I do not, and have never said that memory BW could nto be the issue.. My oppinions are actually falsifiable unlike certain folks around here.
AZN, I simply believe, and have tried over and over again to explain, that it is very much possible that other issues coudl be at play.. I am not saying you are wrong.. just a bit silly with your blind stuborn nature.
I merely think that it is more likely due to drivers, but could also be hardware, maybe even BW.
As to why you are unscientific. You suffer from a severe confirmation bias. There are possible reasons that this behavior is seen that is not bandwidth but you ignore them. Thus you are comparing data sets that change these other possible variables becaseu you refuse to see them. If you were to do things properly you would only bring up data that changes one thing at a time, or at least as few as possible. Comparing a 4870x2 to a 5870 changes several of teh possibilities we have brought up. No matter how unlikely you think they are, it is still makes it unscientific to ignore the possibility by changing them all to try and prove a point. All you are acomplishing is showing that it could still be all of the reasons we have already brought up..
If you did this as your career you woudl ahve no job unless you learned to accept other possible answers. A good scientist always accepts that they could be wrong.. In fact science is about trying to prove you are wrong more so than proving you are right.. If you can't prove yourself wrong then you have a good theory. You do not have a good theory, and I repeat that it is nonsense, because all of the data you use as proof ignores the scientific method entirely.
If yuo want to compare things, choose one of drivers, bandwidth, and core architecture. Leave the others the same, and then compare. Proving that it coudl be memory BW does not prove you right, or us wrong. You do not seem to understand how one goes about dismissing a hypothesis.
For a hypothesis to be valid it there has to be a way to disprove it. To disprove something you have to prove it wrong.. but if there is no way to do this, "There is an elephant beside me," then the claim becomes nonsense. Repeatedly usign an example that does not prove you wrong is a sorry way to go about things...
Awesome, awesome post. I've been reading this thread up to this point, and I just have to throw my hat into the ring on this one. I found this thread by googling '5850 and 5870 at same clocks', because I own both cards (two 5850's, in fact) and wanted to see if others were seeing what I am seeing with the performance of them (i.e. they are very close when at the same clocks).
Daedalus is dead-on right with this post, in terms of what I've been thinking about your *opinions* AzN.
What you are doing *continually* in this thread is erroneous from the standpoint of the scientific method.
You cannot rightfully determine 'a reason' why the 4870x2 is faster than the 5870. It's IMPOSSIBLE to do so, because they are two different cards, and I don't care WHAT you say, they have a TON of differences. Just because you *think* that the only important difference lies in memory bandwidth (which, as has been pointed out to you repeatedly, is not actually true), you DO NOT KNOW THAT. You are GUESSING. Period.
You are violating so many fundamental scientific premises with your entire line of reasoning on this thread, I just felt compelled to chime in. In comparing the 4870x2 with the 5870, you are changing WAY MORE VARIABLES than just the memory bandwidth. Therefore, you CANNOT positively conclude that bandwidth accounts for the difference. Even if it were to 'turn out' that you are actually 'correct', you are still WRONG to claim that you are.
What is going here with regards to the 5850 vs 5870 issue, to me, seems very simple. This is the exact same phenomenon that one can see when comparing the GTX260 192 vs the GTX275 (at 240sp).
If you were to put these cards at the same clocks (they have 100% identical architectures other than SP's), you would find that the 275 would *only* beat the 260 192sp by 25% (based on the extra shaders it has (240-192/192) = 25%) if you specially devised a test where the shaders were the 100% limiting factor to performance.
Much like your pixel fillrate test you continually post here AzN ... it's a test DESIGNED to show a SPECIFIC type of bottleneck.
Out there in the real world, however, testing games and whatnot, you'd quickly discover that DIFFERENT TESTS (indeed even within the same test, it varies from frame to frame sometimes) can cause bottlenecks in DIFFERENT subsystems of the card!
In the real world, you'd likely discover that your 25% shader count advantage would actually give you somewhere between 5% and 22% improvement in frames, and probably the average would be around 10% improvement (off the top of my head).
And that is because, obviously, shader count is only PART of the card. There's many other links in the chain (tmus, rops, threading engine, compression and AA/AF algorithms, caching bandwidth, memory bandwidth etc), all of which are equally capable of acting as bottlenecks, depending on the demands being made on the card at the time.
As far as I'm concerned AzN, you've proven nothing with what you've posted on this thread. So, here's another voice adding to the chorus ... I agree with 'everyone else'. You are exercising 'confirmation bias' in an extreme way, picking and choosing data points that seem to bear out (mostly tangentially, at best) what you have decided you already believe.
And no matter what, anytime on declares that a certain subsystem on a card is 'the bottleneck' on that card ... they are wrong. Because which part acts as the bottleneck depends on the test you run in order to look for that bottleneck.
And lastly, RAID 1 does serve as a reasonably accurate analogy with how the memory subsystem of xfire works. Just as nobody in their right mind would claim that you double the bandwidth when you hook up two drives in RAID 1, one should also not claim that running xfire doubles your memory bandwidth. And this because you will get the SAME (or worse) write performance, it's only on reads that you theoretically could get up to a 100% performance increase.
.
Facebook
Twitter