It was more then 17%. The difference between 666 and 800 is 17% but then you add in he was running in single channel mode that means he was running at half his potential bandwidth minus another 17% on top of that.MagicManICT wrote:Boosting the speed was likely the change that worked. After all, you're talking about a 17% boost in throughput.
Coles should be able to go in and explicitly set single channel mode to test. (It's an ASRock board, and these are designed for custom builders.)
Certainly; but not only bandwidth, since dual-channel operation decreases effective latency as well by allowing multiple outstanding loads to complete in parallel, so I don't doubt that it makes a difference. Nevertheless, I would find it very fascinating if memory operations are the actual bottleneck for client performance.ZeggyZon wrote:Up to a point memory bandwidth does matter.
You are most likely correct. I'm sure your client can use some more optimization but I cant tell in what. When I'm in a village with many details and stockpiles I cant tell what the client is drawing the most. Are there many detailed wire frames under those textures? I ask because I can load up diablo3 at 1600p and run it at 60fps no problem. I know haven is only developed by 2 guys and diablo3 is 100x that but still there is a big difference.loftar wrote:Certainly; but not only bandwidth, since dual-channel operation decreases effective latency as well by allowing multiple outstanding loads to complete in parallel, so I don't doubt that it makes a difference. Nevertheless, I would find it very fascinating if memory operations are the actual bottleneck for client performance.ZeggyZon wrote:Up to a point memory bandwidth does matter.
