49 replies to this topic

[Trixxstrr]

I did some benchmarks to compare frame rates between the 4 combinations of 32 bit, 64 bit, dx9 and dx11.

Wish we had the new maps in training ground to do a proper test since those new maps are killing my frame rates, but I used Viridian Bog, with a Jagermech, running straight line until I hit a wall about a minute in. Used fraps to record stats. Settings on Very High, vsync off.

My rig:
Intel i5 2500K (4 core @ 3.3 GHz)
Geforce GTX 570 (1280 mb vram)
8 GB DDR3 PC3-10600

bits - dx - Min FPS - Max FPS - Avg FPS
32 - 9 - 40 - 86 - 76.158
32 - 11 - 31 - 78 - 66.424
64 - 9 - 37 - 88 - 76.461
64 - 11 - 43 - 79 - 67.160

So DX11 is still giving about 10 FPS drop, but 64 bit doesn't make much difference.

[Greenjulius]

Thank you for testing. We've been seeing a lot of statements flying around about MWO playing way better with the 64 Bit client. I have not witnessed this myself.

I still have random crashes to desktop as well despite running repair a few times too. I think the 64 bit client is a step in the right direction, but I doubt it will be the performance messiah many are making it out to be.

[Demuder]

I have almost exactly the same rig apart from having a GTX560 instead of 570.

I can't say that I noticed any FPS difference with the 64-bit version, but the client is certainly running a lot smoother, loads faster, tabs switch faster, etc.

[o0Marduk0o]

Testing Grounds isn't useful for proper testing. TG performance is much better due to the lack of moving mechs and less weapon effects.

Edited by o0Marduk0o, 22 December 2014 - 03:50 AM.

[That Dawg]

Pity we can't record game tracks, as an example, an old game IL2 Sturmovic allowed us to record tracks, quite small in size and playable.
used to test changes.

I get frames over 100 sometimes in training grounds
I get teens in one of the maps in CW, and average between 25-40 in the rest of the maps.
Hard to blame your rig when there is a such a HUGE variation. Add that, if the server is being fussy.

no change 32 to 64, fps loss and drops to desktop in dx11 vs. dx9

[9thDeathscream]

I have found since switching to the 64 bit client that the frame rate in game is more stable and tends to fluctuate a lot less.

[Angel of Annihilation]

Akulla1980, on 22 December 2014 - 04:45 AM, said:

I have found since switching to the 64 bit client that the frame rate in game is more stable and tends to fluctuate a lot less.

Yeah same. With the 32 bit I would occasionally have my frame rate drop to the 18-20 fps range if there was a lot going on, on screen. Now it never drops below 40 fps which makes for an overall smoother game play experience.

[PorkCereal]

64 b only allows you to use more than 3 g of system memory. The only way it would greatly improve fps is if your video and system shared memory. In most cases of gamers we run a dedicated video card.

[Torgun]

Also found 64bit more stable and overall higher fps especially in CW where I used to have frequent drops under 30 fps. Now I seldom go under 50. And DX9, DX11 for me is still a hog on resources without any noticable improvements.

Edited by Torgun, 22 December 2014 - 05:22 AM.

[Thorqemada]

OP, your minFPS were way higher in the 64 Bit Dx11 Test and that is really what Counts -> increase the FPS in the lowFPS and minFPS scenarios.

Aside that the 64Bit Client runs very smooth in the Public Queue and seems to run way better (compared to the really very bad Performance of CW) the CW-Queue though for CW is lowered Particles and Shadows to High but ihave to get into a game yet to see if it helps enough in the minFPScases (Own particles and Clan Mechs seem to slow me down way to much in CW - to often into red FPS making it impossible to fight back).

Edited by Thorqemada, 22 December 2014 - 07:14 AM.

[Ens]

I don´t experience MORE fps but a much steadier framerate......the drops are almost gone

[Kain Demos]

Trixxstrr, on 21 December 2014 - 12:30 PM, said:

I did some benchmarks to compare frame rates between the 4 combinations of 32 bit, 64 bit, dx9 and dx11.

Wish we had the new maps in training ground to do a proper test since those new maps are killing my frame rates, but I used Viridian Bog, with a Jagermech, running straight line until I hit a wall about a minute in. Used fraps to record stats. Settings on Very High, vsync off.

My rig:
Intel i5 2500K (4 core @ 3.3 GHz)
Geforce GTX 570 (1280 mb vram)
8 GB DDR3 PC3-10600

bits - dx - Min FPS - Max FPS - Avg FPS
32 - 9 - 40 - 86 - 76.158
32 - 11 - 31 - 78 - 66.424
64 - 9 - 37 - 88 - 76.461
64 - 11 - 43 - 79 - 67.160

So DX11 is still giving about 10 FPS drop, but 64 bit doesn't make much difference.

WTF, I regularly see drops as low as 20fps in CW and my system is far, far better than what you have listed.

[o0Marduk0o]

Kain Thul, on 22 December 2014 - 08:41 AM, said:

WTF, I regularly see drops as low as 20fps in CW and my system is far, far better than what you have listed.

Because those FPS values are for testing ground and not a real match, not to say CW.

[Trixxstrr]

Ya it's hard to get a true benchmark test in CW because the match would play out different each time. We would need a demo loop with a bunch of mechs and effects going on to do true benchmark comparisons.

[Kain Demos]

o0Marduk0o, on 22 December 2014 - 09:11 AM, said:

Because those FPS values are for testing ground and not a real match, not to say CW.

I've noticed the drops are the most extreme if you're zoomed in with advanced zoom and have **** loads of incoming LRM/dakka.

[Rhialto]

Trixxstrr, on 21 December 2014 - 12:30 PM, said:

My rig:
Intel i5 2500K (4 core @ 3.3 GHz)

Do yourself a present for Christmas and run that CPU @ 4.2GHz! Many run it higher but 4.2GHz is guaranteed to work! Dunno what motherboard you have but in my case it's was 2 click and done.

My setup is similar to yours but GTX 660 Ti. I may try to run it and compare now that I use the 64bits client but still DX9. I will try to remember disabling 3D first.

Edited by Rhialto, 22 December 2014 - 10:21 AM.

[EgoSlayer]

o0Marduk0o, on 22 December 2014 - 03:49 AM, said:

Testing Grounds isn't useful for proper testing. TG performance is much better due to the lack of moving mechs and less weapon effects - a minimal baseline benchmark.

Actually, testing grounds is the only place to get valid repeatable tests. It's only an indicator for testing relative performance of changes in settings.

It's not a complete indication of in game performance because of all the other variables you mentioned that occur in game with the 23 other mechs on the field, weapon effects, etc. Until we can record sessions and play them back or have demo loops of combat there isn't any repeatable way to test in game performance. But any gains or losses in TG will be reflected in game, it just may be lost/minimized (gains) or amplified (losses) in game with the additional rendering requirements

Edited by EgoSlayer, 22 December 2014 - 12:33 PM.

[Kain Demos]

EgoSlayer, on 22 December 2014 - 12:32 PM, said:

Actually, testing grounds is the only place to get valid repeatable tests. It's only an indicator for testing relative performance of changes in settings.

It's not a complete indication of in game performance because of all the other variables you mentioned that occur in game with the 23 other mechs on the field, weapon effects, etc. Until we can record sessions and play them back or have demo loops of combat there isn't any repeatable way to test in game performance. But any gains or losses in TG will be reflected in game, it just may be lost/minimized (gains) or amplified (losses) in game with the additional rendering requirements

Testing grounds does nothing for me.

No matter what settings I use or what I do, 60fps solid.

Get into a game and its entirely different--regardless of settings from very low to very high I'm still subject to those rock-bottom FPS drops that make things seem like slo-mo.

[EgoSlayer]

Kain Thul, on 22 December 2014 - 12:35 PM, said:

Testing grounds does nothing for me.

No matter what settings I use or what I do, 60fps solid.

Get into a game and its entirely different--regardless of settings from very low to very high I'm still subject to those rock-bottom FPS drops that make things seem like slo-mo.

I've never seen it drop to slo mo, and I run 2550x1440 on an HD 7950. I am GPU bound, my CPU doesn't ever see more than about 30% utilization. But I still never see less than 30FPS, get ~80 in TG, average in game is around 50. Turn particles to low - they are the biggest performance killers.

Edited by EgoSlayer, 22 December 2014 - 12:42 PM.

[HARDKOR]

Is there any benefit to DX11? I didn't see any graphical improvements...