Jump to content

Some Performance Tests


295 replies to this topic

#1 xWiredx

    Member

  • PipPipPipPipPipPipPipPip
  • Elite Founder
  • Elite Founder
  • 1,805 posts

Posted 01 December 2014 - 08:52 PM

I decided to test a couple of things today to see what impact they would have on performance. My system: Intel Core i7 5820K @ 4.1GhZ, MSI GTX 980 Gaming 4G, 16GB DDR4-2666, Samsung 840 Pro SSD

The first thing I did was run numbers on individual maps for a base line. Unfortunately, it appears that a couple of maps simply would not present themselves enough today, so there is a little bit of missing information (not that it is really terribly important).

Posted Image

One of the variables I tested for was the somewhat widespread myth that turning hardware acceleration off for flash actually helps performance a ton because of the HUD being flash-based. I normally have it off anyway, so I turned it on for a couple of rounds. Since I only have 2 results, you could easily dismiss this, but it appears that it does not have much if any impact on game performance. Or, at least it doesn't for those of us with high-end hardware.

The second variable I tested was a new suggestion that popped up regarding the windows system timer resolution. My system is defaulted at 1.0ms, so I had pretty big doubts this would do anything. The results appear to agree (again, there are only a few results, but that is how it appears).

So we can call these two busted when powerful systems are in use.

I also learned a few things, which I kind of already knew, but now have the numbers to back it up. The first is that this game needs a lot of optimization done still. The second is that the map designers clearly do not have much of an idea how their designs perform. The third is that Mining Collective is too heavily-weighted in the rotation (at least, IMO, as I played it roughly 40% of the time while testing today, over the course of about 30 matches). The fourth is that overall performance demands are so high that even with a top-notch system one cannot expect consistent performance and definitely cannot expect a minimum of 60fps without making quality sacrifices. The fifth, which was completely expected, is that dropping with a group and using teamspeak does not contribute to any noticeable slowdowns. The TS overhead is extremely minimal.

There are a couple of anomalies in my FPS that I'm sure are flukes, and I did not remove these from the results, but when I look at the minimums it's pretty clear to me that PGI has a lot of work to do. This kind of performance out of $1500-2000 worth of hardware is crap (as if I needed to beat that dead horse again). The other take away here is that every time somebody says they get a smooth 60fps with everything on the highest settings, they're either full of it or they've got $3k worth of hardware sitting in front of them and all of it is overclocked pretty heavily while being cooled with a wicked custom loop.

#2 Thorqemada

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • 6,366 posts

Posted 01 December 2014 - 10:58 PM

Thx for the Info, i did not expect the Windows Timer Tool do much on a High End System/Clean System with accurate settings but i feel its good to know that there may be a chance where the setting is borked and you get better Performance out of it.

My System is an AMD FX8350 stock + ATI Sapphire R9 280X Vapor-X OC stock and i get around 45 FPS avg at Dx11 Highest Settings 1920 x 1200 VSync On (Cockpit Glass disabled).

When People speak of smooth 60 FPS at highest settings usualy they have either lowered specific settings in the MWO-Options or per config file or per GPU driver settings and they often use Dx9 instead of Dx11 graphics.

Lowering Particles + Shadows (+ sometimes the Resolution) should bring huge performance boosts.

The HUD alone takes up to 15 FPS i have read...

Edited by Thorqemada, 01 December 2014 - 11:08 PM.


#3 Flapdrol

    Member

  • PipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 1,986 posts

Posted 02 December 2014 - 03:43 AM

Very nice, and damn, a drop to 34 fps with an overclocked 5820K. Maybe it's more relevant to look at something like 99th or 95th percentile fps like they use at techreport. Don't know how easy or hard that is though.

Didn't expect the flash thing to do anything. The timer resolution thing apparantly keeps cores from entering the deeper sleep states, so it might be useful on some cpu types if windows is not set to high performance.

#4 xWiredx

    Member

  • PipPipPipPipPipPipPipPip
  • Elite Founder
  • Elite Founder
  • 1,805 posts

Posted 02 December 2014 - 08:16 AM

In most cases, removing the outliers does include those really low numbers. Actually, in all tests it does look like they are clear anomalies rather than sustained dips. They don't appear to be linked to any particular time frame where I would have been in super heavy action or anything. If I redo the averages after taking out the bottom and top outliers, the FPS average goes up between .1 and .5 - not significant.

Looking at frame times is actually rather encouraging. My average frame time is only 11.23ms, where techreport was showing between 16ms and 19ms most of the time.

I can say without a doubt that my experience is improved over my GTX 660 TI and GTX 660 TI SLI systems with a 4.4GhZ 2600K. I am almost always above 60fps (and if we start taking the outlier numbers out, I only dipped below that line twice in 20 tests). Still, a ridiculous amount of horsepower is needed for this game. I don't know what machines they're developing, designing, and play testing on, but they really need to work out all of the kinks. Performance complaints are rampant, most are not solved by little technical oddities, and that is contributing to people leaving the game.

Mostly, I just wanted to test out a couple of things to see if they were placebo effect or not. Turns out, they probably are. If anybody is playing on XP or Vista, I'd check the timing with the Windows Timer Tool. Everybody should already get nagged to upgrade flash, and should do so for more reasons than game performance (security updates!!).

#5 Flapdrol

    Member

  • PipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 1,986 posts

Posted 02 December 2014 - 10:39 AM

When I play I always see a clear variation in fps, very high in the loading screen, lower when the game starts, slowly decreasing when the brawl intensifies, then climbing up agian as mechs start to die off.

The lowest dips are probably only for a fraction of a second when you get hit during the brawl. What's your average fps during the brawl? probably lower than the ~80 overall right?

#6 xWiredx

    Member

  • PipPipPipPipPipPipPipPip
  • Elite Founder
  • Elite Founder
  • 1,805 posts

Posted 02 December 2014 - 11:26 AM

Brawling... well, I consistently played assaults and heavies so I could tell you when I was in heavy action actually. I captured FPS for the first 3 minutes of each match, which for assaults in the 49-65kph range means about 1 minute to get into action, 1 minute of medium-paced back and forth action, and 1 minute of pure brawling (I'm a short-range brawler most of the time). Obviously with the hellbringers that I used for a few matches, brawling comes a little earlier. This makes it somewhat easy, though, to get averages for brawling and action. For this, I'll take the last minute of data from the tests where I had my worst and best average FPS. That happens to be Frozen City Night and Mining Collective respectively.

Last minute FPS avg - Frozen City night: 79.85
Last minute FPS avg - Mining Collective: 91.04

#7 Odins Fist

    Member

  • PipPipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 3,111 posts
  • LocationThe North

Posted 02 December 2014 - 11:57 AM

Thanks for the testing and posting results.

#8 Flapdrol

    Member

  • PipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 1,986 posts

Posted 02 December 2014 - 12:45 PM

View PostxWiredx, on 02 December 2014 - 11:26 AM, said:

Last minute FPS avg - Frozen City night: 79.85
Last minute FPS avg - Mining Collective: 91.04

Not bad.

#9 Durant Carlyle

    Member

  • PipPipPipPipPipPipPipPipPip
  • Survivor
  • Survivor
  • 3,877 posts
  • LocationClose enough to poke you with a stick.

Posted 02 December 2014 - 01:09 PM

I get similar numbers with my system:

Intel i5-4690K CPU (locked at 3.9GHz)
Corsair Vengeance LP 16GB 1600 9-9-9-24 RAM
Zotac GTX 980 AMP! Omega 4GB GPU (standard OC on GPU, memory set at 8000MHz)
Crucial MX100 512GB SSD
Acer XB270H Abprz 27" 1920x1080 144Hz G-Sync LCD
Highest settings in MW:O
Nvidia Control Panel settings geared towards Quality

I get 100+ FPS (some maps are 105 FPS, some are 140 FPS) at the beginning of every match, before everybody starts moving. I get FPS dips like everybody else, sometimes even just looking in a particular direction even if I can't see any other 'Mechs. And yeah, I do get dips into the 40s (don't think I've ever seen below 40 though). But with G-Sync the dips are not actually noticeable during play. Makes me glad I shelled out the extra money for it.

#10 Odins Fist

    Member

  • PipPipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 3,111 posts
  • LocationThe North

Posted 04 December 2014 - 04:47 PM

View PostxWiredx, on 01 December 2014 - 08:52 PM, said:

" This kind of performance out of $1500-2000 worth of hardware is crap (as if I needed to beat that dead horse again). The other take away here is that every time somebody says they get a smooth 60fps with everything on the highest settings, they're either full of it or they've got $3k worth of hardware sitting in front of them and all of it is overclocked pretty heavily while being cooled with a wicked custom loop."


As MWO stands today I COMPLETELY agree with this statement.

In the Past (for those of us that were here since closed beta), things were a little different.

Thanks again for the testing, THUMBS UP.

#11 Lulz Kev

    Member

  • PipPipPipPipPipPipPip
  • 604 posts

Posted 04 December 2014 - 05:47 PM

I've got a pretty pricey rig overclocked i5, overclocked 760, custom loop and still struggle with getting a consistent 60fps.

#12 xWiredx

    Member

  • PipPipPipPipPipPipPipPip
  • Elite Founder
  • Elite Founder
  • 1,805 posts

Posted 05 December 2014 - 09:32 AM

Another thing to consider is the architecture of an Intel chip. In MWO especially, but in pretty much any other application/game as well, we saw a large jump in performance between Nehalem and Sandy Bridge, then a small jump between Sandy Bridge and Ivy Bridge, and another small jump between Ivy Bridge and Haswell. I went from Sandy Bridge to Haswell, and there was a noticeable improvement at the same clock speed with the same GPU. My Haswell-E chip produced the same performance as my Sandy Bridge chip with a 400MhZ deficit when overclocked (2600K @ 4.4GhZ vs. 5820K at 4.0GhZ). It could also have had something to do with the 2 extra cores/4 extra threads, but that is less likely to explain it than architecture improvements.

Previously, I had also tested the effect of RAM bandwidth+latency. Above DDR3-1866, there doesn't appear to be any measurable difference. The difference between DDR3-1600 and DDR3-1866 was very small, within the margin of error. The difference between dual-channel DDR3-1866 and quad-channel DDR4-2666 was within margin of error, too.

#13 Karamarka

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 809 posts

Posted 12 December 2014 - 05:18 PM

View PostxWiredx, on 01 December 2014 - 08:52 PM, said:



I also learned a few things, which I kind of already knew, but now have the numbers to back it up. The first is that this game needs a lot of optimization done still.

This kind of performance out of $1500-2000 worth of hardware is crap (as if I needed to beat that dead horse again). The other take away here is that every time somebody says they get a smooth 60fps with everything on the highest settings, they're either full of it or they've got $3k worth of hardware sitting in front of them and all of it is overclocked pretty heavily while being cooled with a wicked custom loop.


Agreed

It really is a nightmare

Edited by Karamarka, 12 December 2014 - 05:18 PM.


#14 Bill Lumbar

    Member

  • PipPipPipPipPipPipPipPipPip
  • Death Star
  • 2,073 posts

Posted 19 December 2014 - 08:38 PM

View PostxWiredx, on 01 December 2014 - 08:52 PM, said:

I decided to test a couple of things today to see what impact they would have on performance. My system: Intel Core i7 5820K @ 4.1GhZ, MSI GTX 980 Gaming 4G, 16GB DDR4-2666, Samsung 840 Pro SSD

The first thing I did was run numbers on individual maps for a base line. Unfortunately, it appears that a couple of maps simply would not present themselves enough today, so there is a little bit of missing information (not that it is really terribly important).

Posted Image

One of the variables I tested for was the somewhat widespread myth that turning hardware acceleration off for flash actually helps performance a ton because of the HUD being flash-based. I normally have it off anyway, so I turned it on for a couple of rounds. Since I only have 2 results, you could easily dismiss this, but it appears that it does not have much if any impact on game performance. Or, at least it doesn't for those of us with high-end hardware.

The second variable I tested was a new suggestion that popped up regarding the windows system timer resolution. My system is defaulted at 1.0ms, so I had pretty big doubts this would do anything. The results appear to agree (again, there are only a few results, but that is how it appears).

So we can call these two busted when powerful systems are in use.

I also learned a few things, which I kind of already knew, but now have the numbers to back it up. The first is that this game needs a lot of optimization done still. The second is that the map designers clearly do not have much of an idea how their designs perform. The third is that Mining Collective is too heavily-weighted in the rotation (at least, IMO, as I played it roughly 40% of the time while testing today, over the course of about 30 matches). The fourth is that overall performance demands are so high that even with a top-notch system one cannot expect consistent performance and definitely cannot expect a minimum of 60fps without making quality sacrifices. The fifth, which was completely expected, is that dropping with a group and using teamspeak does not contribute to any noticeable slowdowns. The TS overhead is extremely minimal.

There are a couple of anomalies in my FPS that I'm sure are flukes, and I did not remove these from the results, but when I look at the minimums it's pretty clear to me that PGI has a lot of work to do. This kind of performance out of $1500-2000 worth of hardware is crap (as if I needed to beat that dead horse again). The other take away here is that every time somebody says they get a smooth 60fps with everything on the highest settings, they're either full of it or they've got $3k worth of hardware sitting in front of them and all of it is overclocked pretty heavily while being cooled with a wicked custom loop.

I find your post rather funny after other threads you have commented on.....As I have listed my Rig before,

AMD FX-8350 @ 5 ghz very stable
Asrock 990 board
16 gb DDR3 @1866
XFX 7970 3gb card, 1000 mhz Gpu, 1450 mhz on the ram
512 Gb Samsung SSD
two seagate 4 tb drives for back ups/storage

I am seeing very similar fps results as you are in all of my games. This is not a Amd problem, this is a optimization problem with the engine.

Edited by Bill Lumbar, 19 December 2014 - 08:47 PM.


#15 xWiredx

    Member

  • PipPipPipPipPipPipPipPip
  • Elite Founder
  • Elite Founder
  • 1,805 posts

Posted 19 December 2014 - 08:46 PM

Actual numbers or it didn't happen -aka- put up or shut up, dude.

#16 redlance

    Member

  • PipPipPipPipPip
  • 178 posts

Posted 19 December 2014 - 08:51 PM

why does the mech community know more about optimization than pgi?

#17 Bill Lumbar

    Member

  • PipPipPipPipPipPipPipPipPip
  • Death Star
  • 2,073 posts

Posted 19 December 2014 - 08:55 PM

Sorry not trying to offend you, but I don't need to put up the numbers at this point... my word is good on what my system does, and I think Thorgemada kinda just confirmed that is kinda a base line for a very similar system as mine. My settings are all at high to very high, post AA, Dx11 and I haven't even installed the new omega drivers yet, just found out about them. I also have the cockpit glass on. I have not modded my config files, haven't touched anything.

I think this is kinda funny that given what you have said in the other threads and your advice, that you are here posting this. Don't get me wrong, but its really simple, this is not a just an Amd issue, it is a optimization issue with the game. I don't have to throw numbers up, but if I get the time to post some screen shots of what I have claimed, I will just for you. I don't make claims I can't back up with pics or videos regarding what my rig does, why would anyone...be rather silly don't you think?

Edited by Bill Lumbar, 19 December 2014 - 08:58 PM.


#18 Catamount

    Member

  • PipPipPipPipPipPipPipPipPip
  • LIEUTENANT, JUNIOR GRADE
  • 3,305 posts
  • LocationBoone, NC

Posted 19 December 2014 - 09:21 PM

View PostBill Lumbar, on 19 December 2014 - 08:55 PM, said:

Sorry not trying to offend you, but I don't need to put up the numbers at this point...


Nope, you don't. And we don't have to take you seriously.

If you do, however, we just might.

Quote

this is not a just an Amd issue


You're right, it's a set of two problems here: It's a terribly optimized game (for now) and you're exacerbating that by running a system terribly set up for gaming anyways. It's a two part problem for you.

It was for me, too, until my OCed Phenom II became a vastly superior OCed 3570K (the 3770K was a sidegrade for this game, but hey, it was free)

Edited by Catamount, 19 December 2014 - 09:24 PM.


#19 Bill Lumbar

    Member

  • PipPipPipPipPipPipPipPipPip
  • Death Star
  • 2,073 posts

Posted 19 December 2014 - 09:46 PM

View PostCatamount, on 19 December 2014 - 09:21 PM, said:


Nope, you don't. And we don't have to take you seriously.

If you do, however, we just might.



You're right, it's a set of two problems here: It's a terribly optimized game (for now) and you're exacerbating that by running a system terribly set up for gaming anyways. It's a two part problem for you.

It was for me, too, until my OCed Phenom II became a vastly superior OCed 3570K (the 3770K was a sidegrade for this game, but hey, it was free)

Funny, I have very little problems with my rig when it comes to any game I have every played on it. Sounds like the reality is that there are others that are not having serious issues either, and some are, on BOTH AMD and INTEL rigs. You seem to be one of the later folks. I am still waiting on proof that "ALL" Amd rigs need to be overclocked, and have tweaks done to the config. files in order to play this game with out "gimping" a player. Amd has a bumpy history with some of their FX line processors, no doubt. (Fx 8100's would be one for sure) Guess what I skipped that one, and would have to agree with you, it really does have issues. I also have a Fx-6100 that is able to chug right along with a xfx 6870 2gb card, 8 gigs of ddr3 1866 ram, and I think its at 3.4ghz stock. I play that one once in while on a 55" LG at 1080P..... still get better then 30 fps avg, and runs fairly smooth and I am picky.

One of my older Intel rigs Q9550 oc to 4 Ghz and 8 gigs of g-skill dddr2 1066 ram with the xfx 6870 2 gb runs about the same as the FX-6100 Amd build fps wise and smoothness. Both of those systems I have seen drop into the low 20's very briefly... but the norm is 25-35 for me. This was all played prior to 64 bit client and new drivers of course.

Just installed the new omega driver from AMD for the 7970 and dropped in first normal mode game with it, haven't tried it in CW yet. I have already seen gains with the 64 bit client on, and now I have just got a nicer bump in performance in Fps. I would say the new driver has gave me 10 fps bump on top of the bump from the 64 bit client, very nice indeed. :D Other then CW dropping down to 25 fps but normally avg 35-50 before the 64 bit client, after I don't see 25 fps anymore, and avg jumped up to around 45-70 fps in CW.

Edited by Bill Lumbar, 19 December 2014 - 09:48 PM.


#20 Smokeyjedi

    Member

  • PipPipPipPipPipPipPipPip
  • Liquid Metal
  • Liquid Metal
  • 1,040 posts
  • LocationCanada

Posted 19 December 2014 - 09:50 PM

Posted Image

View PostBill Lumbar, on 19 December 2014 - 08:38 PM, said:

I find your post rather funny after other threads you have commented on.....As I have listed my Rig before,

AMD FX-8350 @ 5 ghz very stable
Asrock 990 board
16 gb DDR3 @1866
XFX 7970 3gb card, 1000 mhz Gpu, 1450 mhz on the ram
512 Gb Samsung SSD
two seagate 4 tb drives for back ups/storage

I am seeing very similar fps results as you are in all of my games. This is not a Amd problem, this is a optimization problem with the engine.

Ok bud, I held the 3D MARK DX10 cloudgate world record #1 position for 3 months.......it took custom VRM cooling. lower than ambient air(AC)for room and liquid cooling to hold that position for 3 months.....@ 4960mhz 2600mhz HT+NB BOTH OC'd using GSKILL sniper DDR3 @ 2000+mhz with tightened OC timings......with a ZOTAC AMP GTX 760 OC'd +60core for 1220mhz boostclock and +300mhz for 7000mhz GDDR5 and couldnt come close to these performance numbers that Wired has thrown up here.......Just take your things and walk away. I love amd and will stand my ground but this is sad.

The rig stated in this post is an absolute monster...........I'd bet you race ur civic against a Veyron just cause its your HONDA CIVIC


**EDIT** TOO many of us have been here for years......toying with MWO and any rig of any kind.........helping dozens if not hundreds experience a smoother MWO/
Wired has posted some of the highest performance figures we have all seen here in the forums......96% of his frames are rendered faster than 60FPS> THAT IS GODDAMN INSANE AND NOONE using I7's short of dry ice OC'ing has seen these figures......My amd 8350 doesnt hold a candle in the wind to this 12 thread deathmachine.

Edited by Smokeyjedi, 19 December 2014 - 10:03 PM.






1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users