System Optimisation For Amd Chipsets
#41
Posted 14 May 2014 - 03:02 PM
And then Tek Syndicate ran a test where they did streaming while gaming and the FX8350 beat a few of the intel chips soundly when you didn't have the absolute best GPU at the time.
https://teksyndicate...oth-overclocked
http://www.pcgamesha...chmark-1056578/
Also when the resolutions and details are higher there is MUCH less of a performance delta between systems. The 800x600 test is stupid.
Side note: Tomshardware is GARBAGE and not worth using.
#42
Posted 14 May 2014 - 03:19 PM
http://www.tomshardw...ing,3451-8.html
Don Woligroski of Tom's Hardware said:
Having said that, AMD's FX-8350 provides serviceable Crysis 3 game play. Despite the frame rate valley we experienced in our benchmark run, this CPU achieves smoother performance on average. Perhaps this is something Crytek will be able to address through a future update.
#43
Posted 15 May 2014 - 09:38 AM
Nick Rarang, on 14 May 2014 - 02:58 PM, said:
So did you try to game on a Power6/Power7 system running Red Hat and WINE during the PS3/X360 era?
The Tom's Hardware link from Goose shows exactly what I was talking about - the minute those AMD chips have to do a lot of other work, their lack of perf/clock becomes more apparent. A chip with a 500-700MhZ deficit and 4 less cores has a 30% better minimum FPS. It's an absolutely terrible design. Again, from an earlier post, even AMD knows this and is designing a new architecture completely astray from the modular arch bulldozer/piledriver/steamroller they've been selling. The new design takes cues from AMD's Thuban and Intel's Sandy Bridge.
As a side note: Yeah, Tom's Hardware is mostly garbage. It's too bad HardOCP doesn't get new chips from AMD to test with, or we'd get a more straight-up answer as to how big of a difference there is and in what scenarios it matters. Either way, pretty much any review you look at will tell you CryEngine3-based games need an i5 or better combined with all the GPU power you can possibly throw at it. The difference between DDR3-1600 and DDR3-2400 is a good few FPS, too.
#44
Posted 15 May 2014 - 02:23 PM
Why in the name of all that's holy can't a bunch of devs figure out how to use a GPU in the year of our Lord 2014. After what has seemed like 2 years of pre-alpha.
#45
Posted 15 May 2014 - 07:36 PM
#46
Posted 15 May 2014 - 07:46 PM
#47
Posted 16 May 2014 - 01:15 AM
BF4 runs at 50-60 with everything maxxed out, with the occasional 30 and under in cutscenes. COD: Ghosts has a similar FPS at highest settings with all the bells and whistles.
Every game I have can be played at well over acceptable FPS with this rig in 1080p resolution with everything maxxed out ghaphics-wise.
I guess what I'm saying is I'm angry at myself for sinking so much cash and time in this trainwreck of a game. If it wasn't for the wonderful people I play with I would have quit it a long time ago.
#48
Posted 16 May 2014 - 01:44 AM
Golrar, on 13 May 2014 - 06:20 AM, said:
but you paly game X now.
and in future you may also have replaced your CPU, Since "future" games may easily be 2-3 years before they come out.
So whats the sense of a relying on a possible used feature?
You are trying to tell people they should not buy the sportscar and buy an SUV because in future streets may be so damaged that you can't drive on them at full sports cars speed anymore. But in real now and here streets are fine, and the future is unknown.
#49
Posted 16 May 2014 - 05:52 AM
DarphBobo, on 16 May 2014 - 01:15 AM, said:
Did you not get the memo on how this is a battalion-scale game using a squad-/pedestrian-scale engine? Nomad and Psycho are not Quickdraw sized, you know.
#50
Posted 16 May 2014 - 02:00 PM
Krinkov, on 14 May 2014 - 07:35 AM, said:
So, yeah; About that: If you add
ca_thread0Affinity = 0 ca_thread1Affinity = 0 r_WaterUpdateThread = 0 sys_main_CPU = 0 sys_physics_CPU = 0 sys_streaming_CPU = 0 sys_TaskThread0_CPU = 0 sys_TaskThread1_CPU = 0 sys_TaskThread2_CPU = 0 sys_TaskThread3_CPU = 0 sys_TaskThread4_CPU = 0 sys_TaskThread5_CPU = 0to your user.cfg, and stop with the Lasso tweak, what happens?
If you change them zeros to, say, all even numbers, what happens?
(And why it took me so long to ask … )
Edited by Goose, 16 May 2014 - 02:02 PM.
#51
Posted 16 May 2014 - 03:13 PM
FOR THE REST OF YOU ARGUMENTATIVE SHITHEADS AND VARIOUS OTHERS i find the whole amd vs intel battle a waste of time ...my problem is only with this game and this game only so who gives a **** what your using to play it, im sorry but if i can run any other game including crysis 2 on ultra and crysis 3 on max (which amuzingly is 1 step below ultra settings) and also something that can be more directly compared to mwo which is warface which is a crytek dev'd game that is also online and using cryengine and get 60+fps with a amd fx 4100 clocked up to 3.88 8gb ddr 1600 clocked down to 14**Mhz (forget the exact # but its in the 1400-1490MHz range )for optimal timings and a old ass radeon 7770 GHz edition.... you would think a so called beta online game like this that is using cryengine would run decent as well on even medium settings but yet i bottom out to 20fps in a furball so to speak ...obviously pgi does not know how to deal with cryengine and make a game that uses it regardless of cpu/gpu branding and architecture
#52
Posted 20 May 2014 - 04:46 AM
My system specifications are as follows:
AMD FX-8150 Eight-Core Processor (3.6ghz)
16GB DDR RAM
AMD Radeon HD 7900 Series Graphics
Win 7 64-bit
Game is installed on a SSD
When I run the game on maximum settings I only get about 20-25 FPS, and when I run the game on the lowest settings I also get around 20-25 FPS so reducing the graphics settings is not helping me.
I can run almost any other game on full settings (BF4, COD, RAGE, Skyrim) and not experience any drop in FPS. What is most sad is that the game was actually running better on my 1st generation i5 laptop
Any advice would be much appreciated!
Edited by Nitzol, 20 May 2014 - 04:50 AM.
#53
Posted 21 May 2014 - 05:29 PM
#54
Posted 21 May 2014 - 06:13 PM
Edit:
I also run my game on a 5 gig ramdisk which seems to help too. It removed the pauses caused by looking at new mechs as they came into view.
sys_MaxFPS = xx should be set to somewhere between your max and min frame rate and may take some tweaking to dial in to the right amount. After these changes, I went from 20-30 frames to a steady 40.
r_MultiThreaded = 1
sys_MaxFPS = 40
ca_thread0Affinity = 0
ca_thread1Affinity = 2
r_WaterUpdateThread = 4
sys_main_CPU = 0
sys_physics_CPU = 6
sys_streaming_CPU = 4
sys_TaskThread0_CPU = 0
sys_TaskThread1_CPU = 2
sys_TaskThread2_CPU = 4
sys_TaskThread3_CPU = 0
sys_TaskThread4_CPU = 2
sys_TaskThread5_CPU = 4
sys_flash_address_space = 536870912
gfx_ampserver = 1
gfx_loadtimethread =1
sys_flash_allow_reset_mesh_cache = 1
ca_KeepModels = 1
r_TexturesStreamingOnlyVideo =1
gfx_loadtimethread = 1
gfx_inputevents_triggerrepeat = .025
Edited by Krinkov, 21 May 2014 - 06:19 PM.
#55
Posted 21 May 2014 - 06:25 PM
Krinkov, on 21 May 2014 - 06:13 PM, said:
Edit:
I also run my game on a 5 gig ramdisk which seems to help too. It removed the pauses caused by looking at new mechs as they came into view.
sys_MaxFPS = xx should be set to somewhere between your max and min frame rate and may take some tweaking to dial in to the right amount. After these changes, I went from 20-30 frames to a steady 40.
r_MultiThreaded = 1
sys_MaxFPS = 40
ca_thread0Affinity = 0
ca_thread1Affinity = 2
r_WaterUpdateThread = 4
sys_main_CPU = 0
sys_physics_CPU = 6
sys_streaming_CPU = 4
sys_TaskThread0_CPU = 0
sys_TaskThread1_CPU = 2
sys_TaskThread2_CPU = 4
sys_TaskThread3_CPU = 0
sys_TaskThread4_CPU = 2
sys_TaskThread5_CPU = 4
sys_flash_address_space = 536870912
gfx_ampserver = 1
gfx_loadtimethread =1
sys_flash_allow_reset_mesh_cache = 1
ca_KeepModels = 1
r_TexturesStreamingOnlyVideo =1
gfx_loadtimethread = 1
gfx_inputevents_triggerrepeat = .025
Im going to try this out......lets see if my 8 cores with affinity program core parking off 100% all the time.....90FPS when i take control of mech, 35FPS in the worst of brawls in a LRM storm with lazer light show....... will fare better after this "cfg"
#56
Posted 21 May 2014 - 06:57 PM
Real Testing™!
How did you pick them thread placements? AFAICT, it's ca_thread0Affinity is the busy one, and and not sys_physics_CPU, thus it's ca_thread0 that needs to have a core all to itself.
I also remember sousing out how sys_TaskThread2_CPU and sys_TaskThread4_CPU where somehow "special," and needed more attention then the other four, or even more then threads not-named "ca_thread0Affinity" …
The next quistion(s) becomes
p_num_threads = 1 sys_limit_phys_thread_count = 1 ;help: Limits p_num_threads to physical CPU count - 1 sys_job_system_max_worker = 8 s_NumLoadingThreadsToUse = 2 r_ShadersAsyncMaxThreads = 1 e_AutoPrecacheCgfMaxTasks = 8All 'cept "thread_count" can be set to 4 or 8† for a Piledriver, but even if set to 4, will the be on the right cores without Process Lasso (or Bill2's Process Manager) to "guide" them?
Also: sys_budget_soundCPU of 5 or less seems good to have …
†Er: s_NumLoadingThreadsToUse will only go as high as 5, and sys_limit_phys_thread_count needs to be 0 for p_num_threads to be in play.
#57
Posted 21 May 2014 - 07:53 PM
Goose, on 21 May 2014 - 06:57 PM, said:
I set the cores based on just monitoring cpu usage. The core with physics on it was running the highest so I left it on it's own core. I will try messing with the affinity threads more. Did you try the scaleform bottleneck fix?
#58
Posted 21 May 2014 - 10:16 PM
But how did you figure that was the Physics thread you were looking at? Using a lot of inconsistent results, I'd descided Teh Devs had locked down the threads not-named "Task" to their default assignments, thus the always busy core, the last one on both the more-then-dual-cores I've tested, was ca_thread0Affinity.
Here are the defaults:
;ca_thread0Affinity = 5 ;ca_thread1Affinity = 3 ;r_WaterUpdateThread = 5 ;sys_main_CPU = 0 ;sys_physics_CPU = 1 ;sys_streaming_CPU = 1 ;sys_TaskThread0_CPU = 3 ;sys_TaskThread1_CPU = 5 ;sys_TaskThread2_CPU = 4 ;sys_TaskThread3_CPU = 3 ;sys_TaskThread4_CPU = 2 ;sys_TaskThread5_CPU = 1
#59
Posted 21 May 2014 - 10:56 PM
I'm not going to be much help other than just relaying what I have found works for me. This really isn't my area of expertise. I've just been combing forums and cryteks commands trying to find anything that might help, pretty much throwing everything at it and seeing what sticks.
#60
Posted 21 May 2014 - 11:08 PM
Thanks for reporting anything atoll, man: It was insightful …
1 user(s) are reading this topic
0 members, 1 guests, 0 anonymous users