Core M Air vs Haswell Air

[locksmack]

Hey Guys,

Does anyone know of any comparisons I can read that bench a likely Core M processor in the (possibly) upcoming rMBA against the current MBA Haswell processors?

I can't find a lot, but I am getting the impression that in terms of raw performance, a Core M MBA would be a downgrade compared to what is currently available. That would be a shame.

Cheers.

 

[leman]

It depends on wha you mean by "raw performance". Rendering or video editing: true, Core M might be a downgrade. Internet browsing, office etc. — most likely an upgrade. The Broadwell Core M CPUs are much better at short-time rapid overlocking (i.e. burst performance), what makes them very suitable for common workflows. Professionals that are looking for a more sustained performance will have to look otherwise, but sustained performance has never been the point of the MBA.

 

[motrek]

It depends on wha you mean by "raw performance". Rendering or video editing: true, Core M might be a downgrade. Internet browsing, office etc. — most likely an upgrade. The Broadwell Core M CPUs are much better at short-time rapid overlocking (i.e. burst performance), what makes them very suitable for common workflows. Professionals that are looking for a more sustained performance will have to look otherwise, but sustained performance has never been the point of the MBA.

It wasn't the point before, but people get used to things quick. Until recently the MBAs didn't have especially good battery life, but now it's excellent, and people are always talking about how the idea behind the MBA is long battery life.

If the MBA takes a step backwards in performance (at doing any task) then there will be some people who get upset.

 

[mi7chy]

Intel claims Broadwell architecture is 5% faster than Haswell.

From my experience with fanless Broadwell Core M 5Y10 (800MHz base/2GHz turbo) and active fan cooling i5-4300U (1.9GHz base/2.9 GHz turbo):

- Core M CPU performance is ~20% slower than i5 with Turboboost enabled
- Core M is slightly faster than i5 with Turboboost disabled
- Core M delivers consistent performance with CPU and iGPU utilization maxed out whereas i5 with Turboboost enabled throttles significantly as it warms up

Core M is surprisingly very snappy and usable at 1080p and the benefits of fanless are no noise when doing video chat, better reliability as there's no spinning fan that will clog up with dust over time, thinner profile as there's no need for active cooling and longer battery life. So impressed that I'm try to downclock the i5 to make it fanless but so far even with Turboboost disabled the fan still kicks in although at low RPM. Characteristics of Core M make it ideal for Macbook Air 2015.

 

[motrek]

Intel claims Broadwell architecture is 5% faster than Haswell.

From my experience with fanless Broadwell Core M 5Y10 (800MHz base/2GHz turbo) and active fan cooling i5-4300U (1.9GHz base/2.9 GHz turbo):

- Core M CPU performance is ~20% slower than i5 with Turboboost enabled
- Core M is slightly faster than i5 with Turboboost disabled
- Core M delivers consistent performance with CPU and iGPU utilization maxed out whereas i5 with Turboboost enabled throttles significantly as it warms up

Core M is surprisingly very snappy and usable at 1080p and the benefits of fanless are no noise when doing video chat, better reliability as there's no spinning fan that will clog up with dust over time, thinner profile as there's no need for active cooling and longer battery life. So impressed that I'm try to downclock the i5 to make it fanless but so far even with Turboboost disabled the fan still kicks in although at low RPM. Characteristics of Core M make it ideal for Macbook Air 2015.

Not sure which computers you're using to make these comparisons but I don't think your observations apply to either current MBAs or hypothetical future MBAs with Core Ms.

First of all, current MBAs have a ton of thermal headroom; they don't experience thermal throttling and there certainly isn't any fan noise when video conferencing (beyond the essentially inaudible nominal fan speed of 1200 RPM).

Second, reviews of the Yoga 3 (with a Core M) show that it does experience thermal throttling even though it has a fan.

It sounds like your data was collected using a Haswell with poor cooling (maybe a Surface Pro 3?) and a Core M with excellent cooling (i.e., a huge heatsink, since it doesn't have a fan).

 

[Boyd01]

I have a 2013 11" i7 MBA. It can rip a DVD in Handbrake about 5x faster than my 2008 15" core 2 duo MBP (15 minutes vs 75 minutes, for example). But the fans kick into high gear almost immediately when I start ripping.

Sounds like a fanless Core M MBA is not going to be any match, and that would be a shame.

 

[maflynn]

From what I've read, Core M doesn't really represent any performance gains, and in some cases the apps may run slower. I forget where I read it, but the performance gains appear to be disappointing.

 

[motrek]

From what I've read, Core M doesn't really represent any performance gains, and in some cases the apps may run slower. I forget where I read it, but the performance gains appear to be disappointing.

Core M isn't supposed to be faster. It's an extremely low power part targeted for cheap, fanless tablets. It's supposed to be slower. Not that there's anything wrong with that, but there's really no reason why Apple would want to use it in a MacBook.

 

[cbautis2]

Intel claims Broadwell architecture is 5% faster than Haswell.

From my experience with fanless Broadwell Core M 5Y10 (800MHz base/2GHz turbo) and active fan cooling i5-4300U (1.9GHz base/2.9 GHz turbo):

- Core M CPU performance is ~20% slower than i5 with Turboboost enabled
- Core M is slightly faster than i5 with Turboboost disabled
- Core M delivers consistent performance with CPU and iGPU utilization maxed out whereas i5 with Turboboost enabled throttles significantly as it warms up

Core M is surprisingly very snappy and usable at 1080p and the benefits of fanless are no noise when doing video chat, better reliability as there's no spinning fan that will clog up with dust over time, thinner profile as there's no need for active cooling and longer battery life. So impressed that I'm try to downclock the i5 to make it fanless but so far even with Turboboost disabled the fan still kicks in although at low RPM. Characteristics of Core M make it ideal for Macbook Air 2015.

Unfair comparison. You just compared a higher tier i5 ulv to an entry level core m.

Compare the higher end intel core m 5y71 to the macbook air i5 4250u then you can see it's only less than 10% less than the i5 4250u at all core load while being much faster at single core load

 

[mi7chy]

Unfair comparison. You just compared a higher tier i5 ulv to an entry level core m.

Compare the higher end intel core m 5y71 to the macbook air i5 4250u then you can see it's only less than 10% less than the i5 4250u at all core load while being much faster at single core load

Nothing unfair about it. Actually like the 5Y10 a lot since it's very snappy at 1080p and fanless so ended up buying two Helix2. 5Y71 might not be fanless considering 5Y70 in the Yoga 3 isn't.

 

[powersteer]

Not a great difference actually.

http://browser.primatelabs.com/geekbench3/1673550
vs
http://browser.primatelabs.com/geekbench3/1631808

Single Core: 2128 -> 1955
Multi Core: 3597 -> 3257

Unlikely users can feel the difference. But still a regression though

 

[motrek]

Not a great difference actually.

http://browser.primatelabs.com/geekbench3/1673550
vs
http://browser.primatelabs.com/geekbench3/1631808

Single Core: 2128 -> 1955
Multi Core: 3597 -> 3257

Unlikely users can feel the difference. But still a regression though

The problem is thermal headroom. As long as both processors can stay cool, and they're both plugged into the wall, the Core M should perform as well as a current Haswell, or even a little better.

But if either one gets too hot (i.e., hits thermal throttling), it will run half as fast, or maybe less.

So these benchmark results are not a surprise. If memory serves, Geekbench only needs a minute or two to run. And the Core M score is for a Lenovo, probably a Yoga 3, which has a fan. So the fact that this Core M, with a fan, is able to keep up with a Haswell for a minute or two is not astonishing. And for all we know, the person who ran the benchmark did so in an environment with a cold ambient temperature.

You run into problems if you put a Core M into a machine with inadequate cooling, e.g., the rumored super-thin fanless 12-inch MBA. Such a machine would almost certainly hit thermal throttling within a minute or two of doing something CPU-intensive, and thus run half as fast as a current MBA until it cooled down. And without a fan, it might take a while for it to cool down. I'm not willing to make that sacrifice in CPU power just so my laptop can be a tiny fraction of an inch thinner.

 

[objektør]

The problem is thermal headroom. As long as both processors can stay cool, and they're both plugged into the wall, the Core M should perform as well as a current Haswell, or even a little better.

But if either one gets too hot (i.e., hits thermal throttling), it will run half as fast, or maybe less.

So these benchmark results are not a surprise. If memory serves, Geekbench only needs a minute or two to run. And the Core M score is for a Lenovo, probably a Yoga 3, which has a fan. So the fact that this Core M, with a fan, is able to keep up with a Haswell for a minute or two is not astonishing. And for all we know, the person who ran the benchmark did so in an environment with a cold ambient temperature.

You run into problems if you put a Core M into a machine with inadequate cooling, e.g., the rumored super-thin fanless 12-inch MBA. Such a machine would almost certainly hit thermal throttling within a minute or two of doing something CPU-intensive, and thus run half as fast as a current MBA until it cooled down. And without a fan, it might take a while for it to cool down. I'm not willing to make that sacrifice in CPU power just so my laptop can be a tiny fraction of an inch thinner.

This is why I don't expect a rMBA with Core M, because of thermal throttling.
But can the Broadwell U i5 do the jobb? If it is only marginally faster than the Haswell U i5 why should Broadwell manage intensive tasks over time without throttling if there is no fan? They are both about 15W I believe. Maybe it helps that it has a better GPU but will it be enough for a retina screen? Is there a new way of passive cooling? Otherwise I don't see a fanless rMBA happening this round, which is fine for me. Performance rather than 'thin/thickness'.

 

[cbautis2]

This is why I don't expect a rMBA with Core M, because of thermal throttling.
But can the Broadwell U i5 do the jobb? If it is only marginally faster than the Haswell U i5 why should Broadwell manage intensive tasks over time without throttling if there is no fan? They are both about 15W I believe. Maybe it helps that it has a better GPU but will it be enough for a retina screen? Is there a new way of passive cooling? Otherwise I don't see a fanless rMBA happening this round, which is fine for me. Performance rather than 'thin/thickness'.

Keep in mind that broadwell core m tdp is only 4.5 watts which is the tdp for intel bay Trail-T CPU that is used for Win/android tablet. You're paying for ver close to equal haswell 4250 ulv performance with the power consumption of an atom CPU!

 

[motrek]

This is why I don't expect a rMBA with Core M, because of thermal throttling.
But can the Broadwell U i5 do the jobb? If it is only marginally faster than the Haswell U i5 why should Broadwell manage intensive tasks over time without throttling if there is no fan? They are both about 15W I believe. Maybe it helps that it has a better GPU but will it be enough for a retina screen? Is there a new way of passive cooling? Otherwise I don't see a fanless rMBA happening this round, which is fine for me. Performance rather than 'thin/thickness'.

No, just think of Broadwell as the same as Haswell but it uses 30% less power. If you switched current MacBooks to Broadwell, it would run a little cooler and you might get 1-2 hours of extra battery life but otherwise they'd be the same. You'd still need active cooling for a 15W chip in a laptop.

----------

Keep in mind that broadwell core m tdp is only 4.5 watts which is the tdp for intel bay Trail-T CPU that is used for Win/android tablet. You're paying for ver close to equal haswell 4250 ulv performance with the power consumption of an atom CPU!

No. TDP is Thermal Design Power.

That means if you want to build a computer and use this chip, Intel recommends that you give it enough cooling to handle 4.5 watts.

That doesn't mean it will actually only use 4.5W, and it doesn't mean it will perform well with that amount of cooling.

The Yoga 3 can almost certainly cool 4.5W, probably more since it has a fan, but its CPU still hits thermal throttling pretty quickly and performs like crap vs. a MBA.

 

[mutsaers-vr.nl]

What is Yoga 3 ?

edit : i found it the Lenevo laptop

 

[mutsaers-vr.nl]

so the upcoming broadwell i7-5650U is much slower than the current i7 haswell processor of the MBA ???

 

[cbautis2]

No, just think of Broadwell as the same as Haswell but it uses 30% less power. If you switched current MacBooks to Broadwell, it would run a little cooler and you might get 1-2 hours of extra battery life but otherwise they'd be the same. You'd still need active cooling for a 15W chip in a laptop.

----------



No. TDP is Thermal Design Power.

That means if you want to build a computer and use this chip, Intel recommends that you give it enough cooling to handle 4.5 watts.

That doesn't mean it will actually only use 4.5W, and it doesn't mean it will perform well with that amount of cooling.

The Yoga 3 can almost certainly cool 4.5W, probably more since it has a fan, but its CPU still hits thermal throttling pretty quickly and performs like crap vs. a MBA.

It doesn't even hit its full turbo on a lot of those benchmarks since it BIOS limited! Check all of Lenovo benchmark and you'll see that Lenovo BIOS disables turbo boost (it never even reach their max frequency before throttling) on their i7 laptops when using constant CPU load or concurrent CPU and GPU load such as gaming. That's the reason I never trust any benchmark scores from Lenovo laptops.

What about those very thin bay Trail-T tablets with 4.5 watt tdp? They run very well and can certainly run boost mode all day long while being cooler than krait arm chips which throttles after a while.

Apple as we know don't throttle their laptops as aggressively as a lot of pc manufacturers hence the triple digit temps that some users are reporting.

Core m in the air would certainly run hot like all macbooks under load but would throttle far less aggressive than pcs and as such perform as good or in some cases better than lower end haswell 15 watt ulv (i.e. 4200/10 u or 4250/60u which also throttle heavily on PC laptops but not much on macs)

----------

so the upcoming broadwell i7-5650U is much slower than the current i7 haswell processor of the MBA ???

Nope, it's at least 10% faster since it's tdp is 15 watts which is the same as the current i7 mba

 

[mutsaers-vr.nl]

but fan less ?

Nope, it's at least 10% faster since it's tdp is 15 watts which is the same as the current i7 mba

 

[motrek]

What about those very thin bay Trail-T tablets with 4.5 watt tdp? They run very well and can certainly run boost mode all day long while being cooler than krait arm chips which throttles after a while.

Bay Trail is Atom. Completely different architecture. No point in bringing it up here.

Core m in the air would certainly run hot like all macbooks under load but would throttle far less aggressive than pcs and as such perform as good or in some cases better than lower end haswell 15 watt ulv (i.e. 4200/10 u or 4250/60u which also throttle heavily on PC laptops but not much on macs)

Right, if you drop a Core M into a current MBA then it would likely perform as well as the current Haswells. The current MBAs have more than enough cooling for both chips.

The question that keeps getting asked is about a hypothetical new model of MBA that's thinner and has no fan. If it had a Core M, how would it perform?

The answer is that it wouldn't have enough cooling and if you did anything intensive with it, it would quickly reach a high enough temperature to trigger thermal throttling and then run half as fast as the current MBAs. Or maybe less than half as fast, depending on the cooling.

----------

but fan less ?

No, the 4.5W Core M can't even run fanless in the Yoga 3 without thermal throttling. Of course a 15W chip can't run fanless.

----------

It doesn't even hit its full turbo on a lot of those benchmarks since it BIOS limited!

Forgot to reply to this part.

The Yoga 3's turbo boost (and thus performance) may be BIOS limited but read the reviews--when it's under load, there's a significant amount of fan noise, meaning that the chip is getting hot.

Thus, if it didn't have a fan, it would get up to a temperature where it would get thermal throttled anyway, BIOS settings or not.

 

[mutsaers-vr.nl]

So the anticipated 12" MBA with retina display (see mockup's on internet) will not have the i7 5650U since the rumour goes that the new MBA will be fabless ? That will mean that the performance will go down compared to the present i7 version.

No, the 4.5W Core M can't even run fanless in the Yoga 3 without thermal throttling. Of course a 15W chip can't run fanless.

 

[cbautis2]

Bay Trail is Atom. Completely different architecture. No point in bringing it up here.



Right, if you drop a Core M into a current MBA then it would likely perform as well as the current Haswells. The current MBAs have more than enough cooling for both chips.

The question that keeps getting asked is about a hypothetical new model of MBA that's thinner and has no fan. If it had a Core M, how would it perform?

The answer is that it wouldn't have enough cooling and if you did anything intensive with it, it would quickly reach a high enough temperature to trigger thermal throttling and then run half as fast as the current MBAs. Or maybe less than half as fast, depending on the cooling.

----------



No, the 4.5W Core M can't even run fanless in the Yoga 3 without thermal throttling. Of course a 15W chip can't run fanless.

----------



Forgot to reply to this part.

The Yoga 3's turbo boost (and thus performance) may be BIOS limited but read the reviews--when it's under load, there's a significant amount of fan noise, meaning that the chip is getting hot.

Thus, if it didn't have a fan, it would get up to a temperature where it would get thermal throttled anyway, BIOS settings or not.

Both core m and atom have same TDP even if the burst power consumption in the core m is higher, core m will still run and perform at very close to haswell ulv levels without a fan as intel intended.

Maybe you haven't seen intel showing off their fanless reference tablets while giving benchmark scores much higher than the ones on yoga 3 pro. It's up to OEMs to meet/exceed intel's reference tablets and their performance, and i'm very positive that apple will surely be capable of maintaining i5 ulv performance under long cpu load without using a fan.

 

[motrek]

So the anticipated 12" MBA with retina display (see mockup's on internet) will not have the i7 5650U since the rumour goes that the new MBA will be fabless ? That will mean that the performance will go down compared to the present i7 version.

Nobody here knows anything about future models of MBAs. It's all speculation. But I guess that's what you get for being on a forum called MacRumors.

----------

Both core m and atom have same TDP even if the burst power consumption in the core m is higher, core m will still run and perform at very close to haswell ulv levels without a fan as intel intended.

Maybe you haven't seen intel showing off their fanless reference tablets while giving benchmark scores much higher than the ones on yoga 3 pro. It's up to OEMs to meet/exceed intel's reference tablets and their performance, and i'm very positive that apple will surely be capable of maintaining i5 ulv performance under long cpu load without using a fan.

No, I haven't. Are they demonstrating sustained load? Are these presentations on the Internet somewhere?

 

[objektør]

No, just think of Broadwell as the same as Haswell but it uses 30% less power. If you switched current MacBooks to Broadwell, it would run a little cooler and you might get 1-2 hours of extra battery life but otherwise they'd be the same. You'd still need active cooling for a 15W chip in a laptop.

.

30% less power. Does 'run a little cooler' mean it can run heavy workloads 'a bit longer' before it hits its tdp compared to Haswell? Less energy- less heat?

But a retina screen needs more power. Would this mean it hits its tdp before a non-retina screen would or does the retina needs more gpu-power and not cpu-power and does retina only have impact on gpu and battery life?

 

[cbautis2]

Nobody here knows anything about future models of MBAs. It's all speculation. But I guess that's what you get for being on a forum called MacRumors.

----------



No, I haven't. Are they demonstrating sustained load? Are these presentations on the Internet somewhere?

They have cinebench 11.5 which is one of the longer benchmarks. Yoga 3 pro with a fan was bios/thermal throttled already during the benchmark which yielded only 2.09 pts while reference tablet without a fan didn't even throttle and gave 2.65 points. Macbook air i5 4250u gave 2.5 points on the benchmark.