Has Logic 9.1.3 "actually" helped HT issue on i7 Macbook Pros?

Hi,

I just wanted to share a bit of info on my Macbook Pro i7 2.8 performance.
Underwhelming is the word I would use - at least with Logic 9.1.3.

I have the current "top of the range" i7 machine and note with some concern/alarm that there appears to be something very different about the way the CPU works/is handled by Logic or, perhaps OSX 10.6.6 on a Macbook i7 in general.

Facts:

The Intel i7 950 is a desktop quad core i7 CPU with hyperthreading
When this CPU is presented to Logic 9.1.3, Logic shows EIGHT available cores

the i7 mobile CPU as used in my Macbook pro is a DUAL core i7 with hyperthreading
When this CPU is presented to Logic 9.1.3, Logic shows only TWO available cores -

Meaning surely that the available hyperthreading cores are NOT shown in the CPU meter, on a Macbook Pro but ARE shown on a desktop machine such as the iMac.

So, why does the same version of OSX, the same version of Logic, behave differently in terms of illustrating (utilising as well?) the hypertheading cores of a mobile i7 CPU? Is the CPU architecture different to the desktop version? To the extent that Logic cant "talk" to the HT cores, or is it just that it can't show them in the CPU meter... if the latter is the case, why on earth would that be?

Something is definitely odd about this scenario as Apple (and Intel) certainly do not advertise the HT implementation on the desktop vs mobile versions of the i7 CPU as being completely different?

If it is not a CPU architecture problem, why does Logic not treat the mobile version of the i7 as it does the desktop and show double the number of ACTUAL cores as available to Logic.

All independent reviews of the Macbook i7 substantiate Apple's claims that there are huge performance benefits to the i7 MBP vs the C2D processors at similar or identical clock speeds, but not (at least in terms of my impressions/real world experience thus far) so major with Logic 9?

In *Logic Pro>Preferences>Audio>Devices>Coreaudio,* do you get these options?

Hi - thanks!
Yes, I do have that option and it was set to automatic... showing 2 cores in Logic cpu meter.
I have specified 4 cores now (thanks and I will report back on actual performance change if any!) but again, isn't it odd that the same implementation of OSX and Logic 9 on a desktop CPU vs mobile CPU show 2 cores for the mobile dual and EIGHT for the desktop quad.

To be clear this is booting my MBP using a CLONED system disk from my desktop machine. I have not altered any CPU settings anywhere, I didn't even know you COULD until you helped!
So again, demonstration that OSX or Logic are treating the mobile i7 differently using the exact same OSX and Logic 9 installations.

They are different processors and different machines with different components, each may well need a "personalized" installation... so it may be that cloning your DT system to your MBP might be part of the problem...

Hmmm, sorry but I don't agree (no offence but I have real world experience - please see below).

OSX simply loads the kexts etc that are required, I don't see how my cloning a drive has Logic defaulting to showing two cores (not HT) on the mobile CPU but 8 cores on the desktop quad CPU - unless the CPU driver (kext) is different - illustrating that the behaviour of the i7 mobile HT implementation is somehow different.

For the record, I recently used the same method (cloning) of my system drive from a Core2Quad machine to an i7 - without issue and these are two completely different CPU architectures - not to mention having very different interfaces to RAM such as dual vs triple channel.

Yet, in that instance the core 2 quad showed 4 CPUS in Logic (no hyperthreading - which is correct for the Core 2 Quad) and then showed 8 cores in the i7 (4 + 4 HT) - immediately after cloning without any interference or settings changes on my part.

HT is implemented differently on the mobile processors... Logic does not auto-recognize it. I'm curious to see if you have a performance improvement after setting it to 4 cores.

pancenter-

Yes, it certainly appears so - I guess I knew already that the i7 mobile was never going to "fly" like a desktop equivalent and to be fair, this machine is a quite impressive performer generally.

In fairness also, I may simply not have enough RAM (4Gb - upgrading to 8 ASAP!) and thus I'm probably suffering the already suggested swap-file issues.
The only thing is that on my two most demanding projects Logic is only reporting around 2Gb of RAM in Activity Monitor, thus I had assumed that 4Gb was OK, or at least, not SUCH a big problem.

What seems to be clear is that Activity Monitor's reporting (or Logic's use of memory with regard to EXS24 and Kontakt 4) seems to be either way off or at least rather unhelpful.
I say this because even though Logic was showing around 1.9Gb of RAM in use, I decided to remove a RAM hungry instance of Kontakt 4 to see how Logic behaved - for example, unloading an instance of Abbey Road drums and in doing so, I got a "memory is getting low" message and then the other message about too much RAM being used and Logic closes the project automatically.
On re-opening the project, I succesfully unloaded Abbey Road drums and Logic certainly got more pep back in its step.

My point being that Logic is obviously not happy about the amount of RAM available, but it's nowhere near maxed out in Activity Monitor.
And all of this after heeding some advice to avoid 64bit Logic because of overhead and reliability in the Audio Bridge!
Confused? You will be

As for Hyperthreading (enabling 4 cores inside Logic Audio preferences) I can see/feel no discernible difference in performance with 4 bars in my CPU chart vs 2.
So I guess it is correct that on the mobile i7 there is an overall, quite massive performance difference against an equivalent clock speed desktop i7 and that Hyperthreading on the Macbook Pro within Logic, DOES NOT help improve performance.
Which, I understand from the point of view of battery life, heat dispersion etc. It's just frustrating given that I had hoped for quite a bit more "oomph" out of my new MBP.
But then, I am not alone...

But I'll be b*&^%$£d if I can understand why - if HT works(!) surely it always works when the CPU is busy computing! What's the point in having HT on a mobile CPU if it doesn't hyper-thread such that process heavy tasks like "live" digital audio manipulation benefit from it!?

Maybe it's time to invest in a UAD2?!

Again, out of curiosity, have you run activity monitor while Logic is running/playing to see how it lists the HT usage.

Are you streaming Abby Road drums from the internal drive?

There was much talk in another topic about HT not being able to work directly on a single thread process like a VI channel strip or for that matter, any channel strip. The single thread requirement demands an actual core, not a HT core. n this case, HT can process other operations in Logic that are not single thread related.
I'm wondering if you would see a performance boost with a clean install.

There has never been any definitive answer from Apple.

However... there have been some extended posts regarding HT performance on different machines.

http://discussions.apple.com/thread.jspa?threadID=2417498&start=0&tstart=0

http://discussions.apple.com/thread.jspa?threadID=2529938&start=0&tstart=0

http://discussions.apple.com/thread.jspa?threadID=2445587&start=0&tstart=0

Interesting reading if you've got the time.


Not naming names but occasionally there are Logic programmers responding in these threads. At least one I know of.

pancenter-

Firstly - thanks for the input.

I'm not sure I understand the HT/Activity monitor question: Apologies if this is obvious, but I can't see anything in Activity Monitor relating to HT - it merely shows overall CPU use. Or am I missing something?

I am loading/streaming everything from an internal 1Gb 2.5" 5400RPM drive albeit with my sample library on a separate partition.
I realise this isn't ideal but to be honest, my work is demanding because of virtual instruments rather than from HD audio streams and my experience with almost all modern computers is that (with enough physical RAM!) hard drive speed is really not an issue - other than when initially loading sample data but once IN RAM, the HD is very much secondary to performance.
Even a 5200 RPM 2.5" SATA drive can deal with a heck of a lot of audio data streams - certainly more than I usually demand (less than 10 audio tracks ordinarily).

My understanding of Multi-Threading (not HT) is this:

The OS has to support it.

The APP also has to support it specifically and requires core coding to ensure that the software in question can make best use of actual CPU cores via multiple threads - the OS can only do so much to distribute processing tasks amongst any/all available cores - that is: The OS cannot multithread a none-multithreading APP, it can only distribute the single threads of many (may I call) it "uni-threading" apps.

Such as:
If running APP1 which does NOT support multi-threading and also running APP2 which also does not support MT, the OS will distribute "intelligently" to TWO available cores, not FOUR or THREE or FIVE etc - because the TWO apps in question are only coded to "talk" to ONE thread each.

If a third APP3 is added, which DOES support multi-threading, then this will be distributed amongst whatever number of cores APP3 is coded to utilise. THE OS CANNOT MULTITHREAD ONE APP AMONGST MANY CORES UNLESS THE APP SUPPORTS IT WITHIN ITS OWN CODE.
I don't know if APP3 has to be 2x core coded, or 4x or 8x or whether it simply has to be generically multi-threaded (and the OS takes care of how many threads over whatever cores are available), but that's my understanding.
Certainly, the issues with Logic 9 (prior to 9.1.3) not utilising all available cores on 8 & 12? core Mac Pro's would bear this out... Logic had to be "updated" to properly talk to all (or most!) available cores, OSX could NOT do this of itself.

But I may be wrong?

How HT fits into this is rather beyond me for the moment (and I will look at the links you included)- but in the mobile i7 case HT does not appear to operate as a separate core and may even hinder(?) performance by diverting CPU power over two (one real, one virtual) cores instead of dedicating the one REAL core to solely deal with the task in hand.

I need a drink.

Sid Merrett wrote:
Firstly - thanks for the input.

I'm not sure I understand the HT/Activity monitor question: Apologies if this is obvious, but I can't see anything in Activity Monitor relating to HT - it merely shows overall CPU use. Or am I missing something?

Doesn't activity monitor show all four cores? if Logic is the only app running the HT cores (3&4) should be showing use, some of it will be OS but there should be a noticable difference when Logic is playing a project or at rest.

How HT fits into this is rather beyond me for the moment (and I will look at the links you included)- but in the mobile i7 case HT does not appear to operate as a separate core and may even hinder(?) performance by diverting CPU power over two (one real, one virtual) cores instead of dedicating the one REAL core to solely deal with the task in hand.

I believe some have posted these exact results.. better performance/stability with HT turned off.

Regarding multi-core tasks.. With the OS and App supporting HT, processes that can make use of HT are distributed by the OS, in the case of Logic, real-time processing of a channel strip whether VI, or audio/FX is a single core operation and must use one of the non-HT cores. That was my understanding but I think some users dispute this.... but as far as I know, the channel strip/single core info came direct from the horses mouth... so to speak.

pancenter-