Why use memory risers?

[timmoseus]

What is the advantage of using memory risers? Say I can have "x" amount of RAM without memory risers, and the same "x" amount of RAM using memory risers, what would the performance difference be? Instinct tells me not to use memory risers as the bottle neck would become the original RAM modules on the board. For example, I can run 8GB X 8 PC2 5300F on the board, or 4GB X 16 PC2 5300F using the memory risers. Either way, the board will not use more than 64GB of RAM. 8GB X 16 PC@ 5300F was seen by the board and OS, but only 64GB was usable by the board and Windows 7 Ultimae x64. Bottom line, should I use the 8GB X 8 sticks, or 4GB X 16 sticks using 4 memory risers? Which has the faster performance?

 

[Anonymous09]

If money is no issue, the 4 stick solution is going to be better just by the definition of using riser cards.

 

[timmoseus]

If money is no issue, the 4 stick solution is going to be better just by the definition of using riser cards.

It is not a question about money, these are all donated parts in a refurbishing shop. I'm just wondering if there is any hard specs on using four risers over eight sticks of RAM? Do the risers operate at faster speeds than the individual RAM sticks? It seems to me that plugging a four DIMM riser into one DIMM slot would bottle neck the throughput. Either way, I still have 64GB of RAM and eat up any benchmark tests that I have come across. I was just hoping for any kind of specs as to why I would run 4GB X 16 sticks through four 1-4 DIMM risers as opposed to running 8GB X 8 sticks straight through the native DIMMs on the board.

 

[timmoseus]

Does anybody know anything about this? I still have not been able to find any specs about riser speeds/performance. Dell's official answer is that risers allow one to double memory capacity. I can reach the motherboard's limit without them though, so would it make sense to even install them? Doubling the motherboards max leaves me with an extra 64GB of RAM that is seen by both the BIOS and OS, however is unusable by either. I would really appreciate any thoughts you might have.

I am trying to maximize this old Dell Precision 690 to be my "all-in-one" imaging server, LAMP test server, NAS, VM Linux tester, etc. So far I have two Xeon X5365 3.0GHz CPUs, 64GB PC2 5300F RAM, two 128GB SSDs striped for the primary OS, and four 300GB 15Krpm SAS drives for storage (probably going to stripe them as well since the primary function is a test environment with imaging capabilities).

 

[timmoseus]

So I ran Nova Bench against the two different RAM configurations five times each and here are the results.

64GB (4GB x 16 on risers) PC2-5300F = average Transfer Speed 2640.2 MB/s
64GB (8GB x 8 without risers) PC2-5300F = average Transfer Speed 2689.6 MB/s

So my final conclusion is that memory risers primarily serve to increase capacity only. There actually is a performance drop, as minor as it is, when using memory risers.

The transfer speeds for the five individual tests are as follows (from slowest to fastest):

64GB (8GB x 8 without risers)
2671 MB/s
2689 MB/s
2694 MB/s
2696 MB/s
2698 MB/s

64GB (4GB x 16 on risers)
2611 MB/s
2616 MB/s
2624 MB/s
2675 MB/s
2675 MB/s

This is the best answer that I have. Would anyone else like to suggest other methods?