Slashdot Mirror
Info On Intel Bay Trail 22nm Atom Platform Shows Out-of-Order Design
MojoKid writes "New leaked info from Intel sheds light on how the company's 2014 platforms will challenge ARM products in the ultra light, low power market. At present, the company's efforts in the segment are anchored by Cedar Trail, the 32nm dual-core platform that launched a year ago. To date, all of Intel's platform updates for Atom have focused on lowering power consumption and ramping SoC integration rather than focusing on performance — but Bay Trail will change that. Bay Trail moves Atom to a quad-core, 22nm, out-of-order design. It significantly accelerates the CPU core with burst modes of up to 2.7GHz, and it'll be the first Atom to feature Intel's own graphics processor instead of a licensed core from Imagination Technologies."
Firsties on a firsty with frosty piss
Im looking forward to another atom based hackintosh. Low power file server/media server with iTunes!!
It is Always Reassuring When .... You spend a bunch of money on a new processor and they tell you it is already "Out of Order" from the get-go.
I, for one, will be overjoyed to see the last of Imagination's 'PowerVR' shit, especially on x86, and hope we'll never see the likes of the "GMA500" again.
On the other hand, this report has me wondering exactly what the Atom team is up to. Back when Intel started the whole 'Atom' business, the whole point of having a substantially different architecture, in-order, was to have something that could scale down to lower power in a way that their flagship designs couldn't. Since then, the ULV Core I3/5/7 chips have continued to improve on power consumption, and the Atoms have apparently been sprouting additional complexity and computational power. How much room do they have to do that before 'Atom' evolves itself right out of its power envelope, or Core ULV parts start hitting the same TDPs as higher-power Atoms; but with much more headroom?
I know that it's DDR3 SODIMM but is there any particular reason they're limiting it to DDR3-1333?
Would there be a performance gain if it could utilize DDR3-1600 like how the AMD fusion processors show decent performance gains using higher speed memory? I'm pretty sure that DDR3-1600 SODIMM's are out there.
Runs all your x86 binaries.
By MS' own definition, uefi will support other os options (not guaranteed under ARM).
Has mature, supported foss GPU drivers unlike every android-only ARM SoC.
THE platform for that budget linux tablet that dual boots into MS Office?
A warning for Windows users.
The newer Atom chipsets do not support Windows 7 and intel is dropping support. I posted the story last month but was rejected because it came from Neowin which Slashdot doesnt like. Intel has no plans to support Windows 7 as they want to sell more expensive chips to corporate buyers. I guess they forget consumers hate Windows 8 too.
What, you can properly hyphenate dual-core, but you can't be bothered to properly hyphenate ultra-light and low-power when used as adjectives?
The imagination technology drivers aren't open source, which was a big issue. Moving to an Intel video board means that it will be released as free software (unless Intel changes its policy which is very unlikely). That's a very good news for the open source platforms!
Interesting that the submissions falsely asserts that Baytrail follows Cedartrail incorrectly. Clovertrail is the more direct preceding product in the roadmaps and was released about 3 months ago.
The trouble with Intel is it's competing against LAST YEARS ARM quad core chip, by the time they get it out, the 8 and 16 core ARM chips will be out.
Not only that their main problem isn't the processing power, it's the power draw! They keep talking about idle less than 10 watts, in a market where idle is less than 100mW. Defining idle as 'what Windows 8 does in idle' doesn't work in a market dominated by Android which idles a lot deeper.
So we seem to have these endless puff pieces from them, promises of how amazing the next generation will be. It doesn't look like attack on ARM (which requires weapons AKA competitive chips), rather it looks like defense, to keep the server market from switching over anymore than it has already.
It's more about keeping companies locked into Intel from switching, on the promise of a fix real-soon-now.
The PC market looks like it is down 20 percent from the same quarter last year. And there is no sign that there is anything that is going to change the collapse of the desktop x86 PC market.
On the booming cellphone and tablet markets Intel is effectively a non-entity. Intel has nothing to offer other than hotter, hugely more power hungry, and ridiculously more expensive chips.
Intel's PR campaign that has been going on the past few weeks isn't impressing anyone but existing Intel fans. So far all Intel has been able to demonstrate is they are somewhat competitive versus year old previous gen fab ARM solutions with higher cost and higher power requirements when done by cherry picked friendly people in the computing press.
There is no sign that Intel is going to somehow miraculously transform Atom into a solution that matches the needs of cellphone and tablet manufacturers.
New leaked info from Intel sheds light on how the company's 2014 platforms will challenge ARM products in the ultra light, low power market.
Intel is using the tactic perfected by Microsoft, i.e., compare your product plans from two or so years in the future with the current products of your competitor, and then say how much better your envisioned products are.
.
Intel is behind the 8-ball in the low power market space, and this is nothing less than a move of desperation on Intel's part.
Dual ARM A15 chips destroy any current dual Atom from Intel. The coming quad A15 parts will destroy any Intel ULV i3 part (Intel's crown jewel CPU) that competes in the same space.
However, the A15 design is now years old. ARM is replacing it with a fully 64-bit part that uses only 60% of the same die space in the same process. This means that the ARM part that replaces the A15 early 2014 has either more performance or less energy use- a total nightmare for Intel.
Meanwhile, it is impossible for Intel to 'repeal' the Intel Tax. Intel is addicted to massive profits per chip, and cannot function on the margins made by those that manufacture the ARM SoC parts. Example: Intel is boasting support for 4K video on its next generation CPUs, but 4K support already exists on one of the cheapest ARM chips you can find in a tablet, the Allwinner A10.
When Atom goes 'out of order', it ceases to be an Atom, as is, instead, a renamed version of Intel's current 'core' architecture.Intel going quad with the Atom makes zero sense, when the targeted low power devices try to keep all but one core in idle for power-saving reasons. Intel can already thrash its own future Atom with the earlier mentioned ULV dual-core i3 part, as used in the latest Chromebook.
It gets worse. AMD and ARM are fully unifying the memory space of external memory as used by either the GPU cores or the CPU cores. Intel is going in the opposite direction, attempting to build on-die RAM blocks for the exclusive use of the GPU on versions of its chips aimed at high-end notebooks. This project is dying on its feet as notebook manufactures cannot believe the money Intel wants for this version of Haswell- they know if their notebook customers pay a lot for the product, they demand decent graphics from Nvidia or AMD, not half-working slow graphics rubbish from Intel.
It gets worse. Apple is on the verge of dumping Intel completely for their own ARM SoC designs. The high-end Apple desktop systems that would struggle with current ARM chips hardly make money for Apple anyway compared with the phones, tablets, and Airbooks.
It gets worse. Weak demand in the traditional PC marketplace means that Intel has growing spare capacity at its insanely expensive fabs. It tried to find customers for this free capacity, but Intel fabs are massively customised for Intel's own CPUs, and lack the technical support for other kinds of chips. Intel uses its outdated equipment to make other kinds of parts (like the dreadful Atoms, or the dreadful MB chipsets), but potential customers hardly want to make their new chips on these very old lines.
It gets worse. Global Foundries (AMD's chip production facility- that pretends to be independent) is making incredible strides in attracting business form many companies designing the most cutting edge ARM parts. Samsung's chip business is going from strength to strength. Apple is making massive investments at TSMC. The Chinese fabs are coming along in leaps and bounds.
It gets worse. The GPU is becoming by far the most important part of the modern SoC (system on a chip). Intel's GPU design is a distant fifth to the SoC GPUs from AMD, Nvidia, PowerVR and ARM itself. Of the five, only Intel's GPU still doesn't work properly, and is NOT compatible with modern graphics APIs. Intel has to hack its drivers even to get even a handful of the most popular games running with minimal glitches. Intel GPU = you will have massive compatibility issues.
Where is the Z80 today? The same question will be asked of x86/x64 tomorrow.
My main issue with netbooks was the horrible resolution and the sluggishness.
If, by the end of 2013, they can slim down a Bay Trail-based netbook to 3/4", banish the absolutely awful 1024x600 resolution for 1366x768 or even 1600x900, rev to Windows 8.5, and keep it at $350, I will buy 3 for the price of a Macbook Air.
We are still talking about 5 to 10 watts...
Is it really a challenge tor arm?
If Intel's Atom was actually competitive with ARM in the real world Intel wouldn't be wasting time with this latest PR effort and instead would be putting out press releases with announcements of new customers dumping ARM for Atom.
They aren't.
Intel's absurd boasts about Atom are as believable as their old hilariously fake SPEC compiler scores.
Intel is like Kodak: stubbornly continue business based on old tech no matter what happens in the real world. But there are still lots of fans for old PC stuff so Intel doesn't die. Yet.
http://www.extremetech.com/computing/110563-intel-medfield-32nm-atom-soc-power-consumption-specs-and-benchmarks-leak
"Anandtech did a surprisingly decent comparison [anandtech.com] of the power usage of Clover Trail (Atom Z2760) vs.Cortex-A15 (Exynos 5250)."
The actual 'favorable' test claimed was the Tegra 3 running Windows RT vs a Medfield running Windows 8. With Wifi turned off and the machine left to do nothing, nothing at all, not stream a movie, not run a GPS app, nothing.
This is probably the only case where Medfield can shut down all the silicon blocks needed to compete with Tegra's low power single core.
But you may aswell have compared the 'off' state of both devices and declared it even. If they'd streamed a movie on an Android on a Tegra 3, vs Windows RT on a medfield, the true nature of this is revealed:
The Tegra 3 can run on it's single low power core quite happily, and stream video across Wifi. The Medfield has no such core, it's silicon blocks are on or off, as soon as you do anything it's sucking down the juice. This is why the Razr has terrible battery life. Google [battery life android razr] to see the complaints!
the first Atom to feature Intel's own graphics processor
I have an Atom D510 with integrated Intel graphics.
GMA 3150 GPU and memory controller are integrated into the processor.
Does that count? I bought the motherboard in 2010.
Escher was the first MC and Giger invented the HR department.
as an example, how many easily upgradable Atom boards with ZIF-style socket have you seen?
What makes you think Intel won't do away with upgradable socketed boards on desktops too?
ARM processors are already good enough for 95% of what people do, even on the desktop.
But everybody has a different 5 percent that isn't yet ported to ARM.
Just look at Chromebooks and the near console level gaming available on high end tablets.
Given that the Xbox 360 is seven years old, "near console level" is not saying much. Seven years is just one year less than the gap between the Nintendo 64 and Nintendo DS, which offered near Nintendo 64-level graphics.
My apps are optimized for keyboard, thank you very much. Mice and touch are just extensions to the keyboard interface.
Unfortunately, all the new 10" computers are tablets, not laptops, and these aren't guaranteed to come with a keyboard. How do you plan to adapt your applications to the discontinuation of small laptops?
'Nice' compared to Amarok, Banshee, etc? Or just nice in that there aren't better alternatives?
Last I checked, iTunes was a contender for the best media library available for Windows. Personally I have always found it to be rather lacking, with a small feature list and limited configuration options. I understand that might actually be a feature in itself for a majority of users. In any case, it's been some time since I've used it.
Those who advocate genocide deserve every protection afforded by law, and none afforded by common human decency.
I propose that Intel license ARM technology and then blow everyone off the map with 22 nm ARM chips.
Because it is not a good business decision. That would imply that the whole board would have the shelf life of the (expensive) CPU. That would also imply that for each N board configurations and Y cpus, you'd have N*Y products, instead of a good set of N generics and a small set of Y specifics. The unsold boards that would become obsolete would also have the additional cost of having a valuable, but otherwise useless (because it is soldered in) CPU. Intel's premium market isn't the embedded segment where SbC's are the norm - its highly specialized integrated circuits (such as CPUs) and generic boards. Having them combined by solder would reduce their potential profit, not increase it.
Also, unless the signal integrity issues are truly brutal, it wouldn't be terribly difficult to produce a CPU that is designed to be 'zillion-little-BGA-balls-permanently-attached' for volume constrained embedded applications and also produce a little PCB card that has an array of BGA pads on top and an array of LGA lands on the bottom, allowing you to turn your BGA-only CPU into a socketed CPU at modest additional expense.
Given the uptick in tablets, ultrathin laptops, and 'every CPU manufactured in the past 5 years is faster than I need' cheapy desktops, I certainly wouldn't bet on CPU sockets getting any more common; but it seems unlikely that sockets would be killed entirely in the more expensive areas.