Just please, Devs, don't do what you've done with the iPhone so far. The iPad runs insanely fast right now, but give Devs an idea of how much power they have to work with, and you quickly get apps that push the boundaries of the hardware and ultimately slow everything down. This has already occurred with each iPhone release since the App Store opened. A year ago my 3GS ran games really fast. Then the apps "designed" to take advantage of the OpenCL, mostly run good, but use way too much memory at one time.
Truth be told, not all developers are created equal, and not every app is well done. I love playing Angry Birds, but the guys behind the app really don't understand the hardware they've built it for..its uses over 70 mb of RAM, which is grotesque on these devices, and is immediately obvious why. Gameplay is choppy and aggravating, both on the iPhone and on the iPad. The recent iphone update improved matters, but the launch of Angry Birds leaves a lot to be desired. I cant blame them since they created it without testing on a physical iPad (allegedly), but it's the same game with different PNGs. Shouldn't have been that hard to figure out...
Uhh, gameplay is choppy on an iPad, a device they didn't own for testing before submitting the application? Umm, duh. If you didn't understand that, then all the other stuff you said is crap. And it IS hard to figure out, by the way. It's not just the same game with different PNGs. Just shows you've never developed an iPhone app before.
And what apps take advantage of OpenCL? There are no OpenCL iPhone/iPad libraries.
As an iPhone 3GS owner, but not iPad (we don't even have a price announced yet on this side of the pond), am I being too hopeful? The performance figures do look good for the iPad, such that I wonder if it's all down to hardware. The iPad has clearly sparked a lot of work to the OS, it could be that there are some performance tweaks in there too.
It seems likely that Apple will maintain as much commonality as possible with iPhone so there may soon be an iPhone OS update and it might bring some speed boost. Are there any Snow Leopard or Safari speed-ups that don't yet figure on the iPhone... I hope.
It sounds like the iPad is a terrific combination of software and hardware, and will likely only improve as time goes on.
Just as the the iPhones 3Gs was better than the original iPhone, and subsequent iPod Touches improved on memory and performance in the higher end models, so will the "iPad Family" be positioned for the same kind of speed, performance and RAM upgrades. Time will tell.
Given that just about every computing device made to date include the Macintosh, iBook, iMac, generic PCs, iPods, mainframes etc. etc. has followed that performance path, that is a bit like saying that the sun will probably rise tomorrow.
As I remember reading many years ago about model 'X' of any computer (and their tongue was clearly in their cheek, but ..) "If it works, it's obsolete".
I'm having a hard time understanding why a 2x performance increase is that surprising.
The iPhone 3GS allegedly has a 833MHz Cortex-A8-based processor underclocked to 600MHz. Apple has publicly quoted a 1GHz clock speed for the iPad's A4 processor. Assuming that the article is correct in noting increased memory throughput, the average 2x improvement over a number of tests seems quite plausible.
I know, it seems odd the article didn't mention the huge difference in clock speed. That alone accounts for the speed difference. The RAM just adds some extra performance to push it about 2x faster. Nothing really surprising.
I'm having a hard time understanding why a 2x performance increase is that surprising.
The iPhone 3GS allegedly has a 833MHz Cortex-A8-based processor underclocked to 600MHz. Apple has publicly quoted a 1GHz clock speed for the iPad's A4 processor. Assuming that the article is correct in noting increased memory throughput, the average 2x improvement over a number of tests seems quite plausible.
Same thought for me. If you clock the processor from 600MHz to 1GHz of course it is going to be nearly 2x faster. Duh
In a related topic, the Nokia N900 uses exactly the same Cortex-A8 processor at 600Mhz and people have already figured out how to overclock that to 1GHz.
The other shoe hasn't dropped yet. It's performance per watt that is the indicator of Apple doing something interesting with the A4. It increased the clock rate by 67%, and likely got another 10 to 20% from 64 bit data paths. And is likely fabbed on a 45 nm process. If it offers better perf/watt than Snapdragon (it should as it already performs better per clock and is supposedly fabbed on a more advanced process), but the real comparison is when Samsung (or someone) ships the Samsung 1 GHz Cortex-A8 based SoC.
If it's perf/watt is better than that, then Apple's design work with the A4 is a winner.
So wait, it's a 64-bit ARM processor? I thought ARM was normally a 32-bit processor architecture?
You're confusing the memory data path (rumored to be 64bit) with CPU instruction size (still 32 bit).
And while on the subject, 64 bit insructions in devices that haven't even gotten over 512MB of system RAM is wholly unnecessary. 64 bit increases the number of transistors for a CPU in a pretty negative way for mobile/imbedded devices. What is non-trivial or even beneficial on a desktop or notebook is a very big deal on a phone or something like an iPad.
There is no doubt that 64 bit memory accesses would help performance but so would a lot of other things. Plus I'm not convinced it would double performance even if that was the only thing done.
I'd prefer to wait for more extensive benchmarks especially things like OpenGL. Here is one consideration, much of the performance could be coming from a faster GPU. Think about it the GPU is really wired deeply into iPhones OS. Plus in the iPhone the GPU core runs relatively slow, so they could simply double the speed of the GPU core.
As far as raw processor performance that could be something as simple as a larger cache attached to the CPU memory. A few other improvements might even have been implemented. The whole chip though does sound rather average at the moment, this has me wondering just how much imput PA Semi had in the design.
Dave
As you said it's probably a combination. Just going from 600MHz to 1GHz already accounts for 66% of the theoretical performance increase. The rest could be split across doubled memory bandwidth, maybe doubled L2 cache, and OS optimizations. The iPhone 3GS was running iPhone OS 3.0 not the latest 3.1.3 and the iPad's 3.2 could have more optimizations. Improvements to memory bandwidth would probably actually benefit the GPU more than the CPU.
Although the A4 doesn't really bring anything spectacularly new to the table, it does make sense for PA Semi's first design with Apple. They are using proven technology that they are familiar with in the Cortex A8 and SGX535 and may even be able to get some help from Samsung on the side through their existing iPhone 3GS and 3rd gen iPod Touch SoC contract. They still get some experience re-laying out transistors and finding critical paths as they optimize the design for the 65nm to 45nm die shrink. It should set the stage for something more aggressive in their next design. In them meantime, Apple's probably binning slower A4s for the next iPhone, probably at 800MHz although they may stick to a 32-bit memory bus for power reasons.
Personally, I'm interested in seeing Apple's design rhythm. Apple originally used the same ARM11 SoC for 2 generations and 2 years. Now they are using the same basic Cortex A8 design for 2 generations and 2 years but with what looks like a die shrink in the middle. Is Apple going to try to carry over the same SoC for 2 years like with the ARM11, do Tick-Tock new design and shrink like Intel or try to get ahead of the industry by making a new design every year? They all have their problems, the ARM11 approach being technologically lagging in hardware although Apple could presumably make up for it in software, the Tick-Tock approach being risky since it's dependent on someone else's process Fab schedule which given the TSMC 40nm GPU problems and increasing difficulty of smaller processes should give Apple pause, while a new design every year regardless of process mitigates process risk but has significant development cost.
Impressive speeds marks indeed. The tear-down on iFixit showed an elegant system not just on the outside but also on the inside. Compare that to the tear-down of the JooJoo and you should no doubt understand Apple's lead in engineering and design. The JooJoo's innards look pretty dated by comparison. Apparently the JooJoo's innards aren't the only thinks that suck as the Engadget reviewer complained of major bugs in the system.
There's no competing with the iPad with the current crop of slate computers.
OMFG That has no refinement to it at all. You almost think it has to have some juju in it to even work. You can sure that it wouldn't pass the "green" test.
The video of it show a very frustrating system to navigate. I wish them well in this market as they beat Notion Ink, beat HP Slate and all the others at CES. Also it's not using Windows which I'd say has always been the primary downfall of previous tablets.
Despite apparently using the same single core, Cortex A8 generation of its ARM processor core, the same Power VR SGX 535 graphics core, and the same amount of system RAM, the iPad is around twice as fast as the iPhone 3GS in running native Cocoa Touch apps.
How is that anyway "apparent", AI?
No, no, that's just a perfect illustration why you've lost all serious readers and why you have to advertise yourself on TUAW to keep trolls at least...
No, no, that's just a perfect illustration why you've lost all serious readers and why you have to advertise yourself on TUAW to keep trolls at least...
My iPad is easily 4x faster than my 3G iPhone, if not more. Additionally, my 3G iPhone hangs up for 5-10 seconds on a daily basis, presumably clearing memory or something when new apps are launched; the iPad never misses a beat. The ability to scroll smoothly even when a image-heavy webpage is loading even makes it feel faster than my MacBook Pro at times.
Comments
Just please, Devs, don't do what you've done with the iPhone so far. The iPad runs insanely fast right now, but give Devs an idea of how much power they have to work with, and you quickly get apps that push the boundaries of the hardware and ultimately slow everything down. This has already occurred with each iPhone release since the App Store opened. A year ago my 3GS ran games really fast. Then the apps "designed" to take advantage of the OpenCL, mostly run good, but use way too much memory at one time.
Truth be told, not all developers are created equal, and not every app is well done. I love playing Angry Birds, but the guys behind the app really don't understand the hardware they've built it for..its uses over 70 mb of RAM, which is grotesque on these devices, and is immediately obvious why. Gameplay is choppy and aggravating, both on the iPhone and on the iPad. The recent iphone update improved matters, but the launch of Angry Birds leaves a lot to be desired. I cant blame them since they created it without testing on a physical iPad (allegedly), but it's the same game with different PNGs. Shouldn't have been that hard to figure out...
Uhh, gameplay is choppy on an iPad, a device they didn't own for testing before submitting the application? Umm, duh. If you didn't understand that, then all the other stuff you said is crap. And it IS hard to figure out, by the way. It's not just the same game with different PNGs. Just shows you've never developed an iPhone app before.
And what apps take advantage of OpenCL? There are no OpenCL iPhone/iPad libraries.
It seems likely that Apple will maintain as much commonality as possible with iPhone so there may soon be an iPhone OS update and it might bring some speed boost. Are there any Snow Leopard or Safari speed-ups that don't yet figure on the iPhone... I hope.
It sounds like the iPad is a terrific combination of software and hardware, and will likely only improve as time goes on.
Just as the the iPhones 3Gs was better than the original iPhone, and subsequent iPod Touches improved on memory and performance in the higher end models, so will the "iPad Family" be positioned for the same kind of speed, performance and RAM upgrades. Time will tell.
Given that just about every computing device made to date include the Macintosh, iBook, iMac, generic PCs, iPods, mainframes etc. etc. has followed that performance path, that is a bit like saying that the sun will probably rise tomorrow.
As I remember reading many years ago about model 'X' of any computer (and their tongue was clearly in their cheek, but ..) "If it works, it's obsolete".
It's also a great solution for those looking to multitask on the iPhone.
that made me laugh, thanks RichL
I'm having a hard time understanding why a 2x performance increase is that surprising.
The iPhone 3GS allegedly has a 833MHz Cortex-A8-based processor underclocked to 600MHz. Apple has publicly quoted a 1GHz clock speed for the iPad's A4 processor. Assuming that the article is correct in noting increased memory throughput, the average 2x improvement over a number of tests seems quite plausible.
I know, it seems odd the article didn't mention the huge difference in clock speed. That alone accounts for the speed difference. The RAM just adds some extra performance to push it about 2x faster. Nothing really surprising.
I'm having a hard time understanding why a 2x performance increase is that surprising.
The iPhone 3GS allegedly has a 833MHz Cortex-A8-based processor underclocked to 600MHz. Apple has publicly quoted a 1GHz clock speed for the iPad's A4 processor. Assuming that the article is correct in noting increased memory throughput, the average 2x improvement over a number of tests seems quite plausible.
Same thought for me. If you clock the processor from 600MHz to 1GHz of course it is going to be nearly 2x faster. Duh
In a related topic, the Nokia N900 uses exactly the same Cortex-A8 processor at 600Mhz and people have already figured out how to overclock that to 1GHz.
If it's perf/watt is better than that, then Apple's design work with the A4 is a winner.
Also still waiting on graphics performance too.
So wait, it's a 64-bit ARM processor? I thought ARM was normally a 32-bit processor architecture?
You're confusing the memory data path (rumored to be 64bit) with CPU instruction size (still 32 bit).
And while on the subject, 64 bit insructions in devices that haven't even gotten over 512MB of system RAM is wholly unnecessary. 64 bit increases the number of transistors for a CPU in a pretty negative way for mobile/imbedded devices. What is non-trivial or even beneficial on a desktop or notebook is a very big deal on a phone or something like an iPad.
Bigger isn't always better!
There is no doubt that 64 bit memory accesses would help performance but so would a lot of other things. Plus I'm not convinced it would double performance even if that was the only thing done.
I'd prefer to wait for more extensive benchmarks especially things like OpenGL. Here is one consideration, much of the performance could be coming from a faster GPU. Think about it the GPU is really wired deeply into iPhones OS. Plus in the iPhone the GPU core runs relatively slow, so they could simply double the speed of the GPU core.
As far as raw processor performance that could be something as simple as a larger cache attached to the CPU memory. A few other improvements might even have been implemented. The whole chip though does sound rather average at the moment, this has me wondering just how much imput PA Semi had in the design.
Dave
As you said it's probably a combination. Just going from 600MHz to 1GHz already accounts for 66% of the theoretical performance increase. The rest could be split across doubled memory bandwidth, maybe doubled L2 cache, and OS optimizations. The iPhone 3GS was running iPhone OS 3.0 not the latest 3.1.3 and the iPad's 3.2 could have more optimizations. Improvements to memory bandwidth would probably actually benefit the GPU more than the CPU.
Although the A4 doesn't really bring anything spectacularly new to the table, it does make sense for PA Semi's first design with Apple. They are using proven technology that they are familiar with in the Cortex A8 and SGX535 and may even be able to get some help from Samsung on the side through their existing iPhone 3GS and 3rd gen iPod Touch SoC contract. They still get some experience re-laying out transistors and finding critical paths as they optimize the design for the 65nm to 45nm die shrink. It should set the stage for something more aggressive in their next design. In them meantime, Apple's probably binning slower A4s for the next iPhone, probably at 800MHz although they may stick to a 32-bit memory bus for power reasons.
Personally, I'm interested in seeing Apple's design rhythm. Apple originally used the same ARM11 SoC for 2 generations and 2 years. Now they are using the same basic Cortex A8 design for 2 generations and 2 years but with what looks like a die shrink in the middle. Is Apple going to try to carry over the same SoC for 2 years like with the ARM11, do Tick-Tock new design and shrink like Intel or try to get ahead of the industry by making a new design every year? They all have their problems, the ARM11 approach being technologically lagging in hardware although Apple could presumably make up for it in software, the Tick-Tock approach being risky since it's dependent on someone else's process Fab schedule which given the TSMC 40nm GPU problems and increasing difficulty of smaller processes should give Apple pause, while a new design every year regardless of process mitigates process risk but has significant development cost.
Impressive speeds marks indeed. The tear-down on iFixit showed an elegant system not just on the outside but also on the inside. Compare that to the tear-down of the JooJoo and you should no doubt understand Apple's lead in engineering and design. The JooJoo's innards look pretty dated by comparison. Apparently the JooJoo's innards aren't the only thinks that suck as the Engadget reviewer complained of major bugs in the system.
There's no competing with the iPad with the current crop of slate computers.
OMFG That has no refinement to it at all. You almost think it has to have some juju in it to even work.
The video of it show a very frustrating system to navigate. I wish them well in this market as they beat Notion Ink, beat HP Slate and all the others at CES. Also it's not using Windows which I'd say has always been the primary downfall of previous tablets. But why the hell did they launch the same time as the iPad? It's almost like they wanted to fail.
Despite apparently using the same single core, Cortex A8 generation of its ARM processor core, the same Power VR SGX 535 graphics core, and the same amount of system RAM, the iPad is around twice as fast as the iPhone 3GS in running native Cocoa Touch apps.
How is that anyway "apparent", AI?
No, no, that's just a perfect illustration why you've lost all serious readers and why you have to advertise yourself on TUAW to keep trolls at least...
How is that anyway "apparent", AI?
I that a serious remark? If so, why?
No, no, that's just a perfect illustration why you've lost all serious readers and why you have to advertise yourself on TUAW to keep trolls at least...
Are you including yourself there?