Mobile & Desktop Processor comparison

[PannaKO]

I'm looking at getting a Macbook Pro and would the processor to be as good or better than my last 3.06Ghz Dual Core iMac.

Now, I thought that a mobile i7 processor would be as powerful as a desktop i7 if they ever had the same number of Ghz.
But after trying to research further, I find that this might not be the case.

Am I mistaken to assume this?
If so, what performance differences would be seen during application and how/why?

 

[saturnotaku]

I'm looking at getting a Macbook Pro and would the processor to be as good or better than my last 3.06Ghz Dual Core iMac.

Due to thermal constraints of a laptop versus a desktop, mobile processors are not going to be as powerful. That being said, a current 2.2 GHz MacBook Pro with Retina Display will blow the doors off your iMac. CPU speed as expressed in its numeric form is not indicative of overall performance when comparing different generations.

 

[PannaKO]

Due to thermal constraints of a laptop versus a desktop, mobile processors are not going to be as powerful. That being said, a current 2.2 GHz MacBook Pro with Retina Display will blow the doors off your iMac. CPU speed as expressed in its numeric form is not indicative of overall performance when comparing different generations.

Thanks for your response and the extra bit of information there.

Is that sort of the same reason why, for example, a 3Ghz mobile CPU will be slower than a desktop 3Ghz because they will run in a different way?
I know that mobile CPU's have to run cooler and account for a limited power supply, but I still thought that the GHz power number would have the same meaning because if the mobile CPU was slower they would say it's 2.6GHz instead

 

[cbautis2]

Thanks for your response and the extra bit of information there.

Is that sort of the same reason why, for example, a 3Ghz mobile CPU will be slower than a desktop 3Ghz because they will run in a different way?
I know that mobile CPU's have to run cooler and account for a limited power supply, but I still thought that the GHz power number would have the same meaning because if the mobile CPU was slower they would say it's 2.6GHz instead

There's a reason mobile CPU have high turbo boost and low normal frequency. The turbo boost is completely dependent on CPU power package and temperatures. Some apps don't use all cores or only use all cores but not fully utilizing them and as such CPU can fully utilize its maximum package power if Turbo boost is enabled e.g. at least one core isn't being utilized. Turbo Boost will indefinitely run at highest GHz as long as it meets the maximum CPU package power. If all cores and threads are fully utilized, the CPU will then revert to normal frequency since maximum package power consumption is already reached i.e. all cores/threads are 100% utilized. Try prime 95 app to see that the GHz quickly drops down to normal or 0.1 - 0.2 GHz above normal even though it should've been at maximum turbo speed.

Desktop CPUs can change the package power through overclocking so they're able to maintain their high GHz even when using Prime 95. That's the biggest difference between Desktop and laptop CPU.

That's why Desktop CPUs are preferred for long CPU session since they don't throttle due to power or heat limitations.

 

[shoehornhands]

Depending on whether you're looking at the 13" or the 15" MacBook Pro, the processors in the current MacBook Pro are not too far behind the desktop chips.

Remember, the 13" MacBook pro has a dual-core processor while the 15" has a quad-core processor. Clock-for-clock, there really isn't much difference between the laptop and desktop chips.

Look at the fastest processor option in the 15" MacBook Pro. It's an Intel i7 4980HQ and has the following specs:

# of Cores: 4
# of Threads: 8
Base Frequency: 2.8 GHz
Max Turbo Frequency: 4.0 GHz
Cache: 6 MB
TDP: 47 W

Now compare that to the fastest consumer desktop processor, the i7 4790K (found in the retina iMac):

# of Cores: 4
# of Threads: 8
Base Frequency: 4.0 GHz
Max Turbo Frequency: 4.4 GHz
Cache: 8 MB
TDP: 88 W

Also note that both chips are built on the same Haswell microarchitecture and have all the same technologies so the performance at a given frequency should be very similar.

As you can see, the most significant difference is in the base frequency. For short tasks that require a lot of processing power, the gap between desktop and laptop CPU is even narrower due to the 4.0 GHz turbo frequency. The difference is that the mobile chip isn't going to run at 4.0 GHz for a sustained period of time like the desktop chip (it's not that the chip itself isn't capable of it but more the heat limitations that come with a small MacBook Pro enclosure). The chip will turbo until it reaches its heat ceiling and then start throttling back.

Point being, the more significant difference is with the processors in the 13" MacBook Pro because it has half as many cores (although they are clocked higher). The quad-core chips in the 15" version are awesome chips and in most situations, you'd likely have a difficult time distinguishing them from the desktop versions.

You can see an actual performance comparison here: http://browser.primatelabs.com/mac-benchmarks

At the top is the 27" Retina iMac with the i7 4790K and fifth from the top is the current 15" MacBook Pro with the processor upgrade. As you can see, the iMac is about 10% faster than the MacBook Pro and the iMac is clocked 10% higher (4.4 GHz vs 4.0 GHz).

 

[saturnotaku]

Remember, the 13" MacBook pro has a dual-core processor while the 15" has a quad-core processor.

Dual-core and low-voltage in the 13-inch Retinas, which will also impact performance relative to a desktop counterpart (the 13-inch "classic" still uses a full-voltage dual-core). Only 15-inch versions have full-voltage CPUs now.

 

[UBS28]

Due to thermal constraints of a laptop versus a desktop, mobile processors are not going to be as powerful. That being said, a current 2.2 GHz MacBook Pro with Retina Display will blow the doors off your iMac. CPU speed as expressed in its numeric form is not indicative of overall performance when comparing different generations.

The Retina Macbook Pro has a geekbench score in the range of 13k-14k which is faster than alot of desktop processors including some Retina iMac's. In single core performance, it's even the 4th fastest Mac.

 

[PannaKO]

@cbautis2 - your explanation was a little over my head , but after looking up the terminologies I'm confident I understand it all now .

@shoehornhands - yes, I was planning to get the 2.8GHz rMBP so I'm glad you and @saturnotaku say it's much more powerful than what had before.
Although I do wonder how much a lower base clock speed will affect typical use, I think it'll suit my needs well enough.

Thanks for the responses all, you've helped me make a better informed purchase.

 

[SarcasticJoe]

Thanks for your response and the extra bit of information there.

Is that sort of the same reason why, for example, a 3Ghz mobile CPU will be slower than a desktop 3Ghz because they will run in a different way?
I know that mobile CPU's have to run cooler and account for a limited power supply, but I still thought that the GHz power number would have the same meaning because if the mobile CPU was slower they would say it's 2.6GHz instead

Talking about clock frequencies when you're comparing different chips is really not very smart because different chips in different environments get different amounts of work done per clock cycle. Even manufacturers have started talking about their "IPC", or "Instructions Per Clock", figures and how they've improved it in their new chips.

Memory latency is actually these days a pretty big factor in performance as CPU's operate a lot faster than their memory, which means that when you up the clock frequency, even on the same chip, you don't get anywhere near the same amount of improvement in actual performance. Another thing that has become more important is superscalarity, which basically means that a CPU core can run different stages of different instructions at once (i.e practically several instructions at the same time).

Looking blindly at clock frequencies is like blindly looking at max RPM figures on engines. A motorcycle engine running at 10.000 rpm isn't going to produce twice the power as big diesel powered tractor or lorry engine at 5.000 RPM.

...

There's just one problem with your comparison... The 4980HQ, i.e the laptop chip, is sold to manufacturers in trays of 1000 units at a price of $623 each while the 4790k, the desktop chip, is sold in the same way at almost half the price, $339 each. A better comparison would probably the i7 5930K, which is sold at $583 each in a tray and scores about 30% higher at Passmark (which is probably the best CPU benchmark around).

Here's the laptop part
Here's the proper desktop desktop comparison

 

[snaky69]

Thanks for your response and the extra bit of information there.

Is that sort of the same reason why, for example, a 3Ghz mobile CPU will be slower than a desktop 3Ghz because they will run in a different way?
I know that mobile CPU's have to run cooler and account for a limited power supply, but I still thought that the GHz power number would have the same meaning because if the mobile CPU was slower they would say it's 2.6GHz instead

The days where the clock speed (Measured in Hz) is the only measure of performance are long gone. Ever since the move to small die sizes and multi-core processors, they have been next to useless in measuring performance.

For example, the two MBP's listed in my signature. The first has a Penryn (Intel code name) 2.4Ghz Core 2 Duo. The second has an Ivy Bridge 2.6Ghz quad-core i7.

The difference between the two in terms of clock speed is only .2Ghz, yet my current machine is a tad more than 4 times more powerful.

The difference between desktop and mobile silicon is mostly in power consumed and heat output. A mobile computer has a limited electricity supply and cannot be scorching hot or it won't be used.

With a desktop, you have a case full of air to move around and cool things, and very powerful power supplies, so those are not a limiting factor. For any given generation of processor, the desktop variant will invariably be faster.

With that said, any current generation CPU will blow your old core 2 duo out of the water... unless what you do with a computer is not limited by CPU power(a rule of thumb with computers: if you don't know if you need more CPU power, you don't need more). If you are not CPU bound, you won't notice a difference. The performance will be affected a whole lot more by the fact that you'd be moving to a SSD.

 

[PannaKO]

@snaky69 - thanks for explaining further and advice.

Just to be clearer with my original question, I wanted to compare theoretically the same CPU's in architecture, core count, clock speed, etc., but in mobile and desktop variants and not my old processor.
I guess I commented on my old desktop CPU just in case mobile CPU's were governed by such different rules that I would need to reconsider whether to upgrade to a new desktop or laptop.

If I understand everyone correctly o), of the two theoretical CPU's, the mobile variant could be almost as powerful as the desktop (if all their working conditions were the same), but for the mobile's form factor and restrictions in power and heat.

I consider myself a pretty average user, although I do some video encoding fairly regularly.
But it's good to know, at the sight of current CPU's my old desktop would explode from the oceans with doors fully detached .

 

[yjchua95]

@snaky69 - thanks for explaining further and advice.

Just to be clearer with my original question, I wanted to compare theoretically the same CPU's in architecture, core count, clock speed, etc., but in mobile and desktop variants and not my old processor.
I guess I commented on my old desktop CPU just in case mobile CPU's were governed by such different rules that I would need to reconsider whether to upgrade to a new desktop or laptop.

If I understand everyone correctly o), of the two theoretical CPU's, the mobile variant could be almost as powerful as the desktop (if all their working conditions were the same), but for the mobile's form factor and restrictions in power and heat.

I consider myself a pretty average user, although I do some video encoding fairly regularly.
But it's good to know, at the sight of current CPU's my old desktop would explode from the oceans with doors fully detached .

Based on Geekbench 3 scores here, the mobile 2.8GHz i7-4980HQ (mid-2014 15" rMBP) and the 3.5GHz i7-4771 (late-2013 non retina 27" iMac) perform almost identically: http://browser.primatelabs.com/geekbench3/compare/1079177?baseline=1580504

And yes, GHz isn't everything either, given the same number of cores and threads. Look here (the same i7-4980HQ compared with a 3.1GHz i7-4770S in a 21.5" iMac, both being Haswell): http://browser.primatelabs.com/geekbench3/compare/611544?baseline=1580504

That said, in sustained tasks, the iMac would win by a long shot because the rMBP would run into throttling issues sooner than the iMac.

 

[PannaKO]

@yjchua95 - thanks for the links, interesting results. I assume these are strongly relevant to short tasks that need CPU power and maybe typical daily use?

But, as video encoding is a sustained computing task and though, like you say, it won't keep up with its current desktop counterparts in this regard...
With the desktop I had before being so (sooo) dated, I can hope it's still/will be the case that I can expect the 2.8GHz rMBP to comfortably out perform it despite its mobile limitations?

 

[dusk007]

For any kind of sustained high workload desktops are far better for another reason than just speed. They stay quiet even under full load. Notebooks are quiet when there is little work only very short bursts but load the CPU fully and they are annoying and get hot too.
A Desktop doesn't care even under full load it won't be annoyingly loud and the keyboard stays cool.

Anyway if you have the same generation like 4XXX CPUs they are essentially the same chips only running at different clocks for mobile and desktop. The power curve rises exponentially with clock speed which is why there is this limit at about 3.5-4.5Ghz. That is why Turbos are fairly close. More and more cores just add linearly to the power, which is why the huge server processors have 8-16 cores.
If you don't go iMac you can get a lot more than 4 cores if you have need for them. There is no need to pay Mac Pro prices either.

 

[yjchua95]

@yjchua95 - thanks for the links, interesting results. I assume these are strongly relevant to short tasks that need CPU power and maybe typical daily use?

But, as video encoding is a sustained computing task and though, like you say, it won't keep up with its current desktop counterparts in this regard...
With the desktop I had before being so (sooo) dated, I can hope it's still/will be the case that I can expect the 2.8GHz rMBP to comfortably out perform it despite its mobile limitations?

I did a 4K video rendering once which was sustained for 45 mins on my late-2013 15" rMBP (2.6GHz i7-4960HQ), and I saw that with the fans at full blast and the back end of the laptop lifted up to allow better heat dissipation, throttling was very little. In fact, the lowest clock speed it was operating at was around 3.3GHz, and running at 3.5GHz the rest of the time.

I used Intel Power Gadget to monitor it and also modified a utility called Macoh to produce a graph showing the processor activity throughout the process.

 

[PannaKO]

@dusk007 - thanks for the advice and mentioning Turbo speeds too, I did wonder why they were so similar in speed.

My use is pretty casual so I wouldn't think I need more than 4 cores. My preference would have been to get the new riMac, but I'm not comfortable buying this latest iteration.
I don't know how loud fans can be in current rMBP's under full load, but I'm not usually one to be annoyed by fan noise, however my concern would be more to do with me stressing it too much, too often with heat, causing atypical wear and lifespan shortening.
So, while a the price of a max spec rMBP (virtually over a year old, with broadwell ones likely coming in this summer, which feels too long for me to wait and my use being pretty casual) isn't ideal - I figure it have to do as a stopgap that'll likely last long enough (spec-wise) for a riMac iteration that I have confidence in, where I can then use the laptop as a backup machine.

@yjchua95 - I appreciate you sharing your experience. It's good news, as I see myself encoding only at a max resolution of full HD and less than HD most of the time.

*I was already planning to get something to add height to the MBP mostly for human ergonomic reasons, but also to possibly help with cooling, so I'm glad you think that it makes a difference .