All us overclocking types know that the best results when overclocking the CPU/RAM/system are obtained by running the RAM at the same speed as the CPU's FSB, so the two are synchronised and there's never anything 'waiting' for a clock cycle while the other catches up.
How come this does not appear to be the case with Graphics Card GPUs and their memory? I haven't done any testing myself, (too lazy :smile: ) but I've never seen anyone mention this.
Anyone have any thoughts/Links about why it doesn't matter so much?
Boredom in the workplace eh? a terrible affliction!
This is probably because the GPU have an integrated memory controller, latency is very low and memory chip are near the GPU so they can achieve higher speed and lower latency easily. CPU can't be this close to main memory.
Another factor GPU are lot slower than Video RAM. This mean that is the memory that wait for the GPU. CPU are much faster than RAM, it's why they need CACHE to at least have a decent "working place" speed.
GPU are alse optimized processor for graphic processing, so their circuitry are made to only compute graphics. The GPU memory controllers are surely made to make sure that the GPU will use the maximum bandwidth even in asynch mode. CPU are made to do lot of stuff, so the memory controller and circuitry are made to be balanced, so this ensure good performance in every kind of task.
Lookin' to fill that <font color=blue>GOD</font color=blue> shape hole!
Hmmm. I guess, what with the on-die mem controllers it's more like an Opteron/A64 config, where you effectively <i>don't have</i> a FSB at all, so I guess that's why it doesn't matter? there's no 'Middle Man' between two buses any more?
a good point about the on-die memory controllers then, TheRod, seems the most plausible thing I guess.