VOGONS


Is fake cache a myth?

Topic actions

First post, by Confused UngaBunga

User metadata
Rank Newbie
Rank
Newbie

I'm kind of convinced:

Write Back Cache Modules FAQ Q. What are write back cache modules? A. Write back cache modules are installed in the cache socket […]
Show full quote

Write Back Cache Modules FAQ
Q. What are write back cache modules?
A. Write back cache modules are installed in the cache sockets of motherboards. On some motherboards, they are
socketed, on others they are soldered directly onto the board. Motherboards with these modules are commonly
known and advertised as having 256k of write back cache. These modules do not function as cache memory. They
are blank modules which contain 0k of level 2 or external cache.
Q. If WB cache modules don't function, why are they manufactured on the motherboard?
A. Good question. Technical support representatives from different motherboard manufacturers have stated a few
reasons. First, their motherboards will not function without level 2 cache in the sockets. By installing blank chips, they
can trick the MB into thinking there is cache memory installed. Another explanation is that the CPU can use its
internal level 1 cache in write back mode with the modules installed. Write back mode is faster than the conventional
write through method which increases the performance of the motherboard and CPU by up to 20%. Finally, by
installing blank modules in the component feeders at the assembly line, manufacturers lower production costs.
Reprogramming robots to manufacture no cache and real cache motherboards can cost upwards of $100,000
including down time.
Q. How important is external or level 2 cache memory?
A. It depends upon which CPU you are using. The cache on older CPUs such as Intel's 486 do not support write back
mode. These CPUs need a significant amount of external cache, usually 256k or more, to perform well. Most newer
CPUs such as Cyrix's 486/5x86, AMD's SL enhanced 486, and Intel's Pentium utilize write back mode. They do not
require large amounts of external cache to perform well. Adding external cache to these CPUs will increase
performance however it will not be as significant as adding it to CPUs which use write through mode.
Q. Why did you write this FAQ?
A. The topic of WB cache modules has been very contoversial. People who weren't aware of what WB cache
modules are have purchased motherboards thinking they contained 256k of external cache. Once they found out they
did not have any external cache, they assumed the motherboard manufacturers and resellers were scam artists,
unethical and deceptive. We are committed to our customers and their satisfaction. We want you to be informed of
what you are buying before you buy it.
We appreciate any comments you may have relating to this article.

Reply 1 of 30, by TrashPanda

User metadata
Rank l33t
Rank
l33t

Yes it exists, PCChips used it in a bunch of their boards and it didn't matter if you swapped the fake chips for real cache chips the sockets were either not physically wired to the chipset or the chipset refused to see the real cache chips.

They did it I suspect as a cost cutting measure.

Last edited by TrashPanda on 2022-08-31, 14:22. Edited 1 time in total.

Oh noes, the cap let the shmooo out 😁

Reply 4 of 30, by TheMobRules

User metadata
Rank Oldbie
Rank
Oldbie

Where did that "FAQ" come from? PCChips themselves? The amount of bullshit in there is astounding, making it seem as if the fake modules acted as some kind of required termination. That's just completely false, a motherboard can work just fine without cache in the sockets. Also, that wouldn't explain certain PCChips motherboards which had fake SMD cache modules with traces going nowhere.

Fake cache was very true, and a complete scam from a dishonest company. There's just no valid excuse for it.

Reply 5 of 30, by mkarcher

User metadata
Rank Oldbie
Rank
Oldbie

Currently, you find the FAQ at http://www.cc-solutions.com/tech/wbdef.html (the only hit google finds).

Technically, the FAQ is bullshit with some pieces of correct information in it. The facts about 486 mainboards are:

  1. No standard 486 mainboard needs "dummy cache modules".
  2. Most of the boards you could find the "WRITE BACK" fake cache chips on came with a modified BIOS that printed "256KB L2 cache installed" on every boot. If these modules were validly installed as dummy modules on "0KB cache" low-cost mainboards, there would have been no need to mis-report the cache size to the user.
  3. Possibly some 486 chipsets are unable to run L1 cache in write-back mode in the absence of L2 cache. If that is indeed the case, though, real L2 cache would be needed, and dummy modules wouldn't serve any purpose.
  4. Running L1 in WB mode does in no way reduce the effectiveness of L2 cache. You get acceptable performance from any higher-clocked 386, any 486 or Pentium system only with L2 cache installed. On the other hand, many L1WB capable processors, like the AMD 5x86, came with 16KB instead of 8KB L1 cache. The bigger L1 cache of these late 486-compatible processors indeed allowed certain code to suffer a lot less from missing L2 cache.

Reply 6 of 30, by The Serpent Rider

User metadata
Rank l33t
Rank
l33t
TrashPanda wrote:

if you swapped the fake chips for real cache chips the sockets were either not physically wired to the chipset or the chipset refused to see the real cache chips.

That's incorrect. Most PCChips boards will work fine after "cache" replacement, because some were shipped with working cache. But the nasty thing is - majority of such boards had fake L2 cache soldered directly to PCB.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 7 of 30, by rmay635703

User metadata
Rank Oldbie
Rank
Oldbie
The Serpent Rider wrote on 2022-08-31, 16:17:
TrashPanda wrote:

if you swapped the fake chips for real cache chips the sockets were either not physically wired to the chipset or the chipset refused to see the real cache chips.

That's incorrect. Most PCChips boards will work fine after "cache" replacement, because some were shipped with working cache. But the nasty thing is - majority of such boards had fake L2 cache soldered directly to PCB.

Some had a socket to install a cache module alongside the fake cache

Reply 8 of 30, by The Serpent Rider

User metadata
Rank l33t
Rank
l33t

Yes, those would work fine, of course if all jumpers are also in place and not soldered in one position. But not "alongside", that's true only for one very infamous PCChips board.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 9 of 30, by Confused UngaBunga

User metadata
Rank Newbie
Rank
Newbie

I had the text as a fule for more than 2 years now, and I don't even remember how I got to the site mkarcher references.
I guess some were just "surface mount a pair of black plastic rectangles" but not all were scams?

Reply 10 of 30, by mkarcher

User metadata
Rank Oldbie
Rank
Oldbie
Confused UngaBunga wrote on 2022-08-31, 19:00:

I had the text as a fule for more than 2 years now, and I don't even remember how I got to the site mkarcher references.
I guess some were just "surface mount a pair of black plastic rectangles" but not all were scams?

Everyone who sold boards with the "write back" cache chips as boards having L2 cache was (maybe unintentionally) involved in scamming the customers. There are no chips that are cheaper than cache SRAMs but required to enable L1 write-back on any 486 or Pentium mainboard. Everyone claiming that a chip labelled "write back" is supposed to help get L1 writeback working is (possibly unintentionally) involved in scamming customers, too. The FAQ you quote looks like it is from a mainboard dealer that tries to motivate people to not return their fake cache mainboards by relabelling the fake cache scam that has since been exposed to be "works as intended, you are just to stupid to understand the intention". Maybe the dealer actually believed the stuff, because the mainboard vendor (i.e. PCChips) was so good in convincing them that these chips are not a scam.

Reply 11 of 30, by Horun

User metadata
Rank l33t
Rank
l33t
The Serpent Rider wrote on 2022-08-31, 16:17:
TrashPanda wrote:

if you swapped the fake chips for real cache chips the sockets were either not physically wired to the chipset or the chipset refused to see the real cache chips.

That's incorrect. Most PCChips boards will work fine after "cache" replacement, because some were shipped with working cache. But the nasty thing is - majority of such boards had fake L2 cache soldered directly to PCB.

Agree !! Most like the Pcchip m912/Amptron 6900 v1.4 & v1.7, you could swap the fake with real and they usually worked fine. You are correct.
There were a rare few PcChips/Amptron (like the m919 v3.48/f - Amptron 9700) that have fake onboard with the cache chips traces just circle around each other.
I have one such board. The trick with it is use the proper cache module in the cache slot 😀
Also one Vogon member had an issue with their m912 that even after fake to real cache swap their cache did not work, that particularly rare board did not have a genuine 912 Award bios IIRC (they flashed to fixed it iirc).

Hate posting a reply and then have to edit it because it made no sense 😁 First computer was an IBM 3270 workstation with CGA monitor.

Reply 12 of 30, by pentiumspeed

User metadata
Rank l33t
Rank
l33t

Fake cache existed on mostly junk motherboards like PCchips plus relabeled chipsets. I don't like this as I had hard time remembering what chipset motherboard is using.

Cheers,

Great Northern aka Canada.

Reply 13 of 30, by Romain

User metadata
Rank Newbie
Rank
Newbie

I have this 61C328AH-15
On my PC Chips M918i ...
The ALi chip is really good so that is not really visible to current usage en gaming (I was pleasantly surprised) but ah ... yeah when I see the bench with se same mobo but tunned with real 256kb (av.~+25% perf) ... I just want to trash this mobo ..

But hey, I don't let it because it's a motherboard that I recovered in a dump, in a box where birds had nested etc, a real sh*t - I spent so many hours for the save that in the end I can't bring myself to throw it away or resell it :')

Reply 15 of 30, by Sphere478

User metadata
Rank l33t
Rank
l33t

(At anyone)

I take it that if a bios is fake reporting that cpuz will tell the truth?

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 17 of 30, by ODwilly

User metadata
Rank l33t
Rank
l33t
konc wrote on 2022-08-31, 14:58:

Posts from vogons members showing proudly their fake cache motherboards have also been made!

Also cases where the cache is fake with/O sockets (but still with the traces) and a fake bios check to confirm working and existing cache. Had a M912 for awhile with real cache, great mobo. Reliable and fast, just picky.

Main pc: Asus ROG 17. R9 5900HX, RTX 3070m, 16gb ddr4 3200, 1tb NVME.
Retro PC: Soyo P4S Dragon, 3gb ddr 266, 120gb Maxtor, Geforce Fx 5950 Ultra, SB Live! 5.1

Reply 18 of 30, by rasz_pl

User metadata
Rank Oldbie
Rank
Oldbie
Confused UngaBunga wrote on 2022-08-31, 13:46:

I'm kind of convinced:

By installing blank chips, they can trick the MB into thinking there is cache memory installed.

CPU can use its internal level 1 cache in write back mode with the modules installed

by installing blank modules in the component feeders at the assembly line, manufacturers lower production costs.
Reprogramming robots to manufacture no cache and real cache motherboards can cost upwards of $100,000
including down time

Yes, I totally buy it hook line and sinker. See, they werent, checks notes,

motherboard manufacturers and resellers were scam artists,unethical and deceptive

after all, they simply cared about the user! ;----------D