Hi Paul,

Is there a program that can measure the average fps of FSX/P3D over a set period or something like that?

regards

steve

Fraps can make a benchmark.

Thanks PR. Of course, Fraps is a DX screen grabber so I bet it accesses a frame counter or something without slowing the sim down, that would be handy. What about FS recorder, couldn't that be used to keep repeating an exact flight after changing settings?

Paul,

 

I use TBM=400 with a FFTF=0.1 ans load radius=8.5!

Why TBM 400 will be too big, i notice it was much better with photoreal scenery than if i use 80 or 120?

@Jedi: From the odd snippet I've read here and there,  I think 400 does work with 400, and doing a bit of research - Nick has this to say about it:-

 

"With PURE photo scenery where there is NO textures and landclass anywhere in the scene users may see better results by running a 300-400 Texture Bandwidth Multiplier, UNDERSTANDING THAT FSX handles photo scenery different that textures and landclass, therefore the higher TBM is possible but at the same time the setting Fiber Frame Time Fraction can be increased to place more render time into the larger photo scenery tiles. Values of .50-.75 can be tested for FFTF "

 

I can't tell you about LOD Radius. If textures were involved it may be too high (I limit at the "4.5"), and with DX9 - it is a contributor to OOM's!

 

All the Best, and keep us posted on your results!

 

pj

PJ,

 

Thanks very much for the DX10 How To Guide and your explanatory writeup on bufferpools --very helpful!

 

I am running an 8 MB system with an i5 at 4.2GHz and a GTX 650 with 2GB (which I realize is not a particularly powerful video card ) running at it's native frequency of about 1GHz. So after reading your bufferpools info I switched from usepools=0 to Poolsize=1 and RejectThreshold= 262144. I followed your How To Guide on setting up the GTX 650 with nVidia Inspector. Since I don't have the commercial DX10 fixer, where you gave examples of a number of possible settings (such as for AA) I picked the "middle" value and then manually made the assocaited fsx.cfg entries that you listed. 

 

I don't currently use any scenry add-ons.

 

Anything else you think I should do DX10 wise at this point?

 

Thx,

Al
 

Hi Al;

 

Yup! You could publish your pc hardware in your User Profile, so that it appears under your Avatar (at left), then copy/rename/and attach your fsx.cfg (as Al_fsx_cfg.txt)to a reply, using the "More Reply Options" (bottom right corner), then we can give you any further pointers perhaps... We try not to make guesses on this forum, and your input is needed to do this - ok!

 

All the Best,

pj

Hi Al;

 

Yup! You could publish your pc hardware in your User Profile, so that it appears under your Avatar (at left), then copy/rename/and attach your fsx.cfg (as Al_fsx_cfg.txt)to a reply...

Paul,

Done as requested.

BTW, copying the fsx.cfg file reminded me that it had been tweaked by Bojote's automated FSX Tweaking & Tuning program back in Sept when my processor was an i3 and the GPU a Radeon 6570, so some settings may really be less than ideal. I've also made some of my own changes since then as well.

Thx for your help.

Al

 

Al_fsx_cfg.txt

I deleted all bufferpools mentions in my fsx.cfg and it runs smoother than ever!

 

regards Jeroen

I deleted all bufferpools mentions in my fsx.cfg and it runs smoother than ever!

 

regards Jeroen

Good for you, Jeroen! :Party:

 

__________________________________________________________________

 

Here's your fsx.cfg, Al. You have it set up quite properly!

 

Al_fsx_cfg_chgd.txt

 

All the Best.

 

pj

Good for you, Jeroen! :Party:

 

__________________________________________________________________

 

Here's your fsx.cfg, Al. You have it set up quite properly!

 

Al_fsx_cfg_chgd.txt

 

All the Best.

 

pj

PJ,

Thanks for your time and help!  If I understand correctly, if you specify usepools=1 and don't specify a poolsize, the default of 8MB is used by the system.

Best,

Al

 

 

if you specify usepools=1 and don't specify a poolsize, the default of 8MB is used by the system.

 

Yup - exactly!

PJ,

 

With my i5 CPU at 4.2GHz, based on your experience do you see my GTX 650 as a big bottleneck, or would you say there is a reasonable CPU/GPU balance?  My system is currently limited by its 400W power supply. I think I could upgrade the GPU to a GTX 650 Ti, maybe even a GTX 660. Do you think I'd see a clear improvement with either one of these GPUs?

 

Best,

Al

PJ,

 

With my i5 CPU at 4.2GHz, based on your experience do you see my GTX 650 as a big bottleneck, or would you say there is a reasonable CPU/GPU balance?  My system is currently limited by its 400W power supply. I think I could upgrade the GPU to a GTX 650 Ti, maybe even a GTX 660. Do you think I'd see a clear improvement with either one of these GPUs?

 

Best,

Al

I have a GTX 650ti 1 Gb card and (as stated earlier) removing the bufferpool setting out of my fsx.cfg was the best option.

A solid 30 fps on most airports (30% AI)

 

regards

Jeroen

Jeroen -- thanks for the response.

 

It seems to me that given practicalities like overall system cost that we all have to deal with, what we would like to have for FSX is a reasonable "match" between the CPU and the GPU.  Doesn't make sense to me to spend big bucks on a GPU that is constantly waiting for the CPU, or vice versa.  Neither should be a significant bottleneck for the other. While CPU clock speed (and maybe L3 cache size) seems to be the common measure we use for the CPU, I don't have a good handle on what subset of parameters ( # CUDA cores, base clock, texture fill rate, memory size, memory bandwidth, etc) to look at for the GPU in order to estimate what might be a reasonable match between the two.  I realize other performance factors enter in ( amount of system memory, etc) but a rough "matching" methodology (one that slightly favors GPU performance, perhaps) would be handy when trying to figure out how best to allocate limited funds.

 

One approach might be to use some of the standard GPU benchmark scores that are published for what seems like most of the common GPUs. Then we would "just" need to know how to roughly match the GPU benchmark score to the CPU performance, e.g., a table with entries like: An i7 xxxx at between 4.x and 4.y GHz would be a reasonable match for a GPU with a benchmark in the Range of  Rx to Ry, etc.

 

Al