On Tue, 14 Jan 2014 20:28:47 -0000, dennis@home wrote:

On 14/01/2014 12:07, The Natural Philosopher wrote:
On 14/01/14 08:52, Ian Jackson wrote:
In message om,
"dennis@home" writes
On 13/01/2014 23:02, Uncle Peter wrote:




You could always buy RAM rated much faster than you're going to run it.


Not always..
sometimes the designer will use the minimum hold time for the ram to
allow them to change the address to start the next cycle early.
If the ram is too fast the output will change too soon and put the CPU
into a metastable state.
Then anything could happen.

This disagrees with what Mike Tomlinson has just said when I asked the
specific question as to whether 'faster than motherboard spec' memory
could screw up your computer. A couple of weeks ago I replaced a pair of
512MB PC2-4200 (533MHz) with a pair 1GB PC2-6400 (800MHz). After a week
of faultless operation, the PC suddenly started to play up. Another 1GB
PC2-6400 pair (different make) also gave problems. [As Mike advises,
I'll be carrying out more memory tests.]

I have to say that despite extensive Googling, I have found it very
difficult to find any definitive and authoritative statements about
using 'too fast' memory. When the subject is raised, most discussions
rapidly veer away from the original question, and your are left
wondering whether the answer was 'yes', 'no', 'maybe' - or, more often,
'nobody here knows the answer, so we'll answer a question that no one
asked'.

I will concede the possibility that fast DRAM relies on - say - being
refreshed more often than slower, and might leak charge away at a slower
clock rate.

I will also concede that it may be possible to rely on a propagation
delay to achieve the desired result and getting rid of it could cause
problems, though I am not so clear that anyone actually does this.

So, whilst I will concede a possibility, I am pretty sure (and most
people writing on the subject agree) that the actuality is that faster
RAM wont do any harm *on account of being faster*.

I suspect what dennis is hinting at is an incomplete understanding of
how SDRAM works...

I am not confusing anything.
I stated a fact.

I have fallen foul of a similar problem once when using PLS devices.
They changed the manufacturing process and the system failed as the
setup and hold times no longer worked at the clock speed I was running
the PLS at.
It needed a few straps and a change to the programming to fix.

I can find no evidence online to suggest that 1600MHz memory will fail at 1066MHz for example, but plenty of people saying "of course you can use faster memory at slower speed". Prove otherwise.....

--
Gary's weather forecasting stone:
Stone is wet Rain
Stone is dry Not raining
Shadow on ground Sunny
White on top Snowing
Can't see stone Foggy
Swinging stone Windy
Stone jumping up and down Earthquake
Stone gone Tornado