RAM performance speed

[DeadSirius]

All things being equal and proportionate, does a 1 GB FB-DIMM and a 2 GB FB-DIMM perform at the same speed? If I load up my Mac Pro with 8 x 1GB RAM, does it perform the same/better/worse than if it had 4 x 2GB RAM? (For the sake of discussion, assume that they're configured optimally.)

 

[rgomez]

No. The Mac Pro has 4 channels of memory, then, the optimum configuration is just with 4 dimms. Having more than 1 dimm per channel adds some latency to the mix, as the dimm needs to pass the signals to the next one in the channel. Don't really know if it's noticeable, but that's the theory.

 

[DeadSirius]

So you're saying that (in theory) 4 x 2GB performs faster than 8 x 1GB?

 

[crazzyeddie]

So you're saying that (in theory) 4 x 2GB performs faster than 8 x 1GB?

Yes. Technically speaking it will.

 

[disconap]

Technically agreed. But you are not likely to notice the difference.

 

[wsgroves]

I def. could be wrong but it was my understanding that the more you populated the slots, the more bandwith you would have but with lower latency and the less slots you had occupied would mean faster latency but less memory bandwith.

 

[aiongiant]

that sounds about right

but the real argument is i think

4x512mb faster than 4x1gb?

4x(anything) perform as well as 4x1gb?

 

[CanadaRAM]

that sounds about right
but the real argument is i think
4x512mb faster than 4x1gb?
4x(anything) perform as well as 4x1gb?

Any tiny difference in the latency of addressing a 1 Gb module vs a 512 Mb module will be overwhelmed by the difference in the performance of a machine with 2 Gb RAM vs 4 Gb RAM assuming you are using programs that use more than 2 GB of working space.

If you aren't -- then the whole discussion of whether the RAM is a microsecond faster or slower is meaningless. We are talking about increments of one or two 667,000,000 ths of a second here

The only valid argument is what the speed difference is in addressing 2 modules vs. 4 modules vs. more than 4 modules. Apple has said 4 modules with 256-bit access is faster than 2 modules with 128 bit access. Some have guessed that having additional banks of RAM would take fractionally longer to address, so that 6 or 8 modules may be slower than 4 and only 4 modules.