Replacing fake cache in the 486 build
ฝัง
- เผยแพร่เมื่อ 10 ก.ย. 2024
- Following a comment on the build video for this system that suggested the L2 cache might be fake I just had to test it and yep its fake!
After a bit of googling though it seems that it should be possible to fit real cache on this motherboard so this week thats the job all be it things never work out as simple as you'd like.
I try fitting 1024k and end up with 512k, maybe...
Music in this episode taken from the game Transport Tycoon Deluxe.
I didn't even know fake cache chips were a thing until this :)
Yeah I didn't know about it either until the comment on my other video. I got this system from a former system builder who was buying these boards assuming it had 256k of cache. I must tell him it was fake.
Unfortunately they were REALLY common on inexpensive boards in the day. I got burned twice in the 90's.
Fake chips, crazy stuff!
Why even bother fitting fake chips. Just leave the positions unpopulated.
@@simontay4851 To scam people and make them think you paid for a mobo with L2 cache. If you got a board with no cache chips, you'd quickly notice before installing the board into your machine. I bet a lot of people wouldn't notice the lack of L2 cache anyway, so they probably got away with this shady business most of the time.
"It seems that 256k is probably the sweet spot for a board like this. So, with that in mind, let's fit 1Mb because... why not?" :-) That's the spirit!
I just hope a bios update will let me use all of it.
It's not about sense, it's about bragging rights.
Pricing pressures at the low end of the market in the mid 90s were pretty intense. I remember working at a computer store and a customer walking in and bragging that he knew where to get 486 motherboards for $50. It's a good bet those were PC Chips or similar boards. The fake cache chips just helped make those too-good-to-be-true boards seem even better.
Just goes to show that there is always those at the bottom end of the market that will always try to scam a sell. At least it's possibly to fit the real chips to this board.
pccchips was white van speaker scam of motherboards ;-)
those pc chips boards were real garbage... I remember them well along with the paladium drives
The redhill guide has an article about motherboards with fake cache chips. It is very obvious to customers that something is missing on a board, and therefore that it is straight up inferior next to a board with cache chips. So PC-Chips knew that just by adding dummy cache chips (always soldered on so people would be less likely to find out) theyd attract more customers than if they just left the sockets blank.
They also have a section about the generic chipset on these boards as well.
Hah, that brought back memories, I saw these boards when they were new, they were utter garbage and had such a high failure rate that it's almost a miracle you've found a working one.
Big clue to them being fake chips is that they're soldered in, real cache chips were expensive (which is the reason for the fakes, they sold these boards at a similar price to ones with real chips) so it was almost always socketed.
There were variants of the 'Writeback' scam that used SMD chips, they were even worse as the board didn't even have the tracks to connect the 'chips'.
On those 'blobs' the plastic feels and looks wrong, the markings on the chips are gibberish, 'Writeback' is just crap, I remember one of the wholesalers of the time explaining to me that 'Writeback' meant it used an area of system RAM as cache, completely missing the point of having high speed cache RAM.
A lot of boards were fitted with fake chips and you could just pull them, replace them with real, change a jumper or two and they'd work but there were also a good number of variants where the BIOS was hacked to lie about the cache, you could usually fix that if you could find an unmodded BIOS, but you were SOL if you got conned with one of the boards that didn't even have the tracks to the cache chips routed and connected.
I could name names but at least one of the companies responsible for importing these boards (and plenty of other scams besides) back in the day is still in business and known to be litigous with *very* deep pockets.
You get diminishing returns if you fit more than the optimum amount of cache and there are definitely 486 boards out there that just don't have enough address lines routed to the cache to use more than 256KB or 512KB but it's dirt cheap these days so fit whatever you got.
I have a new bios on its way to me and hopefully that'll sort the issue of me only accessing 512k but I suppose it could also be lack of address lines or just something borked and hense why this particular board got fake chips.
Thanks for the detailed comment. Love reading about the history of these things.
Yes i remember that if they were soldered in they were fake, and the chips looked like bits of plastic. I got caught out once! Well twice actually with cache i bought a PC that didnt have them populated for work once, complained at the poor performance.... Eventually fitted some from a broken machine. I also bought a cpu and board combo over the phone ( in the 90s pre internet) that when it arrived it didnt work and the retailer had to send me a cpu that would work with the write back cache board they had sold me. About 6 months later i sussed it was fake cache, and resold the board... And then onwards ( in the 90s) bought my stuff from the computer fairs at TCR where not only could you see what you were buying but they also tested it in front of you! These days i dont buy anything new I'm not a gamer, and Xeon's are so cheap so build yourself a PC with recycled server parts and a new ssd works for me.
the Transport Tycoon music gives me flashbacks and kicks my PTSD from this game. We had played it to much back then.
Just fell down this rabbit hole and stumbled accross this video.
Vogons is a treasure trove of experience, knowledge, info, kluges, hacks, work arounds, and fixes.
cachechk and similar programs detect the cache size based on timing memory access only. I think what happens here is that your L2 cache is so close to your L1 cache's performance that it sees it as one large chunk of cache. You could try to increase wait states on the L2 to artificially slow it down. Maybe then cachechk will be able to distinguish the two.
Thanks for the tip, I'll give it a go and see if it makes any difference.
You can see at @19:05
L1 20us/KB latency
L2 24us/KB latency
RAM 32us/KB latency
so its picking up cache just fine
@@rasz You are right, it measures it correctly. But that doesn't mean that it distinguishes it based on these numbers. It seems to me as if it sees 20us and 24us as being too close to each other so that it merges both caches into one in the final evaluation.
You know, you're not supposed to permanently bend the legs to fit in pcb...
You use a gripper that pushes on the pins about half way down enough to allow them to be inserted in board. When gripper releases they stay in place as the legs splay back out.
Originally this was the robotic installation method before entire board was wave-soldered. Very neat as it lets the chips stay in place by themselves while you solder.
You're probably correct but I'll just bend the legs 😉.
You can burn the new bios on the board itself. Boot it with the old bios and then hotswap the bios chip and perform a bios update.
I've done this many of times in the old days.
EEPROM programmers can be used for more than programming, they can also be used to check memory, and other chips, are functioning.
Yet another reason why I really need to get one. Would be useful to verify if all my cache chips are working or not.
Came for the build, subscribed for the lovely accent. And the build.
Here is a trick you can use that I learned many years ago... You can fire up another 486 and carefully remove the bios after the board is booted to a prompt and use a different firmware utility to pop in another chip and program it. I have used P1-P3 boards to do the same trick. Had to do this for several boards when we didn't have access to the correct tools. Worked like a champ and continued to do that for years until we picked up a needhams emp10. Good old days...
Very instructive video. I rewatched it after I recently got two infamous PCChips MOBOs: the M912 and the M915.
My M912 was exactly like yours with the UMC stickers but with the difference that only *half* the cache was fake! Both banks were filled with chips marked "TM tech" though bank 1 markings were cruder. And indeed, the jumpers were soldered for 128K instead of the expected 256K that was as well indicated by the BIOS. Also Cachechk detected only 128K.
I guess I'll try to desolder the bank 1 chips to replace them and patch the BIOS as you did.
The M915 banks are filled with the same "write-back" chips as yours so this time, all of them are probably fake. I still need to test that board though.
It's amazing to see how PCChips cheated the market at the time, with boards that had sometimes full working , half working or even not working at all cache!
Incredible how they put fake cache chips on there. And the fact they are hollow like that. LOL
Yeah I thought there might be something in there perhaps tricking the machine into seeing 256k of cache but nope just plastic.
Their Pentium era fakes were even better, with chips routing tracks back to themselves, not even pretending to go to the chipset ;-)
@@rasz Yes, I did see some videos of that aswell. Crazy stuff. I think I heard about this stuff back in the day aswell
And this is why we avoided PC-CHIPS motherboards when new, among other things such as crappy quality.
Used to be all I could afford... Anyway, how long does it take to reboot after a crash? Like I said, i could afford that level of performance.
Well, after all these years, they are still working and many of these board came with a cr2032 battery instead of the usual barrel battery of the time.
Oh man, what a crock. Those damn plastic ACME chips show that fake chip shenanigans aren't a new thing. Like you said, that sticker over the big chip definitely looks suspect as all hell too. Hope to see more on this one if you're ever able to figure it out.
If I can get the bios updated I hope it'll sort the 1024k cache problems. Beyond that though you will be seeing more of this system with a VLB io card and sound card upgrade in the works.
ACME means top of the line for a product. Due to the Roadrunner cartoons people believed it meant a crap product. The joke was the coyote had the best of the best, but was still outsmarted by the roadrunner.
"Who the hell put real cache chips in my fake cache slots?!" -- BIOS
I'd forgotten about the fake cache scams of the 1990s until this video came up as a suggestion. And a quick search seems to suggest there aren't any other videos about this topic.
A place I worked got a cheap local company to supply them with PCs, and we later found a number of them had fake cache chips. They were less convincing than the chips on the board in the video though. The package size looked right, but they only had legs at either end! The bios had also been modified to give a misleading "256K" message at boot time that didn't actually say it was cache.
A friend bought a board from a magazine ad and found it had fake cache, when he complained the company claimed the board allocated some of main memory as cache memory...
It'll be interesting to see if you get anywhere with this board, but I feel it's a lost cause.
Thanks for your detailed comment. Always fun reading about other people's history with such things.
The excuse of it uses part of your memory is a bit lame, suppose it was just a way of trying to riggle out of them obviously knowing it was fake.
I have a replacement bios on route so will be interesting to see if that helps.
Hey, nice videol. 512 kb is quite enough I believe. Congrats on the upgrade and video.
Thanks and while 512 is more than enough I did eventually get the full 1024 working with a bios update.
IIRC there were some boards with fake caches that weren't even empty IC packages. They were just wire links overmoulded with plastic to look like ICs to the uninformed. The giveaway was the leads being round not flat and one presumes the board was specially designed for them with no connections to the fake chips.
At least here I have the option for real cache. Hoping the updated bios will improve things further.
I remember hearing about the fake cache chips "back in the day" but this is the first video I've seen covering them. What a crazy thing! Top detective work there. 👍
Also - unrelated, but what a lovely little board. 30-pin and 72-pin SIMMs together!
I was surprised by just how fake they are which might sound stupid but I expected perhaps something in them, other than plastic.
The board supports either memory type just not both together. The 72pin SIMMs are faster though, at least I think they are.
@@CRG Yeah, I wasn't expecting them to be completely hollow! I'd have thought they'd just put much cheaper chips with the same footprint in their place and then not physically connect them up to anything. Hopefully once you find someone who can flash the BIOS you'll be able to get the new (real) cache chips working properly.
Pcchips and the cache scam indeed. I remember.
Very helpful & interesting video. Recently bought an unbranded board on Ebay. Never knew about the fake cache thing, until I went on The Retro Web to identify it...That's when I saw in the "Known Issues", fake cache was listed. That sent me down a rabbit hole, but looks like I got lucky to find one with real cache. The chips are socketed, which is the biggest positive indicator, but will not know for sure, until I have a chance to pull one of the chips and/or get the rest of the parts in so I can benchmark it. Cheers.
Duke 3D was a bit of a dog on the DX2's. It definitely benefited from the additional speed of the DX4 and 5x86 processors.
It doesn't run bad on this system but yes it would be a lot better on the DX4 or later processors. EDIT: and just in case you're wondering I did get the bios updated which got the full 1mb of cache working.
Just found your channel recently, catching up with your pursuits. I have a similar motherboard as this one, the Pci version (M915i). What a quirky peculiar board it is. I was unable to get 1M to run as well, only 512. Mine had real cache chips with sockets luckily. From the day I started fiddling with it I had all kinds of strange behaviour, required multiple boots for it to recognize new hardware, That cache check program never recognized my cache always gives me the "fake cache on L2" warning. I was convinced that the ebay chips I got were defective ( glad that I didn't accuse the seller before I did more testing). Now the board sits in a box because the last attempt I made it somehow destroyed 2 floppy drives. The first one that stopped working was the most confusing, because that was the last thing I would have imagined to happen. I'm guessing my Bios chip must be corrupt since I did a complete recapping of the motherboard. Hopefully you can find a Bios chip. Great work.
fake cache, that's a trick isn't it! I think that 3 pin header runs in the other order (321) but I'm sure you checked that out.
I'll go check that now....
Yeap I was going to say JP10 is the issue that the pins are 3.2.1
Since that was 3 weeks ago... was that the issue? Nevermind I see this was answered elsewhere.
It turned out to be the bios that needed upgrading.
as for bios updating idea. i had once very crazy way to use a pentium 3 vintage motherboard to program eeproms and i did few other things like programing oryginal bios to a pentium 2 fried board with the little help of bios chip borrowed from a different board with the same chipset and an auto executing floppy, because keyboard controllers of those boards were different
so i would dare to assume any board with flashing support and the same pinout flash chip would work for you
you just boot up the board to dos as normal, swap the chip in the socket and program it with your bin rom. you simply have to override write protection to of unsupported bios by forcing the flashing operation. later you simply turn off the computer and swap bios chips again as they were and you have your programmed rom chip outside the socket ready to put in another computer
fun fact is it also work with serial flash rom chips in newer boards if it is in the socket like on early am3+ board i resurected few of those this exact way at work
the key is to have a board where you can override mismaching or unknown flash roms
That sounds like an interesting method but I've since got a flashed chip for this board. I do want to get an eprom writer though as it'll be useful for other projects too.
The joys of a PCChips board.
did you try changing that jumper? I forget which one it was, it was 3 pins and you made a guess as to which was keyed as 1. maybe it was the other way around?
The issue was eventually resolved with a bios update.
Did you ever fix this issue?
I noticed in your bios settings you had it set to write through but mentioned the ram is write back.....
If it's still not working properly that may be the problem..
Yes I got the issue fixed, it just needed a bios update.
At least on this model the vias of the cache chips actualy are connected to traces. I remember reading that in some models the fake chips weren't 'in-circuit'.
Yes there are some other fake cache motherboards on which the traces just loop round on themselves going nowhere. Thankfully not the case on this board.
Memtest86 was (and still is) an excellent tool for checking memory and cache. Also, change the cache scheme to WT instead of WB
Thanks I'll take a look at memtest. I've been playing with settings in the bios and yep changed back to WT rather than WB (well spotted)
Pretty sure the M912 only has WB support for the internal L1 cache and not the L2, at least that’s the bios setting in my real cache board.
Didn't some of the later 486's introduce write back or something which sped up performance with cache? But I only remember seeing that on Intel 486DX4-100's. Shocking that they put fake chips into these boards like that!
I think your correct as I've certainly seen versions of the AMD DX4-100 with write thru and write back.
Thing I don't get about it is that it probably would have been cheaper to build the board with sockets than fake chips, that said though I suppose a cheap motherboard with supposed cache is tempting to those that just wouldn't know better. I know I'd rather of had sockets.
Even the Intel DX2-66 has a write-back model: SX955. The performance gain is not that big, though...
Very nice video ! Actually those motherboard with fake cache also had a modified BIOS to give the illusion that some L2 cache is installed. I really think you need a BIOS replacement to make your L2 cache active. Those differences you see in benchmarks are way too low to indicate that your cache is enable. I think it's NOT. And the cachechk program is generally accurate. If it says that you have no L2 cache, then you can believe it ! :)
Take a look at the "bios and a gus for the 486" video but yep a bios update sorted the issue. All 1MB is now enabled and usable.
I think you installed all 9 chips as 128kX8, but on position U20 (The TAGram) it must be a 64x8 chip used. The cache status flags are therefor corrupted, so the test software thinks right, that the the cache adresses are valid, but it does not know what data are changed and what not, as the U20 is trashing data. I am too late with my comment, maybe you already fixed this issue
e, but YT algorithm showed me your video just now.
I did get the issue sorted by updating the bios. All 1mb is now available. My guess was that the patched bios for the fake cache was causing the problem but I suppose it might have been the tagram and perhaps a bios update fixed that. Still using the 9 128k chips though.
Just in case you don't know, you CAN use so called "sip collet sockets" for chips if they have very rare form-factor for firm socket beds for them.
Actually i use those sip collet sockets for most of dip chips. One important note - these sockets does not overlap silkscreen on the board, so notch is clearly visible.
I'm so glad we dont have to do this crap any more.... not seen those dip sockets myself for a long time probably the late 90s!
A TL866 together with cheap chinese eprom eraser are a really good investment for retro tinkering ;-)
Yep I keep meaning to buy one but usually another piece of retro hardware catches my eye...
Do need one now but it might have wait until after Christmas.
Great video, yeah problems problems when it comes to 486 cache chips, been there, yeah bios some times gives lots of hassle .
Where did you buy the 1MB cache chips by the way ?,
cheers , Brian
The cache chips came from ebay. Just searched for the cache chips and the 1MB set was actually the cheapest :)
Did you try hitting the turbo button, I had the same issue on my 486 board.
The PC Chips 486 boards are collectible specifically because of the fake cache imo.
Cachechk is not seeing a big enough difference between the 20 at 1,2,4 and 8k and the 24 from 16k to 512k, so its dismissing the 8k as tolerance.
Also, for boards that support a flashable BIOS but have a ROM fitted, the nerve wracking way is to make sure BIOS is shadowed, and then hot swap a flash chip in and write it
I do get there in the end with a new bios written by a friend. If you look for "the new bios and a gus" video it shows the new chip going in and that solves the problem allowing access to the full 1MB.
Not sure id fancy hot swapping a bios chip, that'd be way to nerve wrecking for me haha.
Loved the video, especially when you found the hollow chips! Classic! Did you ever get a new bios onto another EPROM?
Thanks and yes I got a new EPROM which fixed the cache issue, all 1024 now working.
Yes, just watched the sequel. Well done! What a bunch of crooks PCCHIPs were!!
I have a Shuttle 486 board, very much like this one but w/o external cache. I ordered a tag ram to test the compatibility and placed the 24 pin chip too far forward in the 32 pin socket :( took me months to find out why the board stopped booting. Thought I fried it. I like that you got fake ram you can actually remove. I've seen some just soldered or glued to boards w/o even any real traces!
Always need to be super careful inserting parts in these old boards cause there many things can be installed incorrectly. I just got lucky on this particular board and its fake cache. Seemingly there was a version sold with real cache chips so it was easy to sort, well, once I got the new bios in there also.
They could have put empty REAL sockets, instead of FAKE soldered plastic "chips". What a nonsense. Nice build absolutely, only needing to find a VLB hd controller. Cheers, M
Way ahead of you with a vlb controller sitting waiting to drop into this rig along with a nice sound card upgrade.
That wouldn't be very convincing when trying to sell a board advertised as having "real" L2 cache. The reason they (some shady Asian companies) did this was to trick people into thinking they got real cache, since L2 cache was the new hot thing back then.
I seem to remember the cost of an extra 512kb being at least $100 from the store. So faking chips nets them as massive profit.
it wasnt "some", it was PCChips exclusively. This one is PCChips M912 with fake cache and chipset (stickers because it either didnt pass quality and was thrown away at the factory, or pcchips bought bare dies and did packaging/encapsulation themselves).
They were rewarded for their efforts with ECS buyout. As I wrote under 'The Matrox Mystique 1996 Build' 2 months ago, which coincidently also used PCChips board ;) :
"M571 was one of the rare not totally terrible boards made by them, with surprisingly non fake cache :P. At least as long as you avoided chipset buildin SIS graphics. Still it might of been a long con, simply grooming suckers for the next iteration of their scams. Slot 1 line with gems like M717, M747, 748, 751, and all the rest of the Ali/Sis based "BXCel" "BXPro" garbage. Glad its all history now. ECS merged with PCChips and ended scams/non working products. Nvidia bought out Ali. Sis seems to have vanished around 2006 on its own."
At least M912 has real cache routing, they did manufacture models with fake plastic shells routed back to itself, like hilarious M919 :---)
@@rasz you'd never guess I got both boards from the same former system builder.
Found your channel from these 486 vids. Loving it mate. I sold an m912 recently but that did have real cache in sockets on it!
Also it seems you love the TTD soundtrack as much as I do... Recently got my hands on an SW60XG midi board prettymuch just to listen to TTD (While playing ofc :) )
Thanks, that's greatly appreciated.
Hard to beat the TTD sound track. I've been playing that game since the 90s. Can't decide if I prefer the GUS or sound blaster version though. Just so familiar with how it sounds on the sound blaster.
@@CRG My nostalgia hits for the original SB/Adlib that I had in the 90s but it is incredible to hear through a wavetable/Midi device!
Yay, PC-CHIPS :D Good times!
My second machine was an Am486 DX-4 120 with a pc-chips board that had fake cache. Disgraceful practice.
@@cellsplicer2008 Very true!
you mean PC SH_I_T 😒😂😂😒
@@pro5p3c7or1 i had good luck with pcchips, for me.
Maybe those pins (on JP10) aren't 1,2,3 but 3,2,1 ?
Tried it but no difference. The new bios chip should be here at the start of the week then I'll get an update video done.
The M912 v1.4 isn’t a bad board when it comes with the factory installed 256kb of real cache. Found my Fugu Tech branded one for $25 with ram and CPU and it’s very stable with an Intel DX2 66. Only thing is it doesn’t have an option to enable WB L2 cache, only WB on internal L1 cache.
32 pin DIL sockets are easy to get on ebay. Literally dozens of listings. Or Amazon if you prefer!
I did have a look at the time but didn't see any that were readily available but noted for future reference. I really do need to order myself a random selection of sockets for these types of projects.
I remember this from back in the mid 90's. Had to be very careful not to buy fake cache. Cache was expensive back then, especially if it had good timings...scammers made a killing selling fake chips to unsuspecting buyers. Also, the speed increase should be double at least compared to having no cache. Maybe something else is wrong with the motherboard.
what i always wondered was why socket 3 has that estra row of pins around my amd 486 133mhz and same for your board
pentium support?
one more question
was that transport tycon soundtrack during soldering?
The PC Chips motherboard supports the various AMD and Cyrix 5x86 processors. I'm not sure if it supports Pentium chips, maybe the very early 60 and 66mhz chips.
I'll be keeping this system as a 486 though. I've been keeping my eye out for a faster processor. The DX2 66 is a nice chip but I think a nice DX4 100 would help unleash the potential of this rig better.
Oh and yes that was the transport tycoon soundtrack. I spent a lot of time with that game, still play it from time to time. Love the soundtrack.
Some 486s ive had used one of the cache chips as a tag module that indicates how much is installed.
I was able to get all of the 1MB working with a bios update. Should you be interested there is another video showing that titled bios update and gus for the 486.
I have almost the same Motherboard PCChips M912 but V1.4 with (factory) real L2 Cache Chips and an Award BIOS instead of an AMI. The cache chips are even socketed (256K ;-)). Mine has a Sticker on the Chipset, too. But it's labeled "PCChips CHIP 18".
I used to have a Pentium 1 socket 7 board with 256K of L2 cache in a dedicated slot. The system summary reported the L2 cache as pipeline burst. What is the difference between pipeline burst and write back cache.
Those are orthogonal.
Logical:
Write thru means Ram Writes write to both ram and cache at the same time and are as slow as Ram access.
Write back means Ram Writes write only to cache, write to Ram is performed when cache is flushed. In essence you get faster memory writes on average.
Electrical:
Asynchronous cache has set delay independent of cpu clock. 386-486 era.
Synchronous cache has delay counted in cpu clock cycles. Pentium era.
Burst cache means chipset will read whole cache line (32-64 bytes) instead of reading byte after byte, incurring only the initial delay. This requires additional logic in the chipset.
Pipeline burst cache moves burst logic into cache chips. Cache module itself has internal queue of operations and multiple parallel banks.
Those Cyrix CPUs were not your best buddy. Even now I've changed it to an AMD and voila, my intermittent no-boot issue vanished. Apart from my case, you were very lucky to find a newer BIOS version. That ROM looks like UV erasable, but you'll need a higher programming voltage than the mainboard has to offer. If you buy a 29xx series EEPROM and flash it in a more modern board like a 440BX type, you can put it in the 486. Boot from the original then carefully replace it with the blank one. Watch out for pin compatibility. It's a good tactic to drop in an EEPROM instead and keep the previous one for emergencies. You should try to tweak the BIOS settings to run with zero wait states if the hw let's it happen. Also the bus clock divider (if you can stay in the specs). That should give you some more boost. Back in the day even a 5% increase was a big deal, because it made some noticeable performance gain. Now I wonder...should I put in more cache to feed off my 64Megs of RAM or just forget about it. :)
Someone has kindly wrote an eeprom for me with the updated bios, when I get that I'll do a follow up video and also try tweeking the settings as you suggest. Many thanks for the tips.
Is this a problem limited to hardware in Europe and the UK? Or are these boards in North America too? I've never heard of such a thing as fake cache chips, especially on old hardware. It was so much more expensive to make circuit boards in the past than it is now. Very interesting. Cheers.
The boards are from China so I'm sure they would have happily sent them anywhere in the world.
I've read plenty of articles about this board in particular but its not clear if it's limited to Europe and UK only.
@@CRG Thank you for the reply.
Ah from China...that makes a bit more sense. Still what an odd thing to do. I found it to be quite interesting. I love learning new things. Cheers.
Maybe its the mobo? U have a bank of 5 and a bank of 4??? Most memory are usually runs in multiples of 2, so for 512k cache normally uses 4x 128k chips or 8x 64k chips, same with ram sticks, there are even number of chips per stick, whether if your using 1 or 2 sticks of ram. Maybe if thats the issue your having a problem in populating all 9 spaces with 128k chips to get 1meg(1024k) cache but u went over as 9x128k= 1152k so the extra 128k went over the limits?
A little late, buuut:
After I would have found out that the fake cache chips are literally nothing, I wouldn't have bothered de-soldering the rest. I would have just piggy-backed the real chips on them.
Looking at time 13:02, in the chipset settings you have "Alt bit in Tag SRAM" set to 7+1 bits. You don't seem to have a Tag SRAM chip, such as a Q8888, on the board (or even space for one). I wonder if that's the issue and the BIOS setting needs changing.
I later got the issue sorted with a bios update. The full 1mb is now working.
There is a tag chip fitted, it's the 5th chip in the same row. Granted I just used another of the same which is too big but it doesn't affect its operation.
@@CRG Interesting as I have a 486 motherboard I'm fixing which has that same layout of SRAM chips, including a slightly faster 5th chip but also has a special tag SRAM chip, which is what I believe has gone wrong (glitchy outputs on two lines and the chip burn my finger!).
@@decidedly_retro if it's getting that hot it's definitely a bad chip. Hope you can get it up and running.
@@CRG Yeah, the machine runs without the external cache enabled in the BIOS. I’m waiting for a replacement QS8888 chip to arrive from China.
Great video. I felt the pain.
Why not use two 16 pin sockets end to end for the 32 pin chip? Might have had to do a bit of sanding to get them to fit but is should work.
Hey, I said that 6 minutes ago.
@@samuel_towle And he did it in the video!
Does it matter that the tag RAM (U20) is twice as big as it is supposed to be?
No it doesn't matter, the system only uses of it what it needs. Just worked out easier fore this way since the cache chips came in a lot of 10.
Is there a reason why you didn't put sockets in? Whenever I do this kind of work I make sure I always have a socket available.
Didn't have any in stock and the postal service has been really slow recently. Wanted to get it done and there's no harm in not using sockets, if anything it's one less joint. Of course you can counter that with its harder to change a faulty chip but I have the tools to swap anything out later easy enough.
@@CRG Understandable :)
And later on you used socket for easier debugging, so... everything's fine :D
Transport Tycoon Deluxe Song in background! 😅
really nice build
Great vídeo CRG
I recently bought this board off eBay only learning about the fake cache chips issue afterwards. Unfortunately yes mine is the same as yours. But I have decided to treat it as a learning experience and have a go removing them and fitting sockets and some real cache. So I would be interested in knowing if you ever over come the bios issue?
Yep, take a look at the bios and a gus for the 486 video. The latest bios for the board from vogons fixed the problem but you'll need to write it to a new eeprom.
@@CRG my board is actually a slightly different model to yours I've realised but I have today managed to track down the latest bios for it online. However flashing or programming the EPROM is going to be beyond me I expect even if I had the right equipment. Do you think the person who did yours would be interested in helping me out for a fee? Or do you know of a business who offers that service in the UK? Sorry for questions I have spent so long trying to find a service to do that for me today without success. Probably because I'm not searching for the right thing or something
@@VidzVideo send me an email casualretrogamer@outlook.com and I'll see if I can help you.
I noticed that you said the new cache chips were 128kb chips. Please note that it takes eight chips plus one parity chip of 128kb to get 128 kilobyte of cache. Each chip is only single bit in a byte word. I would leave the full set of chips in and change your jumper settings to 128KB of cache. It is just a weird thing that 486 motherboards did in the day.
I know what you mean and I maybe phrased it badly but these chips do total to 102KB so individually they must be 128KB each.
I do get the issue sorted in a future video. The BIOS needed updating to enable the full 1024KB.
Did the updated BIOS make any difference with cache?
You'll have to watch the follow up video to see 😉
Just kidding, yes the bios update fixed the issue and the system sees the full 1MB now.
great analyze, excellent video
Why play audio from transport tycoon deluxe over the whole thing? Quite annoying.
My new videos don't have a music bed but at the time I suppose I was thinking it needed a bed to fill the space between me talking. In hindsight the videos are better without any music so that's the theme going forward. Not much I can do about it now on these older ones, in fact some of the even older videos are terrible with music beds that are far to loud. It is what it is though, all I can do is move forward.
Any idea what jumper 36 is for is there any reference to it in the manual this could be part of the problem as it was shorted with the fake cash chips installed this jumper might not even be populated and shorted on the cash equipped model motherboard removal of jumper 36 is worth a try and would be interesting to know what that jumper does had a system many years ago that had some weird unlisted jumpers that system was a real pain to configure because you were basically guessing anyway looking forward to your follow up video
I tried this and removing jumper 36 seems to reduce the total cache to 36k. Not really sure what its doing, maybe makes a connection for address lines, that's just a guess though. Suppose I could trace it out to see were it connects. If I have time I'll cover it in the next video.
Interesting I look forward to a follow up video in the future
24:19Wish I could "burn myself new cache chips" :P
Well spotted, a little slip of the tongue.
my m912 already came with cache sockets and has 128k cache installed
Yeah some of these boards did have real chips while others like mine didn't. That's why we can swap out the fake chips for the real ones.
how often do you find fake cache in machines?
I've found it in two so far. Not looked at the other machine yet but do need to at some point.
Most EPROM's I have seen on the 486 boards I repair dont have the window for UV erasing under the sticker.
The data sheets usually specify two versions, with or without a window.
My guess is that MB manufacturers dont want the user to erase the EPROM by mistake by removing the sticker and letting UV light over time erase it.
Also EPROMS usually requiers external metod of erasing, UV ligth usually.
There are ROMS obviously and a EPROM with no window is basicly a one time use.
You want to get a EPROM with a window and the proper UV ligth source or get an EEPROM, Electrically Erasable ROM.
Me and my friends have run in to this problem to and it seems better to just buy a new chip and flash it instead.
Also I was wondering did you install 9 cache chips?
Usually there are 4 or 8 of the plus a TAG ram chip.
It looks like you installed a Cache chip on the bank0 with 5 chips, I think the one closest to the fake branded UMC chip should be TAG ram chip, they are like 2 pins shorter and that one usually have no chip next to it.
Might explain some weird behaviors of the L2 cache.
Just adding one thing, tag ram with 8 chips probably is the same package, so same number of pins, its 2 pins shorter if the board uses 4 cache chips.
To the best of my knowledge the tag ram is just another cache chip although usually smaller capacity. For 1024k cache, tag needs a 64k x 8 chip which as you say is shorter. I have it on good authority though that a bigger capacity chip will work fine and it certainly seems to when running just 512k cache (which would only require a 32k tag ram).
I'm hoping to get a new eeprom burnt with the appropriate bios image then fingers crossed it fixes the issues I was having.
@@CRG Yea me and my friends looked in to this and I read some data sheets for my tag + cache chip combo on some of my boards and they are pin compatible from what I can tell. All my boards have smaller tag, usually 16K*8 but like you say its probably for cost reasons not compatibility.
And like you say every 128K of L2 needs 8K*8 of TAG ram so 32K is correct for 512K.
I also want to get a burner, in our group there are a few so we tend to borrow, I need it mostly to flash eeproms for Controller cards, sata and so on for conversion from IDE to SATA etc.
Also a tip, if your out of DIL sockets or need a bigger one then you can get try using female headers, like the 1x40, cut two that has the right number of pins, mount them on your Fake cache chips and then solder them to the board. That way they line up perfectly.
Anyway fun video and I hop you make a follow up, i repair boards and stuff so I fallow most youtubers doing retro computing.
I think I just fixed a Voodoo2 my self. all the chips, both TMU's and the FBI chip had lose pins, one side hade like half of them disconnected XD
Had to dragsolder all of them and I hade on that was stuborn and i could not find it, had artifacts due to one ram chip missing a connection. But fixed now stability testing. Card was totaly dead before.
Fun AND frustrating fixing retro stuff but rewarding when you figure something out dont you think?
Anyway Im gonna subscribe to you and keep track.
I was watching this in the background as I was working but did I understand the video correctly?
They are fake chips soldered onto the board in working sockets?
Nice scam I suppose.
Yes they were fake cache chips installed on the board but replacing with real chips gives you working cache. Well... to get all 1mb working I had to also update the bios but the board does support real cache chips.
You might improve the benchmark results by making the timings more tight :)
Need to have a play with it, any suggestions on what to tweak?
@@CRG not exactly. I don't really understand how it works under the hood, but I know that many 486 boards leave timings available for cache one way or another and this board is no exception.
Try to reduce them as much as possible (this also stands for RAM, as making ram timing faster will also improve the performance)
Also these settings might be hidden or grayed in the bios, either because timings are set to "auto", or because you need to enable "advanced mode" somewhere
@@retrospacenet I was having a play with it last night. If I tighten everything as far as it goes so 2-1-1-1 timings and 0 wait states the system will boot to DOS ok and was stable in benchmarks but did hang about half way through the first level of duke3d. It won't boot to windows with these settings just crashes. I was a bit pushed for time last night but I'll have another play later and see just how fast I can get it and be stable. Still can't get the 1024mb of cache working though. A new BIOS is on route but I'm not sure that'll fix it.
the lower the ns for the cache the tighter the timings on the cache and better the cache performance will be though it's limited for such and old computer and cache built in to the cpu will outperform cache any were else in the system
It's actually possible that the new cache chips you bought are also fake. There's a ton of re-labeled cache chips out there, especially on eBay and other sites.
I got them from eBay and from what seems as if a big reseller so I hope they are real. Certainly listed as tested (not that, that means much). Yet another reason I need to get the finger out and get an EEPROM programmer since that would test the chips too.
@@CRG I lately bought Winbond cache chips from an eBay seller called polida2008. The seller had 99.3% positive feedback, and over 100k sales done. But sure enough, the chips were obviously fake and most of them didn't work at 15ns: i.imgur.com/cFAQ9Rd.png
The seller provided pictures of original Winbond chips, but that's not what I got. Kinda annoying...
@@CRG I´m also planning on buying some cache chips (I aslo have a boards with fake chips), and since your´s appears to be "real", could you please share the link of the ebay seller you´ve bought from? Thank you
@@hugocosta4437 Take a look in the description of my latest video showing the bios upgrade of this machine. I've linked the chips and seller I bought my chips from or just search themicrohut on ebay.
@@CRG Awesome. Thanks
a good 486 board in the day was 500 bucks
The cache is fake, I had been building and repairing pc's for a couple of years when these boards came out.. on some cheaper boards the chips were actually rubber and glued on the board, some descent boards actually came out with empty sockets on to allow you to fit cache and tag ram when it came available, these came out towards the end of the '486's for a short while when the price of ram and fast 15ns or 20ns cache ram rocketed - I seem to remember the cyrix cpu's were less affected back in the day but FPU performance was bad on them and we all know what happened to those after doom and quake came out... on some really chintzy boards the bios was altered to not use the cache no matter what was set by the bios - even if you fit the real cache and fit the right size tag ram for the size of cache you have fitted - edit: you should see a better improvement if you fit an intel or Amd cpu
I did eventually get this machine to see all 1024k of cache, it just took an update to the bios. Benchmarking this board against the few others I have I must say its right up there in terms of performance.
As for the performance of the Cyrix, well, their 486 chips aren't that bad. I know the later Pentium clones are poor in their FPU performance but there is actually a way to slightly improve that. I've been thinking about pulling a video together on how to unlock some features of those chips.
your chips are too high capacity for that old board.
Nice work getting the cache installed. Not 100% convinced that it is a BIOS problem, but definitely worth a shot upgrading it.
With tighter timings, you might be able to eek another 10% out of this system. For reference, I have an Asus 486 system with Intel 486-DX2/66, 32meg RAM and 256kb cache, and S3 804 1meg graphics. With my BIOS optimized settings, I get about 5-10% more performance out of mine.
I do have a EEPROM programmer (TL866-II), and I am in Perth, if that helps at all?
I have been playing with settings and managed to squeeze a bit more performance out of it. I have a new bios on route so certainly can't hurt to try it.
LGR has that same motherboard
his is a little newer
typical PC Chips quality, the fake cache. but the AMI WinBIOS is the only thing i like though. PCChips was definitely a rubbish company for their shit they pulled.
10% faster, but what for the cirix is already a shitty processor.
Any luck 3 years later?
Yes it got sorted not long after this. Just needed a bios update.
@@CRG Great to here, also did you figure out the 1MB cash problem?
Yes that was the issue solved with the bios update. I thought that's what you were referring to with your first comment, but was it something else?
Fake Cache to much Cost. At the next time Cpu with no Cache.
This machine is shockingly similar to necroware's 486: th-cam.com/video/PmM_17JwJtY/w-d-xo.html
Thats not even a real AMIBIOS lol
Or its really, really, hacked up lol
It seems to be at least patched to recognise the fake cache as real but other than that seems fine.
I do go onto install the proper bios for this board. It's layout is the same all be it some of the options are slightly different.