Pilotasso's new mental bonkers Rig finally (OMG what have I done)

[Pilotasso]

Only FC3, A-10 and Mirage. Plan on getting the fast jets above anything else, WWII planes maybe in a bundle with discount.

[BitMaster]

I did the benchmark run with VSYNC + MSAA + AF to OFF, HIGH PRESET elsewhere

 

Start 104/108

Nellis 85

Vegas Strip 60

Vegas Airport 65

 

The fps are +/- 3-5 fps depending on when exactly you take the note, those values are the lower ones

 

FYI: the gpu was at constant 99% load and the system in full oc as in sig

[BitMaster]

Benchmarked again at stock settings for CPU & GPU ( XMP enabled, all cores 4.5GHz )

 

Start 92/96

Nellis 76

Vegas Strip 48/50

Vegas Airport 54/56

[Pilotasso]

this is very interesting. Could a GFX change bring both intel & AMD machines level despite the clock difference? wow, I think it can be done.

Edited May 9, 2017 by Pilotasso

[BitMaster]

I think I need a new GPU :(

 

This is more GPU bound than anything else.

 

I will do another run with CPU in default and GPU in max OC. I expect a big jump in fps towards the "all oc" fps, which in return would mean the CPU at 4.5 is plenty for THIS GPU.

 

Might look totally different with a 1080Ti or Vega, regardless if the cpu is oc'ed or not

 

CPU was about 33-38% throughout, sleek system, all apps closed AV stopped etc..

[BitMaster]

I think, an AMD is a valid option for DCS.

 

Put the money in GPU and RAM and go with a Ryzen ( towards 4GHz )

[*Rage*]

 

Mission start: 85 FPS

Over nellis: 71 FPS

Las vegas downtown: 46 FPS

Las vegas airport: 53 FPS

 

Benchmarked again at stock settings for CPU & GPU ( XMP enabled, all cores 4.5GHz )

 

Start 92/96

Nellis 76

Vegas Strip 48/50

Vegas Airport 54/56

 

Im a little dissapointed to see not much gain from a 7700K at almost the same GHz 4 generations later. Interested to see your overclocked 980 results with the cpu at 4.5 (or 4.6).

Edited May 9, 2017 by ///Rage

[BitMaster]

Well,

as stated, my gpu was 99% , there aint much the cpu can do alone.

I am sure a 1080Ti would score much better with both settings.

Cant blame Intel for the weakness of my Nvidia

 

 

Gesendet von iPhone mit Tapatalk

[Pilotasso]

EDGE has offloaded the geometry onto the GPU and since then the CPU has been under utilized except if you are getting a very high end GFX like a 1080Ti, where Sandy bridge might not have the grunt necessary to feed it anymore.

[BuzzU]

You only need so much. If you can run DCS on high settings and it's smooth. Do you really need more?

[BitMaster]

yes, the more fps the more precise you can steer your craft, more dominant with helis but applies to fixed wing as well.

 

Hovering a heli with 30-60fps and 90-120 fps is 2 different pairs of shoes

[nayu05]

Hello

 

I am quite surprise by your result overall. particularly the 7700k@ one because that is the cpu i owned.

I usually play in VR and even then I feel like my FPS is higher (though, it is harder to check).

 

I know DCS is such a strange beast when it come to performance, so I tried this track in single screen to check.

 

My setup:

7700k@4.9Ghz

DDR4 @4133 CAS17

980TI 1550/7800

 

Result:

Start 140

Nellis 120

Vegas Strip 90

Vegas Airport 105

 

(Do you track those numbers with some kind of tools? I just check the fps with the in game counter so it is very inaccurate. I think only the low point matter tho)

 

I ran the benchmark track as you said, 2560x1440p no AA no AF (I think we are very very far from GPU bottleneck anyway) and high preset on DCS 2.0 and my fps is never going under 90fps (between 95-90fps for a few seconds when the track get close to the center building of vegas)

It is over 100fps all the time otherwise (up to 140fps at the begining)

 

The difference might be that I have very fast optimized ram (4133 CAS 17) and the game scale very well with it (like most single core/cpu limited game) but still, the difference seem like too much.

 

Am I missing something obvious? Are you testing in 16/9?

Wide screen and multi-screen are very taxing on CPU so it could be that.

 

note: I "clean installed" DCS recently so there is no tweaking whatsoever.

 

Thanks for reading.

Edited May 11, 2017 by nayu05

[nayu05]

I played around a bit and realised i was playing with the mirror off (it is a no go in VR) I got around 5fps less on low point (85fps at the lowest point)

 

The gap is closer but still feel a bit too much. Maybe i still underestimate the value of the ram still (like most people i realize)

[BitMaster]

Hello

 

I am quite surprise by your result overall. particularly the 7700k@ one because that is the cpu i owned.

I usually play in VR and even then I feel like my FPS is higher (though, it is harder to check).

 

I know DCS is such a strange beast when it come to performance, so I tried this track in single screen to check.

 

My setup:

7700k@4.9Ghz

DDR4 @4133 CAS17

980TI 1550/7800

 

Result:

Start 140

Nellis 120

Vegas Strip 90

Vegas Airport 105

 

(Do you track those numbers with some kind of tools? I just check the fps with the in game counter so it is very inaccurate. I think only the low point matter tho)

 

I ran the benchmark track as you said, 2560x1440p no AA no AF (I think we are very very far from GPU bottleneck anyway) and high preset on DCS 2.0 and my fps is never going under 90fps (between 95-90fps for a few seconds when the track get close to the center building of vegas)

It is over 100fps all the time otherwise (up to 140fps at the begining)

 

The difference might be that I have very fast optimized ram (4133 CAS 17) and the game scale very well with it (like most single core/cpu limited game) but still, the difference seem like too much.

 

Am I missing something obvious? Are you testing in 16/9?

Wide screen and multi-screen are very taxing on CPU so it could be that.

 

note: I "clean installed" DCS recently so there is no tweaking whatsoever.

 

Thanks for reading.

 

 

In comparison to my 7700k setup your 980"Ti" is like 20% faster...where the other fps come from I dunno tbh. Feel lucky !

 

and yes, the fps are measured by Mark-I Eyeball system with MSI Afterburner. It would maybe suite better to graph it with MSI and send in the graph so we can see the whole run and not some more or less random picked fps from "somewhere" above Nellis, and what is start, 1 sec after begin...3 sec after begin...my fps went up/down +/-10 when the mission starts.

Having the whole curve from A-Z for this benchmark track would allow to layer them on top of each other and compare them completly. STill, your fps seem very very good, so good I almost cant believe it, not saying you dont say the truth ! ItS just very good !

 

The RAM at 4133 do not really gain you any fps imho but make it smoother, snappier and helps TrackIR or VR from what is written here in forum.

 

My CPU was even 100MHz faster and still yours is miles ahead...must be the Ti version of the 980.

 

Time to hunt for a NEW GPU on my end WHEN new maps reach stable state, before that I wont spend any more money on DCS and on this rig.

[nayu05]

Feel lucky !

 

I do! But it is so strange.

 

I really dont know about the GPU, i am so certain we are not GPU limited that it might be an oversight.

 

I will dig it: i will do some test this weekend, stock VS OC (RAM and CPU) and 980ti vS 780ti. I will keep you update.

[BitMaster]

I do! But it is so strange.

 

I really dont know about the GPU, i am so certain we are not GPU limited that it might be an oversight.

 

I will dig it: i will do some test this weekend, stock VS OC (RAM and CPU) and 980ti vS 780ti. I will keep you update.

 

My GPU was at a CONSTANT 99% LOAD throughout the benchmark course. That is a definitiv indicator that I am GPU bottlenecked with the 980GTX @ 4GB and my clocks.

 

Install MSI Afterburner and watch your GPU load, with the 7700k at 4.9 I am pretty sure you can drive your Ti to 99% as well.

[*Rage*]

Its very strange. Almost double FPS. I have a 980ti and an i7@4.6 but not like that!

[Pilotasso]

Maybe the settings on NVIDIA panel could explain that, I got mine on maximum quality with filters set to applications

[nayu05]

Like some of you might have suspected, there is no miracle and it was an oversight on my part.

 

The good news are:

- Nothing wrong with your rig and,

- I will not have to do hours of testing to see what is going on!

 

So I tried this morning to redo the test as I did yesterday, and I got significantly different result, a lot more in line with all the score in the thread.

 

Why though? Well I can’t be absolutely sure but I remember seeing the view distance set on low after selecting the high preset. Witch don’t really make sense I know, and it is specifically why I do remember it (I am almost certain I even switch back in forth several times).

 

I went with it, thinking that is not the first weird thing I saw in DCS (I find the VR preset strange as well).

 

Now, I don’t understand what happened because it is just not the case anymore, with high preset come the high view distance, as anybody would have suspected (and me too I insist!).

 

I did those test again and even check if the number I claimed yesterday were in sync with a test on low view distance, and it is.

 

I join you 3 charts showing my results again, a lot more in line with everyone.

 

High preset + Mirror

High preset

High preset with view distance on low

 

I will methodically test my GPU now and see what is up on this end, I never really bother doing it convince there was no way to bottleneck a GPU on this game on a single screen.

[BitMaster]

Its easy to see if the GPU is at the very end of its performance.

 

Install MSI Afterburner ( actually almost mandatory to troubloeshoot performance issues )

and configure the OSD in the included Riva Statistics Server. Make sure you enable GPU LOAD in the list and whatever else you like to have on the OSD.

[nayu05]

Is it really that simple?

I agree it seem straight forward but how do you explain the actual power consumption of the card?

 

I check the GPU usage and power usage with EVGA precision X (similar to afterburner I suppose) and effectively, my GPU usage was at 99% pretty much all the time.

 

Wanting to actually find the CPU bottleneck point I went on and reduce the resolution, I went down to 900x510 (keeping the ratio similar) and effectively, my GPU usage was around 80% over the desert but went up back to 99% close to Vegas.

In the same time my power consumption went down by half though, and the temperature going with it as well.

 

I am so confuse.

 

I don’t have much knowledge in 3D render so my view don’t really matter, but I really find like there is so many contradiction in DCS behaviour.

 

Still going toward the fact that we are "forever stuck" in a GPU bottleneck situation (with I have still a hard time believing, even with all the “proof” just in front of me) the actual FPS went up while keeping the exact same behaviour.

 

See both run at different resolution:

 

1360x768

900x510

 

So you are right, you would gain fps with a new card (and I would too).

 

What is the game feeding the card with though? I don’t get it, at those resolution it should be a walk in the park (and the power consumption kind of show that) and the CPU should be the limit.

 

It is like the game was artificially feeding the card with something to do, I don’t get it.

 

DCS always confused me performance wise and it seem like despite that many of my conviction build on old testing (pre 1.5) are now wrong, there is one constant: it is weird!

 

I was convince that DCS was a CPU limited game like arma3, apparently it is not the case at all, in fact it might be the most gpu limited game I have ever seen!

I will see how other gpu intensive game behave and see what is up, I generally don’t bother (because it is straight forward, at least I think so) and I might have a lot off preconceived idea that are totally wrong.

For now I will stick to "DCS doesn’t make sense"

 

Anyway, sorry for hijacking the thread, I am sure there is millions of hardware related stuff on the forum and I will definitely look into it.

Edited May 12, 2017 by nayu05

[nayu05]

Just a quick feedback in my quest to find the CPU bottleneck point.

 

It seem to be reachable with everything on low with 900x510 resolution (comparable result at 1400x900)

 

Note that I had to artificially stress my CPU with something else to reach those number, for some reason the fps was capping at 64fps if I lowered the setting too much (while the GPU staying at max clock so I don’t think it come from the GPU power management).

I use a park control software and every power management options are unlock so I don’t think it is from that either.

Maybe something to do with the last windows update and there gaming mode.

I suspect DCS itself, let’s be honest. If I turn the shadow back up the cap go away, the game seem to track if we starve the GPU somehow, and I don’t understand why it would do that.

 

Result All low (shadow OFF) at 900x510 AA0 AF0

 

I think it is the CPU bottleneck point, but i am not even certain, look how the fps is higher in the first part of the 1440p run with everything high but shadow (+200fps) when it never does in lower resolution with everything low.

 

Edit: I am surprise that the FPS are so stable too, it is the limit, but there seem to be no stress at all (maybe the track is too simple).

 

Result All high except shadow (OFF) 2560x1440 AA0 AF0

 

This game is so strange, maybe all of this is for the best and it wouldn't change anything with a more "normal" behaviour but i would like to understand how and why it work like that.

 

Anyway, to go back on track, if you want to actually test CPU on DCS, you have to go much lower than the high preset (and i am more than surprise).

Edited May 13, 2017 by nayu05

[nayu05]

Still digging and it turn out I was not at the CPU limit yet, actually far from it (I was suspicious because of the stability of the fps)

 

In the graphic.lua there is a maxfps value that do exactly what it is supposed to.

The value was 180 and I changed it to 1000 to be sure.

 

Now the true CPU limited chart:

 

And me that was certain it was a CPU limited game :x

 

I know that higher setting might affect CPU too (well I guess there is room for it!) but the GPU bottleneck so quickly… even at 800x450, that i can't check.

 

I will try to dig into the GPU limitation thing, I still have a hard time believing there is that much too do at that low of a resolution.

 

note: the maxfps value doesn't change the low point of the reference run (high preset 1440p no AA no AF) witch make sens.

Edited May 14, 2017 by nayu05

[lazduc]

I just down loaded 1.5.6.5199.291 onto my older system which is a

I7 2600k @ 3.45ghz, 8gb, 2 AMD 6850's(2010 circa) in Crossfire. Only 1gb of video ram.

Crossfire will only use one card's memory. And DCS will only use one card.

 

That old machine runs 1.56.5199.291 perfectly at 40fps and no studders. Its at 1920x1080 @ 60hz. This is the free download version. I cannot tell the difference between it and Lock-on as graphics are concerned running at same settings. Both look good to me.

Laz

[Pilotasso]

Guys, after reading the news about VEGA, the sad fact is that its 2 months away and in very limited numbers. It will be another mess of a launch judging by how countained and refrained enthusiasm AMD has shown at computex. Meaning by the time they get mass shipping I have to add another month to get them here.

This means waiting for the better part of the year. Soooooooooooo...

 

So tomorrow Ill get a 1080Ti. Need to start using this beast as it's meant to be used ASAP.

I'm a tad sad I'm not getting a full red system and it'll have to wait for an upgrade in a couple of years.