Balanced Computer Systems And FSX

[xxmikexx]

Folks,

Any discussion of FSX performance, and of the equipment required to run FSX, is incomplete without a discussion of the concept of a "balanced system". This concept is best understood by examining some extreme cases, beginning with the diskless internet computers of a few years ago. (Remember those?)

For that application the system -- cpu, memory, graphics -- is balanced, and while adding an HDD would not hurt anything but system price, the system would not run any faster as a result of having an HDD added to it. The system was balanced before, but it will be unbalanced after, and money will have been wasted.

Similarly, a mainframe based income tax return processing system is not going to be sped up by the addition of WiftyZ graphics cards to the IRS agents' terminals. The system is disk-bound, so enhancing its graphics will not change its performance.

But what if an application is graphics-bound, as FSX sometimes is thought to be?

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

If the application is graphics-bound then upgrading the graphics will indeed speed up the system. However, if the system is cpu bound, or disk-bound, then upgrading the graphics will have no effect whatsoever.

I have received an offline communication from Paxx in which he indicated an intention to continue to discuss certain FSX-related matters in public in spite of its effect to date on Phil Taylor. If that's fine with site management then it's fine with me too. However, in a balanced system, improving any single component -- memory, cpu speed, HDD speed, graphics card throughput -- will have no effect on system performance, and no amount of Paxx's saying so can rewrite the Laws of Computational Physics.

Let me repeat that: If the system is already in balance, improving any one component will not speed it up. The only way to speed it up is to improve ALL the components.

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Now ... If benchmarks are run that show the RELATIVE IMPORTANCE of the various kinds of components, those statements about relative importance remain true even though installing faster components on some other system won't speed that other system up -- if that other system was in balance to begin with.

So if Paxx were to say "I have proof that adding a faster veeblefetzer to my configuration did not speed it up", I would not be impressed, just as I would not be impressed if he were to report that neither a faster gezorninblatz nor a faster electrowidget made it run faster. All that such statements would prove is that the test system was in balance.

Again, the only way to improve a balanced system is to upgrade ALL of its essential components. The question then becomes the following: What kinds of upgrades should be made to the components -- that is, WHAT IS THE RELATIVE IMPORTANCE of each of the components.

And that is where the Toms benchmarks came in, and that is where I part company with Paxx. The remarks made by Phil Taylor of ACES Studio about the relevance of various benchmark results are bang on because they have to do with UPGRADE PATHS, and with the CONSTRUCTION OF SYSTEMS FROM SCRATCH, and not with the behavior of already-balanced systems, or of systems that are out of balance in a way that is not being tested for by Paxx. They have to do with the relative importance of component types, and not with the "fairness" of Toms benchmarks.

xxxxxxxxxxxxxxxxxxx

Folks, you have to listen carefully to what people say, and you have to analyze what they say. Otherwise you will be easy prey for the following hypothetical news story, which came up when I was in high school fifty years ago ...

As reported in the New York Times ... "Yesterday a race was held between an American car and a Russian car. The American car won."

As reported in Pravda ... "Yesterday an automobile race was held. The Russian car came in second while the American car came in next to last."

[xxmikexx]

P.S.

I should have mentioned that the purpose of the sliders in FS is to balance the sim with respect to the installed hardware. In other words, to help you not waste cpu cycles that won't help if the system is graphics card bound, for example.

Once you balance FSX to your personal taste, the only way to improve its performance will be to upgrade hardware. Yes, the availability of sliders does make it possible for single-component upgrades to improve the overall system throughput. My earlier statements were true in the general case of no-sliders applications and were made to illustrate concepts.

[myocardia]

You know, Mike, I've thought about it, and your FSX system is far from "balanced". Your processor is many times faster than your onboard video, even if it's a single-core Athlon 64. If it's a dual-core, then that goes double. Maybe you should tell your wife that you want a $50 video card for Father's day, even if you have to buy it yourself. Your framerates would more than double (at your current settings), if you have a dual-core, and would be at least 50% higher, if you've got a single-core A64 3500.

As a matter of fact, if you've got a single-core, one of these $29 7300GT's would be plenty: http://www.newegg.com/Product/Produc...82E16814500033 A 7300GT would even work fine with a dual-core 4200+, it just wouldn't be as fast as one of these slightly more expensive 7600GS's: http://www.newegg.com/Product/Produc...82E16814139008

Anyway, I'm not trying to tell you what you ought to do with your computer, just trying to let you know how quick, easy, and cheap it would be to bring yours into balance. I CAN tell you that you would most likely enjoy the time that you get occasionally for a flight more, if you were able to get the same framerates, with the scenery cranked way up.