VOGONS


First post, by CoffeeOne

User metadata
Rank Member
Rank
Member

Hello,

I bought some (20 to be exact) 128kx8 SRAMs from a Chinese seller. My plan was to upgrade 2 486 mainboards to 1MB.
I already posted a picture in "Bought these (retro) hardware today".

OK, packing was horrible, they were just packed in an envelope with some material wrapped around.
1 broke off one pin on one of the 20 chips. But that is the least problem.

On top the all have the exact same printing, looks ok.
On the bottom side there is no printing, that's a bit suspicous.

Can somebody tell me, how real ISSI 128kx8 chips look like?

Ah yes, I should also mention, that I tested them in 2 mainboards (more or less all of the chips, but not the one where something broke off, haha)
Those boards had both had 256kB, one board had 9 times 32kx8, the other 4 times 64kx8 and tag 32kx8. I kept the Tag (on both Winbond 32kx8) and tested with 4 times 128kx8, should be 512kB.
Result: either the mainboard did not boot, or no cache was shown.

For sure I will give the seller a good feedback.

Attachments

  • IMG_7553.JPG
    Filename
    IMG_7553.JPG
    File size
    730.67 KiB
    Views
    419 views
    File license
    Fair use/fair dealing exception
  • IMG_7550.JPG
    Filename
    IMG_7550.JPG
    File size
    689.74 KiB
    Views
    419 views
    File license
    Fair use/fair dealing exception

Reply 1 of 16, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++

Did you install the SRAM chips per the motherboard instructions?

For boards that can take 1MB and you are only configuring for 512KB, some boards have you use every other slot while some will have them all in one row.

Yamaha YMF modified setupds and drivers
Yamaha XG resource repository - updated November 27, 2018
Yamaha YMF7x4 Guide
AW744L II - YMF744 - AOpen Cobra Sound Card - Install SB-Link Header

Reply 2 of 16, by TheMobRules

User metadata
Rank Oldbie
Rank
Oldbie

I don't have my chips at hand now, but if I remember correctly they had some printing on the underside, which said "TAIWAN" and possibly some numeric code.

Another thing to consider is the datecode, your chips have "0523", meaning Year 2005 Week 23. While certainly that could be the case, those DIP cache chips were not that common well into the 2000's, so it may be a case of the original text being reprinted with fake info for sale... for some reason they just love doing that, even on legit working chips, I think it may be to make them look new so they can sell them as "NOS".

I've given up on these chinese cache sellers from eBay, the worst offender was one that put images of actual legit chips on the listing but what I got was a bunch of those fakes with darker text, of which about half didn't work. Nowadays I try to get them from places like utsource, but there's a lot of fishy stuff there also... usually I look for mid-90's datecodes, and prefer "used and tested" rather than "NOS".

Reply 3 of 16, by CoffeeOne

User metadata
Rank Member
Rank
Member

Yes, I believe, did.

The other board is similar but newer: a VL/I SV2GX4 Rev.2.0

Attachments

Reply 4 of 16, by CoffeeOne

User metadata
Rank Member
Rank
Member
TheMobRules wrote on 2022-04-23, 21:28:

I don't have my chips at hand now, but if I remember correctly they had some printing on the underside, which said "TAIWAN" and possibly some numeric code.

Another thing to consider is the datecode, your chips have "0523", meaning Year 2005 Week 23. While certainly that could be the case, those DIP cache chips were not that common well into the 2000's, so it may be a case of the original text being reprinted with fake info for sale... for some reason they just love doing that, even on legit working chips, I think it may be to make them look new so they can sell them as "NOS".

I've given up on these chinese cache sellers from eBay, the worst offender was one that put images of actual legit chips on the listing but what I got was a bunch of those fakes with darker text, of which about half didn't work. Nowadays I try to get them from places like utsource, but there's a lot of fishy stuff there also... usually I look for mid-90's datecodes, and prefer "used and tested" rather than "NOS".

Unfortunately it seems none of the chips work. Half would be maybe still ok.
Honestly spoken I am afraid to make more tests, but both boards still work (when I remove the chips).

Reply 5 of 16, by brian105

User metadata
Rank Member
Rank
Member

The chips look fake, with the usual very bright text and sanded texture that redone tops usually have. I'd return for a refund and a bad feedback if you can't get these working.

I cut my finger on a modem
Presario 5284: K6-2+ 550 ACZ @ 600 2v, 256MB PC133, GeForce4 MX 440SE 64MB, MVP3 chipset, Maxtor SATA/150 PCI card, 16GB Sandisk U100 SATA SSD, 20X DVD-ROM
Presario 2100: MediaGX 133, 24MB EDO, Quantum Bigfoot 2GB

Reply 6 of 16, by CoffeeOne

User metadata
Rank Member
Rank
Member
TheMobRules wrote on 2022-04-23, 21:28:

I don't have my chips at hand now, but if I remember correctly they had some printing on the underside, which said "TAIWAN" and possibly some numeric code.

Another thing to consider is the datecode, your chips have "0523", meaning Year 2005 Week 23. While certainly that could be the case, those DIP cache chips were not that common well into the 2000's, so it may be a case of the original text being reprinted with fake info for sale... for some reason they just love doing that, even on legit working chips, I think it may be to make them look new so they can sell them as "NOS".

I've given up on these chinese cache sellers from eBay, the worst offender was one that put images of actual legit chips on the listing but what I got was a bunch of those fakes with darker text, of which about half didn't work. Nowadays I try to get them from places like utsource, but there's a lot of fishy stuff there also... usually I look for mid-90's datecodes, and prefer "used and tested" rather than "NOS".

I searched on Ebay now and found chips with datecodes:
0116: several different all from China
0927: 3 times China, 2 times Hongkong
9710: all from China
1910: but with 10ns US

EDIT: The US seller is the most funny 😁
https://www.ebay.at/itm/115212702677?hash=ite … jq8AAOSwwrRh60L~
I mean that is most obvious fake, but I would not buy in the US anyway of course.

So which one is good 😁 All fake

Reply 7 of 16, by citronalco

User metadata
Rank Newbie
Rank
Newbie

There are also -12N and -10N chips sold on Ebay.
Yes, they are all fake, so ignore the printing on the case. Upperside and underside. Completely.

Over the last years I've ordered ~100 pieces from different sellers. Not because I'm stupid enough to get fooled multiple times, but because it's damned hard to get L2 cache chips, especially ones with 128kx8.
Usually 3 out of 10 are completely broken (TL866 shows "Bus error" when testing them as 64kx8 chips as 128kx8 are unsupported), and another 2 fail when on the board.
The latter is the most annoying part: I swap them in into a working board one by one, and unzip a huge file. Should run a few minutes. If unzip shows CRC errors the chip is bad.

So you're left with 5 out of 10 working.
I run them stable with 2-1-1-1 at 40 MHz, so they should be at least 15ns parts.

So yes, you can buy them. But half of them are rubbish, and sorting those out can be tedious.

Unrelated to these ISSI chips: Found out that some boards are extremely picky about the TAG chip, regardless of frequency or waitstates: Some work with whatever chip you plug into the socket. For others I had to try several chips from several manufacturers (yes, including non fake chips) until they pass the unzip test. Or even boot.

Reply 8 of 16, by CoffeeOne

User metadata
Rank Member
Rank
Member
citronalco wrote on 2022-05-08, 21:34:
There are also -12N and -10N chips sold on Ebay. Yes, they are all fake, so ignore the printing on the case. Upperside and unde […]
Show full quote

There are also -12N and -10N chips sold on Ebay.
Yes, they are all fake, so ignore the printing on the case. Upperside and underside. Completely.

Over the last years I've ordered ~100 pieces from different sellers. Not because I'm stupid enough to get fooled multiple times, but because it's damned hard to get L2 cache chips, especially ones with 128kx8.
Usually 3 out of 10 are completely broken (TL866 shows "Bus error" when testing them as 64kx8 chips as 128kx8 are unsupported), and another 2 fail when on the board.
The latter is the most annoying part: I swap them in into a working board one by one, and unzip a huge file. Should run a few minutes. If unzip shows CRC errors the chip is bad.

So you're left with 5 out of 10 working.
I run them stable with 2-1-1-1 at 40 MHz, so they should be at least 15ns parts.

So yes, you can buy them. But half of them are rubbish, and sorting those out can be tedious.

Unrelated to these ISSI chips: Found out that some boards are extremely picky about the TAG chip, regardless of frequency or waitstates: Some work with whatever chip you plug into the socket. For others I had to try several chips from several manufacturers (yes, including non fake chips) until they pass the unzip test. Or even boot.

Hello,

I found procedures to test it:
1) I tested in the Asus SV2GX4: Cache configured to 256kB with 4 times 64kx8 and a tag 32kx8, that configuration is 100% working.
Then I replaced only one of the 64kx8 with one of the 128kx8. For (probably) working chips the board show 512k (of course it will not work correctly, but the Bios detects something a least).

2) Test 2: is measuring power consumption in idle. I don't have a SRAM tester, but at my workplace a colleague measured the power consumption for me. An original Winbond 64kx8 takes ~ 1 mAmp.
OK, so all of the 20 were tested. I guess only 2 had something like one milliAmp, some were in the 10 milliAmp range, some ~30 - 50. And there was one chip which was taken 100 milliAmps, 🤣. So again only 5 Volts and ground were connected to the chip.

So from test 1) there were only 8 chips that passed. 12 chips bad. 1 was special, when this chip was in, the board did not boot at all.
So from the 8 I took a set of 4, put it to the board, jumpered the board correctly to one bank 512kB cache. Result: The board showed 512kB, but could NOT boot from floppy. Hangup. Also not with slowest cache timing in Bios. With L2 cache disabled, the board booted 😁

So about test 2: As far as I remember the one chips with the (by far too high) power consumption is the one where the board does not boot at all.
So test 2 is maybe really a great test, I have to repeat it. Obvioulsy 10milliAmps is already far too much for such a 128kx8 SRAM chip. The datasheet say 10milliAmps maximum idle power consumption. So I have to find the one or 2 chips (out of 20, what the fuck), which have 1milliAmp idle only. The 12 that failed test one, I have put to garbage.
So it seems I was unlucky, far away from 50% working.

The seller refunded the money already.

I ordered one more time 20 chips ISSI61C1024 from another chinese seller, maybe I manage to get 8 working pieces at least. Expecations are low of course.

Reply 9 of 16, by mkarcher

User metadata
Rank Oldbie
Rank
Oldbie
CoffeeOne wrote on 2022-05-10, 19:34:

2) Test 2: is measuring power consumption in idle. I don't have a SRAM tester, but at my workplace a colleague measured the power consumption for me. An original Winbond 64kx8 takes ~ 1 mAmp.
OK, so all of the 20 were tested. I guess only 2 had something like one milliAmp, some were in the 10 milliAmp range, some ~30 - 50. And there was one chip which was taken 100 milliAmps, 🤣. So again only 5 Volts and ground were connected to the chip.

Measuring idle power consumption to identify damaged chips is a good idea. ESD damage, or damage caused by putting the chip in the socket 180° turned around often causes excessive current consumption. But the way you did it is not according to spec and might produce unreliable results. CMOS input pins must not be left floating, because it is random whether they are interpreted as "0" or "1", depending on leakage current and residual charge. If you are extremely unlucky, some pin might be in the middle between 0 and 1 and couple to some signal depending on it, starting to oscillate (which will cause high power consumption). Note the conditions specified in the data sheet for measuring idle consumptions. Typically, the data sheet requires to have all input pins within 0.5V or 1V to the rails (i.e. either 0..1V for low or 4..5V for high). And to put the chip to idle mode, /CS must be pulled high.

You can create a kind-of generic test setup for idle power consumption by just pulling all pins high through individual 22k resistors (or some other resistor value in that region). As most chips require an active low enable signal, pulling all input pins high will put them into the idle state. If some of the pins you pull high are actually working as output pins, they sink less than 1mA (which is in-spec for CMOS chips). The current for output pins that are low doesn't pass through the +5V supply pin, so pulling output pins high doesn't affect the consumption on the +5V pin.

Reply 10 of 16, by Imperious

User metadata
Rank Oldbie
Rank
Oldbie

I bought a ton of cache chips from Aliexpress last year from various sellers, they all worked perfectly.
I did get some of these ISSI 1024 chips, so when I get home from work I'll take a photo and upload.

Atari 2600, TI994a, Vic20, c64, ZX Spectrum 128, Amstrad CPC464, Atari 65XE, Commodore Plus/4, Amiga 500
PC's from XT 8088, 486, Pentium MMX, K6, Athlon, P3, P4, 775, to current Ryzen 5600x.

Reply 11 of 16, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

I've had surprisingly good luck with these chips. The first lot I bought in China was in 2006. They are 15ns. I paid quite a bit for those at the time, so they may be legit. In any case, they all worked as intended. I've been running them for 16 years with no problem. I'll check them out again soon to try to determine the authenticity.
The second batch I bought around 10 years ago. They are the 10ns type with the faded piss yellow silkscreen. I think out of the 20 or so I bought, only 3 bad ones. So far none of the others have gone bad, though they definitely aren't 10ns as I can't get the timings as tight as my original 15ns ones.

I think part of the high failure rate might be related to poor packaging. The sellers usually ship them in cheap white stryofoam which has ESD all over it.
The 10ns ones I bought domestically in China came in a nice plastic IC tube.

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 12 of 16, by pshipkov

User metadata
Rank Oldbie
Rank
Oldbie

Unless you got really unlucky with your purchase these 10ns modules are actually quite good.
They scale better than the 12 and 15 ns rated chips for sure - overclock better, hold tight timings better, etc.
But there is a catch in there. Most people assume that they are plug and play but they are not unless timings and frequencies are really low. But even that sometimes is not enough.
You need to experiment with different chips and sometimes rotate them between the sockets to find the right combination that works well.
Most people trip there.
The tech itself back then and the age of the components don't play in our favor.

retro bits and bytes

Reply 13 of 16, by CoffeeOne

User metadata
Rank Member
Rank
Member
pshipkov wrote on 2022-05-11, 07:08:

Unless you got really unlucky with your purchase these 10ns modules are actually quite good.
They scale better than the 12 and 15 ns rated chips for sure - overclock better, hold tight timings better, etc.
......

Hi,
I don't understand that. Why should the "10ns modules" scale better?
According to the datasheet only 12, 15 and 20ns modules exist / were manufactored. So with the 10ns you are only sure, that they were re-labelled.
But yeah, most 15ns modules are re-labelled, too.

Reply 14 of 16, by pshipkov

User metadata
Rank Oldbie
Rank
Oldbie

Will let others comment on the conspiracy theories about relabeling, fakes and such.
I only speak from personal experience.
More 10ns rated chips handle 50, 60 and 66MHz with tight L2 cache timings than 12/15 ns rated ones.
This is what i see from my corner.

(If you have time to kill - some of the links in my signature directly or indirectly outline that)

retro bits and bytes

Reply 15 of 16, by CoffeeOne

User metadata
Rank Member
Rank
Member
pshipkov wrote on 2022-05-14, 03:32:
Will let others comment on the conspiracy theories about relabeling, fakes and such. I only speak from personal experience. More […]
Show full quote

Will let others comment on the conspiracy theories about relabeling, fakes and such.
I only speak from personal experience.
More 10ns rated chips handle 50, 60 and 66MHz with tight L2 cache timings than 12/15 ns rated ones.
This is what i see from my corner.

(If you have time to kill - some of the links in my signature directly or indirectly outline that)

Hi, I will check your link, thx!

About conspiracy theories, I disagree here.
I only found now the attached datasheet, there is dip32 missing, so maybe there was another datasheet only for the dip-32 type. But still I believe, is61c1024 chips with 10ns were never produced. Fastest were 12ns, but these were extremely rare obviously.

Attachments

  • Filename
    is61c1024.pdf
    File size
    127.65 KiB
    Downloads
    3 downloads
    File license
    Fair use/fair dealing exception

Reply 16 of 16, by weedeewee

User metadata
Rank Oldbie
Rank
Oldbie

Just adding my 2cents & personal experience wrt relabeling.

bought some eeproms, all of them were relabeled, some were actually what was on the new label, though most were not, ie 120ns, 150ns, 200ns.
They just want to make them all seem identical, with no regard to the actual specifications of the chips.
It makes the chips seem like brand new, until one does a little deeper inspection like actually checking the hardware ID, reading out the contents, testing the label with acetone(nailpolishremover) or thinner.

Right to repair is fundamental. You own it, you're allowed to fix it.
How To Ask Questions The Smart Way
Do not ask Why !