62]

This is because a set of 256K of 15 nS static RAM was, at the time, the equivalent of $ 200. On the computers I built at the time - when the 80486/DX2 was $ 1,000 when it hit the shelves - I tested the motherboard with and without the cache chips; not only because of the fakes (a limited problem), but mostly because the one of the cache chips was often too slow, and that caused errors - a totally random computer freeze. SRAM cache caused a lot of troubles at the time, even more troubles than dynamic RAM. The fake chips/fake bios was a criminal enterprise at board stockists level, who sold the boards to people who didn't know better, and pocketed the $ 200. Remember that a computer was worth 1000s of dollars, and anyone who bought the latest and greatest run a number of benchmarks and large spreadsheet calculations, AutoCad renderings and Mathlab graphs: the missing cache was easy to spot in the end.

[@moardargons8160]

If you're just doing light word processing (which is what most people used computers for in those days) then you don't need cache, and it's appropriate that manufacturers made it an optional extra. Dishonest manufacturers however....

[@huberthumphry280]

@@moardargons8160 most people using computers in the late 80's and early 90's were not using them for word processing

[@jdmcs]

The characters at the beginning of the BIOS are repeated to make sure that an intact copyright statement appears at the beginning of each ROM chip. For BIOSes designed for 16-bit processors, this usually means the characters are duplicated (in case there’s a low and high ROM chip), and BIOSes for 32-bit processors may be duplicated out quadrupled (in the event that four ROM chips was used).

[@bitsundbolts]

That makes a lot of sense! Thanks for sharing this detail! I guess it was the cheaper method than properly preparing for 32-bit systems. Cutting corner as usual :)

[@BrianG61UK]

@@bitsundbolts I'm puzzled. What do you think properly preparing for a 32-bit system would look like in this case? And would it even make sense in a BIOS for a system with a 16 bit data bus?

[@PlumGurly]

Glad you are good enough to cache this problem.

[@bitsundbolts]

😂

[@glitchwrks]

Encountered these on PC CHIPS boards back in the day. I'd be interested in seeing them decapped!

[@geremychubbuck3730]

Sadly, not unheard of. I have a pentium 90 motherboard with fake cache chips. Unfortunately they are surface mounted so I can't change them. Great video, keep up the excellent work 👍👍

[@bitsundbolts]

Thanks! Eh, soldered fake cache chips are annoying and it's probably difficult to find real replacement chips.

[@rtechlab6254]

Ive seen these with the underside just hollow. I also remember a 486VL board with SMD chips with the traces connecting the chips to themselves and that the fake chips fell off of. I believe it might have been a PC Chips M919

[@charonunderground8596]

Fortunately my first motherboard in a 486 Zida 4DVS had real cache memory. I found this out when I installed another 256kb and the total was 512kb. However, I also had a PC Chips board which was famous for installing fake caches on motherboards and this one even had a special slot for a cache module, which I bought and...still the cache didn't work :)

[@bitsundbolts]

I have one of those socket 3 boards with the strange looking cache module (I also have the module, but I don't know if it is fake or not). Unfortunately, the board is for repair and I wasn't able to fix it yet.

[@gordonwelcher9598]

After this video is completed these chips deserve a thorough Widlarizing.

[@bitsundbolts]

I guess... I need to get... A SLEDGEHAMMER!

[@askjacob]

This fake cache stuff was even around in the early pentium days when they used to have COAST (Cache On A Stick) modules that also were faked in bios, but did nothing

[@bitsundbolts]

So annoying! I wonder how many people got scammed!

[@Knaeckebrotsaege]

That was also PCChips. The M519 with the (in)famous OPTi Viper chipset had no cache onboard and a COAST slot. The COAST slot did work, but if left empty the BIOS would still report "CACHE ENABLED" instead of listing the amount. Wouldn't surprise me if fake COAST modules were stuck into these to make it at least a little more believable

[@the_kombinator]

This is why I stuck to brand name machines for my personal collection of computers - my DX/4 is a SIemens Nixdorf, and the cache is real ;)

[@ILRBW]

I use a few drops of not very liquid silicone oil when temporarily installing radiators. The oil provides thermal conductivity and is easily wiped off afterwards.

[@AndrewFremantle]

Fascinating. I don't remember seeing/hearing anything about fake chips back in the day. I remember chips that *weren't very good*, like Cyrix CPUs and such. But nothing about outright fake chips. The first such I remember hearing about were AMD Durons that were altered to run and report as higher-clocked models.

[@M0UAW_IO83]

Ooof, the Cyrix chips were pretty good but they were cheap so they often got installed on really crappy boards with fake cache

[@sierraboney1394]

@@M0UAW_IO83 Yeah I had a board with what, years later, I found out to be empty WriteBack cache chips, with a DX2/50 - one of those 'It's ST' cpu's - which came up as a Cyrix on bootup. Was somewhat slower than other equivalent pc's I saw at the time!

[@BrianG61UK]

@@M0UAW_IO83 I remember getting a Cyrix 387 equivalent FPU that knocked the socks off the Intel 387 FPU. I think was a short lived win for Cyrix though, because the Intel 486DX was way faster still, and Cyrix couldn't beat it.

[@Lilithe]

I had the exact same thought. They probably took the stock BIOS, searched for "None" and replaced it with "256K" and put the empty chips in the board. I would upload a copy of it to the Retro Web as-is, and a copy where you search for 256K in the file, replace it with "None", test it, and upload it too.

[@bitsundbolts]

The original BIOS is already on The Retro Web with the appropriate note. I will try my best to learn how to undo the modification.

[@Arti9m]

Ha, you are lucky because you at least got some free copper in your fake chips 😃 I have fake chips that have absolutely nothing in them, not a single piece of metal. Also some infamous PC Chips boards have a BIOS that is modified in a much more complex way. Instead of just showing None/256K, when you have no/fake cache and a cache-related option enabled in BIOS, it shows a unique boot message that says "WRITE BACK CACHE ON" or something along those lines. Fortunately, the BIOS still works correctly with reals chips and even displays proper cache size (up to 1M). The model I'm talking about is M915i.

[@bitsundbolts]

Interesting! So what, your cache chips have plastic pins? I guess my chips are one of the better (as in more expensive) fakes then. I have an M919 that needs to be repaired (CPU inserted in wrong orientation). I haven't figured out yet what the problem is. But I do have the cache memory that looks like a COASt module. But that might be fake too - who knows

[@Arti9m]

@@bitsundbolts Oh no, the pins were metal, but they were barely inserted into the plastic and then there's nothing. In fact those "chips" were so bad they melted completely under a heat gun. As for M919, (to everyone who's reading this) plz never put regular COASt modules in there, also plz don't put modules with 3.3V SRAM chips if you use a 5V CPU. It will result in permanent damage.

[@SidneyCritic]

On CuriousMarc they just heated the chips in a frying pan and chipped the plastic encasement off to see the die inside. A sharp wood chisel on the parting line above the legs should work.

[@bitsundbolts]

I will open one of those chips. I don't have the proper tools, but I might just go at it for about 10 minutes with sanding paper...

[@ocsrc]

Those were very common boards. I had one of these and saw many of them in custom made machines.

[@pb_magnet]

Is it possible to install an amount of cache other than 256 or zero? It would be interesting to see if the BIOS is hacked to always display "256K" or would accurately enumerate other amounts.

[@bitsundbolts]

I should be able to get 128K by using half the chips. But from what I understood from someone else that checked the BIOS file, it will continue to report 256K.

[@dmwzr]

The fact there is no download on retroweb could be cause all of this versions are compromised. Like some scammers used to modify a bioses of the boards to sell em with fake chips (most likely as an assembled PC builds) and used to modify a version number/date so no one would guess to update it and fix the cache situation uncovering a scam. Maybe there is really no update.

[@CRSolarice]

Sounds reasonable. Though updating BIOS back then could be less than a smo0oth sailing experience, unless you have done it a few times. Most people didn't even bother to update bios or firmware because it could be risky due to software, hardware and procedural flaws. Nevertheless that would be a tactically sound method of covering the tracks and keeping them covered; until a user or developer who actually used the cache for a certain specific application noticed the difference, perhaps.

[@Ale.K7]

I'm the proud owner of a couple of PC-Chips M919s with fake cache chips soldered to the board 🙂. I got them "recently" (~15 years ago) as "trash", so I can't complain 🙂.

[@ronny332]

Years back I had, maybe still have, a MoBo from PC-Chips with full populated cache chips, but a view from the bottom showed, no leg was connected to anything. I thought it has to be a multi layer board, but even running running diode tests with my multi-meter showed not one connection was existent. Seems to be "normal" to fool people, even back in the day 🙂

[@damianblanchard6700]

I wonder if my 486 DX2 ever had any cache 🙈. It had an award bios…..version unknown. This is why we like KZbin consumer test channels….it keeps the vendors honest, or at least highlights what vendors do

[@datasoftinc]

Georg Schnurer is a legend. He worked at the c't Magazine.

[@SianaGearz]

Worked? Works! He carries vaguely half the magazine on his back basically now, 30 years later.

[@davidgari3240]

I saw the second video first, where you modified the BIOS for some reason. This explains my confusion.

[@neozeed8139]

Wish Id never heard of these bad chips, they were such a plague in the day! It's kind of why coast modules were far more reputable/reliable. After the PentiumPro, and integrating stuff onto the chip it was such a massive plus. Back in the day I did get a cheap 486sx since it had fake cache, it was unstable as heck, but dumping the chips it ran great.

[@bitsundbolts]

Sorry you had to deal with fake cache back then. I agree, integrating the cache into the CPU stopped this method of cheating people. Then they started to fake entire CPUs... Even "new" (but fake) Ryzen CPUs are being sold these days with no die under the heat spreader.

[@neozeed8139]

@@bitsundbolts I had wondered if someone was going to start making ceramic slugs. I've bought motherboards and graphics cards that have mobile parts on them, but they don't try to pass them off as the real deal, but wow. I guess it's inevitable. that scammers got to scam.

[@eformance]

The replicated copyright comes from the time when you had an ODD and EVEN BIOS chip, so the complete copyright appears in each chip image.

[@foxdavion6865]

yeah around the early 00s, fake chips and bad caps seem to be rampant

[@HandFromCoffin]

I'm like 100% sure I built a system in high school with these chips.

[@xyxy1024]

I found some cache chips on a 486 mainboard that I knew were real because I used them as memory in a 8088 project ;)

[@Yrouel86]

I stock up on 256k cache chips from ebay recently and so far of the 50 I ordered I found only 2 bad ones. Unfortunately I remembered too late I could test them with my programmer (TL866CS) so I wasted a bit of time troubleshooting the board after installing all the chips... Seems fake chips were very typical back then, some motherboards with soldered ones didn't even have full packages and they literally had a hole on the bottom and completely bogus markings, for the socketed ones it seems they put a bit more effort to sell the lie. Also in some motherboards with soldered fake SMD chips you can literally see the traces going nowhere but in a loop between the fake chips

[@bitsundbolts]

I had a little less luck with my order. Still, I have enough cache chips for experiments. I do agree, the chips that came on this board look real - they put real effort into it!

[@tylern6420]

This is just insane levels of trolling

[@Constantin314]

well...that was interesting (imagine it said in Seinfeld's voice). those RAM modules that you made look sick

[@galen__]

I remember this era, where even PSUs could have brick material inside to make it seem heavy and higher quality 😂

[@bitsundbolts]

Haha, crazy! They still sell USB chargers with extra weight to make them feel "more premium".

[@tigheklory]

Thanks for this video!! Back around this time (1994) I bought a AMD 386 DX 40mhz motherboard at a computer fair and just buying 1mb of RAM was so expensive it cost more than the motherboard (CPU soldered on), Cache, OPL sound card, graphics card combined! It performed quite well, better than most 386s I see people using on youtube these days. I was able to play Doom on it at a decent frame rate with only a slightly downsized window. My friend had a 486 30mhz which played Doom slower, I wonder if the cache on his was fake?

[@tigheklory]

I meant 33mhz not 30mhz. LOL.

[@lezbriddon]

OMG I had that MB, look at those expansion slots, why cant we have that today for decent expansion. i'm drooling...

[@martinebers8094]

Around 1994 we had a 486DX system that claimed to have 256K cache on-board. We wanted to use it with OS/2. Back then one had to deactivate all caches, deactivate the turbo swich etc to install the operating system (though after installation you could re-enable everything). Afterwards we could re-enable the internal cache, but not the external cache, as the board always complained about "Cache memory bad, do not enable cache". We used it for several years afterward but could never get the external cache working again. However, I could not see any significant performance hit, neither with productivity software nor within games (even when using plain DOS). I think it was a UL486 board. Wonder if the BIOS was modified or if it was legit, but the chips were not/it had been coerced into thinking the chips were legit, or if the system had another issue.

[@ruxandy]

At least you don't have fake cache sockets (with fake traces that go nowhere). Those were even MORE despicable! 😀

[@bitsundbolts]

Somebody was sitting there and designing those traces... Crazy if you think about this.

[@drPeidos]

Great topic. If you ever end up correcting your BIOS so that it correctly detects cache, will you upload your hack to the retro web?

[@bitsundbolts]

Yes, of course! The BIOS of this board is on The Retro Web already (for anyone who wants to take a look themselves). But I'll try to reverse whatever those people have done to this BIOS.

[@williamjones7821]

Years ago, I ordered some 4 Megabyte 30-pin SIMMS (16 megs of RAM cost me $500). They had "fake parity". Packard Bell, at least in the USA, created this "fake parity" RAM. It was not compatible with my existing "real parity" RAM, so I had to return it and get my money back. Several years later, I upgraded to a AMD K5-166 motherboard.

[@bitsundbolts]

Oh, fake parity chips are new to me! Interesting. Thanks for sharing

[@VladislavKusmin]

In fact, these chips were the world's first attempt to download RAM / Cache

[@jesusortega8927]

I have a 486 board that had fake cache chips installed and modified BIOS too. I replaced the cache chips and now I have working L2 cache, but I never updated the BIOS. It seems that the amount of cache shown and the option to switch on/off external cache has been modified, but the cache itself works correctly when the appropiate chips are installed 😅

[@bitsundbolts]

Oh, I did not test the BIOS setting to switch off the cache. That would be interesting to test how well the BIOS hack was implemented. But I have a feeling that it might report correctly as NONE when "External Cache" is turned off in the BIOS since it probably accesses a different routine in the BIOS.

[@jrherita]

2.01r. R is for ripoff :). This is a really fascinating video. Still watching - thanks for posting this!

[@bitsundbolts]

Hehe. Glad you like the video!

[@JimFeig]

Funny how the main memory access time hasn't improved all that much over the years.

[@raymitchell9736]

At this point you have nothing to lose cracking one of those chips open and looking at it under a microscope. I suspect that this was a scam because in the 90's RAM memory was in such demand we had shortages of it... In fact there were companies here in Silicon valley that had gangs breaking in to steal memory at gun point... terrorizing the workers, and trucks oversees were hijacked by pirates from foundries. The fact that you have chips that are completely blank means that this was some kind of inside job. This just seems coordinated and organized. I'll bet I had one of those boards too! To think I was taken in by these hi-tech scammers!

[@bitsundbolts]

Wow, that sounds terrible! I did not expect that people would go that far and break into factories and steal memory at gun point. I'm sure mine are completely empty - just the housing with legs. But I shall open one soon!

[@raymitchell9736]

@@bitsundbolts Yeah, the RAM shortage was real crisis, it put a crimp in a lot of computer manufacturing plans like the Atari with their ST and TT machines, Commodore Amiga, not just the IBM PC's, etc. Again, I can't get over that it was done in such a professional way as to lie about the Cache memory and to make the fake chips, and it is kind of weird that you'd catch it decades later, that's some good detective work! I don't think I would have given it second thought... Just WOW!

[@peterdefrankrijker]

I used to work in a reputable Amsterdam PC company, from 93 to 2000. We made the fastest computers in the country. Except… the models we sent in to the magazines to be tested had special, extra fast cache chips that were different from the ones installed in the PCs that we sold for $$$. Same with the memory boards. And the hard drives, those came in special delivery from WD and were some 50% faster than the regular ones. Oh and by the way… every PC builder did this.

[@bitsundbolts]

Oh, I guess this works until some independent entity buys the product off the shelf and verifies their test results with that reported by magazines. I do remember the occasional newsline popping up where magazines were sent special models that would never be available to the average consumer.

[@herrbonk3635]

1:43 None of these are specifically "cache chip", they are all plain static RAMs (with ca 15-20ns access time).

[@froller]

9:15 This copyright message looks so strange because of some motherboards had 2 ROM chips. One for odd bytes and another one for even ones. This was made to connect 2 8-bit ROMs to 16-bit data bus. Both dumps started from same copyright message. So when they are combined in single ROM image you see duplicated characters.

[@petrkubena]

It could be interesting drilling them and looking what's inside of those fake chips.

[@bitsundbolts]

Way ahead of you :D

[@TommyCrosby]

So, you can download more cache with a BIOS file, amazing!

[@Sigmatechnica]

Nice discovery. I wonder if it would be possible to build an interposer board to use modurn sram chips and give it an ungodly amount of cache :D

[@bitsundbolts]

I think the cache is probably limited by the board. I think it won't go beyond 256k.

[@VladoT]

Having longer sockets for some of the cache chips on this board I assume that it could take 512Mbit chips and have more than 256KB of cache?

[@bitsundbolts]

In theory, yes. But I haven't checked.

[@OzzFan1000]

I have an Asus board I bought where I suspect the BIOS was modified and fake cache chips are being used. I set the project aside several months ago to work on other projects because I couldn't figure out why all tests showed only one cache, including CACEHCHK and SpeedSys from the Phil's DOS Benchmark Pack. With this video, I'm curious if I am running in to a similar situation like what you experienced here. Unfortunately I don't have a chip tester or EPROM burner to troubleshoot.

[@bitsundbolts]

You should be able to use a multimeter (if you have one). It just confirms if the legs on the chips connect to something internally. Otherwise, having a strange behavior in CACHECHK or SpeedSys may be an indicator, but it could be that the cache is just faulty. I think the best method would be a multimeter.

[@RWBHere]

I've owned 486 and Pentium boards with soldered on board 'cache chips' with Copper traces going nowhere. A whole corner of the board was doing nothing electrically.

[@bitsundbolts]

I have one of those boards, but it is a socket 3 board for 486 CPUs.

[@briankleinschmidt3664]

Memory was very expensive in those days. The subterfuge did not go unnoticed, but 20 something college students would not notice. It makes me wonder how they're pulling off the scam nowadays. My laptop says it's an i7, but it could be an i3, and I wouldn't know it.

[@bitsundbolts]

The most recent scam I've seen was a new Ryzen without the chips inside - just plastic and the heat spreader. I am sure there are plenty of scams going on if money can be made.

[@digitaltoaster]

fascinating piece of history.

[@janbrittenson210]

The real cache chips are plain SRAMs in a narrow DIP package; they're all pin compatible (otherwise they wouldn't work in the first place) so you can specify a pin compatible chip of similar speed grade in your tester, and it'll test it just fine. The fact that nothing works pretty much tells you right away that the ones you bought are fake.

[@bitsundbolts]

Well, I didn't buy the fake chips. They came on the motherboard I found at the scrapyard.

[@HeadsetGuy]

If you run Speedsys with the fake cache installed, it will say "The system appears to only have 1 cache??" or something like that.

[@bitsundbolts]

Yes, I think this is what is going to happen - but it might be CACHECHK that has this wording.

[@HeadsetGuy]

@@bitsundbolts You might be right.

[@spirod2657]

I used to assemble PC'S during this era. To thicken the plot further. Real cache chips where so extremely static sensitive that removing cache chips was a huge no no ie how would you know. Especially when they shipped preinstalled. The optional cache chips bought separately did seem more resilient. But always nerve racking. All new PC'S needed to be burnt in tested. Further, Ironically a low failure rate on cache ie systems not crashing was a tell tail sign.

[@rexcameron414]

I have a fake Cache board from the early 90's, keep it as a memento. This con was well known at the time, I guess it's an age thing.

[@batlin]

I really wanted to see inside those fake "chips". Would be good to do a cross-section with a Dremel or other cutting tool, just to prove what is almost certainly the case: the pins go to nothing inside the casing...

[@InsaneWayne355]

Fake cache was quite common at that time. I feel like everybody knows that nowadays.