Crank that S#!t Up
Life After DX 11
I Heart EVGA!
Biostar P45 Overclocking Contest Gr...
Coolermaster Contest Group
NVIDIA SLI GIVEAWAY CLUSTER
Universe At War: Earth Assault Beta...
Fireside Chat with Brandon
AMD Contest Group
Editors Challenge Sponsored by Inte...
Lord Of The Rings Online Beta Conte...
FiringSquad Video Challenge - Preli...
Foxconn Graphic Design Contest
FiringSquad Editors Challenge Round...
||16 entry(ies) in this category
| Help My FPS Has Fallen and I Can’t Get Up! (17 comments )|
by: p4l1ndr0m3 (499) | Posted in cluster Editors Challenge Sponsored by Intel Round 2
Posted 75 months ago ( edited 75 months ago ) in category DEFAULT
|» MEDIA (8)|
I Guess I Can't Walk Through Walls...
Graph Set 1
Graph Set 2
Graph Set 3
Graph Set 4
FEAR: Video Stress Test
CS:S: Video Stress Test
It’s nothing new to PC gamers out there to be able to take their favorite first-person shooters and go head to head with one another online. Made famous by game franchises like Unreal Tournament and Quake, online capabilities are now a necessity for any successful FPS. The ability to play an online deathmatch in your favorite shooter adds a whole new dimension to any game. Custom maps, clans, tournaments, stat tracking and much more are all made possible by online gameplay.
When games are first released, needless to say, they typically rip your computer a new one (just look at Supreme Commander… ). With first-person shooters, you will often find that your performance will dip fairly dramatically when playing a network or online game. While ping and bandwidth would have some negative impact on performance, the fact that your computer and graphics card are required to display many more active players on a 3D plane in real time can be extremely taxing, depending on the game.
(See Video: I Guess I Can't Walk Through Walls...)
I was unsure of what games would give the best examples to test this tax that online first-person shooters take from the average system. So, after much contemplation, I gave in to the voices in my head and chose two of my favorites: Counter-Strike: Source and F.E.A.R.: First Encounter Assault Recon. Both of these titles are convenient due to their large sleepless fan base, making it quite easy to find full capacity servers to grind out some chunks of data.
In order to achieve more timely and consistent results, both Counter-Strike: Source (CS:S) and F.E.A.R.: First Encounter Assault Recon (FEAR) were played on the same map in empty and full servers with a consistent 50 ping or less to minimize fluctuation in my data. My maps of choice were “CS_Office” for CS:S and “Campus” for FEAR I then tested different resolutions and settings over the course of many rounds including Anti Aliasing and Anisotropic Filtering as well as environmental effects (i.e. water reflections, lighting effects, etc), in some of my 1600x1200x32 tests. HDR was enabled only during my testing when the environmental effects where on. All recording was done through FRAPS. Tests were 1 minute in length on CS:S, and FEAR’s tests were 3 minutes in length. For every one test I did in FEAR, I did three times as many in CS:S to compensate for the shortened period of recorded data. This was due to Counter-Strike’s nature of very short, rapid paced rounds whereas FEAR has long continuous rounds. After everything was said and done, I slowly started to regain my sanity and pulled the data together. The graphs show the average frames per second achieved in each condition and the chart gives a more detailed perspective, showing the maximum and minimum frame rate.
•On the Bench•
AMD Opteron 165 Overclocked @ 2.4 GHz
Abit AT8 32X
2x1GB Kingston HyperX PC3200 KHX3200AK2/2G, Timings 2-3-2-6-1T
Ati Radeon X1800XT
Catalyst Drivers 7.2
Creative Sound Blaster X-Fi Platinum
Western Digital Raptor WD740AD
Windows XP Professional SP2
FEAR: First Encounter Assault Recon
(See Picture: Graph Sets 1-4)
(See Picture: Max/Min/Avg Table)
The first thing I noticed when I compiled all of the data was that my average frame rate in a full server on Counter-Strike: Source was pretty much the same throughout all of my tests, no matter what settings I threw at it. So I went back to that server during the wee morning hours when it wasn’t at full capacity. The server holds a maximum of 32 players and it was always full during my initial testing. But, this time the server was only about half full and my frame rate was more in the neighborhood of 80-90 frames per second rather than the 70 I was averaging before. This only proves that, at least in Counter-Strike: Source, that the number of players has very real influence in performance.
I must say, though, after a few hours of running around alone in an empty server, you go a little crazy. I began contemplating ‘nading myself more times than the average male thinks about intercourse... Despite the urge, I persevered and didn’t blow myself up. I finally finished gathering all of my data, but I did one last test; a run through the CS:S and FEAR Video Stress Tests. With max settings on 1600x1200, AA, AF, and environmental effects turned all the way up, CS:S averaged 109.68 FPS and FEAR a mere 40.34. Both maps proved to be fairly accurate benchmarking tools when compared to the stress tests. Of course, each different map in any game will have its individual graphic demands. But to compare settings, one map in each game was not only necessary for practicality’s sake but to provide accurate results as well.
(See Picture: CS:S Video Stress Test
(See Picture: FEAR Video Stress Test
FEAR had a dramatic drop in the frame rate with the environmental effects turned on. The physics engine seemed to have quite an appetite for a graphics card that was last year’s news. In a crowded server, as long as the environmental effects were disabled, FEAR was 100% playable with Anti-Aliasing and Anisotropic Filtering turned all the way up at 1600x1200. Playing with 12-16 other people all but crippled my rig’s performance with everything on. When there was a lot of action in my immediate area, the frame rate was in the low 20’s at best.
CS:S was not only playable, but for the majority of my testing, completely smooth on any setting at any resolution. Despite a few frame rate dips when someone launched a grenade into a pack of file cabinets, Counter-Strike: Source seemed pretty seamless during my testing. On a side note, I must say, the X1800XT certainly exceeded my expectations during most of the testing.
With a better understanding of the effects of online multiplayer gaming in a first-person shooter, we as gamers have an easier task of optimizing our systems for the job. Before my testing began, I didn’t honestly expect to have as dramatic of a difference in performance as I did between a full server and an empty one. Games today are taxing systems like never before. With advanced physics engines, realistic lighting effects and characters so detailed you can see their nose hair, game developers are demanding more of our hardware every day. Revolutionary graphics cards are emerging and more capable, price friendly PC components are becoming more readily available. Between all that and the recent upgrade frenzy cause by the release of Windows Vista, our standard of gaming is going to shoot up by leaps and bounds. Has anyone seen Crysis? Anyone?
|17 User Comment(s) • 9 root comment(s)|
| Oznog (1) Mar 30, 2007 - 11:31 pm|
|Great Stuff my man. Keep up the good work. I program video games and FPS are my favorite to do. I can tell you from experience that 99.9% of the code writers out there are sloppy and careless. They also have no concept of what EFFICIENT code looks like or why it's important to at least TRY to be efficient. I have seen coders grab and manipulate half a MB of data just to retrieve the 16 bytes they needed in the first place. My guess is they figure "Heck, it's a Pentium 4, it can handle it" and just code away to get the job done, not done right. Sure, a few wasted CPU cycles doesn't mean much, but when your entire job is sloppy, and you have 400,000 lines of code each wasting CPU constantly, you find quite quickly that programs bog down heavily. I've seen on many occasions where a few lines of code changed can turn a 100% CPU gulping monster into a 5% CPU sipping kitty cat. Think Efficient, your framerates will reflect it.|
» Login to reply to this
| OgreFade (150) Mar 15, 2007 - 03:41 pm|
|I'm going to play devil's advocate here... what is the article actually telling/teaching me. If I can't upgrade that I should playin empty servers? Or something? |
I see the benchmarks, and I understand the time and effort they take to do.. but what are we learning from them?
I'm not exactly sure.
» Login to reply to this
| p4l1ndr0m3 (499) Mar 16, 2007 - 06:29 am|
|We learn not to ask questions :P Just kidding of course. I was hoping to enlighten those in the dark about why their favorite games dont seem to perform as well on the interweb as they do playing single player, etc. A great deal of people don't understand that playing online puts additional stress on the machine because of the added players...|
» Login to reply to this
| GrapeApe (36) Mar 18, 2007 - 10:26 pm|
|» VPU vs CPU, needs more insight.|
I agree with OgreFade, the article looks like one where you know that there's an issue, you expose that indeed there is a problem, but don't delve enough into the source or remedy other than playing in empty maps.
From the graphs you posted it seems pretty evident that it's the CPU and not the CPU/graphics that is holding back the multiplayer situations, just looking at the impact of the multiplayer impact, versus the extremly stressful 16x12 with AA has less of an effect than more players.
However, that's never really put forth as the main source for the slowdown.
Good work, just needs a stronger finish. You have the question, your explore the question, but it seems to end there.
» Login to reply to this
» Note: You need to be logged in to write a comment!Login here, or if you don't have an account with FiringSquad, register here, it's FREE!
My Media-Blog categories
No categories created yet.