Are newer revisions slower?

potate · (This post was last modified: 07-02-2014, 12:19 PM by potate.)

I've been playing SMG in 4.0-1569 with OpenGl and runs smooth, mostly 60 fps, using HLE audio. Unfortunately I can't get rid of the screen tearing even with vsync on unless I switch to D3D but this is generally much slower.

I read the new june article and I was so stoked to see that OpenGl had a fix and vsync works now in that backend. So I downloaded the latest rev but now SMG is slower, like 42 fps. I also downloaded some revisions since 1981 (this is the first revision since the fix) and the same, 40 fps-ish in game. Using the exact settings as in 1569. Did something happened in the new revisions I'm not aware of?

i7 4770
Radeon R9 290

kinkinkijkin · 07-02-2014, 12:25 PM #2

No, new revisions are not slower. In fact, there was a bunch of lazy testing done, they're now the fastest. There's apparently an issue in recent builds that makes OGL a lot slower for AMD users now, though.

**RachelB** · 07-02-2014, 12:36 PM #3

Slower than what?

**JMC47** · 07-02-2014, 12:39 PM #4

Potate: Yeah, uh, there was a regression caused by a fix for NVIDIA cards. Apparently, Radeons really like Coherent Mapping on. And certain NVIDIA cards really want it off. We did not see the regression coming and it's being researched so that everyone is happy.

kinkinkijkin · 07-02-2014, 12:53 PM #5

Wait, I never noticed a slowdown. I'll have to check, but I'm pretty sure it runs exactly the same for me. Maybe it's a thing with more recent Radeons? I'll test this once I figure out my system interrupts issue, it might possibly help.

**JMC47** · 07-02-2014, 01:09 PM #6

Radeon 5850 and a R290 both have the same exact slowdown, so I doubt you're exempt.

**Anti-Ultimate** · 07-02-2014, 08:27 PM #7

My R9 280x (actually HD7970 but I like to call it R9 280x Tongue

) has the same issue and with the changes reverted it gains ~100-150% more performance ^^

kinkinkijkin · 07-05-2014, 12:20 PM #8

Sorry for taking so long, was bug-squashing.

Can confirm that there is a difference.

Cannot confirm that it is significant.

**JMC47** · 07-05-2014, 04:10 PM #9

The bug should be fixed on latest master and some. Get an 8% boost over BEFORE the regression on SMG1 on my Radeon 5850.

kinkinkijkin · 07-05-2014, 04:14 PM **#10**

(07-05-2014, 04:10 PM)JMC47 Wrote: The bug should be fixed on latest master and some. Get an 8% boost over BEFORE the regression on SMG1 on my Radeon 5850.

Oh, reflects my results exactly. Okay, then.