Jump to content

Rx Vega Clock Problems.


8 replies to this topic

#1 Tharful

    Member

  • Pip
  • Little Helper
  • Little Helper
  • 11 posts

Posted 20 October 2017 - 07:40 AM

Hey guys.
I have a strange issue. In the testing ground my GPU (RX Vega 64) ist working as intendet and clocking around 1550mhz and feeding my 144 fps framecap just fine.... for hours...
Also in the "dropship screen" everything fine and when in the "Game end screen" also...
Bit as soon as i drop out and the match starts and i have control over the mech it clocks down to around 1000mhz and 60-80 frames.
Even when i am dead and spectating its clocking up... ONLY in MP Match and i am controlling my own mech.. then its going down...

No VSync and no matter the graphic settings... even when everything is to the lowest

CPU load is fine (Ryzen 1700)
RAM is almost empty
Temps on CPU und GPU are below/Around 65°C

Every other(newer/more demanding) Game is working fine... and like i said even MWO Testing field!

Already tweaked around in the Radeon Settings and tried diverent profiles but without effekt...
Even reinstalling the Game und all Drivers (Clean/DDU)

Any sugestions?

Edited by Tharful, 20 October 2017 - 07:55 AM.


#2 NARC BAIT

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • 518 posts
  • Twitch: Link
  • LocationAustralia

Posted 21 October 2017 - 12:48 AM

unfortunately, what you see, is 'normal behaviour' .... and has happened to many before you ... and theres no 'real' solution ... these limiters come from within the game code .... and the way it threads, and how ready it can be to deliver the next frame ... most of the games internal timers are synced for either 30, 48 or 60 'updates' per second, these things tend not have any effect, when your not actually 'in control' of your mech ....

try turning the hud off ... you should get an FPS bonus .... another one that amuses me is that showing the scoreboard, will usually increase the load / FPS too .... essentially the game engine in its default configuration cant feed the gpu as many frames as it could handle ... your not insane, but the second you said hardware that isn't what PGI get for free, none of them will have any answers for you ...

I'm in the process of redownloading MWO after doing a fresh install and getting upto the most current windows update, which is meant to have some threading improvements on ryzens ....

youll probably need to mess with some stuff in a user.cfg to non-artificially increase the load ... you mention a 144 hz monitor, but not what resolution your pushing out onto it ... and that's going to be one of the biggest influences overall ...

#3 Tharful

    Member

  • Pip
  • Little Helper
  • Little Helper
  • 11 posts

Posted 21 October 2017 - 11:50 AM

Thx for the answer.
Since its in sutch an isolated and reproducable scenario i thought there is something wrong with my hardware/Settings.
I just dont understand why its ok when i am Spectator and not if a am driving the mech myself....

oh and its wqhd but i already tried playing in HD with same results, only gaining a few top end fps, but its capped anyway.


what about that .cfg? i never dared to change it because... yeah online mp and dont want to get banned and stuff. is it ok for pgi to mess with it? do you have any sugestions?

Edited by Tharful, 21 October 2017 - 11:52 AM.


#4 Nechuchadnezar

    Member

  • PipPip
  • FP Veteran - Beta 1
  • FP Veteran - Beta 1
  • 31 posts
  • LocationWashington, PA

Posted 14 December 2017 - 07:36 AM

Did you ever find anything out about this other than what Narc Bait wrote?

I just purchase a Vega 64 and was very disappointed in how it performs on MWO.

Thank you, friend.

#5 NARC BAIT

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • 518 posts
  • Twitch: Link
  • LocationAustralia

Posted 17 December 2017 - 05:49 AM

the behaviour is not limited to the Vega lineup .... I've had it to one degree or another since I got rid of my nvidia 460 gtx ... I went to a gtx 760 ... suffered underutilization hard .... got a 4K screen ... then a gtx 1060 .... then a ryzen 1600x ... and yeah, back to underutilization ... with my screen I'm pretty much stuck using 60hz, ie 60 FPS, with vsync .... video load 'currently' tends to be between 40-60%, depending on how much of the map I'm viewing .... im thinking on a vega series card, if your not pushing 2K, your probably not forcing it to use enough of its graphics power .... using a 'fullscreen window' its possible to exceed your monitors maximum resolution, its a little bit of a balancing act, and you can totally take it too far .... my 1060 DOES NOT like doing 8K, and up around 16K I was getting about 1 frame every two seconds, not very enjoyable ....

through the use of a user.cfg file, I run a mix between high and low settings, everything in the game settings is set to low, and I turn up the options that I want included, running a user.cfg file to customise things, is completely within the rules, most of the options that would give you an outright advantage, are not modifiable, and those are not the only ones that are locked out from usage, and then, with relation to some values, what you select wont matter, because when you join a game, the server can 'force' keys onto you that suits them .... so, generally speaking, as long as you dont go outrightly mental trying to cheat an advantage, your pretty unlikely to find your way into trouble ... beyond that, theres a rule, and a method you need to be aware of, a user.cfg file is explicitly allowed, and thats where you should keep your customisations, for compliance to the rules, and the method relates to a second file, called 'tuning.cfg' which simply contains the line 'exec user.cfg' .... normally, the user.cfg file is only read once, before you enter your login details, and a good chunk can become irrelevant before you get into the mechlab ... the tuning.cfg file is called, when you spawn or when your released from a dropship ... this will cause a delay at that point, vastly reduced by jumping into the testing grounds, before going into a live game .... please also note, that your user.cfg file is included with any sent through crash dumps ....

to create the files, find the main directory location like "C:\Program Files (x86)\Piranha Games\MechWarrior Online" for the default standalone version, you should have a file called game.cfg present, copy and paste that file into that directory, and it will make a copy, then rename the copy to be user.cfg and repeat for tuning.cfg, be aware, that by default, you wont see filename extensions so easily, and thats why its easiest to copy paste the existing game.cfg file ....

anyway, try adding these to increase the GPU load :
e_Lods = 0
e_TerrainLodRatio = 1
on a ryzen (or intel 4th/5th etc), you might benefit from disabling the jobsystem
sys_job_system_enable = 0
also, you might benefit from keeping all threads on one CCX ... seems to improve my minimum frametimes, but as I'm on a 1600 and I disable SMT, I'm only assigning the game 3 'cores' @ 4ghz ...

for a 1700/1800 with SMT enabled, use 'back' 8 cores
sys_affinity = 65280
for a 1700/1800 with SMT disabled, use 'back' 4 cores
sys_affinity = 240
for a 1600 with SMT disabled, use 'back' 3 cores
sys_affinity = 56


#6 Astennu

    Member

  • PipPip
  • Little Helper
  • Little Helper
  • 22 posts
  • LocationNederlands

Posted 03 January 2018 - 04:09 AM

View PostNechuchadnezar, on 14 December 2017 - 07:36 AM, said:

Did you ever find anything out about this other than what Narc Bait wrote?

I just purchase a Vega 64 and was very disappointed in how it performs on MWO.

Thank you, friend.


I have done some test between a 1080 Ti and a Vega 64 and yes the 1080 Ti is a bit faster but Vega was still more then fast enough to be able to run 4K. The game is mostly CPU Thread bound as in some CPU cores get loaded to 100% cpu load.

If that happens the GPU wont get data fast enough and it will not run at 100%. So the hardware decides it can clock down in order to safe power and still not utilizing 100% of the GPU resources. If the CPU pushes data faster the GPU will clock higher again.

On 4K mine ran max clocks most of the time (optimised vega to not be temperature or TDP limited for the test)
If you have low frame rates with vega you will most likely also have that with a geforce card.

But MWO is more optimised for Geforce cards. I still get very good performance out of my Vega card. nVidia also has some multi threaded driver code that can use 2 cpu cores if needed. AMD does not have that for DX9/11. So nVidia had a little advantage there. But AMD performs better in DX12 titles to bad MWO will never get a DX12 update.

MW5 will have it but might be single player only + COOP and not 12 vs 12 like MWO. (i hope they do port MWO to the new engine tho)

#7 PFC Carsten

    Member

  • PipPipPipPipPipPipPipPipPip
  • Little Helper
  • Little Helper
  • 2,188 posts
  • LocationOn your six

Posted 03 January 2018 - 04:55 AM

What you are seeing with Vega and Geforce cards Is not under-utilization but modern power management. The card clocks only as high as it must in order to achieve near-cpu-limit framerates. MWO is CPU-limited with faster graphics cards in real gameplay as opposed to testing grounds.

You can test this by overclocking your CPU and see how se GPU clock rises in sync.

#8 NARC BAIT

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • 518 posts
  • Twitch: Link
  • LocationAustralia

Posted 03 January 2018 - 08:19 AM

View PostPFC Carsten, on 03 January 2018 - 04:55 AM, said:

You can test this by overclocking your CPU and see how se GPU clock rises in sync.


I see where your going ... but I dont quite agree with your description ...

I'm using vsync, limited to 60hz, and on the monitor I currently have, thats a hard limit, for reasons completely beyond me, more than that makes MWO 'shudder' .... anything else can exceed 60 fps without issue ... but I've given up on trying to fix that, its just a quirk of the hardware I currently have ... 60 fps works out for me, because thats the same rate that I'm recording at ... 60 FPS, at 3072x1726 generally results in about a 30%-40% load on my hardware if I leave it on the low settings, and before long I'm spiralling downwards on the performance numbers ... as you describe, such a low actual load has the GPU lower its clock speeds, it doesnt need to use its full power to draw the scene, and lowers its power usage to increase 'efficiency' ....

and the CPU usage is not a factor at my end, a minimum frame is not caused by a CPU usage spike, of if there is, its so minimal that I cant statistically correlate it .... in fact, I can drop the game to only be running on two of six cores, and play along with no extra issues at all, those cores will jump up to about 60%-80% usage .... PGI claiming 'your limited by something outside of our control' line doesnt hold up to scrutiny at my end ... not that it ever really has ....

I've been stuck on this path ... since my last system (FX/760 @ 1080p) ... and your only real hope is to increase the load on the GPU ... you could try other routes ... but the behaviour generally stay the same .... underutilization lowers the clockspeeds internally on the GPU, and the only solution is to stack more load on it ...

#9 PFC Carsten

    Member

  • PipPipPipPipPipPipPipPipPip
  • Little Helper
  • Little Helper
  • 2,188 posts
  • LocationOn your six

Posted 03 January 2018 - 08:27 AM

MWO does not make good use of cores beyond 2-3. I tried running XMR-mining on 2 of my 4 CPU cores in the background and Framerate did not suffer. It's something very abstractly described as "game loop", which basically is tied to your CPUs IPC×MHz, iow single-thread performance. You won't see anything worthwile dropping to 2-3 cores, but you will probably see a near linear increase in (critical) fps when going from 3 to 4 GHz. At least that's what I'm seeing with my FX 6300 and a R9 Nano.





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users