Harware influence on FSP - test results

[bdlong]

I decided to find out what hardware differences can make an improvement in Frame Rates while playing Flight simulator 2004. I had several things
happen that made this possible.

My power supply .went bad and fried my system board and processor. I decided to spend a little more money and buy an new 939 socket AMD board and a
3200+ processor. I'm hoping that this one will, technology wise, last me for the next few years, before it become obsolete.

A friend asked me to wipe his harddrive and reinstall Windows. While I had his computer, he also asked me to upgrade his memory.

Another friend needed a new system, best price/deal that I could find for him: New case with power supply., Socket A 3300+ processor and motherboard,
($100 after rebate).

Even though it isn't as scientific as I would like to to be, it does show some trends.

Systems:

MachSpeed motherboard, 3200+ AMD 64bit processor

Soyo motherboard, 2400+ socket A AMD processor

Biostar motherboard, 3300+ socket A AMD processor

Components used on all three boards:
XFX GX6600 video card – overclocked at the factor for 600Mhz processor and 1000Mhz memory.

The used the same sticks of memory on each board. It is a mixed assortment of Kingston, Ultra, Spectra, and PNY.

Test setup:

7 minute video of a flight, at night, at JFK. Flight started on downwind and finishes with rolling to a stop at the tiedown.

Software display settings:
Scenery – set to High
Aircraft – set to High
Weather – set to Custom – (sight distance = 80 miles, cloud = 40 miles, 3-d = 100%
Hardware – Target frame rate – unlimited, display 1024*768*16, Render to texture, transform and lighting, anti-aliasing all checked. Filtering is
Bilinear, MIP set to 4, Hardware rendered lights set to 8, global max testure size set to Massive.

The goal was to burden the system enough that I could see what really provides improvement.

I used a program called FRAPS to record the frame rates.

I first wanted to see how much the amount of memory affected things:
The various memory sizes were done on the Socket 939 board.
This memory was all on a 266 speed.

256mb – 34.585 average fps., 50 maximum fps

512mb – 42.409 average fps, 57 maximum fps

1GB – 60.212 average fps, 60 maximum fps

1.25GB – 58.505 average fps, 73 maximum fps

1.5 GB 59.134 average fps, 73 maximum fps

1.75GB 59.559 average fps, 73 maximum fps

2GB 59.838 average fps, 73 maximum fps

The difference from the 1GB in average fps might be related to when I started the Fraps program and stopped it. (I told you this wasn't as scientific
as I would have liked).

I had one 1GB 333 memory stick, (purchased for my friend).

1GB - 66.784 average fps, 82 maximum fps

I decided to try the all three processors with 1GB of 266 memory. (my friend came to get his 333 memory before I could test all three systems with it.)

2400+ socket A 28.070 average fps, 33 maximum fps
3300+ socket A 41.735 average fps, 56 maximum fps.
3200+ socket 939 64 bit 60.212 average fps, 82 maximum fps
You can draw your own conclusions, but here are mine:

1GB of memory is good, 1.25 hits the sweet spot. Anything more than that may not improve your system much.
If funds are limited, instead of buying 2GB of slower memory, buy 1GB of faster memory.

Buy the newer 64bit processors. I was expecting the 3300+ socket A to be a little closer to the 3200+64 bit. After all, I'm running on a 32 bit
Windows with a 32 bit application. From what I have read, the only reason to buy the 64 bit processors at this time is to be ready for future
software - WRONG.
I did note one thing during the testing - with all three system boards and processors, the CPU usage was at 100% during the entire time FS9 was
running with FS9 using anywhere from 94 - 98%. It would probably use 100% if it could. This supports the idea that the processor plays a big role in
the game running smoothly.

Although I didn't test Video cards, from past experience, they do make a big difference. There are better cards available, but I'm happy with my XFX
6600. Price has dropped from the $190 I paid to less than $130. Besides the speed/graphics that it offers, it is one of the few cards that has dual
monitor outputs. I usually run the main view on a 19" monitor and put the GPS and a few other instruments on the 17". This feature only is almost
worth the $130.

Hope this helps someone who wants to spend a little money to improve their system.

Dean

[DBE]

Really great post!

I was wondering if a couple of more Megs of RAM would have any effect on my setup...

[olseric]

Dean-

I expect that the jump you are seeing from the 64bit is not due to the architecture, as you pointed out it's a 32bit program on a 32bit OS,
however, the front side bus.

The problem with the (and I assume you meant Socket A 3200+ since they didn't make an Athlon past that) Socket A is that the FSB is
limited to 200 MHz. Effectively, this creates a HUGE bottleneck in the system. With your socket 939 processor, your FSB is probably,
what, 1600MHz? 2000MHz? Something along those lines.

Nice results though...glad to see someone took the initative to check into all of this.

[bdlong]

Eric,

Thanks for the info, I should have remembered that. I looked at the Front Side Bus among other things when deciding to go with the socket 939. I
just completely forgot about it while running the test.

Dean