DX12 Support in PCSX2
#41
Um... he did say what it means - it's made by the author of the dx12 port by the way - and he also attached the raw data. It's an FPS performance comparison where dx11 is the base reference, that's why it is always 100% so you can see the percentage increase directly.

https://forums.dolphin-emu.org/Thread-un...12-backend

I did not mention it because I thought it was obvious, I guess I was wrong. The source and a build are already there, so you can test it yourself, and find all the mins, maxes, medians, and standard deviations till your heart's content.
Reply

Sponsored links

#42
and your the one comparing emulator

Dolphin is more GPU orientated PCSX2 more CPU orientated
Reply
#43
(12-24-2015, 01:27 AM)tsunami2311 Wrote: Dolphin is more GPU orientated PCSX2 more CPU orientated

That's not true. Dolphin is more GPU demanding yes, but it's CPU demands are pretty close to PCSX2.
[Image: XTe1j6J.png]
Gaming Rig: Intel i7 6700k @ 4.8Ghz | GTX 1070 TI | 32GB RAM | 960GB(480GB+480GB RAID0) SSD | 2x 1TB HDD
Reply
#44
(12-24-2015, 12:42 AM)K.F Wrote: I did not mention it because I thought it was obvious, I guess I was wrong.

It was obvious. The problem is just that the data was manipulated in a way that one can not draw any conclusion from it any more. Why should dx11 be a reference and not for example full speed. This test could've been performed with 2FPS or 2000FPS and in both cases that benchmark doesn't need to be a good measure for full speed performance. Is the used pc and configuration a good measure for api performance? Maybe the results would vary even more but unfortunately there was a cpu bottleneck, disk access or an fps limit. Maybe in dx12 the fps vary much more so that sometimes you get drops to 2FPS and sometimes you get increases to 200FPS and only in average it is better.

Also if dolphin is nearly as unstable as pcsx2 regarding fps the error bars should be large. There is no point in publishing such a figure if you don't give an error estimate.

And putting several times the same to one normalized bar is imo bad style. But that might be preference. I know that some gpu benchmarks do something similar. But most of the time they provide relative and absolute quantities as well as error margins. (most of the time by min/max)

I didn't want to criticize that you put the figure here. I just have the feeling the creator of this graph had some intention before he made the figure. And that the results are not as good (or maybe even better) than presented.
Reply
#45
brag graphs? wtf. measure fps... pls. limit is 60. you gotta score the low and running nice. above is irrelevant. neh? Biggrin

so bore... sorry. Smile
Reply
#46
(12-24-2015, 02:37 AM)willkuer Wrote: I just have the feeling the creator of this graph had some intention before he made the figure. And that the results are not as good (or maybe even better) than presented.

(12-24-2015, 12:42 AM)K.F Wrote: The source and a build are already there, so you can test it yourself

Prove that he did.

(12-24-2015, 02:45 AM)dabore Wrote: brag graphs? wtf. measure fps... pls. limit is 60. you gotta score the low and running nice. above is irrelevant. neh? Biggrin

I'm sorry, I don't understand what you wrote. I'm guessing you say that more than 60 fps is useless? Ok, but on what machine? Better performance means more machines - and more people - can use the emulator.

Did this thread degenerate to the level that I need to explain the importance of performance?
Reply
#47
Dolphin number show that AMD is quite slow. Anyway, raw state are nice.
Quote:Nvidia GTX 770 DX11 DX12 OpenGL
Crazy Taxi 315 434 404
Zelda OOT 54 54 54
Zelda Twilight Princess (HF) 16 23 9
Super Mario Galaxy 72 115 91
Only one game really get a big boost but it remains very slow.

So it would help a couple of games that are draw call limited (ZoE 2 for example) but it won't solve all the texture caches issues. I would prefer a Vulkan solution only (so we can drop dx/gl code).

It depends on what are the priority of the project. From one hand, you can have a 10% speed boost, from the others hand you can have more effect such as lighting / mipmapping / shadow. IMHO, priority must be accuracy.
Reply
#48
(12-21-2015, 07:12 PM)NoleNation Wrote: Yeah I know some guy is experimenting with it on this forum page

https://forums.dolphin-emu.org/Thread-un...12-backend

Read through the comments. A lot of people are having major performance boosts and reduced cpu usage in many games. This is awesome for emulation in general.


Awesome for emulation? Nah not really. Not when you only have about 10% of users using it. Not only is it a small minority but I actually hope it stays that way. I'm still using win7 because win10 is a buggy piece of garbage at the moment and it's causing all kinds of problems in games and other apps for a LOT of people.

I don't ever plan on upgrading to win10 and will stick with 7 until the dinosaurs decide to come back and proceed to roam the Earth. The fact that Micro$haft *REQUIRES* users to own Winblows 10 in order to take advantage of Dx12 is absolutely *****! That alone was enough to piss me off and not support their latest OS. They can take win10 and stick it where the sun don't shine for all I care.

Windows 7 for life!
Reply
#49
(12-24-2015, 12:01 PM)Isamu Wrote: Awesome for emulation? Nah not really. Not when you only have about 10% of users using it. Not only is it a small minority but I actually hope it stays that way. I'm still using win7 because win10 is a buggy piece of garbage at the moment and it's causing all kinds of problems in games and other apps for a LOT of people.

I don't ever plan on upgrading to win10 and will stick with 7 until the dinosaurs decide to come back and proceed to roam the Earth. The fact that Micro$haft *REQUIRES* users to own Winblows 10 in order to take advantage of Dx12 is absolutely *****! That alone was enough to piss me off and not support their latest OS. They can take win10 and stick it where the sun don't shine for all I care.

Windows 7 for life!

Having to code something for 3 or 4 outdated OS would be asinine. It's a waste of time and resources. Not to mention the only thing wrong with my windows install is Nvidia seems to be unable to produce a decent driver. But that really isn't any fault of Microsoft. And performance wise 10 runs circles around 7. As with ANY new piece of software, sure there are some bugs. But within 6 months or so 7 will be totally irrelevant.
Desktop:
Ryzen 7 5800X | MSI MAG CORELIQUID 360R | MSI X570 unify | 32GB DDR4@3600 MHz (14-14-14-34) | EVGA RTX 3070 FTW3 Ultra
1TB + 2TB WD Black SN770 | Windows 11 Pro x64
Reply
#50
(12-24-2015, 12:01 PM)Isamu Wrote: Awesome for emulation? Nah not really. Not when you only have about 10% of users using it. Not only is it a small minority but I actually hope it stays that way. I'm still using win7 because win10 is a buggy piece of garbage at the moment and it's causing all kinds of problems in games and other apps for a LOT of people.

I don't ever plan on upgrading to win10 and will stick with 7 until the dinosaurs decide to come back and proceed to roam the Earth. The fact that Micro$haft *REQUIRES* users to own Winblows 10 in order to take advantage of Dx12 is absolutely *****! That alone was enough to piss me off and not support their latest OS. They can take win10 and stick it where the sun don't shine for all I care.

Windows 7 for life!

Well, they decided to build dx12 around their new driver model. That's why it requires win 10. Vulkan not requiring it will be nice, but from what I hear it will also benefit from wddm 2.
Reply




Users browsing this thread: 1 Guest(s)