12-17-2008, 06:41 AM
(This post was last modified: 12-17-2008, 06:47 AM by BigIg.)
Look at what it takes to emulate PS2 decently. Around a 3GHz Dual Core processor with 2gb of ram and a decent GPU. I figure it'll eventually get optimized to the point that a 3GHz dual core processor should be able to emulate everything at near-full speeds.
Now think of computers in four-five years. I'm sure the "standard" computer will have more than a 3Ghz dual core processor 2gb of ram and a decent GPU. So in four-five years a so called "low-end" PC will probably be high enough to reach good speeds.
As for compatibility, I'm betting in four to five years it'll be nearly as high as PSX emulation.
Of course this is all speculation and PS2 emulation could progress either much more or much less than expected.
as time progresses and more people contribute and add to the project, and increased understanding of how the PS2 works, sure it will get faster and more compatible
but for the time being its truly amazing what this emulator can do, id never thought i could play a game i bought last year on an emulator so soon, well done to you.
Uhm some of the things you said are imprecise, FireFlower. So let's try to clear things up.
About more threads, that's probably not going to happen. It's not a matter of "can be done", it's a matter or "is it worth it?". Keep in mind that adding another thread increases the complexity and requires more syncing. So, unless a sizeable part of the emulation process is found that would be sensible to put in a new thread, it won't be done. As things are now, this doesn't seem likely.
About plugins, I'd like to point out that 99% of the bugs are not due to the plugins. The bugs are in the emulator proper, and some of those bugs are EXTREMELY hard to solve. So there is really no guarantee they'll ever be solved...
And let's not go into the whole undocumented stuff problem XD Like memcards, no documentation, complete reverse engineering.
I doubt clock speeds will increase much in the near future, although more and more cores are being added though. On the compatibility part, even having 50-60% would be more than enough(as long as that percentage includes the most popular games). On the whole, I would like to see more Japanese game support. Btw is there a Japanese PS2 emulation scene >_>?
There are three areas of the CPU that can be threaded additionally to the existing MTGS:
* The IOP (would include the SPU2 processor also)
* DMA Transfers.
* The VU0/VU1 Micro-modes
The IOP/SPU2 usually accounts for about 10-20% of the emulator's workload on average, so threading it would give approximately 6-12% speed increase. Not too impressive. DMA transfer workload varies greatly from game to game. However, running DMAs on a separate thread may speed up the EErecs simply because it would help keep the L2 cache cleaner. VU Micro-mode programs also vary widely from game to game -- many games hardly use the VUs at all while a few use the VUs for everything from simple geometry to complex Artificial Intelligence. For the latter games, having a threaded VU unit would definitely speed things up big-time.
But yeah, adding more threads to the emulator will make it that much more work to debug and maintain. We really need to get it more stable first. :/
Jake Stine (Air) - Programmer - PCSX2 Dev Team
(12-17-2008, 08:23 PM)char Wrote: I doubt clock speeds will increase much in the near future, although more and more cores are being added though. On the compatibility part, even having 50-60% would be more than enough(as long as that percentage includes the most popular games). On the whole, I would like to see more Japanese game support. Btw is there a Japanese PS2 emulation scene >_>?
There's work being done on materials that would allow clock speeds to be pushed much higher, graphene (sp?) transistors have been constructed that can operate at 25 Ghz and up. Getting these sorts of materials into production for consumers is a whole other story
In 5 years, anything can happen. In 2003 the standard desktop had 512 MB RAM and a 2.0Ghz Pentium 4, which is almost painfully underpowered compared to today's hardware. Who can say what will happen?
I'm hoping for at least 4.5Ghz dual cores with RAM speeds scaled to match. Graphics hardware with no less than 256 stream processors and a hardware bus available to move as much data as physically possible. Dedicated hardware links running at some appreciable fraction of the CPU's clock rate. A programming paradigm that seamlessly allows parallel processing on daughter cards (ala PhysX) and graphics hardware at the same time.
No matter what happens, it's going to be awesome. 5 years isn't what interests me, it's 15 and 20 years that I can't wait for. In 1998, could you have imagined what we have today?
"This thread should be closed immediately, it causes parallel imagination and multiprocess hallucination" --ardhi