MacPro Memory Performance real life

[florianuhlemann]

I have just completed a little test series of the latest iMac with the latest UP mac pro.

W3680 with 24GB RAM vs i7-2600 with 16GB RAM

the iMac has an average gain of 25,7% in performance (using xbench as platform) compared with the MacPro.

I'd still just say.. WHY??? dual-channel 1333mhz is faster than triple-channel 1333mhz doesnt go in my head that way. but nevertheless...

Where will I see a difference in performance (real-life) in that matter?

Cheers
Florian

 

[macpro2000]

Really amazing that the iMac has surpassed the king of Apple by that much.

I have just completed a little test series of the latest iMac with the latest UP mac pro.

W3680 with 24GB RAM vs i7-2600 with 16GB RAM

the iMac has an average gain of 25,7% in performance (using xbench as platform) compared with the MacPro.

I'd still just say.. WHY??? dual-channel 1333mhz is faster than triple-channel 1333mhz doesnt go in my head that way. but nevertheless...

Where will I see a difference in performance (real-life) in that matter?

Cheers
Florian

 

[florianuhlemann]

it's just hard to imagine why. if you'd even compare it to amd processors, the performance of DDR3-1333 memory is so different across each platform. with the sandy bridge on top, the mac pro second and no amd processor will reach even part of the performance of the intels. how can the same memory be so much different in speed?

 

[Umbongo]

I'd still just say.. WHY??? dual-channel 1333mhz is faster than triple-channel 1333mhz doesnt go in my head that way. but nevertheless...

Because Sandy Bridge's memory controller and the interconnect that links CPU cores, memory controller and cache are superior to that of Nehalem and Westmere.

 

[florianuhlemann]

Because Sandy Bridge's memory controller and the interconnect that links CPU cores, memory controller and cache are superior to that of Nehalem and Westmere.

so they never used the full performance of it in the first place?!

 

[Umbongo]

so they never used the full performance of it in the first place?!

Nope, which is why often it was below the theoretical limits of the parts. Dual channel sandy bridge gets better memory performance in some areas than quad channel E/EP too due to overheads that are present.

You might find this link interesting:
http://macperformanceguide.com/Reviews-MacProWestmere-MemoryBandwidth.html

The reality is that if you can use the memory capacity then having it beats any optimisations with a lesser amount for real world performance.

 

[derbothaus]

Xbench is crap. Differing results all over the place. Don't count on it for becnhmarks. I just ran a memory test and it told me a G5 is faster.

 

[florianuhlemann]

Hmm. can't say that. Depending on the testing environment and limits, it can be quite useful. I have had similar results with geekbench.