Please explain why older cards perform better?

[dporter22]

Can someone please explain why older nvidia cards perform better than those of the latest generation on Microsoft FSX?

I have researched extensively but cannot find find the reason why an 8800GT outperforms both the 9800 and 260. I know FSX is CPU dependent, but how does a 9800 or 260 with higher core clock , higher shader, higher memory clock, more bandwidth and more stream processors not significantly outperform an 8800?

Tom's articles do not explain what impact each of these specs has on FSX or which is more important. One of the articles even says overclocking has a negative impact on FPS!

I've compared the top cards on Tom's charts for FSX and there doesn't seem to be any trend in figuring out what produces the best rates. Some cards have higher clock speeds and more video memory but are low on the list. Some cards have low bandwidth but high shader clocks and they rate higher.

If I want to purchase a new card, logic says faster/more memory/higher clock speeds/more stream processors is better, but Tom's chart for FSX indicates the opposite!

I'm currently running an 8800 Ultra. How can supposedly much more capable, newer cards in the 9 series and higher perform worse on FSX?

 

[ghmage]

The 8800, 9800 and 260 do not differ in terms of newer or older technology. The model numbers are only marketing.

 

[invisik]

toms charts are becoming more false by the minute these days.
ghmage the 8800 9800 are pretty much similar but gtx260 is a new generation meaning new architect so its new tech.
=]

update gtx200 series is founded by the g80 core but adds a hell lot more raw power, simple put.

 

[br3nd064]

Toms hardware's charts need to be updated. A lot of those benchmarks were made when the drivers for the new cards weren't mature yet.

Also, as invisik said, the 9800's aren't really "new tech", but the gtx 260 and 280 are.

 

[dporter22]

Yes, I understand that the 9 series may not technically be new technology, but my question mainly concerns why "older" cards perform better than "newer" cards.

Specifically, a GTX 260 and a 9800 generally have much higher specs than the 8 series, yet they are both "beat" by a last generation card on FSX.

I'm not an electrical engineer; how can this be? If the chart is correct, video memory doesn't make a difference, core clock speed doesn't matter (in fact, slower clock speeds produce higher frame rates ?!), shaders doesn't matter, stream processors doesn't matter, etc.

I just really don't understand how you can have cards with significantly higher/improved/faster/better specs that don't correlate to performance (at least on FSX and some of the other games on the chart). Why should I pay for an 896 MB card if 512 MB performs better? Why pay extra for a factory overclocked GPU if that will reduce performance? Why buy two cards in SLI if activating SLI hurts frame rates in many games?

What am I not getting? How do I pick a new card if better/stronger/faster/more expensive sometimes means worse/weaker/slower?

 

[megamanx00]

It's probably just different driver versions. Compatibility fixes can lower performance, and new drivers tend to add more overhead. FSX is more CPU dependent than anything, so bug fixes that add CPU overhead can be a big hindrance to it. It just depends.

 

[dporter22]

Different driver versions can't possibly provide so much more power to an 8800 that it beats a gtx 260, can they? If that's the case, we would never need new cards, just new drivers.

Is anyone able to answer any of my questions above? Even if you can't explain the details, how about just an explanation of which specs are actually important in comparing cards? Is it core clock, shader clock, bandwidth? And if one or more of these translates to frame rates, how does a new card with more "power" underperform an older card with lower specs?

Is an extra $40 for a "Maxcore" version of a card with more stream processors valid? Do more stream processors correlate to more higher frame rates?

How about overclocking? What does it do for me in terms of game play? Anything, or is it just marketing?

Is the shader clock speed or core clock speed more important?

Anyone?

 

[NewLCD123]

Perhaps the cpu is way too slow. Ive experienced this myself where upgrading to a "faster" card did absolutely nothing till I upgraded the cpu as well. Ive seen so many people waste $ on a GPU when their CPU holds them back 100%

 

[Dekasav]

It is well known that Tom's didn't use an extremely fast processor...actually, hardly a very fast one at all when it did those tests. They said themselves that the fastest cards, especially in CF and SLI, would be held back by the CPU. I think you're trying to read too much out of charts that are obviously faulty. The GTX 260 is better than the 9800GTX, it's that simple.

Things you want to look for in getting a new card are:
1. Reviews of the card against others. This will show you where a 9800GTX+ is in relation to a 9600GSO.
2. Figure out which card you want (9800GTX, GTX 260, HD 4850, etc.) based on the reviews, then look at the cost of each.
3. Look for/at OC'd versions of each card. Generally, a 10% clockspeed increase will net you less than 10% of an increase. More stream processors are good, higher clocks are good, more bandwidth is good. Really, more is good. Forget the Tom's charts, they're faulty. And don't go by FSX, it's cpu bound. Go by something like Stalker, where it's much more GPU bound (not Crysis, either. Poor coding = bad test).

 

[invisik]

specs dont matter it depends on architect. 2900xt had beastly specs but performed horrible compared to an 8800ultra.

 

[vinegarjoe]

No one mentioned the fact that Flight Sim X sux in it's use of video cards. Same reason why a lot of games run slower in SLI/Crossfire or why fast dual core cpus often outperform similarly priced quad cores.... lazy programmers who don't want to be bothered with writing code so it will not only take advantage of hardware abilities but also be scalable so peeps with old comps can still play it. Though perhaps it's not the programmers' fault, but the bean counters running the software/os companies who want programs done yesterday at zero cost and no quality.

 

[mactronix]

I would say its a mix of some of the things that have been said above. First and most importantly don't just go on Toms charts they are famously out of date and full of errors. I know its handy for your needs that they include FSX but whats the point if the info is suspect ? Try and find some other benchmarks or join a dedicated FSX forum, some of those guys would scare you with the knowledge they have on tweaking and getting the most from the game. (sorry don't play it myself) As far as graphics cards and picking a good one goes, there is no quick way of telling you what to look for. There are so many differentials to look out for that its getting hard for someone who has half an idea whats what to pick through and find the decent cards. If you really want to understand it then take it slow and google around ask on the forum about what does what on a card, don't forget to look at the stickies at the top of the first page. Architecture is really important and a card with lower core and memory clocks can beat one with faster specs because it has more SP's or it could be the other way around depending on the difference in each. The easiest way to get a feel of how well a card performs is to have a look at about 3/4 review sites and compare whats said and take a rough average of whats been shown. Any issues can always be queried here My opinion of good sites are AndandTech, Techreport, X-bit labs, Bit-tec.net and one that compares a lot of cards at a lot of games is Tecpowerup.

Mactronix