54 replies to this topic

[MechWarrior4172571]

Smokeyjedi, on 30 December 2014 - 10:13 PM, said:

That is a solid idea.....why didn't I think of that ......I do notice that there is a mesh bubble back there .......Dremel where are you.........this could also free up some CFM for radiator not having to push through thick steel with mesh holes......Thank you........ And I could add another fan the same as that one right above that one for maximum vrm cooling
(Yes I am looking for any reason to blow this rig sky high..........X99 whispers to me while I sleep )
ok than....../Here we go again.......

Yep, and if your cooler's warranty run out.. you could.. just.. replace the tubing.. and.. make.. them.. longer.. and.. take the whole thing outside of the case.. (another idea).. or/and jury-rig 2 fans together outside of the case so there would be greater tunnel action.. (not really aesthetically pleasing to look at but it would give greater wind-force with the same type of rpm speed/noise fans as before. AH, but then when will you have time to play MWO? duh

[Lordred]

Smokeyjedi, on 30 December 2014 - 07:10 PM, said:

I find that 200X23 = 4600mhz is slower than
228X20 =4560mhz - this would benchmark higher in general.
255X19-19.5-4860-4990mhz is the "sweet spot"** FOR MY MOBO**
260X15-17 BSOD
270X13 BSOD
I have heard of sabertooth boards hitting 280X18-19 for 5040-5200mhz but thats custom loop cooled and military graded nonsense. Hog wash, Lemmi @ em.......LOL

/necro

Get gud son!
321.3mhz Ref clock, still hold 1st place.
297.05mhz Ref clock an a K8!


I kid. But in all honesty, for a FX, Aim for ~2400 on the NB, and aim for 2900-3100 for the HT. Loosen up your memory timing and aim for 2300+

[Smokeyjedi]

Jesus DIED for me, on 30 December 2014 - 09:10 PM, said:

Ok, one thing I noticed that seems wrong with this picture is the CPU heatsink blocking all air movement to the VRM heatsink. What you can do is to remove the rear fan from the water-cooler and place that fan on the back of chassis--this would open up the VRM heatsink to more air movement because the water-cooler is now further away to the wall of the chassis than before... and this just might give you the extra bit of overclocking headroom. You might have to re-orient the cpu block so that the cooling pipes might reach the intended location of the cooler (it would be further away if you go this way.)

Dude 5017 mhz stable on a 970 AM3+ (non gigabit)and torture testing lasted 30 mins before I gave up and called it a win.... I flipped one of the exhaust fans above the RAM and turned it into intake to blow cold air on ram, Moved pull fan from Rad(inside case), to outside of case(required minor modding) and voila' 5.0ghz stable.........Thank you........going from 4890 mhz to 5017mhz bumped my FPS in game a few more.....Min,Max,Avg. A long with no more BSOD in MWO.......Have sucessfully booted into win 7 64 @ 5.2ghz.....but @ 1.52V I said too bad and settled for 5017mhz

[MechWarrior4172571]

Smokeyjedi, on 03 January 2015 - 03:58 PM, said:

Dude 5017 mhz stable on a 970 AM3+ (non gigabit)and torture testing lasted 30 mins before I gave up and called it a win.... I flipped one of the exhaust fans above the RAM and turned it into intake to blow cold air on ram, Moved pull fan from Rad(inside case), to outside of case(required minor modding) and voila' 5.0ghz stable.........Thank you........going from 4890 mhz to 5017mhz bumped my FPS in game a few more.....Min,Max,Avg. A long with no more BSOD in MWO.......Have sucessfully booted into win 7 64 @ 5.2ghz.....but @ 1.52V I said too bad and settled for 5017mhz

Awesome. I am glad that helped out with cooling and overclocking. I do suggest you up other voltages though beside (instead of) the CPU which is pretty obviously maxed out. There might be some OC stability hiding in other voltages. Ram Voltage and tighter timing (less lag = greater stability) might be a thing to look at. RAM is soo tied together to the OC in AMD processors. Play with RAM a little and definitely think about getting a x2 60mm corsair RAM cooling fan or something similar. I can't think of serious OCing without seriously pushing the RAM voltage wise. My ram is rated at 1.75V lol but I keep it at 1.615V and it still gets hot (I disabled over-current and over-voltage protection in the mobo for the RAM--might not be the smartest thing to do in the long run but it gives me the kick and RAM is not CPU, it's built on a lot larger manufacturing process so it's like a tank in comparison) Here is the RAM cooler I use.. http://www.newegg.co...1-008-_-Product It's a little noisy without a resistor and mobo duty cycle adjustment though. There are cheaper and cooler looking ones available, of course,.. http://www.ebay.com/...=item27f56e2dfa But what I wouldn't mind to try is something like this.. http://www.ebay.com/...=item4adaf6b361 Or at the least something like this (as in comparison to my current aluminum heat spreaders).. http://www.ebay.com/...=item2a27f63354 Doesn't hurt to dream every so often.

Edited by Jesus DIED for me, 03 January 2015 - 09:52 PM.

[orcrist86]

As a follow up, I've been poking around the forums on the amd and thread manipulation settings in the user.cfg.

I'm going with this in my cfg.

gp_option_ShowCockpitGlass=0
r_MultiThreaded = 1
sys_MaxFPS = 120
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
p_num_threads = 4
sys_limit_phys_thread_count = 4
;help: Limits p_num_threads to physical CPU count - 1
sys_job_system_max_worker = 8
s_NumLoadingThreadsToUse = 2
r_ShadersAsyncMaxThreads = 1
e_AutoPrecacheCgfMaxTasks = 8

Seems to have added a few frames, weirdly its evened out the cores, all 8 hover between 30 and 50%. GPU still sits in the 50-80% My high end FPS jumped to between 90 and 120 on CW. Low end is averaging around 50. So I picked up an average 10FPS. The FPU thing on AMD is really holding it back apparently, but so long as I can play with a decent frame rate I'm happy. I'll be upgrading my ram next to something clocked higher, and probably adding an SSD. Still running stock voltage and clock speeds.

[Goose]

Even numbers between 0 and 6? ca_Thread0Affinity is sharing with someone? And it works?

Good job sussing out working strings; But I think you wan'a change sys_limit_phys_thread_count to zero …

Edited by Goose, 04 January 2015 - 02:08 PM.

[Exarch Levin]

Quote

The FPU thing on AMD is really holding it back apparently, but so long as I can play with a decent frame rate I'm happy.

More like Crytek's shoddy implementation of FPU operations are really holding it back but I guess the result is the same to the end user; would it kill a game engine maker, especially one that is working now with AMD on Mantle implementation, to code and compile for AMD?

It'd be great if there were some resource that mentioned in detail what each of these CFG commands did and how their values work as these might as well be a mystic incantation otherwise. The Crytek documentation Google gives me often commits the cardinal crime of using the word in its own definition.

[xWiredx]

Exarch Levin, on 04 January 2015 - 09:09 PM, said:

More like Crytek's shoddy implementation of FPU operations are really holding it back but I guess the result is the same to the end user; would it kill a game engine maker, especially one that is working now with AMD on Mantle implementation, to code and compile for AMD?

No, AMD's design is quite flawed, and it has everything to do with sharing certain resources (FPU!) amongst 2 "cores". This is something that has been known for years already. They gimped their own hardware. It has nothing to do with "coding and compiling for AMD". If that were the case, Bulldozer and Piledriver chips would crush Intel chips in a lot more than one or two hand-selected benchmarks.

Yes, most things are compiled with certain flags, and most of them use things native to Intel. In fact, some software has been caught being compiled using flags that would directly be beneficial to Intel. The problem is, recompiling it never yielded much extra performance and Intel chips still came away with faster performance. Cryengine 3 is not one of those things that has been intentionally or even unintentionally skewed, though.

The documentation for many Cryengine cvars is piss-poor. I'll give you that. Probably will end up relying on random google result links for developer forums more than their own documentation.

[ninjitsu]

I'm with xWiredx, AMD's architecture is flawed. AMD is working on a new architecture. Until then, I'm avoiding them.

[Exarch Levin]

Quote

No, AMD's design is quite flawed, and it has everything to do with sharing certain resources (FPU!) amongst 2 "cores". This is something that has been known for years already. They gimped their own hardware. It has nothing to do with "coding and compiling for AMD". If that were the case, Bulldozer and Piledriver chips would crush Intel chips in a lot more than one or two hand-selected benchmarks.

I'm not riding on that bandwagon. Where is any application that was truly optimized to work with AMD's design? AMD released a guide for how to optimize code for their processors ages ago but yet there aren't any that claim to be, so far as I can find. The blame with this lies with AMD not having a compiler that they can point devs to. Intel, meanwhile, does.

If a Bulldozer module were a car, it'd be one that no driver had ever shifted beyond first gear. I'm not saying it could beat Intel to the finish, just that it could improve its own time across the line drastically.

The FPU design has issues. AMD obviously put the cart miles before the horse with that, and the horse is just now starting to catch up with Kaveri/Carrizo. Still, it's not so much "shared" as it is "separate" and a radical departure from the norm like that required AMD to do much more to get devs to optimize their code.


Regardless, PGI's implementation of the Cryengine runs far poorer than it should on any and every system and the visuals accompanying this poor performance are, frankly, worse than those of Crysis. That's really the heart of the issue: PGI needs to fix its game.

[MechWarrior4172571]

Exarch Levin, on 06 January 2015 - 12:14 AM, said:

I'm not riding on that bandwagon. Where is any application that was truly optimized to work with AMD's design? AMD released a guide for how to optimize code for their processors ages ago but yet there aren't any that claim to be, so far as I can find. The blame with this lies with AMD not having a compiler that they can point devs to. Intel, meanwhile, does.

If a Bulldozer module were a car, it'd be one that no driver had ever shifted beyond first gear. I'm not saying it could beat Intel to the finish, just that it could improve its own time across the line drastically.

The FPU design has issues. AMD obviously put the cart miles before the horse with that, and the horse is just now starting to catch up with Kaveri/Carrizo. Still, it's not so much "shared" as it is "separate" and a radical departure from the norm like that required AMD to do much more to get devs to optimize their code.


Regardless, PGI's implementation of the Cryengine runs far poorer than it should on any and every system and the visuals accompanying this poor performance are, frankly, worse than those of Crysis. That's really the heart of the issue: PGI needs to fix its game.

So much potential.. you are correct. Say good bye to the Blue team and say hello to the new ruler of the MWO universe.. http://www.wordtechr...k-with-fx-8370/ [br]

Quote

"...The Stilt, a famous overclocker, hit some incredible results with AMD’s latest FX-8370 8 core CPU. First, for those of you that don’t know, the FX-8370 is essentially a slightly better 8350 that is both cheaper and newer, it’s not much different when it comes to the actual CPU though, other than minor clock rate changes.
Anyway, an incredible 8.722Ghz was achieved using this CPU, and there is a CPU-z window to prove it (below), this clock rate is absolutely insane, almost unbelievable. Now some of you might be thinking that you have heard of slightly higher overclock, and you would be right a 72Mhz higher overclock has been achieved using the FX-8350, but that was with ony 2 cores enabled, this overclock is with all 8 cores enabled. Something that is practically unheard of, having this high of a clock rate on an 8 core CPU, also note that the AMD 8 core CPU’s are true 8 cores and not 4 cores with hyper threading like Intel."

[br]

[Smokeyjedi]

Jesus DIED for me, on 06 January 2015 - 11:46 PM, said:

So much potential.. you are correct. Say good bye to the Blue team and say hello to the new ruler of the MWO universe.. http://www.wordtechr...k-with-fx-8370/ [br][br]

[Bill Lumbar]

Smokeyjedi, on 30 December 2014 - 06:38 PM, said:

gp_option_ShowCockpitGlass=0 (off/on)
r_DepthOfField=0 (off/on)
r_HDRGrainAmount=0.0 (film grain amount)
r_motionBlur = 0
r_MultiThreaded = 1
cl_fov = 70 (Default is 75)


sys_MaxFPS = 144
d3d10_TripleBuffering = 1
d3d11_TripleBuffering = 1
d3d9_TripleBuffering = 1

e_GsmCache = 1
r_FogShadows = 0

q_ShaderWater = 0

r_silhouettePOM = 0
r_UsePOM = 0

*There is a process to calculate this......-using command console.wheres that linkhttp://mwomercs.com/...ost__p__4040171-Winsat MEM is the key young padawan.
sys_budget_streamingthroughput = 21250048
sys_LocalMemoryGeometryStreamingSpeedLimit = 20752
sys_LocalMemoryTextureStreamingSpeedLimit = 20752
sys_streaming_max_bandwidth = 20752




r_WaterUpdateThread = 7


ca_thread0Affinity = 0
ca_thread1Affinity = 4

sys_budget_soundCPU = 7

sys_budget_videomem = 2096

sys_main_CPU = 0
sys_streaming_CPU = 1
sys_physics_CPU = 2
sys_TaskThread0_CPU = 3
sys_TaskThread1_CPU = 5
sys_TaskThread2_CPU = 5
sys_TaskThread3_CPU = 6
sys_TaskThread4_CPU = 7
sys_TaskThread5_CPU = 6
;sys_TaskThread6_CPU
;sys_TaskThread6_CPU
;sys_TaskThread6_CPU
;sys_TaskThread6_CPU
;sys_TaskThread6_CPU
;sys_TaskThread6_CPU

These ^ were to doll out all 12 named threads that crytec uses.................After this there was no need to do so.............
I always have one core max out 55-60% while the rest are pegged @ 80-98% with the occasion 100% on up to 3 cores.....It does dance....I cant control it completely.

But Match after match 300 second benchmark after benchamark I can make my cores peak......

Along side this I run 'Core parking' program to unpart 100% of all cores........all the time.......

I run Project lasso to force realtime affinity to MWOclient -Also use this to limit google chrome from using module 4(7+8cores) cause @ this clock speed online videos crash my rig......been chasing this gremlin for years and I believe it to be the OC period. -fresh install or tried tested and true virus free win 7 64 bit(3 months into new install)

I run FX8350 @ 255X19 for 4860mhz for 24/7 OC........1080P Movies convert+burn to disc in under 10 minutes..........

Gskill sniper 2133 ddr3 @ 2080mhz 10-10-10-30 (10-9-10-30 was twitchy and a diminishing return)@1.65V

Asus M5a97EVO 2558 HT + NB OC with moderate voltage bumps.......to stabilize high FSB on this mediocre MOBO.(custom
VRM cooling fan installed)

Adata SP600 SSD 256GB(mid range non performance based)

Corsair H80 push pull max speed + aggression-!!barely enough!!.....but VRMs being weak make this do just fine. After all I cannot stabilize 5.0+ ghz for MWO.

Zotac AMP 760......2gb.........sure its like SLI GTX460s(had 1 before) but I would have saved that 300$ and the $100-150 I spent getting this thing Oc'd to 4.8+ ghz-( H80+gskill sniper2133) to really get the OC screaming >1866DDR3 is where its @

Now for today so far I bumped my voltage to 1.520V and my multiplier to 19.5 for 4989mhz........Played over 10 matches and haven't BSOD........I even held MWO @ my 144hz monitors refresh.........for a hot second or two.....LOL..........Havent seen higher than 135FPS below 4989mhz................I may have found my absolute limit for Mobo and CPU + RAM + PSU I am literally at the end of the line..........There is nothing I can even do to get more.........aside from a new AM3+ MOBO , not a chance......X99

first string is solid and perfectly enjoyable.........worked well for months........
2 days ago I changed all my settings and tweaked a new way from All med in game settings with MSAA forced VIA Nvidia control panel.....I am so much more happy. Its much prettier and smoother as well as pushing my frames up- the same similar dips happen but less noticeable I am truly happy now with My rig for MWO- MEOW! jedi out

2014-12-25 00:33:24 - MWOClient
Frames: 24293 - Time: 389128ms - Avg: 62.429 - Min: 36 - Max: 131

2014-12-26 20:57:29 - MWOClient
Frames: 15062 - Time: 360000ms - Avg: 41.839 - Min: 0 - Max: 120

2014-12-26 21:21:53 - MWOClient
Frames: 19178 - Time: 298664ms - Avg: 64.213 - Min: 33 - Max: 137

2014-12-27 01:43:18 - MWOClient
Frames: 20893 - Time: 360000ms - Avg: 58.036 - Min: 33 - Max: 133

2014-12-27 01:56:30 - MWOClient
Frames: 16591 - Time: 360000ms - Avg: 46.086 - Min: 28 - Max: 86

2014-12-27 23:55:42 - MWOClient
Frames: 19198 - Time: 360000ms - Avg: 53.328 - Min: 37 - Max: 105

2014-12-28 00:26:42 - MWOClient
Frames: 14015 - Time: 262363ms - Avg: 53.418 - Min: 30 - Max: 113

2014-12-28 00:41:58 - MWOClient
Frames: 21241 - Time: 360000ms - Avg: 59.003 - Min: 32 - Max: 83

-------------------------------------------------------------------------------------
new in game settings....MSAA on 4X(-forced VIA NVIDIA CNTRLPANEL) mostly medium low part. low shaders,low postprocessing.
2014-12-28 11:35:54 - MWOClient
Frames: 24271 - Time: 360000ms - Avg: 67.419 - Min: 36 - Max: 77

2014-12-28 21:01:53 - MWOClient
Frames: 17539 - Time: 264624ms - Avg: 66.279 - Min: 39 - Max: 118-veridian bog

2014-12-28 22:23:10 - MWOClient
Frames: 25070 - Time: 360000ms - Avg: 69.639 - Min: 38 - Max: 138

2014-12-29 21:47:15 - MWOClient
Frames: 30008 - Time: 360000ms - Avg: 83.356 - Min: 35 - Max: 125

*than I bumped cockspeed again* I will add about 10 random benchmarks from matches to show the slight improvement over these*** I really enjoyed this last week in MWO getting to the cusp of my rigs limits.....MWO runs beauty.............after 3.years of tweaking Oc'ing and Help from My techie buddies here in the forums, you know who you are fellas. And you are all maddogs.....

Here is a run on the bog last night and I have not touched any of my config. files and every thing is on high-very high, cockpit glass on, I have just recently turned particles down to med, and only because of my Locust PB's mg flash is way to much to take for me. Someone recommended I turn them down and it would help... I used to run them on very high also.

My FX-8350 is at 4.9 ghz right now 24/7 and my 7970 is overclocked to 1100, 1500 on the ram and this was on 1920x1200 30" hanns-g monitor. My new Ben-Q 32" 1440P is on its way and will be here any time today.

Its great that you and others are modding your files to try and gain more performance from your FX or Intel chips..... But I look at it like I shouldn't have to mod any of the config files, settings in game yes, config files... Hell no. I will bench some more runs to compare mine with yours, this run int he bog was on Windows 7, not 10.

2015-02-05 01:10:38 - MWOClient
Frames: 12173 - Time: 180000ms - Avg: 67.628 - Min: 30 - Max: 108

2015-02-05 01:13:49 - MWOClient
Frames: 13790 - Time: 180000ms - Avg: 76.611 - Min: 46 - Max: 108

Seems like a lot of work for very small results vs. what my FX is putting out, and my in game settings are set higher then yours.

Edited by Bill Lumbar, 05 February 2015 - 07:19 AM.

[Smokeyjedi]

Bill Lumbar, on 05 February 2015 - 07:16 AM, said:

Here is a run on the bog last night and I have not touched any of my config. files and every thing is on high-very high, cockpit glass on, I have just recently turned particles down to med, and only because of my Locust PB's mg flash is way to much to take for me. Someone recommended I turn them down and it would help... I used to run them on very high also.

My FX-8350 is at 4.9 ghz right now 24/7 and my 7970 is overclocked to 1100, 1500 on the ram and this was on 1920x1200 30" hanns-g monitor. My new Ben-Q 32" 1440P is on its way and will be here any time today.

Its great that you and others are modding your files to try and gain more performance from your FX or Intel chips..... But I look at it like I shouldn't have to mod any of the config files, settings in game yes, config files... Hell no. I will bench some more runs to compare mine with yours, this run int he bog was on Windows 7, not 10.

2015-02-05 01:10:38 - MWOClient
Frames: 12173 - Time: 180000ms - Avg: 67.628 - Min: 30 - Max: 108

2015-02-05 01:13:49 - MWOClient
Frames: 13790 - Time: 180000ms - Avg: 76.611 - Min: 46 - Max: 108

Seems like a lot of work for very small results vs. what my FX is putting out, and my in game settings are set higher then yours.

Lol bill, your still way off, Look @ your mins and max. Maybe my numbers have changed too since the patch. funny guy.

[Catamount]

Yeah stutters and serious time spent in the 30s is little better than time spent in the 20s. 40s isn't very good either even though it's what we've had to deal with. These lower framerates hit precisely when performance is most needed, during heavy combat. Getting a 60-70fps average because you got great framerates when literally nothing was going on is absolutely useless. So yes, anything one can do to bring those numbers up is worth it, if one actually cares about one's performance. My faction's place is at stake here. If I can't run a good high-pace light run, or fight against them, because I'm stuttering about or getting 35-40fps when I want 60, that matters, and I'm going to do as much to lift those numbers up as I can. I'm not even high-comp and it matters.

I got mid-40 minimums at worst even before the HUD work, and no one's in-game settings are higher than mine, because mine are maxed across the board. Actually, Bill, since you turn off AA, mine are higher than yours

I know some people turn off damage glow, but I don't. It has utility in spotting mechs in darker maps when a graze with a laser makes them stick out like a sore thumb.

Besides, isn't it a bit illogical to criticize simple CFG changes when you take the time to OC your rig? What takes more time, ocing your CPU and GPU, or dropping some lines into a text file that others have already done the work on for you? It's curious that you'll do one, but not the other. Even if you're like me and having some issues with the provided thread assignments and need to take some time to see why, it's still no more work than OCing is, nor fundamentally different (tweaking numbers for settings, seeing if they make things better or blow up in your face).

Edited by Catamount, 05 February 2015 - 10:04 AM.