Microstutter - fact or fiction? Let's find out once and for all...

Dr_Fartypants · 20 Jul 2008 at 04:32

Bobly said:
No new results?

Unfortunately not

It would be nice to see some 4850 or 4870 crossfire results...

Final8y · 4 Feb 2009 at 10:52

Bump! any chance of making that Microstutter exe available again ?

Lightnix · 4 Feb 2009 at 11:06

Final8y said:
Bump! any chance of making that Microstutter exe available again ?

Just uploading it, gimme a minute.

Edit: http://www.mediafire.com/download.php?2moyugibhez

There you go.

Dr_Fartypants · 4 Feb 2009 at 11:41

Updated the original post with the new url

Final8y · 4 Feb 2009 at 11:41

Thanks a bunch

muppetkiller · 4 Feb 2009 at 13:22

Nice little program. Would be nice to get some people to run this.

steve258 · 4 Feb 2009 at 14:50

doesn't work on my pc - a dos window appears for a fraction of a second then dissapeared when I try to run the exe file.

*EDIT* oops used the wrong csv file

Happy · 4 Feb 2009 at 16:49

Someone needs to run this on a 4870X2 to settle that issue.

Vince · 4 Feb 2009 at 17:02

Another thread revival from the old school.

Either way Im not sure I can trust the app to give us/me a realistic idea of what is really happening. Will be epic fun to run mind... ill get on it when im home.

CreAtiv3 · 4 Feb 2009 at 17:10

Ill take a look at this when I get back tonight.

Shall test World at War, HL2 EP2 and FEAR2 just to mix it up a bit!

Happy · 4 Feb 2009 at 17:17

CreAtiv3 said:
Ill take a look at this when I get back tonight.

Shall test World at War, HL2 EP2 and FEAR2 just to mix it up a bit!

Can you add COD4 to that mix if you have it, since last time a 'Micro-Stutter' argument came up that was the example being used.

CreAtiv3 · 4 Feb 2009 at 17:48

Yea sure can do
Would be interesting to see the differance, if any, between COD 4 & 5.

CreAtiv3 · 4 Feb 2009 at 21:30

And the results are in!

Please note the following settings were used for all tests
1920x1200 Resolution
4x AA set in game, Edge Detect set in CCC & Adaptive AA enabled
16x AF
Maxed graphics options for each title
Vertical Refresh set to off
No special console tweaks used in any game to boost performance, just vanilla settings.

F.E.A.R 2
Started recording at the first conact with mercs.
Couple of hiccups were noticeable before and after the shoot-out but I think this is when the game was calling for NPCs to enter the scene and scripts being ran rather than graphical stutters.

Left 4 Dead
Started recording from the beginning of the No Mercy campaign for roughly 1min. Felt smooth as silk to play and yet this appears to have the highest MS Index...?

COD 5: World at War
Recorded a multiplayer CTF match for roughly 1min.
Felt very smooth with and without Frame Limiting.
Note: Dual Graphics option enabled in Game Options.

Cannot seem to find my COD 4 disk anywhere. Thinking about it I may have lent it to someone about a year ago :eek:

Hope you find these results interesting!

Lightnix · 4 Feb 2009 at 21:33

I think the hiccups you witnessed with F.E.A.R. 2 was the autosave feature, that causes a slight stutter on any machine.

CreAtiv3 · 4 Feb 2009 at 21:36

Ah yes perhaps dint think of that, would make sense as it did seem to only happen before and after some action!

Forgot to mention im running the latest 9.1 drivers on Windows 7 x64. Rest of the rig can be found in my signature.

Happy · 4 Feb 2009 at 22:29

Very good results for COD5, an interesting thread tbh. Hopefully more people will submit their results

Natima · 4 Feb 2009 at 23:26

Looking at some of the earlier results it would seem that the problem has been vastly improved upon.
Still... more results with different driver versions should be interesting.
How about P45 at dual X8 vs X48 at dual X16.

steve258 · 5 Feb 2009 at 00:50

COD5 "Relentless" level, everything at max in game settings 1920x1080. Started benchmark when black background faded at level start. fps went down to 19 when tank blew up. both cards are oced at 700/1000.

Fear 2 - max in game settings, both cards at stock since I can't run this oced without driver stop responding every few seconds. Benchmark started at the start of level when you wake up next to the helicopter. exe renamed to Fear.exe for cf support. It paused for a second every now and again probably due to autosave as suggested.

L4D - Both cards at stock. Benchmark started while playing online co-op during an intense battle. Didn't really notice when the fps dropped to 8 due to all the zombie killing!

results are quite bad I guess it's just the limitations of P45.

Dr_Fartypants · 5 Feb 2009 at 01:21

VinceB1 said:
Either way Im not sure I can trust the app to give us/me a realistic idea of what is really happening. Will be epic fun to run mind... ill get on it when im home.

My programming skills might be crap (well, they are - no question about it), but the calculation algorithm is mathematically sound I guarantee you. I give an explanation of how it works in the readme IIRC, so check it out and if you have any questions I'll try to better explain.

On another note - looks like we're seeing some great results with the 4870x2. Sub 10% values are pretty damn good for any multi-GPU setup, so it's good to see drivers properly addressng the issue now. What driver version were you using, CreAtiv3?

edit - I see you were using the 9.1 drivers

missed that, sorry.

Dr_Fartypants · 5 Feb 2009 at 01:24

Lightnix said:
I think the hiccups you witnessed with F.E.A.R. 2 was the autosave feature, that causes a slight stutter on any machine.

That shouldn't appear in the index, even if it does interrupt gameplay - the largest 2% of local frametime variations are culled to avoid exactly these kinds of infrequent paging/loading based issues. The frametime variation has to be at least fairly consistent in order to contribute to the index.

The idea was to examine the degree of microstutter only, rather than other kinds of longer timescale stutter.