It seems I've read others show 100% CPU use, but when I open Win 7 Task Manager > Performance it is never that high--the combined level is in the 30-60% range combined.  What do others see with this, and with what monitoring app?

Noel, it depends on what AffinityMask you are using and if you have hyperthreading activated. On a non hyperthreaded using default AffinityMask it's fully normal to have combined CPU usage to increase to 100% every 61 seconds. If you will reach the 100% combined load and for how long it will remain at 100% depends on LOD, texture and mesh resolution, what terrain you are flying over, your groundspeed + your CPUs general performance. That's just for basic FSX. Add-on programs like Wx engines will create extra CPU load. Especially at time of injection.

FSX will create a constant 1.5 CPU core load for generating the FPS as long as you're CPU limited. It will also utilise further cores for loading textures, the terrain mesh and the autogen. They are assigned with the Affinitymask. How much load these cores generate depends on what I wrote above. If you are stationary they will generate no extra load. During 'normal' operation they generate up to half a core of load each, and during the refresh of the texture lighting that happens every 61 seconds they generate up to a full core of load each.

So take your 3930K with hyperthreading active and default affinitymask. That gives you 5 terrain loaders. It will always generate 1.5/12=12.5% combined CPU load.

During normal flight it will generate up to (1.5+0.5x5)/12=33% combined CPU load.

And during the light refresh every 61s it will generate up to (1.5+1x5)/12=54% combined CPU load.

So your numbers are normal.

 

I normally assign 9 terrain loaders so my in flight combined CPU load is normally around (1.5+0.5x9)/12=50%.

While it reaches around (1.5+1x9)/12=88% for a while during the light refresh every 61seconds. That makes my CPU nice and toasty whilst flying FSX thou.

 

 

I normally assign 9 terrain loaders so my in flight combined CPU load is normally around (1.5+0.5x9)/12=50%.
While it reaches around (1.5+1x9)/12=88% for a while during the light refresh every 61seconds. That makes my CPU nice and toasty whilst flying FSX thou.

 

OK that sounds nice but what affinity mask value gets you assigning 9 terrain loaders?  Also, when I open Task Manager and view various processes, fsx.exe included, I can select specific cores for each process.  Can you recommend an affinity mask, and how I should assign cores to fsx.exe, or perhaps this should just be set to All and let the affinity mask handle the assigning.  And also, should I assign various non-FSX processes to Core4 or Core5 for example?

It is also worth mentioning that the main FSX Core does not go to 100% if you are using an external framerate

limiter and the environment is not stressing the CPU.

 

Without the external limiter, you should see the first core that is active in the affinity mask, at 100%

all the time.  No affinity mask, that would be Core0.

It is also worth mentioning that the main FSX Core does not go to 100% if you are using an external framerate

limiter and the environment is not stressing the CPU.

 

Without the external limiter, you should see the first core that is active in the affinity mask, at 100%

all the time.  No affinity mask, that would be Core0.

Thank you Bert for that clarification.

It is also worth mentioning that the main FSX Core does not go to 100% if you are using an external framerate

limiter and the environment is not stressing the CPU.

 

When this happens the loading of the ground textures that is conducted on the other assigned cores is hampered. This happens all the time FSX framerate isn't CPU limited or limited by FSXs own internal limiter.

V-sync, PCIe bus, GPU, external limiter, they all slow down ground texture loading when they limit the FPS.

That's why you should use FSXs own internal limiter and not an external one.

 

Just the fact that you set the internal limiter already improves on ground texture loading. But what most people notice the most is that the FPS also drops, that's why many are hesitant to use the internal limiter. These two go hand in hand though and is controlled with the FFTF value. By lowering the FFTF value to around 0.17 you can chose to give up that improvement in ground texture loading to regain that FPS during the time your CPU can't maintain the locked value. Lower FFTF further and you'll actually reach higher FPS than without the internal lock, but at the cost of slower ground texture loading.

When you are able to maintain the internally locked value ground texture loading is improved all the time, regardless of the FFTF value.

OK that sounds nice but what affinity mask value gets you assigning 9 terrain loaders?  Also, when I open Task Manager and view various processes, fsx.exe included, I can select specific cores for each process.  Can you recommend an affinity mask, and how I should assign cores to fsx.exe, or perhaps this should just be set to All and let the affinity mask handle the assigning.  And also, should I assign various non-FSX processes to Core4 or Core5 for example?

 

 

With W7 and FSX fully patched (either SP1+2 or ACCpack) an AffinityMask entry won't help you much. The overall result can be slightly better or slightly worse; the overall gain is very limited. Worst case is a "wrong" entry which would lower performance considerably.

 

What IS recommended, though, is to switch off hyperthreading, as FSX doesn't use it. It would only increase core temperatures.

What IS recommended, though, is to switch off hyperthreading, as FSX doesn't use it. It would only increase core temperatures.

Wrong!

FSX uses HyperThreading for ground texture and terrain loading very well. But not for higher FPS in a complex aircraft, and that's what most people are after. So if you like me use a lot of high res photoscenery it makes a lot of difference.

 

 

Just the fact that you set the internal limiter already improves on ground texture loading. But what most people notice the most is that the FPS also drops, that's why many are hesitant to use the internal limiter. These two go hand in hand though and is controlled with the FFTF value. By lowering the FFTF value to around 0.17 you can chose to give up that improvement in ground texture loading to regain that FPS during the time your CPU can't maintain the locked value. Lower FFTF further and you'll actually reach higher FPS than without the internal lock, but at the cost of slower ground texture loading.
When you are able to maintain the internally locked value ground texture loading is improved all the time, regardless of the FFTF value.

 

It seems like there is an an FFTF value that essentially puts the internal limiter on a par w/ an external limiter.  That is, one can set the FFTF value so low (~.17) that texture loading is impaired just as you are arguing it is when an external limit.  I also notice I can use the FFTF and indeed texture loading is affected even when an external limiter is used, and I always use one because of what you mention, that frame rate performance improves.  The improvement for me is always about 20% in frame rate.  If I set the internal limiter at 30, I get around 25, whereas it sticks on 30 w/ an external limiter.  Actually, I used the old external limiter to get rid of rapid quick jumps in video in my old machine and it worked great for that, and I noticed as I say about a 20% bump in performance.

 

I'll go back and play again w/ internal limiter but I'm not so certain that is all there is going on, but it could be.  Now I set FFTF to .40-.50 and have good texture loading so I'm thinking it will be useful to see if there is perf monitor evidence of higher CPU utilization which should equate w/ better performance at some level so really could be the best evidence internal limiter is best.

 

Thanks for your insights!

 

SABB:  do you use an AffinityMask value?  If so, ??

For what it is worth, I use the internal limiter at 30, FFTF at 0.18.

 

Affinity Mask at 14 for a quad CPU, and NI Vsync at 1/2 of monitor refresh (of 60).

 

This combination gives me fluid flight and no blurries.

It seems like there is an an FFTF value that essentially puts the internal limiter on a par w/ an external limiter.  That is, one can set the FFTF value so low (~.17) that texture loading is impaired just as you are arguing it is when an external limit.  I also notice I can use the FFTF and indeed texture loading is affected even when an external limiter is used, and I always use one because of what you mention, that frame rate performance improves.  The improvement for me is always about 20% in frame rate.  If I set the internal limiter at 30, I get around 25, whereas it sticks on 30 w/ an external limiter.  Actually, I used the old external limiter to get rid of rapid quick jumps in video in my old machine and it worked great for that, and I noticed as I say about a 20% bump in performance.

 

I'll go back and play again w/ internal limiter but I'm not so certain that is all there is going on, but it could be.  Now I set FFTF to .40-.50 and have good texture loading so I'm thinking it will be useful to see if there is perf monitor evidence of higher CPU utilization which should equate w/ better performance at some level so really could be the best evidence internal limiter is best.

 

Thanks for your insights!

 

SABB:  do you use an AffinityMask value?  If so, ??

Do you use the half rate v-sync together with external limiter set at 30 as well?

 

V-sync is a kind of external limiter on its own. Normal v-sync will cap at 60fps while the half rate v-sync vill cap at 30fps. So to have FSXs terrain and texture loading engine to work as intended you have to set the internal lock to 60 or 30 fps as well.

 

I have to say that the tweek that has made the most impact on my flightsimming is the half rate v-sync combined with internally capped to 30fps. The smoothness that comes from the half rate v-sync combined with maintaing and improving the texture loading that comes from the internal lock is great.

 

Try with the internal limiter again and lower the FFTF to regain the FPS if you need it.

 

I personally use Affinitymask 4089 on my Hexacore with Hyperthreading on.

I don't however use any add-on Wx software like active sky. They temporary generate quite a bit of CPU load and the risk is that Windows thinks its a good idea to schedule that to run on the same physical core as FSXs main thread with that affinitymask.

I'd most likely be using either 4081 or 4084 if I had other things running that creates a significant CPU load.

 

Using any of these 3 affinitymasks will as well as giving you faster loading of a flight also give you faster loading of the groundtextures compared to the default 1365. But tied to that improved texture loading is a drop in FPS. FFTF is once again your friend here. You can get your FPS back again by lowering the FFTF value. You get more texture loading at the same FPS by using more cores at a lower FFTF value.

 

Let me show a picture I posted in the photoreal section of the forum earlier. It clearly shows how Hyperthreading is working in FSX. The top part shows how the ground looks when the flight has been paused and all the textures and terrain has been allowed to fully catch up. Below is the effect that a few different affinitymasks and Hyperthreading on/off has on the ground textures during a straigt and level flight at Mach 0.9. You can also see how long time it takes for the ground textures to fully catch up when the flight is paused there.

 

Do you use the half rate v-sync together with external limiter set at 30 as well?

Yes I do.  I have very fluid flight, decent texture loading (comparing only against my Q9650 machine).  So it's hard to say.  I can certainly give another try at using the internal limiter, which is what I first tried after reading various posts on this topic.   

 

 

You can also see how long time it takes for the ground textures to fully catch up when the flight is paused there.

 

Ok, but I don't see the AM value I was advised to use w/ hexacore HT off:  62, not 63.  Try that same test and see what you see.  I'm guessing it will be pretty good, but I hope it's NOT because that means trying one of your AM's should should some serious improvement.  Give that a try and see because as I say, I have pretty darn good total performance in all areas, smoothness, fps, etc.  I think there is some room I hope for texture loading.  That is always a fantastic thing.

AM 63 is the default AM for a hexcore without Hyperthreaading. Its also the affinitymask that gives the maximum texture loading possible. But there is a high risk that you get considerable FPS drops at times. Especially once every 61 seconds during the light refresh when FSX alone can generate 6.5 core load. You only have 6 cores though so there is a potential of loosing 1/3rd of your FPS at times. Wx injection is also a potential for huge FPS losses with stock AM. That's why AM=62 is used to avoid these FPS drops. I't does however also give slower texture loading. Textureloading with AM=62 HToff is ever so slightly slower than AM=249 HTon. With HT on you have a slightly higher risk of FPS drops compared to HToff with one core 'spare' But HTon with the appropriate AM has a lot lower risk for FPS drops than HToff using all available cores with stock AM. Whilst still achieving faster texture loading.

The reason I picked the above 4 examples is that they show the maximum possible texture loading with a quad core HTon/off, and hexa core HTon/off.

Well, the idea of setting setting an internal limiter I think is very misguided, at least on my machine it's just another SB-E chip.  I tried this again with 1/2 refresh vsync and performance was terrible by comparison to what I use, texture loading notwithstanding.  Take a brief recheck on this SABB and whomever else and see for yourself.  Try these settings which are outstanding on my machine--totally fluid flight, IQ crisp  within the confines of the LOD setting-I use 5.5 most of the time:

 

FFTF=40

Unlimited Frame rate internally

1/2 refresh rate vsync

External limiter at 30

SWAP_TIME_OUT=30  (this got rid of spikes when I didn't use external limiter but did use 1/2 refresh)

AffinityMask=62

 

This is apparently not the conventional wisdom here but I defy someone to try these out and then show me settings that come anywhere near this level of total performance.  I'd love to hear it believe me!