Can someone recommend me a AF setting. I have a 6700K and HT is off, I use a lot of Autogen. What should my AF equal? Also, what should texture multi be? I fly mainly NGX and into major USA airports.

Can someone recommend me a AF setting. I have a 6700K and HT is off, I use a lot of Autogen. What should my AF equal? Also, what should texture multi be? I fly mainly NGX and into major USA airports.

HT Off try 0, or 14, or with HT On 85 or 116

 

Recommended HT On and 85.

HT Off try 0, or 14, or with HT On 85 or 116

 

Recommended HT On and 85.

 

 

OK thanks, I'm testing now. 

Another number to try,,,was running AM 1365 per your advice but now you recommend 85 so trying that. Six core 5930k with HT on. Not sure which number is best.

Steve tested many combinations.

The best for 6 Core :

- HT on : 340

- HT off : 30

Thanks but not according to this post;

 

"P3D v3.1, stock scenery and B58, heavy settings, no AA, no vsync, unlimited. Using a 6 core 4.1GHz HT enabled, keeping all processes off the first 5 cores and testing various AMs, putting the top scorer at the top with 100% and marking the rest against that. Making the tests carefully so they are repeatable. Several runs are needed before the PC settles into repeatable results. Tests compared framerate reduced by rate of change in framerate, fps - delta:

100% - 4 LPs on 4 cores AM=85=(01,01,01,01), one per core
99% - 4 LPs on 3 cores AM=116=(01,11,01,00), bunch jobs 2 and 3 on core 2 (LP4 and LP5)
97% - 4 LPs on 3 cores AM=212=(11,01,01,00), bunch jobs 3 and 4 on core 3 (LP6 and LP7)
96% - 3 LPs on 3 cores AM=84=(01,01,01,00), one per core
96% - 5 LPs on 5 cores AM=341=(01, 01,01,01,01), one per core"

Paul, that post is attempting to show how a 4 core works best. You have a 6 core CPU, and Paul, I'm pretty sure I told you to try 340 and you posted shots of the performance graphs and I said they looked good. A 4 core is not a 6 core and demands different numbers.

If you look at the bit pattern of 85, it's the same bit pattern as 340, but on the 6 core we can move the pattern up by one core to avoid core zero. It's the bit pattern, not the number that's interesting.

 

I want back to check and found you were trying 1364 which is 5 cores and should produce a very similar result to 341 in my post you copied above. So instead use 340 to use 4, since a lot of testing has shown more than 4 to be less efficient. You worried about some processors looks dormant, they were simply the Logical partner LPs of allocated cores, and should look dormant, else they are sharing core throughput. I said you could try that as opposed to 4 processors so you could see if there was a difference.

"A Frame-Time Analysis of P3D v3 -- Effects of CPU Affinity, Frame Lock, and HT"

Effects of CPU Affinity: Affinity masks are a binary pattern of ones and zeros. Ones represent ON processors. Processors can be cores with HT Off, or with HT On present two logical processors (LP) per core for faster task switching. Rightmost bit is "LP" or "core" zero.

Frame Lock: Frame lock makes each next frame in an accurately set out time period. 30fps would be one frame every 1/30s or every 0.0333s. Unlimited setting produces each next frame according to the time it took to draw the last frame. So the frame rate lock is there to help improve the accuracy of the "simulation". In order to do that it requires at least one frame ready to display at the correct time. However, almost every time will not be correct and the stored frame is used up. This means to use the frame lock successfully, with Unlimited set we need to be seeing double the frame rate we want to lock at. Unlimited frame rates showing only just above the intended locking rate will lose the buffered frames and take ages to replenish those frames. A powerful machine is required for this mode of operation. Even so, in unlimited mode with all the different frame times, the calculations within the physics of the sim average out to fairly close and are not an issue for generally flying in the sim.

and HT: HT on halves process switching time across the CPU, but apps like P3D need managing how they split themselves over many cores, P3D and FSX will keep splitting even though it is inefficient, and they do not understand the difference between an HT core and a non-HT core. HT off simply dumbs down the whole thinking process in order to avoid too many jobs. HT off loses out in terms of process switching, and HT off presents less processors to the app so the app splits into less jobs. Too often I see "HT off and no AM", this may get you out of trouble without understanding binary numbers, but it's certainly not a tweak.

Paul, that post is attempting to show how a 4 core works best. You have a 6 core CPU, and Paul, I'm pretty sure I told you to try 340 and you posted shots of the performance graphs and I said they looked good. A 4 core is not a 6 core and demands different numbers.

If you look at the bit pattern of 85, it's the same bit pattern as 340, but on the 6 core we can move the pattern up by one core to avoid core zero. It's the bit pattern, not the number that's interesting.

 

I want back to check and found you were trying 1364 which is 5 cores and should produce a very similar result to 341 in my post you copied above. So instead use 340 to use 4, since a lot of testing has shown more than 4 to be less efficient. You worried about some processors looks dormant, they were simply the Logical partner LPs of allocated cores, and should look dormant, else they are sharing core throughput. I said you could try that as opposed to 4 processors so you could see if there was a difference.

 

 

Thanks but when then does your text say this "Using a 6 core 4.1GHz HT enabled, keeping all processes off the first 5 cores and testing various AMs, putting the top scorer at the top with 100% and marking the rest against that."

Paul, I'm using a 6 core for demonstration purposes so that I could add a fifth.

 

 

If you look at the bit pattern of 85, it's the same bit pattern as 340, but on the 6 core we can move the pattern up by one core to avoid core zero.

 

Just to be clear, 85 does NOT free up Core 0, 340 will free up Core 0 ... on a 5 or more core processor.

 

Cheers, Rob.

exactly. so if we take 85=01,01,01,01 and put two zeros at the front (on the right) to represent one more HT core, we get 01,01,01,01,00 which in decimal is 340.

...In the case of 680=(00,10,10,10,10,00) which is the bit reversed 340, can make a subtle difference to the way the jobsheduler allocates processes. The jobsheduler looks at the LP and moves on by one to find the next least used LP, that could be on the next core, or simply the neighbour LP on the core. That's why I'm quoting all sorts of numbers. Please no-one mail me to say have tried 680 on a 4 core HT disabled, since will allocate only one core. lol.

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.

Right I have done my best to try and understand all this. But it's 15 pages of I do this I do that. 

 

I am running 4790k (& GTX970) and I think my AM (1 and 4 maybe?) has been set historically from one of Rob A's optimisation guides. Does anyone here have a better setting I could use please?

 

Many Thanks,

Dom