Well, I read about an idea once on here about Dolphin loading the full game in RAM to provide a speed boost, rather than Dolphin interacting with the GC or Wii ISO via HDD. GC games are only about a couple or so GB I believe, and the average amount of RAM on a modern PC built post 2009 is about 4-8GB of DDR2/DDR3 memory. The idea is to keep the interaction between the application on RAM, the CPU, and the GPU, seldom having the need to transfer data from mechanical storage solutions. Opinion(s)?
OpenGL performance regression on (ATI/AMD) Radeon GPUs since 4.0-1778
|
11-05-2014, 06:58 AM
(11-05-2014, 06:58 AM)ExtremeDude2 Wrote: That would possibly speed up load times but nothing else. I believe the performance would be more general than that. The GPU and RAM use Solid State memory, and have almost a direct access to the processor. And the performance gain would not be exclusive to AMD, Nvidia, or Intel graphics solutions. 11-05-2014, 07:33 AM
The Wii doesn't process data directly on the disk, it loads just the bit it needs into RAM, and then does stuff to it. As well as that, it expects a certain delay between asking to load the data and the data actually showing up. I'm pretty sure the devs have had to artificially slow down loading to stop some games crashing, so moving the whole game to a ramdisk is sort of unnecessary.
OS: Windows 10 64 bit Professional
CPU: AMD Ryzen 5900X RAM: 48GB GPU: Radeon 7800 XT
To all users who have benchmarked the builds and can't reproduce the issue:
Are you using the optimal GPU driver settings for Dolphin? * In the Catalyst Control Center, you should set the Texture Filtering quality To 'Very High' (the default is 'High') * Surface Format Optimizations should be turned OFF (the default is ON) * OpenGL Triple Buffering should be set to OFF (default is OFF) * All other settings should be set to 'Application Controlled'. You should NEVER force Aniso Filtering or Anti-Aliasing through the driver. If you're running Windows 7, you should aso disable desktop composition (DWM, Aero Glass, UxSms) NOTE: It's not possible to do this in W8, 8.1 or W10 TP. DWM is always enabled and can't be turned off without breaking the OS.
How to do a proper Dolphin benchmark run: a short guide
------------------------------------------------------------------------------ When benchmarking Dolphin, you should be aware that the real-time shader cache compilation / caching can cause massive framerate fluctuations and make the results invalid. Before doing a benchmark, you'll have to "warm up" the cache by doing 3 runs, so the shader cache is fully built up and copied to system RAM. - In the first run you'll experience severe stuttering / pausing and frequent disk accesses as the shaders are compiled and cached to disk. - In the second run, you won't experience stuttering issues, but Dolphin will still access the disk frequently to load the already precompiled shader cache. - The third run will be smooth and the framerate will go up a bit, as the OS has already copied the shader cache to system memory for fast access. After running Dolphin 3 times to build up the cache, you may start the real benchmark by doing at least 5 additional runs, since the performance can vary between runs. Record the results of each run and post them here in this thread.
And, finally, Part 2 of the benchmark results:
SMG_observatory_121s_complete_test OpenGL @ 4xIR, EFB to Texture, borderless fullscreen, framelimit OFF. FPS average of 7 runs: --------------------------------------------------------------------------------------------------------------------------------------------------------------------------- 4.0-1778 (non-coherent with GL_ARB_Buffer_Storage) = 34 fps 4.0-2010 (non-coherent with GL_AMD_Pinned_Memory) = 40 fps 4.0-1769 (coherent mapping) = 43 fps ...and just for comparison - Direct3D with the same settings at 6xIR = 53 (!) fps - a HUGE difference. |
« Next Oldest | Next Newest »
|
Users browsing this thread: 1 Guest(s)