Tweaking, Needed or not needed in P3D?

[swiesma 625]

2. The clock frequency is only one of the parameters that determines the power of a CPU. Why is a i7-5960X @3.5GHz regarded as more powerful than for example a i7-6700K @4.2GHz or a i7-4790S @4.0GHz?

 

And "more powerful" always depends on the use case / benchmark / software ;-)

[Yogipriester 11]

And "more powerful" always depends on the use case / benchmark / software ;-)

In principal I do agree. That is the reason why I wrote: "The clock frequency is only one of the parameters that determines the power of a CPU."

 

Unfortunately most magazines do not use FSX or in my case P3Dv3.1 to benchmark the various CPUs. They use other programmes or special benchmark programmes to compare. A long time ago CPUs and Graphics Cards were compared at Flightsim (if I remember correctly) using the simulators of that time. The results tallied somehow with other tests that used "modern" test programmes.

 

Anyways, looking at the results of those tests I do not think that the i7-6700K or the i7-4790s will beat the i75960X when using them in FSX or P3D. Do you?

[swiesma 625]

In principal I do agree. That is the reason why I wrote: "The clock frequency is only one of the parameters that determines the power of a CPU."

 

Unfortunately most magazines do not use FSX or in my case P3Dv3.1 to benchmark the various CPUs. They use other programmes or special benchmark programmes to compare. A long time ago CPUs and Graphics Cards were compared at Flightsim (if I remember correctly) using the simulators of that time. The results tallied somehow with other tests that used "modern" test programmes.

 

Anyways, looking at the results of those tests I do not think that the i7-6700K or the i7-4790s will beat the i75960X when using them in FSX or P3D. Do you?

 

I am completely with you.

It always depends on the architecture of the processor, the clock speed, cache size and a million other factors....

 

I personally would alwys go for the newer architecture which would be Skylake in this comparison:

http://cpuboss.com/cpus/Intel-Core-i7-6700K-vs-Intel-Core-i7-5960X

 

I do not think that P3D or FSX are highly threaded and therefore there are not really much performance gains by the additional cores.

And when you look at the price..... Well....

 

Here it is clear: Single core performance: 6700k, Multicore usage: 5960k

http://www.technikaffe.de/cpu_vergleich-intel_core_i7_6700k-518-vs-intel_core_i7_5960x-438

 

For FSX / P3D I would always go for the 6700k.

[SteveW 1,792]

Comparing 5960x in single thread to the 6700K may be a mistake, unless compared at the same heat. The bigger chip with more cores will get hotter, and so sports a slower turbo speed. Consider that at 4GHz the 4790K is 2530, the 6700K is 2325, but the 3960x is 1992 at 3.3 GHz but runs all day at 4GHz with out specialist cooling. So at 4Ghz that puts out 2414. 4960x is 2074 at 3.6GHz, 4GHz is childs play for that and puts out 2304.

 

cpubenchmark singleThread

 

The big improvements in real throughput have been in motherboards, memory, GPU, and hard disk.

[Raven9000 272]

Except most of those benchmarks mean squat in real world performance... and even less when thrown in with other factors such as different motherboards, ram, video cards, etc., etc.,etc..

[SteveW 1,792]

In reality, benchmarks do actually provide a basis for understanding the performance of a chip.

[swiesma 625]

Comparing 5960x in single thread to the 6700K may be a mistake, unless compared at the same heat. The bigger chip with more cores will get hotter, and so sports a slower turbo speed. Consider that at 4GHz the 4790K is 2530, the 6700K is 2325, but the 3960x is 1992 at 3.3 GHz but runs all day at 4GHz with out specialist cooling. So at 4Ghz that puts out 2414. 4960x is 2074 at 3.6GHz, 4GHz is childs play for that and puts out 2304.

 

cpubenchmark singleThread

 

The big improvements in real throughput have been in motherboards, memory, GPU, and hard disk.

 

Still no big gains by more cores in gaming, especially non-multithreaded applications like FSX / P3D.

[SteveW 1,792]

Still no big gains by more cores in gaming, especially non-multithreaded applications like FSX / P3D.

That's why we consider the single threaded throughput so important.

 

 

However, P3D and FSX run best on 4 cores, add in some addons and the next 6 core is looking very tasty.

 

 

 

In terms of tweaks, these simulators have scope for hundreds of settings but are set up fairly well to work out of the box. Even so, there will always be a few things that can affect certain circumstances and require tweaks, or rather could be deemed necessary settings. I've collated a few notes for P3D:

 

Frame Rate and VSync settings with NI:

For regular PC systems, to obtain reasonably high fps combined with smoothness, try this; a standard setup of P3D v3.1 frame control on a typical 59/60Hz dvi-d monitor. Possibly the closest we'll get with P3D to the beloved 1/2 refresh FSX exclusive mode:

 

Using Nvidia Inspector (NI) choose 29 fps or 30 fps for the "Frame Rate Limiter" item. Setting the frame rate limiter with 3.1 in NI is like having a lower refresh monitor mode. However, the limiter settings don't all work on all systems, so check you get fps showing near to the setting you choose in NI.

 

Now run P3D v 3.1 and in Display settings we have two options;

 

1. Set VSync=On and set Triple Buffer=On, set Unlimited frame rate,

or

2. VSync=Off and TP=Off, set Locked Frame Rate to 29/30.

 

Using example 2, the pre-emptive setting. requires unlimited frame rates showing in the region of at least 45-90fps to be reasonably successful. Seems counter-intuitive, but just when look ahead frames are needed, the sim is running too slow to build them up again. Example 1 can be utilised by less powerful systems and still maintain a reasonably smooth 30fps.

 

 

The Prepar3D tuning guide:

 

The Prepar3D tuning guide

 

In the guide the Affinity mask of 14=(1110) is recommended. However this is perfect for Dual Core with HT enabled (HT=On), or Four core systems with Hyperthreading disabled (HT=Off).

 

If we enable Hyperthreading (HT=On) with no AM in use with Prepar3D, the JobSheduler will be assigning another major simulator thread to a logical processor on the same core occupied by the primary simulator thread. That results in the main thread sharing throughput of the core unnecessarily with one of its secondary threads. Note that FSX is the same. The simulator will count the available LPs and split itself over those running threads on every LP. This will use also use up considerable VAS. If more than four are available, only the first four are meaningful. Jobs beyond 4 are discarded unless they finish first, they can only serve benefit if something interrupts the others. Sweet spot is four jobs, on a 4GHz CPU, 3 or 5 jobs perform worse, 4 is best.

 

Systems with 4 core HT=On should use AffinityMask=116=(01,11,01,00) with addons on core 0

Systems with 4 core HT=On should use AffinityMask=85=(01,01,01,01) or 170=(10,10,10,10).

Systems with 6 core HT=On can use AM=340=(00,01,01,01,01,00).

Systems with 8 core HT=On can use AM=5440=(00,01,01,01,01,00,00,00).

 

Dual core CPUs should have HT enabled if they support it, and an AffinityMask of 13, or 14 specified.

 

example added to Prepar3D.cfg:

[JOBSCHEDULER]

AffinityMask=170

 

Those without HyperThreading CPUs (or HT disabled) need not apply an Affinity Mask. Even so, an AM can be applied to partition the chip, perhaps reserving one core free of the simulator and it's multi-threaded resources, so that other processes starting up can have a core free, although Windows JobSheduler will do a good job distributing work away from the main simulator thread once it's up and fully running.

 

Systems with 4 core HT=Off should use AffinityMask=0=(1111) or

Systems with 4 core HT=Off should use AffinityMask=14=(1110) with addons on core 0.

Systems with 6 core HT=Off can use AM=30=(011110).

Systems with 8 core HT=Off can use AM=120=(01111000).

 

AffinityMask=0 is equivalent to no [JOBSCHEDULER] section specified in the cfg.

 

Overclockers: Reducing the amount of ones "1" in the mask decreases heat. 4 core HT enabled - how about AM=52=(00,11,01,00). With the four core, if we want to go cooler than 116 try 52=(00,11,01,00). This only uses two cores, but splits into three jobs, and performs very well, not much less that 116, but better fps at higher GHz at the expense of loading textures. Slightly lower fps, slightly more heat, for increased texture loading AM=60=(00,11,11,00) which splits the sim into four jobs, moving the second job onto the main thread core.

 

 

Prepar3D.cfg:

 

Tessellation set low in the Display Settings is 60, but can be set around 50-20 for better performance, even 20 should be better than no Tessellation at all.

[GRAPHICS]

TESSELLATION=1

[TERRAIN]

TESSELLATION_FACTOR=25 (50% or below for performance, above for quality.)

 

[Display]

TEXTURE_BANDWIDTH_MULT=180 (Smoother texture handling with medium+ autogen.)

TEXTURE_BANDWIDTH_MULT=10 (Smoother frame rate with very low or no autogen.)

UPPER_FRAMERATE_LIMIT=20 (Or use 0, zero=Unlimited fps, when testing frame rates and performance. With Locked frames we need to see unlimited frame rate 2x - 4x the locked rate setting.)

TEXTURE_FILTERING=8

MSAA=4 (set Transparency 4xSGSS AA with NI or...

MSAA=2 ...set Transparency 2xSGSS AA with NI.)

Adding SGSSAA is only required for certain aircraft so always turn it off to check if your aircraft is affected. Otherwise, only use MSAA in P3D Display Settings.

 

Not a recommendation, but for experimentation with shadows, cascade count and texture size:

[GRAPHICS]

SHADOW_NUM_CASCADES_LOW=3

SHADOW_NUM_CASCADES_MID=4

SHADOW_NUM_CASCADES_HIGH=5

SHADOW_NUM_CASCADES_ULTRA=6 (Decrease cascades and...

GROUND_SHADOW_TEXTURE_SIZE=4096 ...increase texture size = fps boost with similar shadow quality.)

 

[Weather]

OPTIMIZE_DENSE_CLOUDS=1 (Discussed on Prepar3D forums. May help with some clouds.)

 

[sIM]

OPTIMIZE_PARTS=1 (Discussed on Prepar3D forums. May help with complex aircraft and other objects.)

 

[uSERINTERFACE]

SITUATION=C:\Users\[you]\AppData\Local\Lockheed Martin\Prepar3D v2\Prepar3D_Default (The default flight file.)

 

 

Prepar3D Learning Center.chm

 

For more tuning tips run the P3D Learning Center:

 

C:\Program Files (x86)\Lockheed Martin\Prepar3D v3\Learning Center.chm

 

Open (double-click) Learning Center.chm, in the left panel go to: Performance, Tuning Guide.

[simmerhead 885]

I only use the Affinitymask tweak to free up core 0 when using TrackIR and several simconnect apps. P3D runs OK with relatively high settings on my i7 2700K & GTX580 combo - as long as I keep shadow settings to a minimum. Compared to FSX it is a dream come true!

[specialist 51]

I did away with AffinityMask and FFTF, and now my P3Dv3 is running much smoother and sharper. Not sure why I thought I needed those tweaks, but I guess after so many years with FSX, all I did back then was tweak, tweak, and more tweak. Less is more when it comes to P3Dv3, at least for me.

[Bobsk8 12,546]

I did away with AffinityMask and FFTF, and now my P3Dv3 is running much smoother and sharper. Not sure why I thought I needed those tweaks, but I guess after so many years with FSX, all I did back then was tweak, tweak, and more tweak. Less is more when it comes to P3Dv3, at least for me.

 

Best thing I like about P3D is it doesn't need tweaks to run great....