On the two reviews the numbers are :
Core M idle 4.2W load 19.1W delta 14.9W
Beema idle 5W load 25.8W delta 20.8W
Check out the CPU package power number like this: http://www.notebookcheck.com/fileadmin/Notebooks/Sonstiges/Prozessoren/Broadwell/dota2.png
For a short period TDP can exceed 5W and then it goes down to ~5W, hence why the clock speeds are low as well in games or MT. Beema doesn't suffer from this, it is a true 15W product.
It depends how the TDP is set by the OEMs, the HP Envy has been tested with Furmark + Prime 95, the 14.8W delta let no doubt that there s more than 5W drained by the CPU, about 10W actualy.
For comparison the Beema is at 14W while a Btrail 3520 (7.5W official TDP...) is at 11W in the same test, the three chips are in the same range TDP wise, the Core M bring nothing in respect of what was available at lower cost set apart a better GPU than BT s within Intel offering.
With a product like this tdp seems wrong and gives, as we can see in another thread, consumers wrong signals. Its marketing. But its also results as shown not only by bm but the subjective comments following the devices using it for office.
I am not buying this cpu. Its imo stupid not to have 2-3 times performance just adding a small fan. Its priced for performance and imo in portable device i need solid consistent performance or i could just use a tablet or phone.
Furmark is not representive of actual use.
This is a product intended for light office use and it delivers in a small tdp here. I have no doubt that with a fast ssd attached this product feels faster than bt for that purpose.
With a product like this tdp seems wrong and gives, as we can see in another thread, consumers wrong signals. Its marketing. But its also results as shown not only by bm but the subjective comments following the devices using it for office.
I am not buying this cpu. Its imo stupid not to have 2-3 times performance just adding a small fan. Its priced for performance and imo in portable device i need solid consistent performance or i could just use a tablet or phone.
As much as Sandy Bridge created the entire Ultrabook category, it was only with Haswell that Intel got the balance between power consumption, and performance. I think we're going to see something similar with Broadwell. It will create an entire new category of devices, but only Skylake, or even Cannonlake will strike the right balance between performance, power consumption and affordability.
I'd say that's a good way to think of it. My guess is that it'll truly be viable once the big core is part of a single die SoC.
Why do you think it isn't viable right now? I don't hear anyone saying tablets aren't viable, and those have >2x worse performance. That of course doesn't mean products won't improve with time and new architectures.
Why do you think it isn't viable right now? I don't hear anyone saying tablets aren't viable, and those have >2x worse performance. That of course doesn't mean products won't improve with time and new architectures.
Going off of previous discussions, Intel17 is likely talking about power consumption, and how the chipset wastes a lot of power."Viable" doesn't just mean "high performance". Cost is also a key factor.
Ultrabooks were also viable with Sandy Bridge, but you can't really compare the kind of performance and battery life you had with it and what you have now with Haswell. It's night and day from the Ultrabook POV. I expect the same to happen with Broadwell and Skylake/Cannonlake.
I don't know how Ivy Bridge was supposed to compete with ARM. IIRC, in 2013 Intel announced the Y-series to go into thinner but no fanless designs, but the real improvement was always supposed to be Haswell. Now Broadwell improves significantly upon Haswell to allow for fanless designs with great battery life and performance, with transistor up to platform level innovations. Bay Trail was also a major advance, but it was plagued by BOM issues (which Intel paid for, fortunately) and TTM, but nonetheless it let Intel get a foothold in mobile and every new generation will only get better and more widespread used.
Problem is intel is getting into what I call the "next chip syndrome". Always seems like the next chip coming up will be the ARM killer, but when it finally gets here, it just is not quite there for some reason. The problem is, with each new generation, and especially with all the delays, it is allowing ARM to become more entrenched and develop faster chips as well
The problem is straight forward imo but not nice to look at; The leaking stays more or less the same vs 22nm as i can see. And thats even with finer finfet tech.
The performance in this power evelope looks fine to me but the leaking is not going away and i think we will have to get used to it until we get some radically new process tech.
Looking at the battery life of the devices actually made me think arm is probably right about the little big design as the little cores non ooo a11 will leak far less. The first iteration was buggy but arm customers will have to use process nodes that is traditionally more leaking than Intel as i can see. I think the results here indicate the decision to go little big is right. But ofcource we need time to tell that.
Frankly I doubt anyone is dreaming of killing Arm (or Intel!). They're both quite happily making massive profits
Apple to orange comparison and you know itto be clear
ARM's profits: ~600 million USD this year
Intels profits: ~15.6 billion USD this year
Apple to orange comparison and you know it
Even if i repeat myself. Leakage goes up not despite finer tech but because of it. If you reduce gate length and in addition keeping Ion the same and thus increasing fin-height -> leakage goes up.The leaking stays more or less the same vs 22nm as i can see. And thats even with finer finfet tech.
Right, going planar to FinFET will drastically reduce (subthreshold) leakage. This holds until the next shrink after you moved to FinFET. Technically you can still improve by going from FinFET to a gate-all-around (GAA) technology.The first iteration was buggy but arm customers will have to use process nodes that is traditionally more leaking than Intel as i can see.
Even if i repeat myself. Leakage goes up not despite finer tech but because of it. If you reduce gate length and in addition keeping Ion the same and thus increasing fin-height -> leakage goes up.
Right, going planar to FinFET will drastically reduce (subthreshold) leakage. This holds until the next shrink after you moved to FinFET. Technically you can still improve by going from FinFET to a gate-all-around (GAA) technology.
Back on topic, Core M looks like a disaster. Meanwhile 64bit ARM SoCs (Apple A9x/Exynos) achieving similar performance at apparently much lower power on an inferior 20nm planar process node.