This thread? 256AK vs 257AK Cache

No, but I had completely forgotten I had started a thread on this 2 years ago! Sorry about the double thread.

There was a thread recently whereby someone bought a motherboard from amoretro on eBay (vogons user FBG?). The response was something to the sort of the seller upgrading the cache to '257' pieces for the extra dirty bit and being faster than '256' chips.

Maybe reading this book can help?

http://books.google.rs/books?id=-7oOlb-lCpMC& … epage&q&f=false

I only found that Pingaloka replaced his cache chips due to a guy named Fabian recommending it.
Re: Mi future 486...hopefully!

Yes, Fabian is FGB and is amoretro on eBay. This is the quote I was looking for, thank you. And here is is,

Pingaloka wrote:

replaced the standard 256KBit cache modules with 257KBit modules that have the "dirty bit" (This is courtesy of Fabian. I never heard about 257 modules neither the "dirty bit") Supposedly it makes the board run faster?

Fabian, maybe you can shed some light on this?

From the book:

One very simple method of implementing a copy-back cache would be to
write every Valid line which was being replaced back into main memory,
whether or not it had actually been written to by the processor. This would
make the cache waste a considerable amount of bus bandwidth with un-
necessary main memory write cycles, due to all the evictions of lines which
had not been written to by the CPU. Another problem with this method is
that all line replacements would take twice as long as would line replacements
in a write-through cache, since write-through line replacements require
only a main memory read cycle. To avoid this burden, the cache is usually
implemented with a means to signify whether a line in cache is more
current than the main memory location it represents. The simplest method
is to use another bit for each line in the cache, and this bit is called the Dirty
bit. Data which has been written in the cache, but has not been updated in
the main memory is tagged as Dirty by the cache controller's setting this bit.
Like the Valid bit, there is usually a Dirty bit for every line in the cache (Figure 2.l4).
During a cache miss cycle, the line to be replaced is examined, and, if its
Dirty bit is set, the current contents of that cache line are evicted back into main memory.

With '256' and '257' cache, you can set the TAG as 8+0, or 7+1 in the BIOS, where the +1 is the dirty bit. Usually, you run 7+1 for write-back cache and 8+0 for write-through cache. Does the '257' pieces handle this dirty bit any differently than the '256' pieces?

Don't know.
Did you try some benchmarking?

Are you sure a W24256-15 actually exists? The large datasheet sites don't seem to have them.

The existing W24256 is the -70ns variant. Appears not suitable as cache chip.

W24257-15 does exist. It is functionally identical to W24256 except access time and power consumption.

Yes, a W24256 exists. I looked at the datasheet this morning. It exists in -35ns and -15ns parts.

Datasheets for both the W24256 and W24257 appear to be identical, except that the 257 is available with better access speeds. There is no mention of any integrated dirty bit.