User(s) browsing this thread: 1 Guest(s)

Post Reply 
 
Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
SSE4 vs AVX
Author Message
Alexander Offline
★ Blah ★
***

Posts: 577
Joined: Sep 2011
Location: Scottsburg, Indiana
Reputation: 16
Post: #1
SSE4 vs AVX
I'm going to be conducting a few tests between SSE4 and AVX because I was interested in seeing if there is any difference.

Note : Running on GSDX11 Hardware Mode @ 1920 x 1080 resolution.

So far, AVX has seemed to be a bit doing a bit better than SSE4. The test was on Kingdom Hearts 2 Final Mix.

SSE4 - About 280FPS Average (270 ~ 310)
AVX - Usually around 320FPS Average (290 ~ 350)

But the problem is that the framerate fluctuates far too rapidly for me to test well, so I'm going to try a more demanding game... ZOE 2 will be my test subject.

I'll post further results in a few hours. (Btw, playing without the framelimiter can be kinda fun. In-game fights are such a...seizure XD)

If you have any recommendations for benchmarking please let me know.

[Image: signat11.jpg]

My Guide to PCSX2 for Windows (outdated, but still generally accurate)
(This post was last modified: 05-30-2012 06:09 AM by Alexander.)
05-30-2012 06:01 AM
Visit this user's website Find all posts by this user Quote this message in a reply

Sponsored links

smartstrike Offline
---
****

Posts: 2.498
Joined: Dec 2008
Location:
Reputation: 43
Post: #2
RE: SSE4 vs AVX

Main Hub:i5-4670(3.4Ghz),HD7770,Win8.1 SL/x64 Laptop:i7-Q720(1.6Ghz),HD5650,Win7Pro/x64
Tablet:ifiveX2,ARMv7-RK3188(1.6Ghz),ARM Mali-400,2GB DDR3 RAM,16GB Storage,Android 4.4.2(KK)
05-30-2012 06:05 AM
Find all posts by this user Quote this message in a reply
xstyla Offline
hobby coder

Posts: 820
Joined: Aug 2011
Location: just a box
Post: #3
RE: SSE4 vs AVX
(05-30-2012 06:05 AM)tallbender Wrote:  already discussed
http://forums.pcsx2.net/Thread-SSE4-X-OR-AVX

3 post is not what i call discussed. Laugh

@op a real benchmark would be to take the source code and implement a routine to measure the frame times precisely. and cheat or somehow setup a scene in the game where the content is constant. means a good ingame sequence. also start and end are supposed to match for both tests.

there you go. try that. Wink
05-30-2012 06:19 AM
Find all posts by this user Quote this message in a reply
Alexander Offline
★ Blah ★
***

Posts: 577
Joined: Sep 2011
Location: Scottsburg, Indiana
Reputation: 16
Post: #4
RE: SSE4 vs AVX
(05-30-2012 06:19 AM)xstyla Wrote:  3 post is not what i call discussed. Laugh

@op a real benchmark would be to take the source code and implement a routine to measure the frame times precisely. and cheat or somehow setup a scene in the game where the content is constant. means a good ingame sequence. also start and end are supposed to match for both tests.

there you go. try that. Wink

Well, I did test it at the same savepoint standing idle for a while, but it's kind of hard to test it in-battle since battles are never exactly the same. But yes, I make sure to be strict with the controlled variables - just like any science experiment in high school... only without all the cool chemical reactions Sad

Also I forgot to mention that I'm using 5241 SVN.

[Image: signat11.jpg]

My Guide to PCSX2 for Windows (outdated, but still generally accurate)
(This post was last modified: 05-30-2012 06:59 AM by Alexander.)
05-30-2012 06:57 AM
Visit this user's website Find all posts by this user Quote this message in a reply
rama Offline
PCSX2 coder
******

Posts: 6.848
Joined: Nov 2008
Location: Germany
Reputation: 108
Post: #5
RE: SSE4 vs AVX
You need to use a FPS measurement tool that averages the numbers more, try FRAPS for that.
Then you'll want to check several scenes, that all seem to stress different parts in GSdx.
Make sure you choose scenes that go slow enough (below 100fps) and that show a high use percentage in the GS window title bar.
05-30-2012 11:01 AM
Visit this user's website Find all posts by this user Quote this message in a reply
Alexander Offline
★ Blah ★
***

Posts: 577
Joined: Sep 2011
Location: Scottsburg, Indiana
Reputation: 16
Post: #6
RE: SSE4 vs AVX
Dark Cloud 2 (turns out to be a heavier game to emulate than the graphics make it look)
AVX : Roughly 145fps
SSE4 : Roughly 145fps

Final Fantasy X-2 (Was tested in main menu only due to me not having any actual saves and going through the opening cutscene twice would be crazy).
AVX : About 310fps
SSE4 : About 310fps

Final Fantasy XII
AVX : Stays quite steady at 131 - 136fps
SSE4 : Stays quite steady at 131 - 136fps

* All tests were conducted after loading the exact same save and standing in place of start for about 20 seconds. This was just about the only way I knew of to test this accurately.
** I can't account for anything besides performance here. Any quality differences etc are left unchecked. I've never seen anything different between the two myself.

It seems to be accurate to say that AVX and SSE4 are nothing different besides the software rendering mode, just as everyone thought. I, personally, will be using AVX since there's nothing different and if I run into trouble with Hardware Rendering, such as in FFX-2 cutscenes, the software rendering after pressing F9 will be better. Basically, I say if your computer has AVX and SSE4, use AVX.

I conducted this test to put hard evidence behind what everyone thought in the first place. I don't like accepting any opinion without hard proof behind it, and it seemed that no tests were really conducted between the two, so I did it. If anyone wants any other comparitive tests done, I'd be glad to do them. I have a range of games to test with.

I didn't end up testing with ZOE 2 because I can't seem to find my disc now and I never burned the ISO onto my computer...so yeah. Got plenty of other games to test with though, and I'll find that game soon.

Also, I just realized that Kingdom Hearts is quite easy to emulate for it's level of graphics...

[Image: signat11.jpg]

My Guide to PCSX2 for Windows (outdated, but still generally accurate)
(This post was last modified: 05-30-2012 11:07 AM by Alexander.)
05-30-2012 11:03 AM
Visit this user's website Find all posts by this user Quote this message in a reply
Alexander Offline
★ Blah ★
***

Posts: 577
Joined: Sep 2011
Location: Scottsburg, Indiana
Reputation: 16
Post: #7
RE: SSE4 vs AVX
(05-30-2012 11:01 AM)rama Wrote:  You need to use a FPS measurement tool that averages the numbers more, try FRAPS for that.
Then you'll want to check several scenes, that all seem to stress different parts in GSdx.
Make sure you choose scenes that go slow enough (below 100fps) and that show a high use percentage in the GS window title bar.

Okay, will get that then.

If I want to get below 100fps then I'm going to have to play a game more intensive than say KH, because it just never takes that kinda power.

When I find ZOE 2 I think it'll be a good subject - I get good fps out of it, but it stays at about 80fps, and in battles it can drop below the framelimiter...or at least it did back on 0.9.8 last year (haven't tried it in current SVNs)

I might conduct more tests on this soon, but really I think I might switch over to testing something more meaningful. Looking at the numbers, I can already kinda tell that everyone's right. I'm sure the devs know more about it than I do and trusting their opinion is best - I just wanted to run a few quick tests to make sure and to put real evidence behind it, so that when anybody asks you could give them a bit of proof in addition to your answer...I don't think this needs to go into deep testing to prove.

[Image: signat11.jpg]

My Guide to PCSX2 for Windows (outdated, but still generally accurate)
(This post was last modified: 05-30-2012 11:44 AM by Alexander.)
05-30-2012 11:40 AM
Visit this user's website Find all posts by this user Quote this message in a reply
Alexander Offline
★ Blah ★
***

Posts: 577
Joined: Sep 2011
Location: Scottsburg, Indiana
Reputation: 16
Post: #8
RE: SSE4 vs AVX
Oh, and devs, I'd like to report that PCSX2 plugins crash after changing video plugin - even if you restart games and everything. You have to close PCSX2 entirely and reopen it. Not sure if that's already well known or not, but it's been like that for a while and I'm sure it is.

[Image: signat11.jpg]

My Guide to PCSX2 for Windows (outdated, but still generally accurate)
(This post was last modified: 05-30-2012 11:56 AM by Alexander.)
05-30-2012 11:55 AM
Visit this user's website Find all posts by this user Quote this message in a reply
Bositman Offline
Angel of Death
********

Posts: 13.447
Joined: Aug 2005
Location: Athens,Greece
Reputation: 246
Post: #9
RE: SSE4 vs AVX
Changing plugins on the fly, as the warning dialog says, has a high chance of instability. I don't think that that can be helped.

[Image: newsig.jpg]
05-30-2012 12:09 PM
Visit this user's website Find all posts by this user Quote this message in a reply
Yungblaqs Offline
Newbie

Posts: 3
Joined: Jul 2011
Location:
Reputation: 0
Post: #10
RE: SSE4 vs AVX
(05-30-2012 12:09 PM)Bositman Wrote:  Changing plugins on the fly, as the warning dialog says, has a high chance of instability. I don't think that that can be helped.

Do you think it would be possible to write in code to PCSX2 to auto restart if the plugin was changed or would that be a big headache?
05-30-2012 12:39 PM
Find all posts by this user Quote this message in a reply
Post Reply 





Current time: 11-29-2014, 09:17 AM