Com-Central.net

Testing CFX

With all sliders set to minimal managed to hold a fixed 20fps most of time... CPU usage was at 100 % continuosly also used about 600 meg of page file

Set all sliders to max (includin terrain etc )... 16 fps at ground level and locked at 20 fps around 2000ft Alt... CPU usage still at 100 %

So think hard before you drop serious bucks on a Graphics card its CPU horesepower you need

Test Rig Athlon 64 FX53 clocked @ 2600 + 6800 ultra + 1 gig RAM

This may vary if you are parked up at a very scenery intensive airport when the GPU has lots of rendering to do..more testing req.

Sly

Thanks Sly! Did you try the things suggested in This POST?

Yep Jerry

I can get it looking good and flyin fine, But just wanted to know which side of the equation ( CPU / GPU ) was having a throttling effect.

It was late when finished but should have mentioned tests were with Capt, Sim C130 @ Lyneham Airbase.. gonna rerun at stock Aircraft and extreme types of location eg.. empty air strip, Heathrow etc


post em soon

sly

What's interesting is that although FSX is not a multiithreaded app, you can force it to (at leat appear to) load balance between the two CPU's on a dual-core or HT system. At least on my P4HT which shows CPU0 and CPU1.

With it running, pop up Task Manager, find the FSX process, right click and set the affinity. de-elect CPU0, leaving CPU1 enabled, select OK, then right click and set the affinity again, and re-enable CPU0.

Check your CPU performance graph and you should see a split between both CPU's instead of CPU0 cranking out at max and CPU1 being mostly idle.

However.....is there an actual performance gain? So far it doesn't seem like it. Could be just a placebo effect.

MMM tried same sort of test using stock plane and unlimited frames... pulling 35 frames plus..... so it does seem that AC model complexity has a significant effect.

Some interesting stuff over at AVsim forums topic #364105

Seems like one possability is that FSX uses any spare CPU capacityy to fetch , place and render scenery ahead, therefore will always use 100% of cpu? which leads us to the possabilty of upgrading you Graphics card and getting worse visuals ( no preloaded scenery etc so fuzzy ) but more frames?

Only slider that has any real effect on frames is Autogen, Even running 1600 x 1200 AF and AA seems to have litttle effect..... so the GPU seems to handle anything the CPU can throw at it..

So I stick by the thought that processor power is the key not graphic cards

If I get time gonna swap GPU cards and check differences.. would be interesting to see how 6800 U, 5900, TI4600 and Ti 4400 compare on same set up.


Sly

I have FS X running on my new computer now

AMD Athlon 64 4800 dualcore 939 Socket
1 Gigabyte RAM DDR
1 6600GT Graphics Card
Windows XP 64 Professional
Thermaltake fan for CPU
www.thermaltake.com/pr...114-01.asp

www.thermaltake.com/pr..._index.asp

around 16 FPS but can be as low as 5 FPS at Seattle Airport, I tried overclocking the processor and brought it up to 2800 megahertz from the stock 2400 of the 4800 939 but it did not make much difference to frame rates, though it seemed to load somewhat quicker, so I reset the chip back to the standard non overclocked values. It looks to me as if this programme's performance once you have a fairly decent but far from the top of the range processor for example AMD Athlon 64 3200 754 with 1.5 Gigabyte RAM and 6800GS 512 Megabyte GPU which I have in my older machine and I am also running FS X on, is determined by the GPU ie Graphics card, in that the 4800 is only running at just over fifty percent capacity in even the most difficult visuals. With this in mind, I would go for SLI or Crossfire with best graphics cards you can afford as suits your choice and a base model dual core processor eg lowend of the Conroe and AMD Athlon 64 ranges and if you allready have a machine I would give consideration to upgrading your graphics card[s] on your existing rig rather than purchaseing a new processor / motherboard.

Best and Warm Regards
Adrian Wainer

Well,

Not having a lot of time to play/experiment with FSX I can't comment fully but initial observations suggest SLI or Crossfire may not be the answer.

My rig has the 7950GTX2 cards fitted and in either 'single' card mode (I say single even though it's a dual card) or in SLI mode, I get about the same frame rates. (40-60)

I won't say FSX doesn't like 'four' graphics cards as it runs quite nice but maybe a couple of 7900 in SLI mode may be better 'bang for bucks'

- Kitform

Well,

Not having a lot of time to play/experiment with FSX I can't comment fully but initial observations suggest SLI or Crossfire may not be the answer.

My rig has the 7950GTX2 cards fitted and in either 'single' card mode (I say single even though it's a dual card) or in SLI mode, I get about the same frame rates. (40-60)

I won't say FSX doesn't like 'four' graphics cards as it runs quite nice but maybe a couple of 7900 in SLI mode may be better 'bang for bucks'

Is it possible you are running in to the problem, that games need a specific Quad SLI profile and if they do not have it, they only use one of the four cards?

Best and Warm Regards
Adrian Wainer

My initial tests seem to support this.

Would be nice to test a 7900 single against a 7950GX2 to see if the program takes advantage of the dual cards.

One thing also I would have to say is that in the Athlon 64 4800 I am running it under Windows XP 64 Professional which is a 64 bit operating system, and I am just getting used to this operating system which is not usually used for gaming. I am going to take a look at the NVIDIA site with a view to install specialised drivers. Are you getting forty to sixty frames Kitform, that sounds pretty good for a start as best as I remember hollywood film runs at thirty something frames as once one is the thirty region the human eye does not sense stuttering.

Best and Warm Regards
Adrian Wainer

Does anybody know if FS X is a 32 bit or 64 bit programme, I am not writeing this on my home computer but I think my machine which is runing the Windows XP 64 Professional assigned FS X this header

(x86)

which is reserved for 32 bit applications, which suggests to me that Microsoft is overplaying what Vista will do for FS X and did FS X really need Vista to exploit Dual Cores or did MS just decide that would be a nice way of forcing people to acquire Vista?

Best and Warm Regards
Adrian Wainer

- Kitform

My initial tests seem to support this.

Would be nice to test a 7900 single against a 7950GX2 to see if the program takes advantage of the dual cards.

I am not well up on this stuff but I think what might be getting the good results for you is when you are running one of the single paired cards combi in SLI? that is if FS X does not suppport Quad Sli?

Best and Warm Regards
Adrian Wainer

Hi Adrian

Your results seem exactly the opposite of those I found.. My single core CPU (FX53) is running at 100% regardless of settings. Is it possible that your load of 50% is the result of one core running at 100% and the second core at minimal? I believe that FSX only uses one core.

Easy way to check whether the fps are limited by CPU or GPU is wind up your screen res and see if it affects frames, a better way is perhaps increase graphic settings but certain sliders also have a definate loading increase on the CPU, most notably cloud draw distance and autogen..

Certainly on my rig all other sliders can be maxxed at virtually no cost of frames, loading time another story though

Sly

- 401RCAF_Sly

Hi Adrian

Is it possible that your load of 50% is the result of one core running at 100% and the second core at minimal? I believe that FSX only uses one core.

Yes you're right about FS X using only one core, a Microsoft person told me that FS X will not utilize multicore unless it is run in Vista. No, I do not believe that it is what it was saying i.e. giveing a combined figure for the two cores, since there are individual figures for the two cores.

Your FX53
www.xbitlabs.com/artic...x53_2.html

looks quite similar in a number of respects to my

Athlon 64 4800

www.xbitlabs.com/artic...-x2_4.html

just a guess but I would suggest a possability is that the five to ten percent that the second processor is dealing with might have a disproportionate effect when it is dealt with [ combined ] in a single processor situation.

Best Regards
Adrian Wainer



Best and Warm Regards
Adrian Wainer

I hear what you're saying,

But everywhere I looked, Avsim etc the general consensus is that CPU the limiting factor. As per my initial findings and indeed as all MS flightsim series has been.

Good luck if yours is limited by the graphics card, But I will reiteratte my warning (to others) that if you shell out big bucks on a Vid card set up you probably will not see the returns you are expecting.

Sly

Anyone else's cpu running at less than 100%? (without silly screen resolutions like 2880x1840)