Most of Atom's advantage has to do with its relatively enormous memory bandwidth compared to mobile SoC's. When looking at DMIPS/MHz of relatively cache-bound tasks (Dhrystone, for instance), Atom's clock-for-clock performance is about the same as a Cortex A8.
Atom, when used in a netbook, can of course afford to use high-power memory modules as well as high-powered multi-channel memory interfaces. Most dual-core SoC's announced (except OMAP4) use single-channel, 32-bit, LP-DDR2.
I believe the reason it's called the A15 is that it's a new tier. That is, it won't replace the current A9 (or future revisions, A10 for instance) that is targeted at smartphone-level performance and power.
The numbers released for the A15 do not fit into a smartphone's power envelope. DMIPS/Watt is significantly lower than that of the A9 and it's going to target leaky, high-speed processes first.
Eagle (A15) is ARM's push towards netbooks, tablets and notebooks (and possibly servers). Sure, some people may try to fit it into a smartphone chip (I'm looking at you, nVidia) but the thing's a beast both area and power wise. The damn thing takes 12 stages just to fetch and dispatch an instruction with another 3-12 to execute and write-back. That's a 24-stage pipeline.
For reference, the A9's integer pipeline is 8-stages. The first Pentium 4's had a 20-stage integer pipeline.
I believe the ARM license went with the xscale team over to Marvell. Intel can, of course, purchase another architectural license. But MS isn't going to *stop* supporting x86; they're just adding ARM.
While I'm not going to comment on the specific claim of a 900MHz ARM vs a 1.5GHz x86, you can cherry pick different implementations all you want to show different results. There are no ARM implementations with the amount of parallelism of a Sandy Bridge core (yet).
However, looking at similar implementations targeting similar performance goals, for instance, a Cortex A8 compared to Atom (both are dual-issue, in-order, with the Atom having a slightly longer integer pipeline), there have been studies comparing different workloads on a Cortex A8 to Atom.
For tasks that relied on heavy memory, Atom of course, won out. But for tasks that were more CPU-bound (Dhrystone), a Cortex A8 achieved somewhere around 1.47 DMIPS/MHz compared to Atom 330's 1.17.
The A9 should significantly improve on this and at 1GHz, will likely be as fast if not faster than a 1.6GHz Atom at less than half the power.
The A15 should close the gap even further and rival low-tier Core CPU's. I'm not sure what nVidia's going to make.
All in all, ARM implementations of far less complexity and power consumption are achieving performance on par, if not better, than what Intel can offer....for now at least.
I'm not sure how you arrived at that conclusion. Most of the benchmarks of real-world games -- while they are a bit rare -- show the Tegra 2 as being slightly faster; take Quake 3, for instance. Neocore is particularly slow on Tegra 2 because it's a benchmark made by Qualcomm that targets tile-based rendering and also offloads a lot of the FP workload to the CPU (where Tegra 2 is particularly bad).
Realistically, Tegra 2 is pretty much a beast of a chip, but I'm curious about its power numbers.
It'd take a lot for WiFi to be "ubiquitous" to the same level as 3G is now. In major metropolitan areas, it's possible but people outside of cities still want their mobile data.
I think net neutrality as it's currently being proposed simply treats a symptom, not the cause. The cause is that telecom and cable companies have virtual monopolies due to access to public property that is granted to them exclusively. This allows them to pull whatever shenanigans they please; not just prioritized service. They have fiber wire just laying there unused because they don't feel the need to compete -- there's no one else around after all. They can price their plans ridiculously high (Verizon Wireless, I'm looking at you) without fear since they have sole access to a creme de la creme slice of spectrum that no one else is allowed.
If we truly want change, the various levels of governments need to pool some public funds into developing line and spectrum sharing technology. Plenty of non-profit organizations out there (EFF for instance) can provide technical input and ways to improve it in the future.
Then and only then will there be true competition when it comes to the internet. Anyone can start a cable company if they choose to invest in the routing/server infrastructure. Communities would be free to setup their own service with one giant pipe to the nearest hub. Competition will keep prices down and force companies to improve service instead of sitting back and raking in cash (and charging all sorts of nebulous, borderline scam fees).
I keep wondering about this. Why do laptops not come with a switch to cut off the charge when the battery is, say, 99% full? Is it purposely so that the battery will die faster? The laptop works without a battery (while plugged in) so the regulator obviously is capable of handling both AC and battery levels of voltage.
So why are they constantly destroying batteries while plugged in?
You have to wonder, with all of these devices that use the rather spread-thin (no pun intended) "free" spectrum, just how much bandwidth is left in your average area? What with everyone's WiF, cordless phone, microwave, bluetooth, etc. devices running constantly, are we reaching a point where "free" consumer spaces simply need more bandwidth?
Given an option between an API that gives me optimized graphical UI functions and one in which I'd have to create myself, I think I know which one to go with.
I believe, as a developer of apps, he means the API's he has available to use those effects in *his app*. Not just the UI overlay that Samsung or HTC provides.
The iPhone4, as regarded by most of the tech community, was a bad design choice. That doesn't mean it's a consistent thing for Apple. The 3Gs, for instance, had form and function that was top of its class.
In fact, even in RF tests, it had one of the highest signal-to-noise ratios both when holding lightly as well as while "death gripping".
For the most part -- with a few exceptions -- I'd say Apple's recent products in the past ~4 or so years have more or less followed along the lines of "it works well and is pretty to boot". There are, of course, exceptions; such as the iPhone4.
You must've missed the other part of the argument: a crowd of technical people trying to make a consumer product. Engineers are great at what they do: making things work. But they aren't necessarily good at things like figuring out what sells and what the average person would want. They also tend to be less receptive towards suggestions from people they deem "beneath them" (read: marketing).
So say you have this brilliant idea about how to make an portable music player. You put in a chip with an insane graphics processor but a dog slow CPU but from an engineering perspective, "an extra 1/10 sec lag for a menu choice isn't much, but this thing can render incredible UI graphics!".
The marketing or artsy guy comes along and says "that's great, but 99% of the people will see that 1/10 sec of lag, and conclude "this thing is slow"."
At Microsoft, that guy would be told to go back to drinking latte's.
Apple leaves the hard engineering up to other people, be it the Darwin group, Samsung or Infineon (soon to be Qualcomm, apparently). Because those are the types of companies where the product's technical abilities is by and large the selling point and those companies should be run by engineers.
For a company that tries to make consumer goods, it can't be run adequately by engineers. And that's what Microsoft is trying to do.
See that's the thing. People -- including myself -- don't *want* to manually tag stuff. It's the difference between a computer geek and a computer user. The user just wants everything done for him/her with the possibility of tweaking how the machine makes its decisions.
I wouldn't put it past Google to work on advanced file auto-classification algorithms that will one day make "oh I gotta put this in my music folder" obsolete.
Of course, then there's the "wait, don't name that 'porn' automatically!!" aspect....
One of the important UI changes about a touch-only interface is that things such as managing the filesystem, arranging folders and icons, etc. are too cumbersome to do in the traditional navigator window type of interface.
iOS just gets rid of it altogether whereas Android limits you to handling files via applications. Unless they've managed to come up with a proper auto-categorization and file organizer -- such that I don't need to go through folders to get to a media file I want to play -- this will still be a cumbersome desktop OS with a touch UI "layer" on top.
It's a huge advantage but it's only one part of the game. Many would argue that overall strategy is far more important than micro-management.
One of the top protoss players (WhiteRa) in SC2 (former SC1 player) has a relatively low actions-per-minute (APM) count compared to most players. Yet he still comes out on top by a lot.
Being able to multitask and micro-manage is definitely an advantage but a far more important ability is being able to plan on the larger scale. I've never seen an AI capable of harassment techniques, guerrilla warfare and exercising map control. Multi-pronged attacks are also something that it should theoretically be great at but it never really tries.
While I love the new combat system, the AI is simply incapable of playing it well. It would seem that even in this day and age, the idea of forming a front-line to protect your ranged units is something a computer can't grasp. I don't think we'll need to worry about Skynet anytime soon.
On a local level, it's pretty much just that -- mainframes with light clients.
The vision many have for "the cloud" though is that you access it from clients from anywhere, not just your company's intranet. And access from any device would be identical. With wireless internet becoming ubiquitous, this could actually happen.
Slashdot Mirror
Most of Atom's advantage has to do with its relatively enormous memory bandwidth compared to mobile SoC's. When looking at DMIPS/MHz of relatively cache-bound tasks (Dhrystone, for instance), Atom's clock-for-clock performance is about the same as a Cortex A8.
Atom, when used in a netbook, can of course afford to use high-power memory modules as well as high-powered multi-channel memory interfaces. Most dual-core SoC's announced (except OMAP4) use single-channel, 32-bit, LP-DDR2.
I believe the reason it's called the A15 is that it's a new tier. That is, it won't replace the current A9 (or future revisions, A10 for instance) that is targeted at smartphone-level performance and power.
The numbers released for the A15 do not fit into a smartphone's power envelope. DMIPS/Watt is significantly lower than that of the A9 and it's going to target leaky, high-speed processes first.
Eagle (A15) is ARM's push towards netbooks, tablets and notebooks (and possibly servers). Sure, some people may try to fit it into a smartphone chip (I'm looking at you, nVidia) but the thing's a beast both area and power wise. The damn thing takes 12 stages just to fetch and dispatch an instruction with another 3-12 to execute and write-back. That's a 24-stage pipeline.
For reference, the A9's integer pipeline is 8-stages. The first Pentium 4's had a 20-stage integer pipeline.
I believe the ARM license went with the xscale team over to Marvell. Intel can, of course, purchase another architectural license. But MS isn't going to *stop* supporting x86; they're just adding ARM.
While I'm not going to comment on the specific claim of a 900MHz ARM vs a 1.5GHz x86, you can cherry pick different implementations all you want to show different results. There are no ARM implementations with the amount of parallelism of a Sandy Bridge core (yet).
However, looking at similar implementations targeting similar performance goals, for instance, a Cortex A8 compared to Atom (both are dual-issue, in-order, with the Atom having a slightly longer integer pipeline), there have been studies comparing different workloads on a Cortex A8 to Atom.
For tasks that relied on heavy memory, Atom of course, won out. But for tasks that were more CPU-bound (Dhrystone), a Cortex A8 achieved somewhere around 1.47 DMIPS/MHz compared to Atom 330's 1.17.
The A9 should significantly improve on this and at 1GHz, will likely be as fast if not faster than a 1.6GHz Atom at less than half the power.
The A15 should close the gap even further and rival low-tier Core CPU's. I'm not sure what nVidia's going to make.
All in all, ARM implementations of far less complexity and power consumption are achieving performance on par, if not better, than what Intel can offer....for now at least.
I'm not sure how you arrived at that conclusion. Most of the benchmarks of real-world games -- while they are a bit rare -- show the Tegra 2 as being slightly faster; take Quake 3, for instance. Neocore is particularly slow on Tegra 2 because it's a benchmark made by Qualcomm that targets tile-based rendering and also offloads a lot of the FP workload to the CPU (where Tegra 2 is particularly bad).
Realistically, Tegra 2 is pretty much a beast of a chip, but I'm curious about its power numbers.
It'd take a lot for WiFi to be "ubiquitous" to the same level as 3G is now. In major metropolitan areas, it's possible but people outside of cities still want their mobile data.
They don't have the spectrum to do it. Their "partner" is Verizon because Verizon won out a lot of the 700MHz spectrum blocks.
I think net neutrality as it's currently being proposed simply treats a symptom, not the cause. The cause is that telecom and cable companies have virtual monopolies due to access to public property that is granted to them exclusively. This allows them to pull whatever shenanigans they please; not just prioritized service. They have fiber wire just laying there unused because they don't feel the need to compete -- there's no one else around after all. They can price their plans ridiculously high (Verizon Wireless, I'm looking at you) without fear since they have sole access to a creme de la creme slice of spectrum that no one else is allowed.
If we truly want change, the various levels of governments need to pool some public funds into developing line and spectrum sharing technology. Plenty of non-profit organizations out there (EFF for instance) can provide technical input and ways to improve it in the future.
Then and only then will there be true competition when it comes to the internet. Anyone can start a cable company if they choose to invest in the routing/server infrastructure. Communities would be free to setup their own service with one giant pipe to the nearest hub. Competition will keep prices down and force companies to improve service instead of sitting back and raking in cash (and charging all sorts of nebulous, borderline scam fees).
I keep wondering about this. Why do laptops not come with a switch to cut off the charge when the battery is, say, 99% full? Is it purposely so that the battery will die faster? The laptop works without a battery (while plugged in) so the regulator obviously is capable of handling both AC and battery levels of voltage.
So why are they constantly destroying batteries while plugged in?
With almost all of its revenue coming from internet ads, what kind of sales team does Google need in Manhattan that warrants $1.9 billion?
Redundant invention is redundant
You have to wonder, with all of these devices that use the rather spread-thin (no pun intended) "free" spectrum, just how much bandwidth is left in your average area? What with everyone's WiF, cordless phone, microwave, bluetooth, etc. devices running constantly, are we reaching a point where "free" consumer spaces simply need more bandwidth?
Given an option between an API that gives me optimized graphical UI functions and one in which I'd have to create myself, I think I know which one to go with.
Would that be why they just released Facebook-for-music on iTunes?
The problem is Samsung in this case. Most if not all HTC and Motorola high-end phones have 2.2 already and have no GPS issues.
I agree with a lot of the Android criticism but those issues you listed are specific to Samsung.
I believe, as a developer of apps, he means the API's he has available to use those effects in *his app*. Not just the UI overlay that Samsung or HTC provides.
But go on knee-jerking.
The iPhone4, as regarded by most of the tech community, was a bad design choice. That doesn't mean it's a consistent thing for Apple. The 3Gs, for instance, had form and function that was top of its class.
In fact, even in RF tests, it had one of the highest signal-to-noise ratios both when holding lightly as well as while "death gripping".
For the most part -- with a few exceptions -- I'd say Apple's recent products in the past ~4 or so years have more or less followed along the lines of "it works well and is pretty to boot". There are, of course, exceptions; such as the iPhone4.
You must've missed the other part of the argument: a crowd of technical people trying to make a consumer product. Engineers are great at what they do: making things work. But they aren't necessarily good at things like figuring out what sells and what the average person would want. They also tend to be less receptive towards suggestions from people they deem "beneath them" (read: marketing).
So say you have this brilliant idea about how to make an portable music player. You put in a chip with an insane graphics processor but a dog slow CPU but from an engineering perspective, "an extra 1/10 sec lag for a menu choice isn't much, but this thing can render incredible UI graphics!".
The marketing or artsy guy comes along and says "that's great, but 99% of the people will see that 1/10 sec of lag, and conclude "this thing is slow"."
At Microsoft, that guy would be told to go back to drinking latte's.
Apple leaves the hard engineering up to other people, be it the Darwin group, Samsung or Infineon (soon to be Qualcomm, apparently). Because those are the types of companies where the product's technical abilities is by and large the selling point and those companies should be run by engineers.
For a company that tries to make consumer goods, it can't be run adequately by engineers. And that's what Microsoft is trying to do.
See that's the thing. People -- including myself -- don't *want* to manually tag stuff. It's the difference between a computer geek and a computer user. The user just wants everything done for him/her with the possibility of tweaking how the machine makes its decisions.
I wouldn't put it past Google to work on advanced file auto-classification algorithms that will one day make "oh I gotta put this in my music folder" obsolete.
Of course, then there's the "wait, don't name that 'porn' automatically!!" aspect....
One of the important UI changes about a touch-only interface is that things such as managing the filesystem, arranging folders and icons, etc. are too cumbersome to do in the traditional navigator window type of interface.
iOS just gets rid of it altogether whereas Android limits you to handling files via applications. Unless they've managed to come up with a proper auto-categorization and file organizer -- such that I don't need to go through folders to get to a media file I want to play -- this will still be a cumbersome desktop OS with a touch UI "layer" on top.
It's a huge advantage but it's only one part of the game. Many would argue that overall strategy is far more important than micro-management.
One of the top protoss players (WhiteRa) in SC2 (former SC1 player) has a relatively low actions-per-minute (APM) count compared to most players. Yet he still comes out on top by a lot.
Being able to multitask and micro-manage is definitely an advantage but a far more important ability is being able to plan on the larger scale. I've never seen an AI capable of harassment techniques, guerrilla warfare and exercising map control. Multi-pronged attacks are also something that it should theoretically be great at but it never really tries.
I totally agree. The amount of vertical space needs to be tighter.
While I love the new combat system, the AI is simply incapable of playing it well. It would seem that even in this day and age, the idea of forming a front-line to protect your ranged units is something a computer can't grasp. I don't think we'll need to worry about Skynet anytime soon.
On a local level, it's pretty much just that -- mainframes with light clients.
The vision many have for "the cloud" though is that you access it from clients from anywhere, not just your company's intranet. And access from any device would be identical. With wireless internet becoming ubiquitous, this could actually happen.
Both have the exact same chip inside. There's no such thing as a "discrete graphics processor" in mobile phones. It's all one chip.