Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
8-Core vs. Quad-Core vs. Snow Leopard
- Thread starter Apple Genius
- Start date
- Sort by reaction score
In reference to SL, Yes, some aspects will improve due to the ability to multi thread.But it will depend on specifics though. Particularly where user input is involved for the completion of a task.
You worded it generally enough that it's hard to argue but the general sentence leaves me with the wrong impression. SL won't improve multitasking much if at all. Grand Central is only a framework. It would be like installing fast-fourier-transform link libraries on your system and expecting all your math apps to run faster. If your math apps don't use that library they do you no good at all. None of the apps we have today will multi-task better or faster on Snow Leopard.
If Apple tweaks their scheduler it can only really improve HT scheduling. So in like 20% of the cases where virtual cores on the same physical core were getting schedules better executed on two physical cores we'll see a 10% to 20% increase at the most. So it's basically like a 15% increase 20% of the time and ONLY when the CPU is the bottleneck in the operations being performed - which is like 10% of the time?
15% of 20% of 10%. LOL - the new scheduler won't make a significant difference and Grand Central zero in all current applications.
And no, Apple Genius, we won't be able to sue them. LOL
he is wrong, read the whole thread..
by the way, sbb155, if you would really know anything about audio you would know that SSD has bad write speeds which are significant for audio applications, in recording situations very much over a good read speed... i personally would not suggest an ssd for an audio drive...
Quad + SSD // Octo + Raid0 HDD..
for audio ? pff... read up on it...
I was trying to keep it simple.You worded it generally enough that it's hard to argue but the general sentence leaves me with the wrong impression. SL won't improve multitasking much if at all. Grand Central is only a framework. It would by like installing fast-Fourier-transform link libraries on your system and expecting all your math apps to run faster. If your math apps don't use that library they do you no good at all. None of the apps we have today will multi-task better or faster on Snow Leopard.
If Apple tweaks their scheduler it can only really improve HT scheduling. So in like 20% of the cases where virtual cores on the same physical core were getting schedules better executed on two physical cores we'll see a 10% to 20% increase at the most. So it's basically like a 15% increase 20% of the time and ONLY when the CPU is the bottleneck in the operations being performed - which is like 10% of the time?
15% of 20% of 10%. LOL - the new scheduler won't make a significant difference and Grand Central zero in all current applications.
And no, Apple Genius, we won't be able to sue them. LOL
Maybe a bit too much. Your right, that SL offers a frame work for 3rd party developers. That's really it from what I've been able to glean past the marketing. It's not a magic wand that's going to make everything a user does go faster. It's just not possible in the first place, as not all applications can run concurrent threads. And for it to happen to those it can, they must be developed with the new API's, or have found a way to do it already (some have).
I see mostly potential, but not massive increases across the board. In areas Apple could control, certain aspects of the OS and it's associated features, yes, where ever possible. Evolutionary, though. Maybe 20%, and that's for memory bound (enough BW to feed it).
3rd party developers are on their own. They now have a frame work, but it won't write the code for them, and it's not fool proof either (poorly written code is still possible, and all the deleterious effects). Good for fence sitters that haven't updated their offerings, but applications that it really matters, seem to have done it already.
So when SL drops, those that are expecting their system to hit warp speed, will be in for a disappointment IMO.
And he still didn't change his name
I know... How do I contact the admin to request his name change? "Apple Genius" to "Not so Genius"?
I know... How do I contact the admin to request his name change? "Apple Genius" to "Not so Genius"?
Nah clearly he's refering Apple as the "genius" in his name
It will be obsolete as soon as the next Mac Pro revision is released. But weather or not it will still do what you want in 3 years time depends on your needs. Mac Pros are way more powerful than the average computer, and unless you work in an environment where time is money (eg video effects rendering) it should be more than enough for many years. I went to a recording studio a year or two ago and the producer there still had an original Powermac G5 - proof that many professionals don't need the latest and greatest
Link to explanation of Grand Central Dispatch
Here is a link to an interesting and understandable explanation of Grand Central Dispatch on Apple's website
http://images.apple.com/macosx/technology/docs/GrandCentral_TB_brief_20090608.pdf
My take from this is that there is some limited performance increase out of the gate, since the os dispatch engine has been modified. Of course, the apps will still need to be modified to fully leverage the capabilities.
From a programming point of view, the changes needed appear to be very simple. When I did parallel programming on supercomputers, the focus was either on vector (array) processing, or trying to divide an conquer. This seems to have a lot of promise. The programmers won't have to worry about hardware assumptions, thread management, etc. just the division of processing.
I expect that the intensive Mac applications will be adapted rather quickly. Performance of an application is a driver in purchase decisions for intensive applications. Yes, there is a lot of work to be done, but it seems to be a lot more straightforward than I expected.
Here is a link to an interesting and understandable explanation of Grand Central Dispatch on Apple's website
http://images.apple.com/macosx/technology/docs/GrandCentral_TB_brief_20090608.pdf
My take from this is that there is some limited performance increase out of the gate, since the os dispatch engine has been modified. Of course, the apps will still need to be modified to fully leverage the capabilities.
From a programming point of view, the changes needed appear to be very simple. When I did parallel programming on supercomputers, the focus was either on vector (array) processing, or trying to divide an conquer. This seems to have a lot of promise. The programmers won't have to worry about hardware assumptions, thread management, etc. just the division of processing.
I expect that the intensive Mac applications will be adapted rather quickly. Performance of an application is a driver in purchase decisions for intensive applications. Yes, there is a lot of work to be done, but it seems to be a lot more straightforward than I expected.
Wrong
2.93 quad is WAY faster than 2.26 octo
check out post #61
https://forums.macrumors.com/threads/738138/
thxs
2.93 quad is WAY faster than 2.26 octo
check out post #61
https://forums.macrumors.com/threads/738138/
thxs
For a SINGLE CORE render, sure it was faster (one core at 2.93 > one core at 2.26). When using more than one core, the 2.26 is faster based on the chart you are referring to.
and i can guarantee, that for the $$,
MOST of the time, the 2.93 quad + SSD will be faster than 2.26 octo with HD on MOST programs.
Who does work with "cinebench"???? duhhhhhhhhhhhhh
if you're going to spend the money, get a 2.93 quad + SSD and be prepared to be happy
Show me one place that Apple has ever claimed that multithreaded apps will be developed. Are you saying that Apple is responsible for the work of every third-party developer out there?
Good luck with that suit.
Are you reading this chart wrong?
The 2.26GHz octo-core with a score of 20,138 clearly outperforms the 2.93GHz quad-core with a score of 15,182.
This is consistent with other tests. When using highly multi-threaded applications the 2.26GHz octo-core will be 30-50% faster than the 2.93GHz quad-core.
This is of course the other way around if you run single-threaded applications.
But that was said time and again.
No one. Cinebench is not an application.
The application is Cinema4D. A lot of people use Mac Pros for 3D computer graphics with applications like Cinema4D, Maya, Houdini or MentalRay.
AND DID THE 2.93 QUAD HAVE AN SSD IN THOSE TESTS? NO
DID THE 2.93 HAVE THE SAME AMT OF RAM? I DONT KNOW
AND ARE MOST PROGRAMS MULTICORE? NO
TRY THEM IN THE REAL WORLD
MOST APPLICATIONS, most of the time are faster with the 2.93 quad.
I don't think we're seeing your point sbb155. The benchmarks we have today are not based on Snow Leopard and the single core performance is the only test where higher clocks come into play.
I really think it comes down to the individual and their apps and workflow. For some a fast clockseed is going to be key and for others the ability to have more instructions in flight for dispatch is more key.
I think that if someone has the money for an Octo they should go there even if they have to accept a lower clockseed because the crystal ball is easy to see here. In just a couple of years quad cores will the the baseline and performance systems will be 6/12/8/16 cores.
sbb155, I am not disputing what you're saying.
Today only a small portion of applications use multiple threads effectively, but the ones which do are running faster on a 2.26GHz octo-core even today.
So the overall performance today will depend on what applications you are using.
Yet the discussion is not about purchasing something today and never upgrading the software ever. The discussion was about the Mac Pro being 'obsolete in 3 years', or not.
Intel can't crank up CPU GHz any more and every application will eventually have to use multiple threads to increase performance.
Assuming the OP is going to upgrade his software packages in the future, it is very reasonable to expect that these future versions will make use of multithreading at which point the octo-core Mac Pro will give him more performance in the future. And as a result be less obsolete.
But this of course assumes the OP intends to keep his Mac Pro for 3+ years.
It is clear that he will be better off to sell the Mac Pro in a year and buy a then current new Mac Pro. But that will also be more expensive overall.
Regarding the SSD I am not sure we're comparing apples to apples here.
Clearly the quad-core as well as the octo-core can be upgraded with a SSD at some point. And then again the 2.26GHz octo-core with SSD will be faster than the 2.93GHz quad-core with SSD.
However, the quad-core cannot be upgraded to an octo-core later.
The decision regarding the number of cores must be made today.
Today only a small portion of applications use multiple threads effectively, but the ones which do are running faster on a 2.26GHz octo-core even today.
So the overall performance today will depend on what applications you are using.
Yet the discussion is not about purchasing something today and never upgrading the software ever. The discussion was about the Mac Pro being 'obsolete in 3 years', or not.
Intel can't crank up CPU GHz any more and every application will eventually have to use multiple threads to increase performance.
Assuming the OP is going to upgrade his software packages in the future, it is very reasonable to expect that these future versions will make use of multithreading at which point the octo-core Mac Pro will give him more performance in the future. And as a result be less obsolete.
But this of course assumes the OP intends to keep his Mac Pro for 3+ years.
It is clear that he will be better off to sell the Mac Pro in a year and buy a then current new Mac Pro. But that will also be more expensive overall.
Regarding the SSD I am not sure we're comparing apples to apples here.
Clearly the quad-core as well as the octo-core can be upgraded with a SSD at some point. And then again the 2.26GHz octo-core with SSD will be faster than the 2.93GHz quad-core with SSD.
However, the quad-core cannot be upgraded to an octo-core later.
The decision regarding the number of cores must be made today.
I keep noticing one trend with discussion about threading. People tend to get fixated on single or multi-threaded apps.
The difference with Snow Leopard is that have new concurrency API at both high and low level. So Apple has really "eaten their own dogfood" here so to speak and with Snow Leopard even a bunch of single threaded apps will still perform better than on previous OS.
I think apps that run in the background and apps that work well in tandem with other apps will see a nice benefit even if they are only single threaded.
The difference with Snow Leopard is that have new concurrency API at both high and low level. So Apple has really "eaten their own dogfood" here so to speak and with Snow Leopard even a bunch of single threaded apps will still perform better than on previous OS.
I think apps that run in the background and apps that work well in tandem with other apps will see a nice benefit even if they are only single threaded.
or just pure dissatisfaction over the fact that they purchased quad core and in about 2 months time when SL comes out it will prove them wrong the notion that quad core is "faster" then octo.
Indeed. I guess it's just up to the individuals style but I want as many "pistons" working for me as possible even if they developer less horses per cyclinder.
Judging from the benchmarks I'd guesstimate that from within the same processor architecture it would take a clockspeed advantage of 30+ % for a Quad core to negate the advantage of an Octo core. The 2.96Ghz and the 2.26Ghz are really close. Snow Leopard may necessitate a Quad being 40 % higher clocked to compete.
Very true. I'm glad that Apple is giving the user the option of 4 or 8 cores, and varying clock speeds.
And if Gulftown becomes an addition to Gainestown rather than a replacement, we could see 4, 6, 8, and 12-core Mac Pros in 2010. Each with 1 or 2 speed options.
Looks more like about 70% to me (20138/15182·2.93/2.27), at least for the multicore render.
And, again, for the same amt of money...
a 2.93 quad + ssd + RAM will be faster for most tasks most of the time, then the stock 2.26 octo with a regular HD.
the premium for the octo is ridiculous compared to what you can get a in a souped up 2.93 quad.
Actually, the quad can be upgraded to octo... it is called ebay... mac pros have great resale value at any time...