bb said:
Just a couple of days ago, I decided to upgrade. I installed a 1GB DDR
PC3200 Geil memory card, and seems to me that instead of speeding
things up, they have slowed down
A7N8X-E Deluxe
AMD Athalon XP2800+ Barton core
2x 512 MB Xerox
1x 1GB Geil DDR SDRAM
Windows XP Home
Geforce FX 5200
The first thing to do, would be to measure the slowness.
I looked at the geilusa.com website, and I can see one
1GB module there which is PC3200 with 3-4-4-8 timings.
A slow module, will determine the speed for all the modules.
1GB modules with tighter timing, would cost a fair bit more.
I used a copy of memtest86+ version 1.40, from memtest.org
and tested my 2x512MB dual channel config, to simulate your
possible memory configurations. Memtest86+ has a bandwidth
indicator in the upper left hand corner (three indicators,
third one down).
AthlonXP-M 2200MHz 200x11 (= 3200+)
2x512 Ballistix memory, dual channel
Memory 2-2-2-6 Bandwidth 1485MB/sec
Memory 3-4-4-8 Bandwidth 1346MB/sec
You might compare your 2x512 + 1x1GB configuration to
that first, and give us a number. If you get a reading
of around 900MB/sec or so, then for some reason you would
be in a single channel mode. (Also check the BIOS screen
during post, and see if it says "dual channel" or not.)
It would help to know your CPU speed and bus speeds, to
allow relative comparisons to the above results.
You can also measure various configurations, and post some
results. Try your 2x512MB in dual channel and set the timing
to match the timing that the BIOS is using for 2x512+1x1GB.
Try a single stick of 512MB (which is a single channel config,
and should give ~900 at CAS2 timings), and compare it against
a 512MB plus the 1024MB in single channel mode (the two slots
closest to one another). Try to maintain the same memory
timings between tests, to rule the timings out as the source
of the difference.
The CPUZ tool is useful. You can get it from cpuid.com .
It has a memory tab, which shows the current settings.
It also has a "report" generator, which will do a raw dump
of the SPD in each DIMM. If you are convinced your new
config is performing worse than its 3-4-4-8 timing,
you might post the contents of the SPD for the three
DIMMs, so we can look at them.
Mushkin used to do testing of motherboards (and the results
are now lost, as they could not be archived, and Mushkin saw
fit to dump their results). They used to claim that some
chipsets were a lot slower when a 1GB module was used. They
never offered any explanation as to why this should be so
(and their tech people should have understood why). I don't
recollect them mentioning any visible timing parameters
being a contributing factor, and personally, I cannot imagine
what hidden monkey business could cause further degradation
than just the memory timing parameters alone.
If you have some other freely downloadable software that I
could test, I can change the multiplier on my mobile, to
make my CPU match yours. I can then test and see what I get
with my 2x512MB config. The only thing I cannot currently
do easily, is change memory config, as the case is a bugger
to get open (old fashioned case with goofy home made cooler
fitted to it).
HTH,
Paul
