Hi all,
I've been sent recently a few fake 6502, probably from chinese ebay/aliexpress/wathever.
They have been sanded and original markings removed from both sides, then repainted and laser etched with funny markings, pretending to be MOS 6502A.
One of them is a real NMOS 6502 (but likely not a MOS one), another is clearly some 65C02 variant (it's easy to test if it's a CMOS or NMOS chip even with a multimeter).
Now, does anyone know what was the testing procedure to "bin" the 1 MHz vs 2 MHz parts? I know most parts will run fine even at 2 MHz, but out of curiosity, I'd like to understand
how they were tested back in the day. Obviously they weren't run for days at 2 MHz. I guess it was a combination of higher frequency and error-free runtime and maybe power consumption?
Anyone with a reliable source?

Thanks
Frank IZ8DWF

Put it into circuit and run up the speed until it becomes un-reliable.

The following is by no means rigorous, just my own personal experience. I find the NMOS units will generally run at twice (or sometimes even more) their rated speed. The CMOS ones will rarely got more than 1.25 time their rated speed.

So if an NMOS ones becomes unreliable at 2.5 MHz, it is likely a 1MHz part. I have one MOS unit that is marked at 2MHz and it does not become flaky until over 5MHz. However, I have a bunch of 4MHz marked CMOS units and none of them will work reliably over 5MHz.

Someone may have a better way, but actual speed testing is what I've done. The circuit I use is a small computer I've tested with a WDC W65C02 to over 16MHz and I usually run a program under EhBASIC of some sort.

_________________
Bill

Welcome Frank!

It's an interesting question, what the speed qualification test was, back in the day. I don't know. But the test would have been at higher than the target speed: maybe 2.5MHz, maybe more. Also, it would probably have been done at reduced VDD: at or below the lowest specified voltage. It probably would not have been at elevated temperature, which is one reason for overdoing the frequency.

Of, course, it's bad economics to set the margin too high, but it's also bad business to sell parts which fail in the field.

Thanks for the answers so far! Yes. having also first hand experience on how reliably one can overclock known parts also helps a lot.
Interesting suggestion about the reduced Vdd!
I have (of course) a small 6502 SBC, with RAM/ROM and I/O so I could easily tweak the clock generator to produce different frequencies.
Another thing I was wondering is whether they had to use a full "computer" with some real code or just a simple
clock + nop generator socket (as simple as it can get) and just checked the CPU wouldn't hang or jump in the address sequence.
If a complete test setup was needed, that was a serious effort to bin all the mass production of CPU (and I/O chips) back in the day.

Frank IZ8DWF

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

I guess that explains why I was able to run 14MHz W65C02 to 29.5MHz.
Bill

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

29.5MHz seems achievable with the batch of W65C02 I purchased recently from Mouser. It is 7 out of 7 tested operating at 4.75V, room temperature, but only when driving the few components (RAM, CPLD, CF disk) on board. It will crash when plugged into a backplane.

Reliable operation at 25.175MHz would allow W65C02 to operate as a VGA controller.
Bill

I don’t know how it was/is done in the professional semiconductor manufacturers facilities.

But as the frequency is increased, before the point of ‘total’ failure, some instructions may no longer work reliably. So unless you have a reasonably wide margin, yes you would need to test that it can do more than NOPs.

Mark

For "real" results this is the way to go. ROM, RAM and I/O are essential for real world applications and greatly affect system speed. There is fast RAM out there and there a methods I've been investigating to make slow I/O work on fast machines, but ROM has been the limiting factor in any of the systems I've made over the last few years. That being said, I would bet that even very fast RAM, ROM and I/O (let's say tested at 20MHz) would limit the speed of a CPU that tested at say 5MHz on a direct wired NOP generator to something less. Just the bus loading alone would manage that.

_________________
Bill