Wrong. It has 2Cores/4Threads (hyperthreading makes 2 cores appear and work as 4 when needed) while crappy Celeron has 1 core per thread for 4 threads in total.
(
xkcd.com)
You seem to have hyperthreading and number-of-cores completely back to front.
Start by comparing like-with-like: The Celeron N2940 with 4 cores and the N2840 from the same 'family' but with only 2 cores.
See
Intel's pages on the N2940 and
N2840 and click on the '# of cores' label for help:
Cores is a hardware term that describes the number of independent central processing units in a single computing component (die or chip).
The N2840 has 2 cores - 2 independent CPUs - each running at 1.83 GHz. Each core can only run 1 thread (i.e. 1 sequence of instructions) at a time, so it is 4 cores, 4 threads.
The N2940 has 4 cores -
4 independent CPUS - each equivalent to one of the cores in the 2840. Its
not the same core subdivided into quarters rather than halves. Software that is written to take advantage of multi-threading (that's important) will be able to use all 4 cores to work on different parts/steps of the problem simultaneously and get the job done almost twice as fast as on a N2840 (it can't be fully twice as fast because there are other bottlenecks and system processes competing for CPU time).
Lots - most - software isn't really optimised for multi-threading, and will only use a single thread, so it won't see any advantage of 4 cores over 2 cores (although the system as a whole will benefit from better multi-tasking) but video processing is probably the biggest use-case for quad-core and beyond and usually makes the most of it (certainly, Handbrake is one of the few things that lights up all the cores of my i7).
Hyperthreading is a trick that lets 1 core run 2 threads using 'idle time' on the various components that go to make up a modern CPU core. Obligatory car analogy: its like having a single car production line producing two completely different car models at the same time - much more efficient than stopping the line, changing model, and re-starting (c.f. regular multi-tasking) but nowhere near as fast as having a completely separate production line for each model (c.f. multi-core). Its also going to be very, very dependent on the nature of the code running, and there are other ways the core could be using that idle time (speculative execution etc.) to accelerate its single-thread performance.
The processor in the MBA only has 2 physical cores, but uses hyperthreading to make that look like 4. I.e. 2 cores, 4 threads. That is
not going to give the same performance as a proper 4 core, 4 thread chip with comparable cores.
As you say, comparing different CPU models is hard (and Intel's branding scheme is deliberately designed to make objective comparisons difficult) but the individual cores in a 1.4GHz i5 are not going to have vastly more raw grunt than a 1.8GHz celeron. Refinements in core design produce incremental improvements in work-per-GHz and power consumption (1.4Ghz now as good as 1.8 GHz then seems feasible). Effective use of multiple cores can easily double performance on multi-threaded tasks
I do realize my MBA is 2 years old, but still... If I buy the current MBA, will it be much faster?
Probably not, and certainly not 'night & day'. Video transcoding really is the killer app for quad- or better core processors. If the difference between 4 hours and 7 hours is costing you money then you've got a business case for a quad i7 machine. Otherwise... just leave it running overnight (Handbrake has a handy 'suspend computer when finished' option).
Yeah, "4 'real' cores" beats "2 slightly better cores + hyperthreading" is pretty much to be expected. However, I'm surprised by the
huge margin by which your Lenovo beats the Air, so I would look into other contributing factors (Time Machine backup running in the background? SSD on the Air getting full up vs. 'virgin' ssd on the PC? Compressing to/from an external drive?)
In contrast, I've recently done some M2TS -> MP4 transcodes using Handbrake, with similar settings, on both my 2011 2.2GHz i7 MBP and a 2014 Lenovo W540 i7 2.5GHz, getting about 60 fps on the Mac and 70 fps on the Lenovo, which is about what you'd expect from a 10% clock-speed bump and Sandy Bridge -> Haswell.
(Plus, while the W540 certainly gave better specs-per-buck, its a horrible plasticky, sharp-edged lump with a horrible screen alongside the MBP).
[doublepost=1457701355][/doublepost]
Wrong. It has 2Cores/4Threads (hyperthreading makes 2 cores appear and work as 4 when needed) while crappy Celeron has 1 core per thread for 4 threads in total.
(
xkcd.com)
You seem to have hyperthreading and number-of-cores completely back to front.
Start by comparing like-with-like: The Celeron N2940 with 4 cores and the N2840 from the same 'family' but with only 2 cores.
See
Intel's pages on the N2940 and
N2840 and click on the '# of cores' label for help:
Cores is a hardware term that describes the number of independent central processing units in a single computing component (die or chip).
The N2840 has 2 cores - 2 independent CPUs - each running at 1.83 GHz. Each core can only run 1 thread (i.e. 1 sequence of instructions) at a time, so it is 4 cores, 4 threads.
The N2940 has 4 cores -
4 independent CPUS - each equivalent to one of the cores in the 2840. Its
not the same core subdivided into quarters rather than halves. Software that is written to take advantage of multi-threading (that's important) will be able to use all 4 cores to work on different parts/steps of the problem simultaneously and get the job done almost twice as fast as on a N2840 (it can't be fully twice as fast because there are other bottlenecks and system processes competing for CPU time).
Lots - most - software isn't really optimised for multi-threading, and will only use a single thread, so it won't see any advantage of 4 cores over 2 cores (although the system as a whole will benefit from better multi-tasking) but video processing is probably the biggest use-case for quad-core and beyond and usually makes the most of it (certainly, Handbrake is one of the few things that lights up all the cores of my i7).
Hyperthreading is a trick that lets 1 core run 2 threads using 'idle time' on the various components that go to make up a modern CPU core. Obligatory car analogy: its like having a single car production line producing two completely different car models at the same time - much more efficient than stopping the line, changing model, and re-starting (c.f. regular multi-tasking) but nowhere near as fast as having a completely separate production line for each model (c.f. multi-core). Its also going to be very, very dependent on the nature of the code running, and there are other ways the core could be using that idle time (speculative execution etc.) to accelerate its single-thread performance.
The processor in the MBA only has 2 physical cores, but uses hyperthreading to make that look like 4. I.e. 2 cores, 4 threads. That is
not going to give the same performance as a proper 4 core, 4 thread chip with comparable cores.
As you say, comparing different CPU models is hard (and Intel's branding scheme is deliberately designed to make objective comparisons difficult) but the individual cores in a 1.4GHz i5 are not going to have vastly more raw grunt than a 1.8GHz celeron. Refinements in core design produce incremental improvements in work-per-GHz and power consumption (1.4Ghz now as good as 1.8 GHz then seems feasible). Effective use of multiple cores can easily double performance on multi-threaded tasks
I do realize my MBA is 2 years old, but still... If I buy the current MBA, will it be much faster?
Probably not, and certainly not 'night & day'. Video transcoding really is the killer app for quad- or better core processors. If the difference between 4 hours and 7 hours is costing you money then you've got a business case for a quad i7 machine. Otherwise... just leave it running overnight (Handbrake has a handy 'suspend computer when finished' option).
Yeah, "4 'real' cores" beats "2 slightly better cores + hyperthreading" is pretty much to be expected. However, I'm surprised by the
huge margin by which your Lenovo beats the Air, so I would look into other contributing factors (Time Machine backup running in the background? SSD on the Air getting full up vs. 'virgin' ssd on the PC? Compressing to/from an external drive?)
In contrast, I've recently done some M2TS -> MP4 transcodes using Handbrake, with similar settings, on both my 2011 2.2GHz i7 MBP and a 2014 Lenovo W540 i7 2.5GHz, getting about 60 fps on the Mac and 70 fps on the Lenovo, which is about what you'd expect from a 10% clock-speed bump and Sandy Bridge -> Haswell.
(Plus, while the W540 certainly gave better specs-per-buck, its a horrible plasticky, sharp-edged lump with a horrible screen alongside the MBP).
