Techguy is correct though. He may be right in his calculations and what he is saying, but FSX is unpredictable. What we think we know about FSX is guess work. Yes it may be partially correct, but nothing is 100% solid. The only way we can actually prove what he is saying is through testing. We know the data for PCIx2.0, we don't know for PCIx3.0. We know there's a limit to BW of the PCI slot, but the question is, does FSX use all that bandwidth in the most dense areas where the graphics is needed most.

I just want to point out one thing. In the very first comment, NickN stated this

Also, don’t forget to test it in combat. the theory is great, but it must always be tested on different systems to be sure the design is in fact on the right track

We all know how well Nick understands FSX, more than we all do.

I recently purchased a 570 and have clocked it to 875/1750/4400, at which point it performs approximately equal to (or better) than a 580, excepting again for memory bandwidth and frame buffer limitations.

That would be nice but that would really be a stretch, maybe closer to 480 teritory but certainly not a 580. You can clock the 570 to 1100 and it still wont make up for the lack of proccessing units, L2 cache, besides 30% less fillrate. GPU Clock rate just doesnt change very much as far as FSX is concerned.

You're the one not familiar with science, because you fail to discern when you need to test a theory and when you don't. In the case of autogen, you know the object counts from the fsx.cfg, you know the object vertex counts, the data type sizes and the frame rate. You also know the PCIe bandwidth available, so you know "a priori" what the maximum number of autogen objects you can render. So argue Phil Taylor's reasoning (he explains it in his blog), but please stop repeating that a test is needed when its clearly not.For your own educational benefit go visit the "a priori" page on Wikipedia so you can learn that "a priori knowledge or justification is independent of experience". If you still disagree, you are welcome to argue the roots of epistemology with my dog.Cheers,- jahman.

+1Simple math, and that really cant be argued with (which I highly doubt Nick was suggesting).

Reviews I have read all say that the 570 performs better than the 480 in majority of benchmarks, but the 480 beats it in some benchmarks too. So really the 570 is big brother to the 480. And 480 is below the 570 in hierarchy.

Not for FSX, but close.

That would be nice but that would really be a stretch, maybe closer to 480 teritory but certainly not a 580.

The 570 is already 480 speed.

You can clock the 570 to 1100 and it still wont make up for the lack of proccessing units, L2 cache, besides 30% less fillrate. GPU Clock rate just doesnt change very much as far as FSX is concerned.

The peak pixel fillrate numbers in your graph are incorrect. Each SM can only pass 2 pixels to the ROPs per clock, so there are essentially an extra 50% ROPs. The 580 has meager functional unit increases over the 570, they're both based on the same GF110 silicon. 580 has 16SMs comprised of 512 SPs and 64 TMUs and 6x 64-bit GDDR5 memory controllers. Compared to the 570's 15 SMs comprised of 480 SPs and 60 TMUs and 5x 64-bit memory controllers. That and the clockspeed differentials are the only differences between the two cards. I'll grant that the 580 has more video RAM and a bandwidth differential that cannot be overcome, but in terms of processing power the difference is rather close. Since I've overclocked my GPU by nearly 20% and my RAM by almost 16%, it is now faster than a 580 except when limited by the aforementioned video RAM and bandwidth bottlenecks.

Techguy is correct though. He may be right in his calculations and what he is saying, but FSX is unpredictable. What we think we know about FSX is guess work. Yes it may be partially correct, but nothing is 100% solid. The only way we can actually prove what he is saying is through testing. We know the data for PCIx2.0, we don't know for PCIx3.0. We know there's a limit to BW of the PCI slot, but the question is, does FSX use all that bandwidth in the most dense areas where the graphics is needed most.

Chris,A theoretical upper-bound on performance is a priori correct without any need for further testing.In the case of FSX autogen and PCIe, Phil Taylor (who I believe knows more about FSX internals than anyone else) demonstrates (as in mathematically proves) that the maximum requirements of FSX autogen exceed PCIe 2.0 bandwidth. There's nothing to test here at all. Of course it is possible PT's calculations are wrong, but in this case we would need to show where the theory is wrong and how it needs to be corrected.Now if you want to know how much BW autogen actually consumes, of course you will need to test for this, but I can guarantee the actual BW will be below the theoretical upper boundary.Perhaps its easier wth a simpler example: A HDD has a certain data density of X bytes per track and a rotational speed of Y r.p.m.. Well, using a formula you can state a priori what the maximum (upper-bound) sustained transfer rate will be for this particular HDD. Now if you want to know the actual transfer rate you will have to set-up a test bed, run the experiment and measure the result. But you know a priori that the result you measure will necessarily be below the theoretical upper-bound.There is a difference between a mathematical proof that yields a priori knowledge and a hypothesis that needs to be tested to be demonstrated, and good science is having the knowledge and experience to discern which is which.Cheers,- jahman

Chris,A theoretical upper-bound on performance is a priori correct without any need for further testing.In the case of FSX autogen and PCIe, Phil Taylor (who I believe knows more about FSX internals than anyone else) demonstrates (as in mathematically proves) that the maximum requirements of FSX autogen exceed PCIe 2.0 bandwidth. There's nothing to test here at all. Of course it is possible PT's calculations are wrong, but in this case we would need to show where the theory is wrong and how it needs to be corrected.Now if you want to know how much BW autogen actually consumes, of course you will need to test for this, but I can guarantee the actual BW will be below the theoretical upper boundary.Perhaps its easier wth a simpler example: A HDD has a certain data density of X bytes per track and a rotational speed of Y r.p.m.. Well, using a formula you can state a priori what the maximum (upper-bound) sustained transfer rate will be for this particular HDD. Now if you want to know the actual transfer rate you will have to set-up a test bed, run the experiment and measure the result. But you know a priori that the result you measure will necessarily be below the theoretical upper-bound.There is a difference between a mathematical proof that yields a priori knowledge and a hypothesis that needs to be tested to be demonstrated, and good science is having the knowledge and experience to discern which is which.Cheers,- jahman

You have not established that autogen transfers via PCI-e are an upper bound on performance, though. Just because you (and PT) say something does not make it so. You have to adjust autogen settings and test at different PCI-e speeds to *prove* that it does have an impact on performance. You can choose to believe whatever you like, but I will not join you in that belief without proper testing for verification.

You have not established that autogen transfers via PCI-e are an upper bound on performance, though. Just because you (and PT) say something does not make it so. You have to adjust autogen settings and test at different PCI-e speeds to *prove* that it does have an impact on performance. You can choose to believe whatever you like, but I will not join you in that belief without proper testing for verification.

You a right! I'm a total noob re: video rendering software. The one that makes the claim is Phil Taylor, and he does so in a manner consistent with what is considered mathematical proof. All his formulae are there for peer review. Anyone can go to his page, review his calculations and say wheter he goofed or not. So unless you or someone elñse disproves PT's calculations, the proof stands.Additionally, you could even claim that it's not possible at all to establish a theoretical upper-bound requirement for FSX autogen, but you can only do so with a cogent argument as to why this is not possible, an argument that will necessarily take you straight to PTs formulae where you would have to prove no formuale exist for calculating the max autogen BW.What cou can't do is blanket state that all theoretical limits have to be tested. In my previous post I provided an easy example of an a priori theoretical BW limit for HDDs that doesn't need to be tested. Do you agree with that example?Cheers,- jahman.

You a right! I'm a total noob re: video rendering software.

Video rendering software? Is that what you consider FSX to be or are you trying to establish credibility of your own knowledge via some other source? Either way, don't make this personal, it's childish.

The one that makes the claim is Phil Taylor, and he does so in a manner consistent with what is considered mathematical proof. All his formulae are there for peer review. Anyone can go to his page, review his calculations and say wheter he goofed or not. So unless you or someone elñse disproves PT's calculations, the proof stands.

I know who Phil Taylor is. You're missing the point entirely. What you have posted is not proof of anything in and of itself. If Phil Taylor has some data to post by way of an application profiler which can demonstrate that PCI-e transfers are holding back FSX performance, then I will believe it. Something like Microsoft's Pix utility should do the job.

I think we should end this argument here. It's getting a little out of control now. The OP simply wanted to know what the performance in FSX maxes was like with a 980x and a gtx580. He didn't ask for a full-blown argument over whether the upper bounds exceed PCIe 2.0 bandwidth. I think we should rap this one up by saying that FSX is unable to be maxed with decent performance as of yet. However it will run very high settings superbly.

Since I've overclocked my GPU by nearly 20% and my RAM by almost 16%, it is now faster than a 580 except when limited by the aforementioned video RAM and bandwidth bottlenecks.

Faster? Sure...coughBelieve what you want but your are only fooling your self.Even if a 570 had all the goods that its missing from the 480 and matched 580 clock speeds it wouldn't be close, no where, no how and especially where FSX is concerned.

Faster? Sure...coughBelieve what you want but your are only fooling your self.Even if a 570 had all the goods that its missing from the 480 and matched 580 clock speeds it wouldn't be close, no where, no how and especially where FSX is concerned.

Seeing as how I review GPUs in my spare time, I'm going to go ahead and say: "I know more than you" and end this discussion right here. I mean, if this is your best rebuttal then I'm wasting my time. Numbers don't lie.

Seeing as how I review GPUs in my spare time, I'm going to go ahead and say: "I know more than you" and end this discussion right here. I mean, if this is your best rebuttal then I'm wasting my time. Numbers don't lie.

Good luck holding that job my friend. :0

What is with you two?! Seriously! You're acting like a pair of children. Grow up. If Techguy reviews GPUs, I'm fairly confident he can tell whether or not his OCd 570 is performing close to a 580. Stop arguing and start acting like responsible adults

If Techguy reviews GPUs

Thats a BIG wow my nanny friend.I think we all know what it means when someone tries to boast like that, "I review GPUs" " "I know more than you" - Give me a frigin break LOL.This is FSX, GPU clock doesn't account for squat, what does is in the ->accurate "graph" posted above that speaks volumes to the why the differences in these cards when used for FSX.