Pancenter wrote:
tbirdparis wrote:
In any case, I've filed a bug report with Apple on this and am waiting on word on the matter from some engineers who may be in a position to actually chase this up. I don't care whether I am actually shown 4 CPU bars in Logic or 2. I just want to know one way or the other if Hyper Threading, an advertised hardware feature of the mobile i5/i7 in the new MBPs, is actually being used by Logic, or not. Doesn't matter to me in the slightest if I only ever see 2 CPU bars, I don't need to actually see that there are 4 virtual cores being used to process threads. But I am interested in making sure that the hardware is being used to its full extent, that's all.
Thanks for doing this, I admit to being curious as to how HT cores can be used in Logic, especially on single thread processes like a single channel strip with VI and FX. Maybe it's done at the OS level, which would be a feature of Snow Leopard as it sure didn't work like that on Leopard.
pancenter-
I think it's even simpler than that. With Hyper Threading, as far as the computer is concerned, it actually simply
has 2 cores per physical core that actually exists. This is because of something to do with something technical-sounding like 'long pipelines' or something like that.. So, a core gets given threads to process, and it pretends to be 2 cores instead of one because it can divide the threads one at a time down two separate pipelines or something like that. So in effect, no it's not quite as fast as having two real actual cores doing the crunching.. but, it absolutely acts as if it were two separate cores, meaning that it can take on a VI and a plug in at the same time, dividing them up as though they had their own core to run on. So in terms of parallelism, it's functionally identical to having multiple real physical cores with all the advantages that this brings. It's only a little slower than the equivalent amount of cores doing the same thing if they were all actual physical ones, because CPU time is being shared. But apparently because of this long pipeline thingy, this split processing is much faster than you'd think it would be if you're just imagining a CPU sharing its time evenly between one thread and the next.
Look, my understanding of it is rudimentary to say the least. But the basic principle is such that Logic Pro on an i5 or i7 MBP should absolutely, definitely behave like a quad core, with of course a slight penalty vs an equivalent CPU that actually
is a quad core. In any case, let's wait and see what happens on this front. People seem to have been getting performance results in Logic on the new i7 MBPs that aren't as much faster as they should be compared to a Core 2 Duo,
if Hyper Threading was really being used. So this is why my suspicion is that the architecture of these CPUs is slightly different to Xeons and so Logic isn't tweaked yet to take advantage of it. OSX 10.6.3 definitely does, because you open Activity Monitor, and you see the four CPU bars right there on your screen. My feeling is that there's just an adjustment or two needed to enable the full use of Hyper Threading by Logic (which may come in a Logic update, or a Pro Kit or Pro App Support update), at which point we'll see a jump in Logic performance that reflects the same jump you already see in simple benchmark tests. I'm almost certain that this must be the case, because if the i7 is reporting x amount of a speed bump over the previous core 2 duos, and yet the performance increase specifically for Logic is nowhere near in that same range.. then this must be why.
So I've done the bug report and also mentioned it at Logic feedback, and again I encourage anyone else who is interested in this to do the same. I asked for them to at least put out a KB article explaining Logic and the mobile i5/i7, so at least we know what the story is.
Like I said, you look at the benchmarks and they don't add up. In simple terms - the MBP i7 shows a much bigger number when you run a geekbench CPU test on it compared to a Core 2 Duo. And yet that huge speed difference does not seem to correspond to the amount of extra plug ins etc that Logic can run, so that tells me something is missing in the equation. Another giveaway for me is that in the past, Apple have often put up a comparison chart showing real-world(ish) figures of how the performance has improved for all of the Pro Apps when they release a new series of Mac Pro or MacBook Pro. For this release, the usual comparison chart showing how many reverb plug ins or whatever in Logic Pro is suspiciously absent. So, they either just overlooked it or decided not to put that up on the site this time.. or, they intentionally skipped over it because they
know that for the moment, Hyper Threading is an advertisable feature for this series of machine at least in terms of being used by OSX, but that Logic Pro currently doesn't use it. So rather than putting up a lukewarm comparison chart that only showed a slight edge in Logic performance (which would have looked completely out of whack with the quite significant boost these machines show in overall speed), they may have decided to just not mention it at all for now.. until the code adjustments are made such that Logic can actually use the Hyper Threading feature which is probably a little different in architecture to how it is in the Xeons. Anyway like I said.. we'll see. I still like my i7 and I personally haven't even had the chance to see how it performs yet at all.. so when I do, I might find that it really already is a lot faster, although it seems that people are discovering the contrary to be true.