980ti SC+ to 1080 FTW Performance changes or lack there of?

[theskyisthelimit 48]

I decided to throw this in its own thread, since its pretty specific..

 

Basically I used fsmark07 and other means to gauge a baseline before switching to the evga 1080 FTW to see if i would get a jump in performance.

 

Well, there was NO CHANGE, maybe a slight loss..  (3dmark firestrike ultra went from say 4100 to 5500, so i do see the expected normal 3D increase, but nothing for p3d)

 

My fsmark07 120sec run went from 24.8 fps to around 23.5 as well.. the initial paused screen went from 33 fps to 31ish.  I also tested oculus rift / flyinside performance, there was hardly any perceivable change.. panning may have been slightly smoother, but still stuttered (no matter the settings), see my other thread for those type of VR details.

 

I saw others online who have made the swap and have had similar experiences.. seems either p3d just doesnt utilize the full benefit of the 1080.. or the nvidia driver has a bug or both.  (or its the usual cpu locked application problem thats plagued the sim for years, despite recent updates)

 

When i ran my testing, i saw the GPU hit a max of 80C and 60% utilization, while the prepar3d.exe averages 35% during the test give or take.

 

I tried setting the NI profile to maximum performance per someone elses suggestion I found.. i also cleared the shader cache and i renamed the prepar3d.cfg file to have it rebuild it, with no change once I put all my sliders back to their settings (which are already pretty high to begin with, including reflections, cloud shadows etc).

 

Just curious if anyone else has found a solution on this.

 

I'm debating if this doesnt have a fix, to return the card and get another 980ti for SLI, mainly for VR.. not sure if that would even make a difference in the VR world, but i have no comparibles to check.

 

Thanks in advance

[Biggles2010 342]

You've probably considered this already, but most of the new cards come with manufacturer specific software which allows overclocking etc.

 

They usually have a facility to control the max power the card is allowed to use, which you can adjust. Your situation sounds like throttling of some kind.

 

With all the pressure from the eco warriors these days, is it possible that EVGA has set a default power setting at less than what is required for full performance.

[ppo 67]

You've probably considered this already, but most of the new cards come with manufacturer specific software which allows overclocking etc.

 

They usually have a facility to control the max power the card is allowed to use, which you can adjust. Your situation sounds like throttling of some kind.

 

With all the pressure from the eco warriors these days, is it possible that EVGA has set a default power setting at less than what is required for full performance.

 

Absolutely. It is best if you use the manufacturer utility. In some of my tests with my new GTX1080 strix I overclocked with the Nvidia Inspector utility. The base clock overclock ok, but not the Power and Temperature; as a matter of fact it reduced the GPU clock to levels that I have never seen before: like 200mhz for current speed. Had to reboot and everything went back to normal.

I now use exclusively Asus gpu tweak and reach speeds upwards of 2020 for the clock. And temperature is in the sixties with the fan at 60 percent.

The fan control utility is also something that you want to tweak.

What speed did the 1080 reach in your tests?

[vgbaron 3,486]

Perfectly normal. Furmark, 3d mark,etc all take advantage of the current state of the art methods to test hardware. Modern game titles have been designed with these methods in mind. 

 

P3D is based on the ESP engine, which is antique by current standards - big kudos to LM for getting as much out of it as they have but until that core code is completely rewritten you will never see that kind of performance jump. Small improvements, yes but even though there is still room for improvement by upgrading hardware, that window is small. 

 

IMHO, in the OP's case, it's six of one and half dozen of the other as to whether or not two 980's will outperform a single 1080 in P3D.

 

Just MHO,

 

Vic

[willy647 193]

P3D is mainly cpu driven.

You will see improvement with a faster cpu, not by swapping a 980ti with a 1080.

 

There is a very simple test : run P3D in a window, and resize that window (from full screen to smaller)

If there is no significant fps change, then a faster gpu will get you no benefit at all.

[downscc 1,327]

I wouldn't trade my 980Ti for a 1080; however, the 1080Ti might be a different case but not expecting a quantum leap.

[Pugilist2 878]

I have a 980ti and will only upgrade if Nividia releases a 1080ti.  I agree with Willy, as P3D is more cpu driven so you would see more improvement by upgrading your cpu.  I upgrade my cpu, memory, and video card all at the same time every two years, but depends on your budget.

[SimonC 116]

Did you try turning everything to max in P3D and testing that way? Usually higher resolution, AA/AF (+lots of clouds), shadows, water detail, reflections, HDR use more GPU than say bunch of autogen. That might leverage GPUs differently.

[theskyisthelimit 48]

Did you try turning everything to max in P3D and testing that way? Usually higher resolution, AA/AF (+lots of clouds), shadows, water detail, reflections, HDR use more GPU than say bunch of autogen. That might leverage GPUs differently.

 

Yeah, I tried cranking these up.. it lowered the paused fps from 31 to 20 on the 1080.. I didnt try the same test on the 980ti though.

 

I would have expected at least a 5-10% increase going from 980 to the 1080, but then again, i'm comparing the jump I had from 780 to 980.. i was hoping for 5 fps difference, doesnt sound like much but would have been.

 

I"m doing similar max out tests in other VR programs to see at what point things become stuttery, then try the same test on the 980ti.. if they still equal out then i guess ill get rid of the 1080, if the 1080 wins on all other arenas, then maybe hold on to it.

 

BIggest hope was that it would help with VR and flyinside, hence i wish i had another 980 to test the sli and compare as well.

 

I might consider doing a clean install on different drives just to perfectly rule that out, but i'm sure its the CPU locked P3D at fault here (unless my p3d install is just mangled because i have so many things installed and maybe one is killing frames).

[Slayer 190]

Low GPU usage is sometimes an indicator of an overburdened CPU ( an overloaded CPU can't feed the GPU so it has a lot of wasted time waiting). Maybe take a look at Robs P3D guide and see which settings affect CPU and which the GPU and adjust from there.

[rcladin 4]

Yeah, I tried cranking these up.. it lowered the paused fps from 31 to 20 on the 1080.. I didnt try the same test on the 980ti though.

 

I would have expected at least a 5-10% increase going from 980 to the 1080, but then again, i'm comparing the jump I had from 780 to 980.. i was hoping for 5 fps difference, doesnt sound like much but would have been.

 

 

Adding a better GPU won't give you more frames unless you're previous GPU was saturated with load.  

 

Typically you'll just be able to run more eye candy, like better AA or turning water or shadows up some more, you should be able to get a decent boost on GPU specific tasks that P3D does, but you're overall fps will be the same.

 

if you want fps boost, then improve the cpu where possible is the most direct approach, overclock it etc if you haven't already.

[TheDanger 182]

Use GPU-Z and monitor the graphs after flying around somewhere intensive for a bit.

 

If the Load and Memory graphs are close to maximum then you will benefit from an upgrade.

 

If they are halfway or less then its a waste of money.

[SimonC 116]

Well, this is also what I was getting at. Be CPU limited (CPU maxed, GPU not), getting faster CPU will give you more FPS. Other way applies of course too.

 

Best thing is always to balance the two. So you get best of both worlds. In my experience, doing somewhere about 70% on the GPU in most of situations is the best option. But I don't use that many shadows. That way, if clouds get really dense, your FPS won't drop too much. Of course the more GPU stuff you use, the more this overhead will be needed. Als using stuff like Sparse Grid Supersampling in the Nvidia Inspector will push the GPU usage a LOT.

 

Setting this up is only a matter of trial and error.

[z06z33 763]

The problem is CPU's arent really getting faster in per core performance which is what p3d and FSX need. In fact my 4790k is just as good as a 6700k in the reguard according to benchmarks. I'm running a 2 year old cpu and upgrading would be pointless as there is no benefit to performance. The X99 ram is faster (DDR4) but not enough to justify a new build.

Intel or AMD need to start getting the clockspeeds higher maybe 6-7ghz instead of sticking around at 3-4ghz that we'be been at for the past 10+ years. Yes I know new cpu's are more efficient but I had I pentium 4 that was 3.2ghz out of the box 13 years ago.

[AviatorMoser 353]

The problem is CPU's arent really getting faster in per core performance which is what p3d and FSX need. In fact my 4790k is just as good as a 6700k in the reguard according to benchmarks. I'm running a 2 year old cpu and upgrading would be pointless as there is no benefit to performance. The X99 ram is faster (DDR4) but not enough to justify a new build.

Intel or AMD need to start getting the clockspeeds higher maybe 6-7ghz instead of sticking around at 3-4ghz that we'be been at for the past 10+ years. Yes I know new cpu's are more efficient but I had I pentium 4 that was 3.2ghz out of the box 13 years ago.

 

A Pentium 4 at 4.0 GHz is still much slower than a 3.0 GHz 6700K. Clock speeds aren't really an indicator of processor power anymore. Rather the the instructions per cycle is more important today. The new Intel chips today are able to complete many more instructions each cycle, which makes them quicker today although the number of cycles per second hasn't really moved beyond 4 GHz for a while.