• Hello, Guest! Welcome back, and be sure to check out this post for more info about the recent service interruption and migration.

A 'once and for all' declaration regarding L88M mask MC68040 CPUs

jessenator

Well-known member
I've searched a few threads and wanted to make a place for a 'once-and-for-all' declarative thread regarding these final incarnations of the MC68040 CPU.
I've read many things and heard many testimonials, but I wanted to definitively understand what people mean.


Most of the time, this is what I see written about them:
they run cool to the touch
Which I take to mean, "these can run without a heatsink" and also "there's little perceptible temperature differential when 'touched'"

Is this truly their behavior, globally? Were they meant to run like this, or are the takes like this simply the anecdotes from various users? I'm not looking to cast blame, I just want there to be more of an empirical reference point for these, because, you hear things and things get passed around.


And if this (above) is taken to be true, is it enough to declare CPUs that are marked such, but do not run this way as 'fakes' or 'counterfeits?' AND MORE TO THAT POINT: what is an unacceptable range of temperature operation?

I have some E42K mask versions that would probably run ~50–60C at stock speeds without a heatsink, but would be interested to know the range in the real world. Modern CPUs operate around that at load or higher, even. I've seen in the Freescale documentation something close to 101C being the upper limit, but I don't recall for certain. Obviously, we don't want to run in that range, but within an acceptable one.


Ideally, I would appreciate the insight of anyone who worked at a Motorola fab during the period, but on the other hand I'd also understand if any such person would prefer to keep quiet on the matter .
 
Last edited:

Fizzbinn

Well-known member
Amen.

I have a bunch of 68040s from various systems and a couple bought off eBay as well as a Flir thermal imaging camera for my iPhone. I'm thinking about setting up some sort of test bed (perhaps my Quadra 605) and swapping in all the 68040s CPUs I have (including "suspect" L88M and E42K eBay CPUs)

XC68040
E31F - .8um masks:
Have 1 E31F @ XC68040HRC25 - In C650 25MHz
Have 1 E31F @ XC68040RC33M - In Q630 33MHz
Have 1 E31F @ XC68040RC25M - In Q610 25 MHz

E23G - .65um masks (not sure about this mask/die shrink?)
Have 2 E23G @ XC68LC040RC25 - In Q605 25 MHz
Have 1 E23G @ XC68LC040RC33B - In LC580 MB 33MHz

MC68040
E42K - .65um masks
Have 1 E42K @ MC68040RC33 - from eBay

L88M - .57um masks
Have 1 L88M @ MC68040RC33 - from eBay

My gut (which admittedly doesn't have a perfect track record) has a hard time believing L88M CPUs are actually "cool to the touch" when in active operation, cooler than early mask versions yes but cool...? Of course without a known legitimate L88M I'm not sure what I could prove.
 

jessenator

Well-known member
not sure about this mask/die shrink?
This is from the German AmigaWiki, (and until it's corrected in the same manner I'm hoping for the rest of it, it's the best we have), and they mention that full 040s had different masks from the LC and EC versions. Here are the full version masks from that wiki:

MaskYearXC/MC
D43B 1989 0.8µm XC
D50D 1989 0.8µm XC
D98D 1989 0.8µm XC
E31F 1989 0.8µm XC
E26A 1992-19950.65µm XC
E42K 1995 0.65µm MC
K63H 2000 0.57µm MC
L88M 2002 0.57µm MC
 

jessenator

Well-known member
Bumping this after a couple of months.

I know folks who've worked there might still be under NDA (IDK how the current IP holders feel about those kinds of things), but feel free to PM me, or PM me for an email address and I'll keep you confidential. Again, this is to simply to clear the air and put the hearsay to rest.
 

trag

Well-known member
Would there be any practical difference between a K63H and an L88M? I guess if they were built on different processes, even though they're both at .57um feature size. Some processes were better at limiting static leakage than others, but I don't think that was a big issue until things got about 10 times smaller.

He may have arrived too late (at Motorola), but I think the person's brain you want to pick is @MOS8_030 .
 

trag

Well-known member
Well, a little searching turned this up:

MC68040 product built with the Rev. A mask, K63H, will have an "A"
suffix at the end of the device number (MC68040FE25A), as opposed to
the no-suffix for E42K non-rev mask product (MC68040FE25). Also,
the device marking will contain "0K63H" as the mask revision. All
other marking on the device will remain unchanged.

from https://www.nxp.com/docs/en/product-change-notice/PCN5606.txt

Which at least suggests that if you think you have a K63H, it should have an A at the end of the part number.

A little searching on Ebay turned this up, which, as far as I can tell, matches none of the above list of revisions, suggesting the list may be incomplete, or...

https://www.ebay.com/itm/201603003217
 

Fizzbinn

Well-known member
Another interesting site:

Lists full 040s separate from LC/EC versions and says the full 040 “E31F” is .65 μm. That does make some sense to me since that’s (XC68040RC33M, 02E31F) what’s in my Quadra 630 and Apple didn’t use a heat sin… I still intend to find some time to do some FLIR tests with CPUs I have.
 
Last edited:

jessenator

Well-known member
That's interesting that they didn't use heatsinks on those late 040 performa/quadra machines...

I mean, it could be they're just binned 40MHz chips and can run just fine at 33 (essentially underclocking it) without one. I haven't tried running my MC qualified RC40 chips at speed for long without a heatsink, and I still can't find my stinking IR thermometer to even get a cursory reading...

I was chatting briefly with Bolle on IRC today about the chips I have which are supposedly L88M and we haven't yet come to consensus about their veracity. However, he shared with me a photo of known-genuine specimens which have that wacky "MC68040LRC40" sku, so it's a thing, just not documented in Moto/Freescale/NXP files that are publicly accessible.

Here are the geuine ones (ignore the 060 in there...)
IMG_20211013_133452.jpg

And here is what I've got.
 

Fizzbinn

Well-known member
Here’s mine:
CFBC647F-F81A-4C44-9FAE-34E7DF43173E.jpeg

I got it off eBay in a listing of “Motorola CPUs” that also included XC68040, XC68LC040, and MC68LC060 processors that all appeared similarly used/possibly pulled from telco networking gear. As nothing was advertised as “MC68040 L88M” it doesn’t seem like the seller would have been knowingly passing off remarked CPUs.
 

jessenator

Well-known member
A little searching on Ebay turned this up, which, as far as I can tell, matches none of the above list of revisions, suggesting the list may be incomplete, or...
https://www.ebay.com/itm/201603003217
…The markings on this one look chunkier than the chips that I think have original Motorola labels, although I guess different production facilities/runs could look different.
That's has to be a fake. It's got the letters in the mask rev swapped; and what's more, that's an LC040 mask they swapped! E23G is the LC mask, and that listing shows G23E which doesn't come up anywhere! The true mask shows up on a CPU (to be replaced) from Q605 here: https://68kmla.org/bb/index.php?thr...8m-mask-revision-from-ebay.36759/#post-397616

LC040 mask rev shown here where Fizzbin found it:
Another interesting site:



And back to the real L88M temp claims. @Fizzbinn et al. have you at least used a fingerometer to test how your L88M chips radiate? Future FLIR tests notwithstanding ;) @Bolle sorry for the tag, but I wondered if you'd ever done the same fingerometer test. I know everyone's got different tolerances, but for me it was 120 F / 49 C where I'd pull my finger away (at least, that's when I'll pull my finger away from a stream of hot water from the tap where I measured it to be ~120 F).

As imprecise as that is, the original/parroted claim is "to the touch" so...


Also, that stupid xenforo-imgur-gallery-iframe-thingimajig doesn't show up on desktop, but posted just fine on mobile last night? But here's mine for better visibility with a bbcode tag:
zgLxRkth.jpg
 

Fizzbinn

Well-known member
And back to the real L88M temp claims. @Fizzbinn et al. have you at least used a fingerometer to test how your L88M chips radiate? Future FLIR tests notwithstanding ;) @Bolle sorry for the tag, but I wondered if you'd ever done the same fingerometer test. I know everyone's got different tolerances, but for me it was 120 F / 49 C where I'd pull my finger away (at least, that's when I'll pull my finger away from a stream of hot water from the tap where I measured it to be ~120 F).

As imprecise as that is, the original/parroted claim is "to the touch" so...
My ”L88M” is certainly not “cool to the touch”, has anyone seen/experienced that first hand?

My expectation that I want to confirm via FLIR testing it that there is a measurable temperature range between the 3 known 68040 process sizes running in the same test bed (Q605) at the same speed (25MHz).

.80 μm - hottest - Q700 XC68040RC25B D50D
.65 μm - hotter - Q630 XC68040RC33M E31F
.57 μm - hot - eBay ”MC68040RC33A L88M”

I will also measure others, to see where they fit in.
 
Last edited:

Fizzbinn

Well-known member
Last edited:

jessenator

Well-known member
According to the polish site I linked to above their is a “L89M” .57 μm LC version.
Beat me to it :) Yeah, the EC and LC versions are different mask revisions to the RC and LRC 040s.

Is it exclusive to 33Mhz?
Not sure what you mean, the CPU speed is set by the external clock chip.
Could be he means the stamped/rated clock on the package. The 040 was seen from 25–40 MHz in Mac applications, but I'm not sure of the full product lineup by the time NXP and Freescale took over. I can only assume they still binned some twitchy parts down to lower speeds so they could still move product.
 

jessenator

Well-known member
The date table might be in weeks for some and not for others, so IDK what's going on there, but here are the tables from the Polish site

FULL 040s:
MaskProcess (μm)FabProduction years (if known) yy/wk?
D43B0.8MOS891/03-91/15
D50D0.8MOS891/50
D98D0.8MOS892/09-92/3-50
E31F0.65MOS1193 - 2-Jun-95
E26A0.65MOS8 (???)92-95
E42K0.65MOS112-Jun-95 - 13-Aug-00
K63H0.57TSC813-Aug-00 - 23-Nov-02
L88M0.57MOS11Nov 23, 2002 - 2004 (?)

EC and LC 040s
MaskProcess (μm)FabYears of production if known
D39H0.8MOS892
E23G0.65MOS119339 - 02-Jun-95
E71M0.65MOS1102-Jun-95 - 17-Jan-00
J46X0.57TSC817-Jan-00 - 07-Sep-2002
L89M0.57MOS11From 07-Sep-2002
 
Top