Will GSDX update for DX12?
#31
(03-19-2015, 06:46 PM)xemnas99 Wrote: PCSX2 is a very CPU demanding application.
whoops, a misunderstanding. I meant the factors needed on reducing the cpu overhead. I will edit my posts to make it a little bit understandable.
We're supposed to be working as a team, if we aren't helping and suggesting things to each other, we aren't working as a team.
- Refraction
Reply

Sponsored links

#32
I think the only thing we know for sure is that lower graphics driver overhead would give pcsx2 more cpu time.

Who knows what kind of difference it would actually make.
Reply
#33
The purpose of the new API is to reduce CPU usage to submit data to the CPU.
1/ They remove error checking (good for us, it will reduce latency) but I'm sure error checking is not the biggest part of the driver. Some stuff is already done to reduce this cost anyway
2/ They move the ultra-optimized driver layer that create command into the application.
=> application will need to do the driver job
=> painful to implement but you have the chance to hand tailor this layer for your application instead of a generic implementation (note driver are already MT)
This part gives a huge speed up if you know what the application will do. For example you know texture that are often use on the scene so the important texture to cache.
However there is a terrible issue for PCSX2, we don't know what the application (game) will really do. At least we know a bit GSdx behavior so maybe you can improve it but it won't be easy.

The new API won't remove barriere and flush when we toggle hardware configuration, neither removes texture transfer.
Reply
#34
I'm not sure what you mean by "driver are already MT." You mean Direct3D 12 driver, right?

According to this picture, the UM driver and KM driver of Direct3D 11 aren't multithreaded:
[Image: 2806.cpucompare.png]
Quote:If you’re a gamer, you know what 3DMark is – a great way to do game performance benchmarking on all your hardware and devices. This makes it an excellent choice for verifying the performance improvements that Direct3D 12 will bring to games. 3DMark on Direct3D 11 uses multi-threading extensively, however due to a combination of runtime and driver overhead, there is still significant idle time on each core. After porting the benchmark to use Direct3D 12, we see two major improvements – a 50% improvement in CPU utilization, and better distribution of work among threads.
http://blogs.msdn.com/b/directx/archive/...tx-12.aspx
Reply
#35
(03-19-2015, 09:26 PM)xemnas99 Wrote: I'm not sure what you mean by "driver are already MT." You mean Direct3D 12 driver, right?

Remember that that test was running on an Intel GPU.
I think amd and nvidia's drivers are probably somewhat better in that aspect.
Reply
#36
(03-19-2015, 05:03 PM)Blyss Sarania Wrote: But it will because the base GS emulation is done on the CPU itself.

How much is the actual base? Do you know some further information about that? I always thought that the base gs emulation would be more of data transfer character and not real calculations...
Reply
#37
(03-19-2015, 11:13 PM)willkuer Wrote: How much is the actual base? Do you know some further information about that? I always thought that the base gs emulation would be more of data transfer character and not real calculations...

I don't know tons of specifics, but the GS is a chip in the PS2(you know this of course) and as with all chips in the PS2(e.g. EE, VU) it has to be emulated first on the CPU. The GS output is then rendered with Direct3d/OGL. But the chip itself is emulated on the CPU. I do know from experience that pushing the resolution WILL tax the CPU.

The GS% in the title bar is CPU side usage. E.g. if it's 99%, your CPU is limiting the emulated GS. I've asked the devs this and they confirmed.

In my experience, some games use the GS itself for geometry calculations(or at least, some type of calculations) and others favor the VUs.
[Image: XTe1j6J.png]
Gaming Rig: Intel i7 6700k @ 4.8Ghz | GTX 1070 TI | 32GB RAM | 960GB(480GB+480GB RAID0) SSD | 2x 1TB HDD
Reply
#38
(03-20-2015, 01:06 AM)Blyss Sarania Wrote: In my experience, some games use the GS itself for geometry calculations(or at least, some type of calculations) and others favor the VUs.

I'm not sure the GS can do stuff like that.
Reply
#39
(03-20-2015, 01:50 AM)dogen Wrote: I'm not sure the GS can do stuff like that.

Yeah, I don't know exactly what it's doing, just that it's doing something. I've not looked at the code to see. But it does do something and some games use it and some don't.

E.g. most of Suikoden III hits the VU tremendously hard(99% VU%) where GS% is 0-3%. Other games push the GS to 99% and barely touch the VU.
[Image: XTe1j6J.png]
Gaming Rig: Intel i7 6700k @ 4.8Ghz | GTX 1070 TI | 32GB RAM | 960GB(480GB+480GB RAID0) SSD | 2x 1TB HDD
Reply
#40
(03-20-2015, 02:21 AM)Blyss Sarania Wrote: Yeah, I don't know exactly what it's doing, just that it's doing something. I've not looked at the code to see. But it does do something and some games use it and some don't.

E.g. most of Suikoden III hits the VU tremendously hard(99% VU%) where GS% is 0-3%. Other games push the GS to 99% and barely touch the VU.

Some games use just the ee fpu for vertex stuff.
Reply




Users browsing this thread: 1 Guest(s)