More details are starting to emerge about Intel’s next line of processors. Code named Broadwell, the processors, which are likely to be released in the second half of 2014, are shrunk down versions of Intel’s current Haswell range. Like Haswell, the Broadwell range will come in several types including the Broadwell-D for desktops, the Broadwell-U for Ultrabooks and, of most interest to us, the Broadwell-Y for tablets.
Intel uses what it calls a tick-tock development process. In one cycle the company releases a new architecture with new designs and new features (this is the tick) and then during the next cycle it shrinks down the current devices using new manufacturing processes (the tock). Broadwell is a Haswell “tock” that will be manufactured using a 14nm process down from Haswell’s 22mn.
Intel is enjoying some success in the tablet space with companies like Asus and Samsung both offering tablets with its processors.
Unfortunately TDP (and its friends like SDP) is a measure of the amount of heat the processor will give off while running. It can’t tell us the power efficiency of the processor, but it does show that the Broadwell-Y is designed for fanless implementations with an eye on low power consumption. It is thought that Broadwell-Y processors will have two 64-bit cores (dual-core), use Intel’s GT2 GPU and support up to 8GB of memory.
The question is therefore, is Intel’s Broadwell-Y processor a threat to ARM? Unfortunately the answer is complex. In the smartphone market, which is where ARM has it’s most success, the Broadwell-Y is no threat. The current range of processors from the likes of Qualcomm are already quad-core and have a lower TDP than the Broadwell-Y. MediaTek has released its octa-core processor and all the big names are working on ARMv8 64-bit processors which should ship before Broadwell. Of course the performance of the Broadwell-Y isn’t yet known but it is unlikely that it will find its way into smartphones.
For tablets the Broadwell-Y will certainly be an attractive option for high-end devices. Most likely the next generation of the Surface Pro (the Surface Pro 3?) will use a Broadwell processor, either a -U type as in the current Surface Pro 2 or possibly the -Y type which would allow Microsoft to make its tablet thinner. Intel is enjoying some success in the tablet space with companies like Asus and Samsung both offering tablets with its processors. If the Broadwell-Y performs well in terms of computing power and energy usage then more manufacturers could start to offer Intel powered tablets. However it is worth noting that at the moment these Intel powered Android tablets use Atom processors which are a very different, and separate, range from the Haswell/Broadwell series.
For smartphones and tablets it is ARM who has the dominant position with Intel trying to gain ground. But there is one area where Intel is still very much king and ARM would like to make an impact, that is in the server business.
With the advent of the 64-bit ARMv8 architecture and AMD’s plans to release it Hierofalcon 64-bit ARM-based SoC in the same time frame as Intel’s Broadwell, ARM is gunning for the server room. In the server room every Watt counts in terms of electricity bills, cooling requirements and ecological impact. The use of processors with lower TDPs could reduce the total cost of ownership per server. This makes ARM an interesting prospect for servers. With support for operating systems like Linux already very well established, an ARM server becomes a drop in replacement for an Intel server. But with Intel pushing the TDPs of the Broadwell series lower and lower, the likelihood of ARM making a massive entrance into the energy efficient server market is decreasing.
Intel, ARM and AMD all have big plans so it will be interesting to see what 2014 brings, especially for tablets.
Like this post? Share it!
Intel’s atom chips seem to preform better with less cores than their arm equivalent so it will be interesting to see how this new breed compares.
Personally I prefer arm though, I prefer their model , letting more people male their chips and allows others to compete with Intel . at last the WinTel dominance is broken !
The Intel atom chips have threading so the dual cores can do basically as much as 4.
The Clovertrail version of the Atom using the Saltwell core which are the older chips have hyper threading. The newer Bay Trail chips do not have hyper threading so a dual core runs two threads and a quad core four threads.
Bay Trail is not Broadwell. :) Broadwell is the next generation of Intel’s laptop-desktop processors. The low-power Y variant will compete with the high-performance ARM chips of 2014. Broadwell chips do have hyperthreading, meaning that each core can dynamically switch between tasks while idling. Not quite as good as more cores, but it arguably has more usage scenarios.
Bay Trail is the main CPU to go against ARM. Broadwell would only be used in the higher end tablets as it has a higher performance and higher power dissipation than Bay Trail. Bay trail and it’s variants can be used in lower cost tablet’s and smartphones. There’s no plan to use Broadwell in smartphones because the power envelope will not allow it so Bay Trail is the main ARM competitor. And it doesn’t have hyper threading but it doesn’t need it because they just add more cores. These days the cores are much smaller in die area than the GPU. By a lot so it’s easier to just add the cores
True, true. See my other comment. :) However, the article was about Broadwell-Y being a factor in the ARM vs x86 race, not Bay Trail.
As always, moarh competition → better products. Intel’s processors seem to perform well, so I wouldn’t mind having one in my phone
The ones they have used in android devices have always had two major problems:
Graphics performance and app support.
Fortunately graphics performance is easily fixed. Making apps work with x86 should be easy, right?
Intel tries to improve power efficiency through better manufacturing process, and that will be more expensive.
ARM achieves similar or better power efficiency through better design with much simpler manufacturing process and that will make its chips much more affordable.
Intel still has a long way to go before catching up with ARM.
I guess they won’t be expanding their manufacture of ARM chips beyond the ones being made for Altera, and will continue to try and push x86/x64 chips into territory dominated by ARM. When the announcements about Intel making ARM chips came last October, it nearly sounded like they were giving up on the Atom (at least for Android devices).
You are confusing the issue. The ARM chip Intel is manufacturing has nothing to do with tablets and phones, laptops, etc… It is an FPGA and has no competition to anything Intel sales. It’s hardly an ARM chip. It’s an FPGA that uses an AMR controller.
Apple make their own processors so Intel are out of the game there.
For Android, which for so long has used ARM, I would have thought the Broadwell processors will have too much power for the tablets’ OS so it won’t know what to do with it. And they still use too much power – so we will see tablets that still have reduced battery life with no real performance gains.
I do, however, think process shrunk Atoms are the way forward for Intel in the Android smartphone and tablet space – these could be even lower power use, with potentially higher performance.
Microsoft Surface tablets use ARM processors by definition.
That leaves a niche for Broadwell on Microsoft Surface Pro tablets – these could do with all the performance and battery life gains they can get. But how small is that market? And is it’s future rosy or uncertain – not something I would gamble millions of Intel’s dollars developing a processor product for.
Could this mean new Intel powered fanless Chromebooks?
I think the Atom line in the consumer space is under threat from the Broadwell Y processors. ARM processors usually can idle at <200mw per core, and run upwards of 750mw per core fully loaded. They are still in the wrong scale for a comparison.
honestly, Intel’s desktop line is in trouble. At the high end there are xeons and they are untouchable still, but the 4770k without overclock, still performs only about 3 times as fast per core as Apple’s A7. ARMv8 coming next year will all likely perform around that level, if not higher, but will come with four to eight cores .. meaning you could feasibly see a phone performing at 2/3′s or more than a fully decked out 4770k that is barely a year old.
Has mobile processing really come that far? That’s hard to believe that they can go toe to toe with Intel’s flagship Haswell chip. I’d have to see hard numbers to believe that.
Yes, mobile processing has really come that far. To be clear, nothing new needs to be done. Current designs at 28nm are already capable of performing at this level. Assuming any improvement over the next year, they will beat that estimate.
Are you seriously using geekbench for comparisons on CPU performance?
You actually think a mobile chip thats smaller than your fingernail is almost as powerful as a desktop chip that has 10x the surface area? Have you compared the amount of time it takes for a desktop website to load on your phone compared to your pc? And do you think they can run full fledged desktop operations like excel? Stop comparing clock speed and look at everything else that goes into a chip. And for the record a Tulatin Pentium 3 at clocked at 400mhz has the equivalent processing power as your little apple chip. Yes and that chip is from 2001. Mobile chips have made great strides in processing power but they are no where close to there desktop equivalents.
I don’t think Robert has a full understanding of the issue.
No I didn’t say it is currently as powerful, I said it will be as powerful in the near future as very expensive desktop chips from (at that point) the recent past. But you are really failing to understand…
As a direct counterexample to your whole line of objection, consider the Tegra K1. This is a dual core chip that is paired with a GPU, it’s the GPU that is interesting. It is, according to Nvidia, the same Kepler you already know and love from desktops, but instead of being 1350 or so cores (in the 660ti), it is 192 of them. It performs at 360 gflops instead of the 660ti’s 2.5 teraflops. It also runs at 2w instead of 150w cores plus 150w for it’s pcie bridge. 2w versus 300w — and it is the same archtecture? Sounds like bs.
Well, it is BS. Nvidia had to convert large sections of the core over to 32 bit (SP from DP) in order to pull this off, and tweak some other features in the architecture too .. the point is though, that the desktop paradigm wasn’t power-optimized at all, and you really could have had a 660ti that ran at ~12w + 150w = <170w, instead of a total of 300w peak (pci plus power connectors on board). This in turn means that they could have offered you a lot more muscle at 300w. Like, 10s of tflops… but unfortunately, they hadnt learned the lessons necessary to do that yet, when they designed the desktop kepler, because they hadn't already had years of experience designing lower power parts.
Intel is in a similar situation. It is hemmorghing cycles on a bloat of wattage that doesnt go into computation. Dont get me wrong they are a good company, but they have some lessons to incorporate yet and right now yeah, a 88w part from intel is not so impressive next to a ~4w total draw mobile phone… I mean overall it is still more powerful, but efficiency is horrible.
This is ridiculous, you’re pulling numbers out of your ass. The 4770K is an extreme performer ain’t no tablet chip hitting anywhere near the levels of performance this thing puts out for a LONG ass time.
in terms of the comparison of the intel chip to the arm chips in relative processing power, the current benchmarks for the Apple A7 in Geekbench 3 is at ~1400 single core, while the intel chip is at about 4500. The A7 is also only dual core, meaning ARMv8 designs coming this year will likely far exceed the multithreaded processing power.
I’m not making anything up, you are just surprised with how powerful ARM processors are nowadays.
“In one cycle the company releases a new architecture with new designs and new features (this is the tick) and then during the next cycle it shrinks down the current devices using new manufacturing processes (the tock). Broadwell is a Haswell “tock” that will be manufactured using a 14nm process down from Haswell’s 22mn.”
This is wrong. Broadwell is a “tick”, and a “tick” is a process shrink.
Cheap intel chips are usually around $75 to manufacturers. Cheap arm chips are around <$5 to manufacturers. The only reason intel did so well in the past is due to the WinTel monopoly.
Intel isn’t looking to go mobile because they are broke, but rather because it is another revenue source. And I have no doubt that the Broadwell Y series will be priced competitively, though I would agree that it is likely going to cost more.
But Intel has the cheaper Bay Trail to use for lower priced and lower power.
Right. Atom is designed to be a very low power chip for smartphones and tablets, where as Broadwell-Y is basically a low-power laptop chip designed to provide ultrabook power and will probably be only used in high-performance tablets.
Intel can actually build chips cheaper than most because they own the entire cost from design to wafer fab to assembly. I doubt there are many people that can make money at $5 for a chip that anyone would want in their tablet or phone. I would be you an Apple A7 is not a $5 chip. There are cheaper alternatives in China but no one really wants to sell an SOC for $5. It takes some real estate on silicon to have the integration a modern SOC need’s and at $5 for the cost of the die, test and assembly you better have a cheap ass wafer fab.
The listed price of the Atom Z3740 is $32. Volume discounts reduce that to $25. The price of a competitive hi end snapdragon or apple A7 is about $20…. and they are more power-hungry than Silvermont Atoms (believe it!!) and they don’t run windows.
Now do the math again using true data.
man, the amount of misinformation or ignorance about cpu’s is unbelievable.
Speaking of “misinformation or ignorance”…
Are you the same “Guest” that compared a “cheap” $75 intel chip to a $5 Chinese ARM chip??
I want to believe that 2014 is Intel’s Android breakout year. Really I do. But after all the false starts I am going to see it. Show me.
False starts? Saw a 8″ Toshiba Encore with the Atom Z3740 and this thing packs some serious punch.
I have tested S800 3D gaming battery life (5″ Android smartphone, 11.4 Wh battery): it lasted almost an hour and consumed ~12.4 W. It also throttled a lot. And btw. – the older (and weaker) S600 lasted longer and throttled less – all in all much better experience.
I have also tested Z3740 10″ tablet in 3D gaming. It lasted over 5 hours with 34 Wh battery and therefore consumed ~6.5 W. This was a Windows 8.1 tablet that could run Mass Effect 3 and League of Legends at 25 and 29 fps respectively.
Of course, these are not directly comparable: different platforms, architectures and HW equipment, but it should be enough to give you an idea. I recommend TomsHardware and AnandTech for better and more precise tests – they also have many non-gaming tests which show Bay Trail is more efficient in regular usage.
Considering these Bay Trail devices are constantly sold out (for example the ASUS T100 was the 5th best selling tablet on Amazon about 2 months ago), I would guess Intel is actually doing very well regarding tablets. And these are just the more expensive Windows tablets (license costs) – Android tablets with Bay Trail should come soon. Considering Broadwell-Y will be more efficient and have more performance, it seems that Intel is becoming more than competitive.
Make a hybrid with Broadwell Y and 16 hrs batterylife plss!
This article incorrectly describes “tick” and “tock.” The “tick” is the process node change, the “tock” is the architecture change. Haswell is a “tock.” Broadwell is a “tick.”