So I managed to get a free 512M PC2300 card@142 MHz (infineon). Added it to my 1G PC3200 (200MHz) card (corsair). I now have 1.5G, and the net speed actually dropped below the lowest (128MHz) as measured by one of those hardware test programs (can't remember the name right now). Don't really see an observable difference either way. Anyone have any good test or battery of tests in a suite to verify one better than the other? Most google searches turned up forums suggesting 1G at higher speed is better than 1.5G at lowered speed (I can see the logic at low memory requirements). Experience/thoughts appreciated. (Ran a few hogs like photoshop and full screen camtasia vids simultaneously; dropped mem availability to 900M, but no performance degradation noticable).
the performance difference will be marginal at best, its not significant enough to be wasting time running benchmark programs.
For the fastest and most efficient use of RAM, you really need to have "equal" size and speed of memory. Putting together "unequal" size of RAM ( such as 1 gig + 512mb ) will cut performance by at least 10%.
I have always heard (yet never seen evidence to prove or disprove) that if you have differing RAM speeds, your system will operate at the slower of the speeds. Further, if you have a 200mhz fsb and run 142mhz RAM, in theory you are slowing down the fsb because the slower RAM is causing a bottleneck. If it was not broken before, you may want to reconsider whether you want to continue with this "fix" or just put it back to how it was before. I use Prime95 as a system burn-in/ system check when I build systems. From what I have been able to find out there, it's the best system test out there, and it's free; just google it.
That is correct. Also in order to take advantage of the dual channel design, memory modules must be matching in capacity in paired memory banks (usually color coded on the motherboard). Thus the original poster might have the worst case scenario of slower speed and non-matching memory modules in paired memory sockets. It also pays to note that there is no increased performance from buying faster memories in case the FSB cannot scale up with it. So with 200 MHz FSB stick with 200 MHz memories. Depending on the architecture there can be substantial improvement with increasing memory speed as the it is the biggest limiting factor of software execution speed (pretty much all software is data starved due to memory being much slower than cpu). However with multicores and hyperthreading this is less so (other core or thread executes while data is being fetched). And the question is pretty much moot on a trading computer unless building a very high end system. It's more cost effective to run different apps on different computers if changing keyboards and mice is not perceived to be a problem (one software package with 4 monitors on one computer, another with 4 monitors on another, as opposed to running all in just one box with 8 monitors).
bl33p is right. but for someone running a dinosaur of a computer (pc3200 ddr ram = probably a old pentium 4 or athlonxp system), the few mhz in speed difference will be most likely unnoticeable. The hours you will waste running test and benchmarks, could be spent working at Macdonald's so you could get a new computer with your 1st paycheck instead.
FSB redesign is coming in the 32 NM chips in a year or so. Could be interesting. http://pcworld.about.com/news/Dec172004id118986.htm