News - Next-gen Intel Broadwell SoC reduces Haswell power-draw by 30pc

[HEXUS]

Can be configured for same performance, lower power.

Read more.

[cheesyboy]

How much more excited would we all be if that were 30% more PERFORMANCE for the same power draw?

[kingpotnoodle]

How much more excited would we all be if that were 30% more PERFORMANCE for the same power draw?

Me personally not very, my computers are all more than fast enough for their purpose but the electricity bill is looking a bit painful these days and cool running with long battery would be incredibly useful to me.

[crossy]

How much more excited would we all be if that were 30% more PERFORMANCE for the same power draw?

Erm, try thinking that one through. If Intel can deliver a processor that takes less power to deliver the SAME performance then that's got to be a more efficient design. So - assuming it's physically possible - you may be able to overclock to the same thermal output as the old designs, but end up with correspondingly better performance.

That said, we've already got enough cpu performance to be able to comfortably run an OS within another OS (virtualization), so I'm quite happy to take the same level (or slightly increased?) performance level and pocket the power savings. Heck, the money saved might even be enough to put a down payment on improving the other hardware in the rig...

I did read something the other day saying that cpu and graphics performance these days are very good, and it's "peripheral" factors like memory bandwidth and IO responsiveness that are more of a limiting factor to overall system performance.

[cheesyboy]

Me personally not very, my computers are all more than fast enough for their purpose but the electricity bill is looking a bit painful these days and cool running with long battery would be incredibly useful to me.

Laptop, yes.
Desktop, no

As far as energy bills go, it's all something of a myth. If a 100w CPU can deliver 30% more grunt than a 70w CPU, then it can do the task 30% faster which means using the same electric (or do the task at 70% capacity, which will reduce energy draw too). Low-energy idle states are preferable, of course, but that doesn't need to come at the cost of performance.

Erm, try thinking that one through. If Intel can deliver a processor that takes less power to deliver the SAME performance then that's got to be a more efficient design. So - assuming it's physically possible - you may be able to overclock to the same thermal output as the old designs, but end up with correspondingly better performance.

Ah yes, just like we can clock the much-more-efficient Haswell up to correspondingly better performance than Ivy and Sandy, you mean?

[Tpyo]

Ah yes, just like we can clock the much-more-efficient Haswell up to correspondingly better performance than Ivy and Sandy, you mean?

It would be nice if they at least spent the extra dollar/chip on the thermal compound, at least for the 'K' and extreme desktop variants. It might not be their aim to push performance, but it doesn't exactly hurt them to let people try (and makes for ridiculous OC competition speed headlines for minimal extra effort on their part). If they aren't releasing higher performance chips anyway, it doesn't even steal from the higher end of their own market.

[scaryjim]

I don't know if anyone's bothered reading the title, but this clearly says Broadwell SOC. Most desktops don't use SOCs. So this is specifically targeted at the mobile space, where 30% lower power is actually pretty significant.

They can't do it in the performance space because of the physical realities of silicon: more voltage = more heat = more leakage = decreasing returns on performance/watt. So it's a question of how little voltage you can use to make the chip run stably at a given clock speed. Besides, there's a limit to the usefullness of increasing performance at the top end now - particularly in the consumer space where 99% of the market is well served by Intel Core i3s and quad-core AMD APUs....

[cheesyboy]

Yes, scaryjim, but if you bother reading the actual article, and the linked article where CEO Brian Krzanich talked about the 30% saving, he is clearly talking about the whole architecture including desktop.

The SoC bit was just a demo system (28pc demonstrated)

[scaryjim]

And if you read Krzanich's comments carefully, he's clearly talking primarily about mobile:

The CEO also believes the PC is in the process of reinventing itself .... He also showed a demo laptop running the next-generation 'Broadwell' architecture, explaining that the 14nm process and general architecture improvements enable a 30 per cent decrease in power when compared to equivalent Haswell chips.

So maybe it's true that they'll show a 30% reduction across the whole range, but Broadwell is clearly targeted at mobile, which is the driver for power reduction rather than performance improvement. In fact, both Intel and AMD are moving towards targeting mobile computing rather than desktop. They've both invested a lot in redesigning their "small-core" offerings to improve performance/watt to the point where you can get a "full" PC experience from a very low power chip.

Of course, that does leave the enthusiast feeling somewhat short-changed. AMD have failed to advance performance significantly due to a major redesign of the architecture not performing as well as they may have liked, and Intel therefore haven't felt any need to push performance at the enthusiast level significantly. But then again, both companies care about their bottom line and both have decided that mobile is the way to go, so fair play to them - and Intel particularly - for engineering their way to significantly lower TDPs...

[cheesyboy]

Broadwell is clearly targeted at mobile, which is the driver for power reduction rather than performance improvement.

Yes

Of course, that does leave the enthusiast feeling somewhat short-changed.

Indeed, so perhaps I speak for us all when I say how we would be....

...much more excited ..... if that were 30% more PERFORMANCE for the same power draw

Which is, approximately, where I started

BTW, I would contend the generally held belief that it's AMD's fault that Intel aren't pushing at the top end. The enthusiast market is shrinking; a reflection of a maturing market where the core audience is getting older (How many on Hexus are under 30?), and smaller. There are also "enthusiasts" digging their own grave by buying into the energy-efficiency propaganda.

Whilst the enthusiast market is shrinking, the mainstream is also moving away from pure performance - portable and "good enough" tech is the growth area. Intel are doing what they have to to diversify away from a receding market.

[Dottorrent]

Intel, make a new Sandy Bridge. Then, I'll consider upgrading.

[kingpotnoodle]

Laptop, yes.
Desktop, no

For me (and I'm not alone) its a yes on both, there are many reasons to construct desktop type machines to a lower power budget (heat, noise, cooling requirement, form factor), a desktop computer isn't always about absolute maximum performance with the power draw irrelevant.

As far as energy bills go, it's all something of a myth. If a 100w CPU can deliver 30% more grunt than a 70w CPU, then it can do the task 30% faster which means using the same electric (or do the task at 70% capacity, which will reduce energy draw too). Low-energy idle states are preferable, of course, but that doesn't need to come at the cost of performance.

Sometimes something of a myth, but if Broadwell also reduces idle power (not unreasonable to expect) then that calculation falls over. If a higher TDP CPU can do the same work in less time for equal total electricity and then return to an equally low power idle state then it is good, but often to sustain higher clocks often means disproportionately more power and there is an optimum efficiency point on the curve to hit, it doesn't always make sense to just push the clocks higher and higher.

Also if the higher TDP CPU requires a screaming fan to stay cool, more chassis fans, pushes power supplies into less efficient usage etc then not so great. There is a lot more to a system than the CPU to balance up.

[mikolaj87]

I'm glad to read about lowered power consumption but I'm very concerned that current trend is going more and more towards SoC. I just cant imagine that one day I wont be able to replace my CPU :/.

[cheesyboy]

Sometimes something of a myth, but if Broadwell also reduces idle power (not unreasonable to expect) then that calculation falls over. If a higher TDP CPU can do the same work in less time for equal total electricity and then return to an equally low power idle state then it is good, but often to sustain higher clocks often means disproportionately more power and there is an optimum efficiency point on the curve to hit, it doesn't always make sense to just push the clocks higher and higher.

Idle power draw, whilst better if lower, is really not worth worrying about.

See here;
http://www.anandtech.com/show/7003/t...4560k-tested/2

Total system idle power for Haswell i7 is 34w. What's 30% of that - around 10w. That means it takes 100 hours (4 days) of idle use before Broadwell would save you 1KWh. And 1KWh is about, what, 12p?

I'm not exchanging a 30% (or any significant) performance hit for that!

[scaryjim]

... so perhaps I speak for us all when I say how we would be....

...much more excited ..... if that were 30% more PERFORMANCE for the same power draw

Well, you don't speak for me - I've only ever bought one "enthusiast" CPU, a Q6600, and that was in January 2008 just as the Q9xxx series was being released. I've never found the need to shop out of the lower couple of brackets for CPU performance. 30% less power potentially means mainstream laptop performance in a tablet/handheld device. And I find that pretty damn exciting

BTW, I would contend the generally held belief that it's AMD's fault that Intel aren't pushing at the top end.

Oh, I don't think it's *just* AMD's fault. There are lots of other factors. But when AMD were winning the CPU wars Intel put in a huge effort and made a genuine step-change in CPU performance - not just 30% at the same TDP, but up to 100% performance increase at around half the TDP. Now, it's questionable whether they could ever get the same gains again, given how efficient the architecture they're starting from is, but it's a demonstration of what you can do with a large R&D budget and a strong driver to improve your CPU business. Intel haven't had that strong driver since Core 2 was released, and their performance related improvements have plateaued. It's hard to view that as a coincidental correlation...

EDIT to add:

I'm not exchanging a 30% (or any significant) performance hit for that!

Well, no, you're not - that's the point. You're getting exactly the same performance - no hit at all.

[CAT-THE-FIFTH]

TBH,I think people are reading way too much into the PR.

The screen,and chips associated with GPS and wireless functionality consume the most power in a phone or modern tablet,and battery technology is not really keeping pace either. The SOCs used actually in most cases,are down the power consumption list. These are the areas which still need the most work,and for desktop users your usage habits and lifestyle and far more draining on your power and energy usage.

Simply,using a sub-notebook or tablet for most of your day to day media consumption and e-mails would save a load of power,but OTH as with everything be careful of spending pounds to save pennies!