Imagine how smooth Doom 3 could be if Intel would finally address the effective memory bandwith issue.
But go ahead and waste your bucks (or your daddies)
on pooly designed hardware.
hheellpp - please assume that I do know that a 2GHz CPU can do more computations than a 1GHz clocked one.
The point is that they are only able to finish the workload in the time wich is given by the effective memory bandwidth.
When the CPU waits 200 cycles on average to go to memory and back it performs exactly the same amount of payload effective operations than a CPU with half the clockspeed wich still waits 100 cycles to go to memory and back.
DMA is limited by the RAM speed as well.
The @home applications may be more memory bandwidth agnostic than everything but they weren't my intention when I bought my gear.
The reason a processor with half the clockspeed shows the same performance is that both processors just wait for the magnitude slower clocked memory.
Why waste 1000$ for the current top Intel Chip?
A 500MHz CPU does the same job when working with the same RAM/Chipset.
Try this: throttle down the CPU clock while keeping memory at full speed (if that is possible with your system;) - you won't notice a performance drop for quite a while...
Since there is about a magnitude difference between CPU and memory speed there is only one thing making sense to me - bringing them closer in geometry and clockspeed.
see this approach with nice piccies
Slashdot Mirror
Imagine how smooth Doom 3 could be if Intel would finally address the effective memory bandwith issue. But go ahead and waste your bucks (or your daddies) on pooly designed hardware.
The point is that they are only able to finish the workload in the time wich is given by the effective memory bandwidth.
When the CPU waits 200 cycles on average to go to memory and back it performs exactly the same amount of payload effective operations than a CPU with half the clockspeed wich still waits 100 cycles to go to memory and back.
DMA is limited by the RAM speed as well.
The @home applications may be more memory bandwidth agnostic than everything but they weren't my intention when I bought my gear.
The reason a processor with half the clockspeed shows the same performance is that both processors just wait for the magnitude slower clocked memory. Why waste 1000$ for the current top Intel Chip? A 500MHz CPU does the same job when working with the same RAM/Chipset. Try this: throttle down the CPU clock while keeping memory at full speed (if that is possible with your system ;) - you won't notice a performance drop for quite a while...
Since there is about a magnitude difference between CPU and memory speed there is only one thing making sense to me - bringing them closer in geometry and clockspeed. see this approach with nice piccies