Linked by Thom Holwerda on Fri 24th Jun 2016 22:10 UTC
Games

The long awaited Dolphin 5.0 release is finally here! After nearly a year of bug-hunting and handling the release process, everything has come together for our biggest release yet! The three previous releases followed a very distinct pattern: sacrifice performance, hacks, and features in exchange for higher accuracy. As such, Dolphin 3.0, 3.5, and 4.0 progressively grew slower. But thanks to the cleanups put forward throughout those releases, Dolphin 5.0 is the fastest Dolphin has ever been! By removing all of those hacks and outdated features while cleaning up the codebase, Dolphin has reached a new level of efficiency, powered by a revitalized dynamic recompiler. On the GPU side, OpenGL and D3D11 have seen tons of optimizations and accuracy improvements, and have been joined by a brand new D3D12 backend for huge performance gains. If there's a CPU or GPU extension that can make Dolphin faster, we take advantage of it.

Dolphin is an incredibly impressive project - not just from a technological standpoint, but also from an organisation one. They post regular, detailed development updates, have in-depth release notes that are still entirely readable for laypersons such as myself, and you always learn a ton of new stuff following the project's progress.

A great example of how to run a project like this. Don't forget to check out the release video with tons of side-by-side examples of the long list of improvements.

Thread beginning with comment 630920
To view parent comment, click here.
To read all comments associated with this story, please click here.
raom
Member since:
2016-06-26

Low-clocked xeons aren't useful in dolphin and neither is SLI. I'm sure you know that. A 5ghz skylake should about do it.

Reply Parent Score: 1

Alfman Member since:
2011-01-28

raom,

Low-clocked xeons aren't useful in dolphin and neither is SLI. I'm sure you know that. A 5ghz skylake should about do it.


Yea, I've never used dolphin, but I've experienced this elsewhere.

While the industry has been pushing multi-core processors, it only helps up to a point and in many cases starts to hurt performance. It's often the case that single threaded performance is compromised as the number of cores goes up. Sure enough, this is the case for the macpro line, where the lowest models have a significantly faster speed (4 core * 3.7GHz) while higher spec models use slower core speeds (12 core * 2.7GHz).

One of the programs I'm writing for a client is a multithreaded CPU-bound process that ran well on my local 4 core/8 thread i7-3770 3.4GHz-3.9GHz machine but actually stuttered on a 14 core/28 thread E5-2695 2.3GHz-3.3 GHz machine running in the field. While my program was delegating work to all the cores, the granularity of tasks was such that individual cores could not keep up in real time. I needed to make changes to the multithreading granularity to compensate for the slower core speed.

Reply Parent Score: 2

sergio Member since:
2005-07-06

Low-clocked xeons aren't useful in dolphin and neither is SLI. I'm sure you know that. A 5ghz skylake should about do it.


Interesting, didn't know about that. Taking into account Xeons are idling 80% of the time when I run F-Zero GX, I'm almost sure the problem here is Dolphin's lack of SLI support.

Just for the records, I have D500 not D300. I know it's not a good GPU for gaming at all, but It runs The Witcher 3 1080p perfect under Windows 10... so I think Dolphin is the problem here (OSX is not because F-Zero GX sucks on Windows too).

Thanks for the info man! ;)

Reply Parent Score: 2

FlyingJester Member since:
2016-05-11

SLI is extremely tricky to take full advantage of in the first place. When emulating another GPU, I would be very surprised if it could be used very well.

Emulation is inherently not very parallelizable. Asking it to be somehow parallel on the GPU is kind of ridiculous.

Reply Parent Score: 1