4850 Memory speed....vs Core

[nanohead]

OK. I have one 4850 who has lame memory somewhere. So looking through the OC threads, a pattern seems to have emerged where high Cores are with lower memory, and higher memory is with lower cores. This is my non statistical, caffeine influenced analysis, but it seems to be somewhat consistent.

The great question is, IS there any benefit to memory speed increases over core?

If I try to raise both, 3Dmark will blow up with some DirectX error. I'm now experimenting with lowering memory, and raising core. I suspect that the VRM section is only capable of satisfying a total voltage demand, and maybe if both core and mem are raised substantially, the whole thing falters... I dunno....

Right now, I've reduced the mem to 1000, and raised core to 740 and it crashes. So it seems that 730/1050 is the max for my pair at least at the moment

 

[Ketxxx]

They in a PCIE or PCIE 2.0 slot? I found my 3870 had rubbish memory clocking ability in a PCIE slot, but moving to PCIE 2.0 it now clocks pretty well. I think it has something to do with the backward compatibility and PCIE not being able to provide as much power via the slot as PCIE 2.0 can.

 

[nanohead]

All PCIe 2.0. Not sure how that would affect memory speed, but its worth a shot to play around with it for sure. I'm using a DFI LANParty 790FX board, and you can specify power to each PCIe slot, as well as set whether its running at either 1.1 or 2.0 PCIe spec. I might play around with it a bit. Worst that can happen is I'll cause a fire and have to buy all new parts