5:4 251FSB or 1:1 240ishFSB?

As you all know, due to the limitations of RAM speed, it's extremely rare to find a 250MHz stick that's capable of tight timing. Opinions needed here:

Assuming a rock solid stable system (Prime95 48 hours+) in both cases and the tightest mem. timing (2:2:2:5)

5:4 251FSB/200 Mem or 1:1 240ishFSB/240ish Mem.?

I think the 1:1 will be better after all.
prbably the cpu marks will be lower but I assume it's fast enough

try to bench it yourself

For cpu dependant tasks, the extra clock rate of the 5:4 configuration will dominate. For applications that are sensitive to memory performance, the 1:1 setup would be preferable. Nearly all applications are more sensitive to cpu clock than they are memory performance. But as Suma pointed out, we aren't going to stop you from performing some tests and actually learning something.

Well....going to the bench it is then......Do the OC Community like PCMark2002, or SANDRA better?

As far as a performance test, PCMark's memory test is pretty good. But you have to understand that you are testing only memory substem performance. To put the results in perspective you need also run some 3D stuff and some real applications.

I run sandra only to verify clock rates and check other settings it shows reliably. As a benchmark it is well neigh useless, but if may help you spot a problem if the scores come out really wacky.

Run PCMark2002. It is a good test.

Run 3DMark2001SE. It will help to put the PCMark results in perspective. 3DMark2003 is so insensitive to core performance that it is poor for this use. I also run Quake3 Arena timedemos and do a SETI benchmark unit to further illuminate which setup is faster, and where. If you do much video or audio encoding you may wish to use that as part of your evaluation.

But the important thing is to realize that your are building a computer system, not just a memory subsytem. The most important tests are those that correspond to what you do with your machine, and for which tasks speed matter to you. Myself I play entirely too much Quake3, and Q3 needs speed. So I tend to optimize my system for best Q3 results, and let the other results fall where they may.

larva said:

As far as a performance test, PCMark's memory test is pretty good. But you have to understand that you are testing only memory substem performance. To put the results in perspective you need also run some 3D stuff and some real applications.

I run sandra only to verify clock rates and check other settings it shows reliably. As a benchmark it is well neigh useless, but if may help you spot a problem if the scores come out really wacky.

Run PCMark2002. It is a good test.

Run 3DMark2001SE. It will help to put the PCMark results in perspective. 3DMark2003 is so insensitive to core performance that it is poor for this use. I also run Quake3 Arena timedemos and do a SETI benchmark unit to further illuminate which setup is faster, and where. If you do much video or audio encoding you may wish to use that as part of your evaluation.

But the important thing is to realize that your are building a computer system, not just a memory subsytem. The most important tests are those that correspond to what you do with your machine, and for which tasks speed matter to you. Myself I play entirely too much Quake3, and Q3 needs speed. So I tend to optimize my system for best Q3 results, and let the other results fall where they may.

Click to expand...

That sound like a solid benching guide line and reasonable, thanks!