🚀 go-pugleaf

Good boy wrote:
> 
> Could anyone tell me how to check a DRAM is 60ns or not?  Any utilities
> available?  Or just by looking at the RAM chips?-
well I notice there's a couple 'standard answers'.....look at the NUMBERS

NOT!

In the US you can call 800-726-7827 and ask about memory test products
from a company called RAMSTAR. Seems to me their standard DRAM tester
cost about $250 several years ago. I have one.

Even more important would be to request copies of their 'Tech Reports'
on memory and speed. And misc other topics.

In general, YOU CAN NOT TRUST THE SPEED NUMBERS PRINTED ON THE CHIPS.

The only GENERAL exception is when you buy them NEW from a reputable
supplier. If you DON'T know where they come from, you gotta MEASURE
them. The numbers lie, more often than not. I've tested THOUSANDS of
chips. From just about everywhere, including HONG KONG.

A manufacturer would USUALLY only run ONE production line at a time,
and produce ONE basic chip geometry. If you called him up to order
300,000 chips, and wanted 70ns, and he was running a 60ns basic part,
you'd get your chips, marked 70ns, and costing a little less than the
60ns part, BUT IT WOULD BE THE EXACT SAME PART. Marked 70ns instead
of 60. It cost too much, and was inefficient to run different product
versions. Usually they'd upgrade all lines to the same mask-set
revision level as soon as they could.

But more importantly, not ALL 70ns (or whatever) chips were the same
speed.

I personally HATE Micro Technology. They are out and out liers.

By and large the Japanese parts would test 15-20ns faster than they
were marked. The Korean parts maybe 5-15ns faster than the markings.
Micron would not only test from -5ns to +10ns (right, slower than
they were marked) but the only bad memories I EVER found on motherboards
were Micron. You figure it out.

Then there were the Hong Kong parts. Sometimes defaced with a couple
scratched lines right down the chip, over the markings. Then cleaned
off and re-labeled. Sometimes with NO regard to their actual measured
speed.

Now, you might fault my tester. Maybe IT was 10ns off. But since I'm
talking RELATIVE speeds, the general conclusions should still hold up.

Anybody wanting just a few parts tested, I'd be willing to plug them
in and tell you how fast - usually a sample is sufficient, as long as
all the parts in a batch have the EXACT same markings. At 70ns I'd
say 100 pieces would distrubute about 80% within 10ns of eachother.
For example, if the peak was 60ns, 80% of them would be between 55ns
and 65ns. At that level, maybe ONE would be 45ns, maybe ONE would be
70ns, USUALLY pretty much a bell-shaped curve.

BUT I've also seen batches where the PEAK was a twin, with the middle
chopped out. Somebody obviously did some selecting of parts.

Bill
Tucson, AZ  USA