Dolphin, the GameCube and Wii emulator - Forums

^See what I mean?

You guys keep saying that when efb to ram is ticked, it will significantly affects the fps in games with ram or memory speed when paired with the Core i7. How about providing some proof with screenshot or video to back up what you guys said or claims?

Hey xtreme2damax if you're reading this could you please provide him^ with proof. That way he doesn't have to take my word for it.

@tuanming
Integrated memory controller frequency affects it as well.

Some hard numbers would be nice to see. When you say "significant" I think "40 FPS becomes 60 FPS" or "not-fun-to-play becomes smooth" or something like that, which is why I'm interested in this(and I'm sure this discussion will help someone else at some point). I'm also sure it depends on the game as not all games which need EFB to RAM would need it for as much stuff.

Fuzzy terms and computer stuff do not mix. Charts or examples of different games with different ram speeds, however, are great.

BTW, Sandy Bridge CPUs should have more memory bandwidth than older i-series. I saw a chart or article somewhere about it.

Quote:BTW, Sandy Bridge CPUs should have more memory bandwidth than older i-series. I saw a chart or article somewhere about it.

That depends. Bloomfield cpus (9xx series) should still beat sandy bridge in memory throughput/read/write.

Quote:I'm also sure it depends on the game as not all games which need EFB to RAM would need it for as much stuff.

It's essentially all dependant on how many efb copies are needed to render each frame.

For anyone interested here are some benchmarks for 1866MHz 9-9-9-24 CR 1 vs 1600MHz 8-8-8-24 CR 2 on a Sandy 2500k OC to 4.3GHz

1866 --------------- 1600
21,897MB/s ------ 19,642MB/s Read
21,847MB/s ------ 21,307MB/s Write
23,042MB/s ------ 20,695MB/s Copy
40.1ns ------ ----- 45.0ns Latency


The 1866 speed required 1.304 vcore idle and 1.336 vcore load
The 1600 speed required 1.232 vcore idle and 1.264 vcore load (might be able to get that down a little more, I had it running at around 1.220 to 1.246 then my mobo wigged out on me and changed settings)

1866 speed made the processor hit 81c during peak Prime95 load
1600 peaked at 73c during Prime95

(01-31-2011, 12:07 PM)remedy2 Wrote: [ -> ]1866 speed made the processor hit 81c during peak Prime95 load
1600 peaked at 73c during Prime95

!

Excuse my lack of overclocking knowledge, but why would vcore need upped when using higher speed ram. Wouldn't only the vtt need raised? I wonder what the heat/voltage differences would be at stock.

High memory speed = higher clocked memory controller

higher clocked memory controller = higher voltage demands

higher voltage = higher power consumption

higher power consumption = more heat