3 Nov
2000
3 Nov
'00
9:59 a.m.
From: Richard Erlacher <richard(a)idcomm.com>
The Z80 core is about three times the size of the 650x
core and really
can't
do anything the 650x core can't do quite a bit
faster. If you clock the
two
That may be true, I don't care. It's a "but, who cares" thing. If a
manufacturer used z80 in a product and wanted an improved product
do you think they will drop z80 or find a way to to package it tighter.
Gate arrays are not relevant to the typical
microcontroller application.
The gate arrays are handy when an MCU is not fast enough, but for short
runs, where the MCU's shine, the FPGA's and CPLD's are much slower to
develop.
You only have worked with low volume stuff then. In the high volume
world it's common for vendors to suff parts into ASICs as a cost
reduction
or product design protection.
The single-chip examples I think about when I'm
considering what to use
are
frequently the Scenix SX family. They're as fast
as a core in an FPGA,
and,
Who cares. I didnt' exclusivly say speed was the issue. You did.
Allison
3 Nov
3 Nov
12:20 p.m.
please see remarks embedded below.
Dick
----- Original Message -----
From: ajp166 <ajp166(a)bellatlantic.net>
To: <classiccmp(a)classiccmp.org>
Sent: Friday, November 03, 2000 4:59 PM
Subject: Re: Monitor for iSBC 8024
Speed isn't the point behind using the SX. The SX will accomplish what a
gate array would do in a large-volume application, though its cost is in the
low-volume range, i.e. the parts are cheap regardless of your volume,
because they're already high-volume products before you buy them. The speed
is what allows you to implement complex functions with the relatively
inexpensive part at speeds achievable only with gate arrays. That puts the
mcy in the role of a logic element. If you code it to behave as an 8255,
then that's what it is. The same component could function as a USART or
some other "LSI."
From: Richard Erlacher <richard(a)idcomm.com>
Since cores were not germane to the topic at hand, you're right. However,
if a manufacturer uses a Z80 and wants more performance, the thing's still
made, so all he has to do is buy higher performance devices and adjust
whatever timing parameters are involved. If he designs a new product with
the thing, however, he gets what he deserves. That goes for using any CPU
rather than an MCU.
The Z80 core is about three times the size of the
650x core and really
can't
do anything the 650x core can't do quite a bit
faster. If you clock the
two
That may be true, I don't care. It's a "but, who cares" thing. If a
manufacturer used z80 in a product and wanted an improved product
do you think they will drop z80 or find a way to to package it tighter.
>Gate arrays are not relevant to the typical microcontroller application.
>The gate arrays are handy when an MCU is not fast enough, but for short
>runs, where the MCU's shine, the FPGA's and CPLD's are much slower to
>develop.
$100K is a bout the minimal cost for a gate array application. An MCU
application doesn't require that sort of up-front outlay in general. FPGA's
are in between somewhere, but higher than the MCU's. That's why several
vendors have core+ products now. You design the FPGA around the core in an
array including whatever memory and peripheral functions are required. That
makes sense on the way to a full ASIC design, in the event volume warrants
it. It's not like the '80's any more. If your product is profitable enough
to be worthy of any sort of volume production, someone else will steal it if
you don't use the most cost-effective production methods. Therefore, you
don't waste money on MPU's if an MCU will do the job for less cost.
You only have worked with low volume stuff then. In the high volume
world it's common for vendors to suff parts into ASICs as a cost
reduction
or product design protection.
You're onto something there, Allison, I normally work the first few monents
of the product design cycle, i.e. from concept to first functional
prototype. I've seldom worked on the final "productized" version, though I
frequently see "fixes" or upgrades on the way to a new product version.
However, that's not what the topic is.
The single-chip examples I think about when
I'm considering what to use
are
frequently the Scenix SX family. They're as
fast as a core in an FPGA,
and,
Who cares. I didnt' exclusivly say speed was the issue. You did.
Allison
5:17 p.m.
Just some data; sometimes data helps.
Year Chip #transistors Die Size
---- ---- ------------ --------
74 8080 4,800 20 mm^2
75 6502 4,000 21 mm^2
76 Z80 8,500 18 mm^2
-Mike C
On Fri, 3 Nov 2000, ajp166 wrote:
From: Richard Erlacher <richard(a)idcomm.com>
The Z80 core is about three times the size of the
650x core and really
can't
do anything the 650x core can't do quite a bit
faster. If you clock the
two
That may be true, I don't care. It's a "but, who cares" thing. If a
manufacturer used z80 in a product and wanted an improved product
do you think they will drop z80 or find a way to to package it tighter.
Gate arrays are not relevant to the typical
microcontroller application.
The gate arrays are handy when an MCU is not fast enough, but for short
runs, where the MCU's shine, the FPGA's and CPLD's are much slower to
develop.
You only have worked with low volume stuff then. In the high volume
world it's common for vendors to suff parts into ASICs as a cost
reduction
or product design protection.
The single-chip examples I think about when
I'm considering what to use
are
frequently the Scenix SX family. They're as
fast as a core in an FPGA,
and,
Who cares. I didnt' exclusivly say speed was the issue. You did.
Allison
4 Nov
4 Nov
1:43 a.m.
Pleasesee embedded remarks below.
Dick
----- Original Message -----
From: Mike Cheponis <mac(a)Wireless.Com>
To: <classiccmp(a)classiccmp.org>
Sent: Saturday, November 04, 2000 12:17 AM
Subject: Re: Monitor for iSBC 8024
Just some data; sometimes data helps.
Year Chip #transistors Die Size
---- ---- ------------ --------
74 8080 4,800 20 mm^2
75 6502 4,000 21 mm^2
76 Z80 8,500 18 mm^2
-Mike C
That's interesting! I was comparing the quoted sizes of a couple of HDL
versions available on the web. The smalles 650x I found was about 3200
gates, while the smallest Z80 was listed at 9200. It's difficult to say
whether that's the actually-used gate count, or whether it's the gates used
as a consequence of the implementation. If a logic cell is used, regardless
of how much of that LC is used, the "gates" contained within it are gone,
whether it's a nand gate or an adder, or whatever.
The counts listed within an IP item description usually are reflective of
what's used rather than what's consumed, so it can be quite misleading.
Sometimes they specify how many cells it uses in a given array series, which
gives a better count of consumed resources if one's looking to use that
particular family. If one wishes to use something else, however, it's
tempting to use the number of cells x the number of gates each cell
represents, which is often "optimistic" with respect to what's likely to be
the actual result of a fit.
Eric Smith reminded me some weeks back that the Z80 has two complete
register sets, which pretty quickly leads me to conclude it's justified in
consuming considerably more resources in an array than the 650x core. One
might conclude that the 8080 should lie between the two, since it has more
registers than the 650x, yet fewer than the Z80. The transistor count is
not quite reflective of the register resources, but it does appear that the
8080 has more transistors used than the 650x.
On Fri, 3 Nov 2000, ajp166 wrote:
> From: Richard Erlacher <richard(a)idcomm.com>
>
>
> >The Z80 core is about three times the size of the 650x core and really
<snip>
8814
days inactive
8815
days old
3 comments
3 participants
participants (3)
-
ajp166＠bellatlantic.net -
mac＠Wireless.Com -
richard＠idcomm.com