Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

adjeff8

macrumors 6502
Nov 18, 2012
466
4

Erasmus

macrumors 68030
Jun 22, 2006
2,756
298
Australia
What did you expect? Mobile CPUs aren't going to get vastly more powerful until they get to 8 real cores. Everything else is just twiddling 'round the edges.
 

IllmasterMath

Cancelled
May 16, 2012
114
10
What did you expect? Mobile CPUs aren't going to get vastly more powerful until they get to 8 real cores. Everything else is just twiddling 'round the edges.

Not necessarily true. TSX, which will be introduced with Haswell, offers better multi-core scaling with minimal effort on the part of developers. I'm not sure if this is where the ~10% figure is coming from, but my understanding is there will be significant improvements in performance and power efficiency.

http://www.anandtech.com/print/6290
 

leman

macrumors Core
Oct 14, 2008
19,202
19,062
What did you expect? Mobile CPUs aren't going to get vastly more powerful until they get to 8 real cores. Everything else is just twiddling 'round the edges.

And how should 8 'real cores' improve the performance of, say, an office application, a web browser or a PDF viewer? Unfortunately, more cores won't speed up my LaTeX compiled documents either... There is only that much that can be parallelized and I personally would prefer improving the efficiency of the CPU instead of simply slapping more cores onto it.
 

Erasmus

macrumors 68030
Jun 22, 2006
2,756
298
Australia
And how should 8 'real cores' improve the performance of, say, an office application, a web browser or a PDF viewer? Unfortunately, more cores won't speed up my LaTeX compiled documents either... There is only that much that can be parallelized and I personally would prefer improving the efficiency of the CPU instead of simply slapping more cores onto it.

It won't. But if there were ways of easily increasing performance of a single CPU core, they would have been done by now. So short of ditching Silicon, and moving to something more interesting, single core performance gains will be stuck with tinkering with algorithms, which are unlikely to provide large performance gains. Certainly not between consecutive CPU generations.
 

leman

macrumors Core
Oct 14, 2008
19,202
19,062
It won't. But if there were ways of easily increasing performance of a single CPU core, they would have been done by now. So short of ditching Silicon, and moving to something more interesting, single core performance gains will be stuck with tinkering with algorithms, which are unlikely to provide large performance gains. Certainly not between consecutive CPU generations.

Alas, you are right...
 

yusukeaoki

macrumors 68030
Mar 22, 2011
2,550
6
Tokyo, Japan
If you are expecting for power, go for desktops with desktop CPU.
iMac and Mac Pro should be for you then.

MBPs and MBAs are only a "mobile" solution.
 

Erasmus

macrumors 68030
Jun 22, 2006
2,756
298
Australia
If you are expecting for power, go for desktops with desktop CPU.
iMac and Mac Pro should be for you then.

MBPs and MBAs are only a "mobile" solution.

Hardly. They just suck more power.

However, on review of the numbers:

A pretty much top of the line quad i7 desktop CPU has a base clock of 3.6 GHz, and can turbo up to a whopping 3.8 GHz if it's only operating on a single thread.

A pretty much top of the line quad i7 mobile CPU has a base clock of 2.8 GHz, and can turbo up to a pitiful 3.8 GHz if it's only operating on a single thread.

However, when loaded fully with 8 threads, the mobile CPU will probably be operating at something around 3.4 GHz, and using ~45W, as opposed to the desktop CPU running 6% faster, and using 130W.
 

yusukeaoki

macrumors 68030
Mar 22, 2011
2,550
6
Tokyo, Japan
Hardly. They just suck more power.

However, on review of the numbers:

A pretty much top of the line quad i7 desktop CPU has a base clock of 3.6 GHz, and can turbo up to a whopping 3.8 GHz if it's only operating on a single thread.

A pretty much top of the line quad i7 mobile CPU has a base clock of 2.8 GHz, and can turbo up to a pitiful 3.8 GHz if it's only operating on a single thread.

However, when loaded fully with 8 threads, the mobile CPU will probably be operating at something around 3.4 GHz, and using ~45W, as opposed to the desktop CPU running 6% faster, and using 130W.

If you compare with a hackintosh or Mac Pro, a mobile CPU literally does not stand a chance.
 

Erasmus

macrumors 68030
Jun 22, 2006
2,756
298
Australia
http://www.techpowerup.com/179631/I...ock-for-Clock-with-quot-Ivy-Bridge-quot-.html
Read this article, in layman's terms what does it mean. Sorry for my illiteracy.

It doesn't really mean much. It just means that on average, the slightly different instruction set on what this site claims is a Haswell CPU is marginally faster than the Ivy Bridge instruction set. Or in layman's terms, the Haswell CPU usually does slightly more work in the same time than Ivy Bridge under the same conditions.

What would be far more interesting is a comparison of power draw at that clock speed. If it uses significantly less power (as it is supposed to), it could mean it can overclock itself (ie Turbo Boost) higher than Ivy Bridge can, thereby increasing performance.

----------

If you compare with a hackintosh or Mac Pro, a mobile CPU literally does not stand a chance.

True, but only if you consider multi-CPU, or severe overclocking. And then the costs start to climb quickly. And if you stick with just overclocking, you'l still struggle to push much over 5 GHz, even with great cooling.
 

dusk007

macrumors 68040
Dec 5, 2009
3,411
104
Not necessarily true. TSX, which will be introduced with Haswell, offers better multi-core scaling with minimal effort on the part of developers. I'm not sure if this is where the ~10% figure is coming from, but my understanding is there will be significant improvements in performance and power efficiency.

http://www.anandtech.com/print/6290
I doubt TSX helps much in consumer workload. Locking isn't really the primary issue. That is a server thing. MC performance issues on consumer hardware are usually because one thread simply has too much work and not because threads are waiting for unlocks on shared data all the time.
 

ohbrilliance

macrumors 65816
May 15, 2007
1,010
355
Melbourne, Australia
I'm curious, are any of you finding yourselves constrained by the speed offered with current gen quad-core processors? If so, what kinds of tasks are you running to feel that the CPU is slowing you down?
 

Erasmus

macrumors 68030
Jun 22, 2006
2,756
298
Australia
I'm curious, are any of you finding yourselves constrained by the speed offered with current gen quad-core processors? If so, what kinds of tasks are you running to feel that the CPU is slowing you down?

Real time processing of video. Especially with code one has written themselves.

In fact, pretty much any area of Science or Engineering could always benefit greatly from more CPU beef. Especially any kind of finite element modelling, like stress analysis, fluid dynamics, or electro-magnetic models. Luckily they are also usually easy to multithread, but still, even simple examples can take hours to solve on even the best CPUs. Then, if you need to solve it in real-time, you can pretty much give up on anything short of a quantum computer. Which don't exist yet.
 

calvol

macrumors 6502a
Feb 3, 2011
995
4
Turbo's eat gas, as seen with Sandy and Ivy, anytime the weak GPU HD3000/4000 is under load, turbo kicks in and battery life falls.

Will be interested in GPU performance of Haswell, to see if HD5600 GPU is really 60% better, because HD4000 is a dog. If not better, will get a 15" with discrete GPU.
 

bill-p

macrumors 68030
Jul 23, 2011
2,889
1,550
It doesn't really mean much. It just means that on average, the slightly different instruction set on what this site claims is a Haswell CPU is marginally faster than the Ivy Bridge instruction set. Or in layman's terms, the Haswell CPU usually does slightly more work in the same time than Ivy Bridge under the same conditions.

Haswell would run the same "instruction set" as Ivy Bridge, because it is still an x86 CPU after all.

I think you're trying to say that Haswell can execute more instructions per clock cycle (IPC). Like you've said it yourself: do slightly more in the same time.

What would be far more interesting is a comparison of power draw at that clock speed. If it uses significantly less power (as it is supposed to), it could mean it can overclock itself (ie Turbo Boost) higher than Ivy Bridge can, thereby increasing performance.

TDP with Haswell actually remains the same or higher than Ivy Bridge:
http://www.cpu-world.com/news_2012/2012121601_Specifications_of_mobile_Haswell_CPUs.html

The reason is because Intel is trying to integrate more from the chipset into the CPU. So they're reducing power consumption of the chipset rather than of the CPU.

Plus you can't expect 50% GPU performance improvement to come without any power consumption penalty.

I have said this a while before, but Haswell is not a magical fix for whatever performance problem Ivy Bridge has now. It's a revolutionary cycle, not an evolutionary cycle.

If you're looking for drastic power consumption and performance improvements, wait until Intel releases their 14nm parts (Broadwell) in 2014.
 

Erasmus

macrumors 68030
Jun 22, 2006
2,756
298
Australia
Haswell would run the same "instruction set" as Ivy Bridge, because it is still an x86 CPU after all.

I think you're trying to say that Haswell can execute more instructions per clock cycle (IPC). Like you've said it yourself: do slightly more in the same time.

What I'm trying to say is that the algorithms it uses to do its calculations in the most efficient manner possible have been tweaked slightly.

TDP with Haswell actually remains the same or higher than Ivy Bridge:
http://www.cpu-world.com/news_2012/2012121601_Specifications_of_mobile_Haswell_CPUs.html

The reason is because Intel is trying to integrate more from the chipset into the CPU. So they're reducing power consumption of the chipset rather than of the CPU.

Plus you can't expect 50% GPU performance improvement to come without any power consumption penalty.

I have said this a while before, but Haswell is not a magical fix for whatever performance problem Ivy Bridge has now. It's a revolutionary cycle, not an evolutionary cycle.

If you're looking for drastic power consumption and performance improvements, wait until Intel releases their 14nm parts (Broadwell) in 2014.

I'm not really sure how Intel calculate their TDP values any more. My quad 2.2GHz i7 is at this very moment using 43.5 W, (less than its TDP of 45W) yet has all 8 virtual cores (loaded with 8x yes>/dev/null) running at 2.5 GHz, far above its default clock.
 
Last edited:

TallManNY

macrumors 601
Nov 5, 2007
4,742
1,594
I have said this a while before, but Haswell is not a magical fix for whatever performance problem Ivy Bridge has now. It's a revolutionary cycle, not an evolutionary cycle.

If you're looking for drastic power consumption and performance improvements, wait until Intel releases their 14nm parts (Broadwell) in 2014.

Do you have that sentence above backwards? I believe revolutions are considered dramatic and evolutions are considered gradual. Are you saying that Haswell is not a big change for Laptops? But that Broadwell is going to be where things really change?
 

dblissmn

macrumors 6502
Apr 30, 2002
353
107
The big story is the Haswell integrated graphics will be dramatically faster.

So the 13-inch models will be able to do what a discrete-graphics chip computer now does, and the 15-inch models will be able -- if Apple adjusts their firmware -- to do all kinds of good things without ever switching the discrete chip on.

The result is a big boost in performance for 13-inch models and longer battery life and cooler running temperatures for 15-inch models. In both cases increased integrated activity is probably offset by reduced strain on the CPU.

If you have a 15 and don't use the discrete chip as it is (e.g. not much serious graphical stuff, no external display) you probably won't notice much difference. Likewise if you have a 13 and aren't pushing the graphical capabilities much.
 

theuserjohnny

macrumors 6502
Jul 7, 2012
450
7
Yeah, but a 50% bump in the GPU?

The bump in integrated graphics isn't that big of a deal for the 15" considering the fact that they come with their own GPU.

The jump in integrated graphics which is expected to be roughly 50% is more of a big deal for Ultrabooks/Macbook Airs which don't use a GPU.
 

bill-p

macrumors 68030
Jul 23, 2011
2,889
1,550
What I'm trying to say is that the algorithms it uses to do its calculations in the most efficient manner possible have been tweaked slightly.

Algorithms are on the coding side. The CPU can only run whatever instruction it is fed. Namely... addition, subtraction, multiplication, division, read things here, write things there, move things from this to that... CPUs don't have "algorithms" built-in unless you consider basic mathematical operations "algorithms".

Haswell is just faster at doing those basic operations than Ivy Bridge.

I'm not really sure how Intel calculate their TDP values any more. My quad 2.2GHz i7 is at this very moment using 43.5 W, (less than its TDP of 45W) yet has all 8 virtual cores (loaded with 8x yes>/dev/null) running at 2.5 GHz, far above its default clock.

TDP is Thermal Design Power. On one hand, it's just how much thermal the CPU is expected to push out, so power consumption is actually higher than that, still. You're looking at roughly 50W. It's the absolute worst case scenario with the processor completely bogged down.

I'd guess your Core i7 quad at 2.2GHz is a 2720QM (2011 MBP model). That one can actually Turbo Boost to 3.0GHz, so you're just a bit shy of maximum load.

In any case, the story is that Haswell doesn't improve on power consumption.w

Do you have that sentence above backwards? I believe revolutions are considered dramatic and evolutions are considered gradual. Are you saying that Haswell is not a big change for Laptops? But that Broadwell is going to be where things really change?

Unless you're saying human evolution (from "monkeys", according to popular theories), doesn't mean as much as human revolution (winning a conflict of political interests), I think it's fair to say "evolution" is the dramatic change.

And yeah, I am saying that Haswell isn't a big change for laptops. It's a big change for those who are reliant on the integrated graphics solution that are looking for more performance in that area, but it's not a huge change anywhere else.

The manufacturing process is still the same 22nm that's used with Ivy Bridge, for one thing.

Broadwell moves to a new manufacturing process (14nm). Typically, when that happens, things improve dramatically. That's why Ivy Bridge (22nm) is dramatically faster and more efficient than Sandy Bridge (32nm).
 

dusk007

macrumors 68040
Dec 5, 2009
3,411
104
Unless you're saying human evolution (from "monkeys", according to popular theories), doesn't mean as much as human revolution (winning a conflict of political interests), I think it's fair to say "evolution" is the dramatic change.
No you do have it backwards. Evolution may lead to significant change at some point but it is gradual. It builds upon the old and only changes a little each generation.
Revolution is chop of everybody's head we are doing it completely different than it used to be. Revolution is something entirely new. It is a big sudden change.

Evolution is anything that is better than the previous however little. Revolution would be if they pull graphene transistors or something actually new out of their butts. Diamond chips or something that would be easier to cool.
 

Ploki

macrumors 601
Jan 21, 2008
4,308
1,558
No you do have it backwards. Evolution may lead to significant change at some point but it is gradual. It builds upon the old and only changes a little each generation.
Revolution is chop of everybody's head we are doing it completely different than it used to be. Revolution is something entirely new. It is a big sudden change.

Evolution is anything that is better than the previous however little. Revolution would be if they pull graphene transistors or something actually new out of their butts. Diamond chips or something that would be easier to cool.

Actually revolution = not necessarily better, just new.
Evolution doesn't necessary have to be slow you know? It just happens that evolution of life was slow.
Graphene transistors would be in fact evolution. Except if you would change everything you know about a computer...
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.