Posts: 1.165
Threads: 16
Joined: Mar 2009
Reputation:
0
Location: Alexandria, VA
SSE4a is not SSE4/SSE4.1
SSE4a only includes a few instructions from SSE4, none of which are useful for emulation purposes.
Posts: 298
Threads: 16
Joined: Mar 2009
Reputation:
0
Location: Russia Moscow
thanx for the advises.
Will i get some reasonable fps if i buy discrete sound-card? Few years ago build-in sound could lower performance in games, what about now?
Phenom II X4 940 3 Ghz / 8 gb RAM 800 / Geforce GTX 460 / win7 64
Posts: 240
Threads: 1
Joined: Jun 2009
Reputation:
0
Almost no change in performance what so ever as far as I have seen, switching between my integrated RealTek HD-audio and my Creative X-Fi Gamer (the one with dedicated 64 mb ram).
Posts: 1.452
Threads: 64
Joined: Nov 2008
Reputation:
17
Location: Meshoppen, PA
06-22-2009, 12:58 PM
(This post was last modified: 06-22-2009, 01:08 PM by Air.)
Soundcard type/quality is irrelevant for PCSX2, which must do all mixing in software (via your cpu). In fact, fancy "quality" sound cards are becoming increasingly irrelevant under Windows too -- Vista itself has no support for hardware soundcards at the DirectX/Xaudio2 level, and does everything in software regardless of how fancy your soundcard happens to be. There are in fact a lot of good reasons for this so don't get mad at Microsoft. I coded audio mixers and midi/module players for years and hardware-accelerated soundcards have always been a royal pain to work with. And with modern mutli-core CPUs, the overhead of a mixing is miniscule at worst.
Even in Pcsx2, the mixer is practically nothing for overhead. The problem is the emulation stage, which must do quite a bit of volume and reverberation logic on a per-sample basis (48000 times a second). Most mixers for PC games can mix in large "chunks" that occur about 80-120 times a second, and the CPU can handle that much more efficiently. Even on an especially fancy PC game that uses 64 voices and reverb, overhead wouldn't be more than 5% -- and that's 5% that can typically run in parallel with the gfx rendering thread (which always has periods where it stalls and waits for the video card to enter a ready state for the next set of gfx commands). So in other words, 0% overhead on any modern multi-core or hyperthreaded cpu.
Jake Stine (Air) - Programmer - PCSX2 Dev Team
Posts: 1.452
Threads: 64
Joined: Nov 2008
Reputation:
17
Location: Meshoppen, PA
I guess you missed the part where I said hardware accel is fairly irrelevant. Sure game developers can waste their time implementing a more complex API like OpenAL for little or no benefit, just to satisfy your ego and poor monetary decision making when it comes to upgrading your computer. Or they can use a very simple Xaudio2 API which provides a high quality software mixer, and which gives them infinite more options for streaming audio manipuations, and is also less prone to latency than using a hardware accelerated soundcard. In fact, to some extent software mixing can actually be faster than hardware. Many modern games do a lot of realtime audio streams and for those they usually have to manually upload each stream to the soundcard anyway, as it plays. Since soundcards are typically not plugged into the PCI Express bus slot, the uploading process can be quite slow and fairly system-intensive.
Also if you keep spamming the forums with one word/one sentence snappy replies you will get a warning.
Jake Stine (Air) - Programmer - PCSX2 Dev Team
Posts: 298
Threads: 16
Joined: Mar 2009
Reputation:
0
Location: Russia Moscow
we are falling into the offtopic...
so, AIR, there is no way to implement some small benchmarking feature to detect the bottlenecks?
Phenom II X4 940 3 Ghz / 8 gb RAM 800 / Geforce GTX 460 / win7 64