Is there a core clock/memory ratio

[dwell]

I don't know much on the subject, but I noticed that core clock and memory in cards are always in ratio where this memory is always ~1.3x as fast as the core.

Can throwing that ratio off hurt performance? I'd image is the clock were faster than the memory, performance would take a hit, but what about the other way around? I'd imagine it wouldn't because the driver is the core, not the memory.

Just asking because I can get my core up to 600Mhz but the memory does 800 (1.6Ghz effective) and it's pushing the ratio.

 

[Matthias99]

The 'ratio' shouldn't matter very much, since they're not clocked in any particular synchronous way (unlike CPUs, where both the RAM and CPU core run at a fixed multiplier of the FSB). If you up either one, performance should go up.

Most cards have fast enough memory at stock that it can feed the GPU enough data for it to stay busy. So upping one or the other a whole lot may produce diminishing returns -- if the GPU core is too fast, it won't be able to send/receive data fast enough and will waste many cycles, and if the RAM is too fast, it will just spend a lot of cycles sitting idle while the GPU works on data.

 

[chizow]

Performance should never "take a hit" if you're increasing one or the other, or both, but as Mathias said, you'll see diminishing returns if you're increasing the one that isn't the bottleneck. Depends on the chip architecture and # of pixel pipes/VS etc. in a lot of cases, but increasing bandwidth tends to give diminishing returns sooner than increases in GPU clock speed.

As for a 1.3 ratio, I don't think that really holds up does it? Seems in most cases, manufacturers work with the limitations of the current yield of GPU and then pair it up with the fastest or most affordable memory to try and hit theoretical bandwidth numbers.

 

[BassBomb]

There is no ratio.

My RAM is 2x my GPU