New PowerVR chip may give Apple's future iPhone HD, OpenCL
Hidden among the many announcements at the Consumer Electronics Show in Las Vegas is a new PowerVR mobile graphics chip that could power advanced 3D, HD video and even general computing on an eventual generation of iPhones.
Imagination Technologies has quietly slipped out word this week of a new mobile chip known as the PowerVR SGX543.
The silicon is the company's fifth generation of graphics accelerators for cellphones and other handhelds and contains a much more efficient instruction set with particular optimizations for rendering graphics shaders, or the visual effects that apply to pixels onscreen and geometry in 3D views. Imagination reckons that its new technology is about 40 percent faster for "shader-heavy" software and can push as many as 50 percent more triangles when handling 3D.
In practice, the technology is powerful enough to push 35 million polygons per second and 1 billion pixels per second and can thus easily drive HD resolution video output, including when 3D is involved.
Moreover, the new SGX chip is very efficient and very scalable, according to the company. That same performance is achievable at a clock speed of just 200MHz, making it relatively power-efficient, but can be improved further still with new multi-core support that would let a device maker join two or more cores together to ramp up the speed in the devices with the physical space and battery power to use the hardware.
But the most important aspect of the PowerVR chip may be the universal nature of its processor. The GPU is capable of handling Apple's now open and virtually finished OpenCL standard for speeding up general computing and letting it move to non-CPU technology such as graphics chips, even on mobile devices.
Imagination has lately been hiring OpenCL engineers to ensure its hardware supports the compute standard.
As of the present, it's unknown whether or not the OpenCL move is being made with Apple's iPhone and iPod touch in mind; the existing iPhone has used an older PowerVR MBX core for its graphics in two generations and isn't currently known to be making the leap to SGX. However, Apple is now a major PowerVR licensee and has boosted the number of shares it holds in Britain-based Imagination, all but committing iPhones to using future PowerVR advancements.
SGX543's inventor pledges to provide more information about the product at the Multicore Expo in March.
Imagination Technologies has quietly slipped out word this week of a new mobile chip known as the PowerVR SGX543.
The silicon is the company's fifth generation of graphics accelerators for cellphones and other handhelds and contains a much more efficient instruction set with particular optimizations for rendering graphics shaders, or the visual effects that apply to pixels onscreen and geometry in 3D views. Imagination reckons that its new technology is about 40 percent faster for "shader-heavy" software and can push as many as 50 percent more triangles when handling 3D.
In practice, the technology is powerful enough to push 35 million polygons per second and 1 billion pixels per second and can thus easily drive HD resolution video output, including when 3D is involved.
Moreover, the new SGX chip is very efficient and very scalable, according to the company. That same performance is achievable at a clock speed of just 200MHz, making it relatively power-efficient, but can be improved further still with new multi-core support that would let a device maker join two or more cores together to ramp up the speed in the devices with the physical space and battery power to use the hardware.
But the most important aspect of the PowerVR chip may be the universal nature of its processor. The GPU is capable of handling Apple's now open and virtually finished OpenCL standard for speeding up general computing and letting it move to non-CPU technology such as graphics chips, even on mobile devices.
Imagination has lately been hiring OpenCL engineers to ensure its hardware supports the compute standard.
As of the present, it's unknown whether or not the OpenCL move is being made with Apple's iPhone and iPod touch in mind; the existing iPhone has used an older PowerVR MBX core for its graphics in two generations and isn't currently known to be making the leap to SGX. However, Apple is now a major PowerVR licensee and has boosted the number of shares it holds in Britain-based Imagination, all but committing iPhones to using future PowerVR advancements.
SGX543's inventor pledges to provide more information about the product at the Multicore Expo in March.
Comments
In fact, the only thing that was newly announced at this CES was the specific "SGX543" unit. The PowerVR "SGX" line has been known about for a few years and has just recently (finally!) been integrated into system-on-a-chip platforms from T.I., Qualcomm, etc. It was actually demo'd at the ARM tradeshow in 2006. Unfortunately, unlike the PC market, the embedded moves slow in getting new stuff to market. The ARM Cortex-A8, which will most likely find it's way into the new iPhone, was also licensed since 2006.
That's it! I'm not shaving till I get this in my next iPhone.
I still don't believe we'll see the fruits of OpenCL and the real reasons for Apple's investment in the company until the 2010 iPhone. This year's new iPhones (maybe new form factor) may have new processors but it just seems to quick that they'd get the OpenCL benefit.
Unfortunately, unlike the PC market, the embedded moves slow in getting new stuff to market. The ARM Cortex-A8, which will most likely find it's way into the new iPhone, was also licensed since 2006.
Well, the Palm Pre has an ARM Cortex A8 as well as a PowerVR accelerator.
Anyways, having worked for several years in the smart phone development industry and with those SOCs, I believe that people seriously underestimate the amount of work that goes into getting new platforms like the iPhone or the Palm Pre up and running, both HW and SW. It is absolutely mind boggling how complex those devices are under the hood and to make everything work smoothly and without glitches is a herculean task.
Well, the Palm Pre has an ARM Cortex A8 as well as a PowerVR accelerator.
Anyways, having worked for several years in the smart phone development industry and with those SOCs, I believe that people seriously underestimate the amount of work that goes into getting new platforms like the iPhone or the Palm Pre up and running, both HW and SW. It is absolutely mind boggling how complex those devices are under the hood and to make everything work smoothly and without glitches is a herculean task.
This is all well and true but really don't you think Apple has been working on a SoC for a long time now? In fact I wouldn't be surprised if Apple started the process not long after starting the iPhone's development. Frankly it should be obvious to everyone that the current processor doesn't really deliver the performance that Apple would want, this they would have seen early on that a follow on processor would be needed.
If you buy any of the above you could see where a SoC research and development for the next iPhone processor has likely been under development for at least two and a half years. So yeah this a big undertaking but one that likely has been underway for a long time. A new device with the fruits of this effort inside is very likely this year.
Now I'm not going to guess as to what cores and GPUs it will have but you can't dismiss recently revealed tech. From Apples standpoint what ever they offer up will be faster and likely lower power. So maybe it ends up being nothing more than an incremental iPhone upgrade.
The other thing is that Apple has varying needs with respect to handhelds so I still have this idea that there are more than one SoC underway. To this many will respond that it is way to expensive. While not cheap I do believe Apple has the device volume to justify the effort. If you look at the iPhone Nano as a real product it should be obvious that it simply won't have the battery power to drive a high performance chip nor the screen realestate to make it worthwhile. So an Appple chip would go into a low end iPhone and iPods. The higher end devices would get a faster more robust device.
So is a custom SoC from PA possible this year? Certainly it just might not be what everybody wants.
Dave
I still don't believe we'll see the fruits of OpenCL and the real reasons for Apple's investment in the company until the 2010 iPhone. This year's new iPhones (maybe new form factor) may have new processors but it just seems to quick that they'd get the OpenCL benefit.
I'm inclined to agree. Rome was not built in a day.
Actually, calling it a Netbook would be misleading. Just imagine a multi-core Netbook capable of running hi-res games, running next-generation, Javascript-heavy web 2.0 apps, and running current generation iPhone apps. (...and imagine a slightly updated version of iPhone OS X on it)
The other thing is that Apple has varying needs with respect to handhelds so I still have this idea that there are more than one SoC underway. To this many will respond that it is way to expensive. While not cheap I do believe Apple has the device volume to justify the effort. If you look at the iPhone Nano as a real product it should be obvious that it simply won't have the battery power to drive a high performance chip nor the screen realestate to make it worthwhile. So an Appple chip would go into a low end iPhone and iPods. The higher end devices would get a faster more robust device.
So is a custom SoC from PA possible this year? Certainly it just might not be what everybody wants.
Dave
I wonder if those many have extensive history with Apple. Chip design isn't new to the company. Back in the days of NUBUS Macs Apple designed a large portion of their ASICS and supporting chips which made Macs better than off the shelf parts but also made them expensive.
What we're seeing is a return to custom parts yet as you say the volume necessary to warrant going down that path. No one understood the PA Semi acquisition at first. All the press could think of was "Apple's going to use the PWRefficient chip in iPhones" which was laughable.
If it's not clear to people where the next gold rush is let's make it clear. It's mobile computing and I'm not talking about laptops. I 4 years the nation will see Wimax and LTE vying for customer dollars and providing wireless bandwidth that is faster than many home broadband connections.
Retail stores will feel the pinch. I think the only music retailers to survive will be the larger Brick n Mortar. Content will be available from thousands of sources in some areas 24/7
Apple is investing in Cloud (yeah yeah I know) infrastructure and aligning with ARM and Imagination to deliver low cost devices.
OS X is being made more efficient (removing PPC and Carbon) and smarter about hardware (GrandCentral and OpenCL)
Apple has 25+ Billion and momentum to position themselves.
If it's not clear to people where the next gold rush is let's make it clear. It's mobile computing and I'm not talking about laptops. I 4 years the nation will see Wimax and LTE vying for customer dollars and providing wireless bandwidth that is faster than many home broadband connections.
I don't quite agree with you there. Notebooks will still be very relevant. Handheld devices are good for consuming content, not so much for creating it. The desktop market will probably shrink quite a bit though. Also, while LTE will be fast, and all devices will likely have an LTE card, fiber optic connections (which should be far more ubiquitous that far in the future) will still be king in the home and will be necessary for a long time.
OS X is being made more efficient (removing PPC and Carbon) and smarter about hardware (GrandCentral and OpenCL)
Removing PPC support has nothing to do with making the OS more efficient. Software benefits from supporting multiple platforms since it leads to greater levels of abstraction which can reduce the number of bugs. Mac OSX currently runs on x86. ARM, PowerPC, and possibly a new secret architecture locked in Apple's labs. The OS is likely abstracted from the hardware enough at this point, and since PowerPC is basically dead as far as Apple is concerned, it just makes sense not to maintain it and the test machines any longer.
Regarding Carbon, while the UI stuff is being retired, Carbon is still used to implement a lot of OS functionality (Audio, the file system, etc). It's going to be a while before enough of Carbon is exposed by CoreFoundation-like frameworks to even deprecate it (and even then it might still run a lot of the same code), so don't expect it to be gone in 10.6 or even 10.7.
I still don't believe we'll see the fruits of OpenCL and the real reasons for Apple's investment in the company until the 2010 iPhone. This year's new iPhones (maybe new form factor) may have new processors but it just seems to quick that they'd get the OpenCL benefit.
You're wrong.
Something folks here [most assuredly due to their lack of OEM background experience] doesn't seem to realize is that most announcements come after 18 months of partner development efforts.
Removing PPC support has nothing to do with making the OS more efficient. Software benefits from supporting multiple platforms since it leads to greater levels of abstraction which can reduce the number of bugs. Mac OSX currently runs on x86. ARM, PowerPC, and possibly a new secret architecture locked in Apple's labs. The OS is likely abstracted from the hardware enough at this point, and since PowerPC is basically dead as far as Apple is concerned, it just makes sense not to maintain it and the test machines any longer.
Regarding Carbon, while the UI stuff is being retired, Carbon is still used to implement a lot of OS functionality (Audio, the file system, etc). It's going to be a while before enough of Carbon is exposed by CoreFoundation-like frameworks to even deprecate it (and even then it might still run a lot of the same code), so don't expect it to be gone in 10.6 or even 10.7.
Let me clarify. Removing PPC support means one less platform to have to test and debug. I do not meant to postulate that removing PPC support intrinsically makes OS X better/worse but that it's one less entity to manage. It'll be nice from an end user perspective to not have to download Intel/PPC binaries in some apps.
Carbon is certainly going to be around but the train has moved on. 64-bit Cocoa or bust. I'm not a Cocoa is God guy but I want one major framework to be promoted for clarity's sake.
You're wrong.
Something folks here [most assuredly due to their lack of OEM background experience] doesn't seem to realize is that most announcements come after 18 months of partner development efforts.
I do find it intriguing that people assume that a press release is the first time licensee get word of a new product. I imagine designing a SoC is a lot like designing software. You work on alpha and beta product and tweak your design all the way up to the final release.
That's it! I'm not shaving till I get this in my next iPhone.
Well.....well I'm not going to take a shower either then.......ever!
You're wrong.
Something folks here [most assuredly due to their lack of OEM background experience] doesn't seem to realize is that most announcements come after 18 months of partner development efforts.
So it is highly likley that the iphone 3g's sucessor will feature this new hardware?
So it is highly likley that the iphone 3g's sucessor will feature this new hardware?
Doubtful that the 54x series will make the nextgen iPhone.
Guesses are that we'll see a PowerVR SGX 53x series mated to a faster ARM core in the
next revision.
The 2010 iPhone will be a powerhouse. We could see a custom chip based on ARM Cortex processing and PowerVR SGX 54X chips.
Apple could also deliver a Netbook class processor in 2010 with a ARM Cortex MPA9 and a PowerVR SGX 54x running full on OS X. That would be a quick little device.
So it is highly likley that the iphone 3g's sucessor will feature this new hardware?
You misunderstand me. They've been working on it for 18 months. The 2009 successor will be OpenCL enabled.
I wonder if those many have extensive history with Apple. Chip design isn't new to the company. Back in the days of NUBUS Macs Apple designed a large portion of their ASICS and supporting chips which made Macs better than off the shelf parts but also made them expensive.
It is not a given at all that custom chips will make a device more expensive. Rather there are many ifs that have to be summed up. In the end fully custom is the way to go in most cases, the biggest if being your volume.
It should be noted that Apple has the volume when iPhone and tablets are combined. Easily in fact as the volume there far outstrips the computer line up.
What we're seeing is a return to custom parts yet as you say the volume necessary to warrant going down that path. No one understood the PA Semi acquisition at first. All the press could think of was "Apple's going to use the PWRefficient chip in iPhones" which was laughable.
I'm not aware of such reporting as it was made clear by Apple that PPC wasn't in the equation.
If it's not clear to people where the next gold rush is let's make it clear. It's mobile computing and I'm not talking about laptops. I 4 years the nation will see Wimax and LTE vying for customer dollars and providing wireless bandwidth that is faster than many home broadband connections.
This I agree with, with one caveat that being the right product coming to market. I believe Apple is on the right track here but they need a whole family of pocketable and portable devices.
Retail stores will feel the pinch. I think the only music retailers to survive will be the larger Brick n Mortar. Content will be available from thousands of sources in some areas 24/7
Apple is investing in Cloud (yeah yeah I know) infrastructure and aligning with ARM and Imagination to deliver low cost devices.
Mobile Me has been great for me even if there has been a few glitches. I like the way my iPhone works with my MBP via ME. The ability to get to Mobile Me from just about anywhere else is very handy, this from a die hard linux fan.
It is not just Mobile Me either, as one of the original motivators was iTunes. If Apple gets involved similarly with books and magazines we will be all set.
OS X is being made more efficient (removing PPC and Carbon) and smarter about hardware (GrandCentral and OpenCL)
Apple has 25+ Billion and momentum to position themselves.
I just wish that they avoid squandering that money on low pay back efforts and focus on devices at a price that can't be beat. One of Apples biggest problems is their sometimes excessive focus on minimalism. They need to offer value in these new markets to dominate.
Dave
Let me clarify. Removing PPC support means one less platform to have to test and debug. I do not meant to postulate that removing PPC support intrinsically makes OS X better/worse but that it's one less entity to manage. It'll be nice from an end user perspective to not have to download Intel/PPC binaries in some apps.
Fat binaries are certainly a waste of space so that is an improvement right off the bat.
As to building an OS for different ISAs well that in an of itself seems to be a good thing. It certainly has helped bring Linux to the point it is today. Then again the Linux crowd doesn't do fat binaries. My position is that supporting Mac OS on differrent ISAs is a positive thing for Apple in the same way it is positive for the Linux crowd.
In Apples case a portable OS makes for the flexibility to do things like iPhone. I would be surprised if we don't see MacOS embedded in more products from Apple in the future. More importantly the cross building and optimisations act as fertilizer to grow a better mainline Mac OS kernel. Note I'm playing a little loose here with the term Mac OS as we all know the portable devices are different.
Carbon is certainly going to be around but the train has moved on. 64-bit Cocoa or bust. I'm not a Cocoa is God guy but I want one major framework to be promoted for clarity's sake.
I do find it intriguing that people assume that a press release is the first time licensee get word of a new product. I imagine designing a SoC is a lot like designing software. You work on alpha and beta product and tweak your design all the way up to the final release.
Yeah this is bothersome
People should realize that these are major deals or developments that take a lot of time to get ready for public consumption. Further, iPhone is basically generation ONE and was developed over a long period of time. Due to gen one performance issues I would suspect that the process to develop a custom chip started soon after the engineering started on the iPhone. I'm really holding out for mid summer but it could be earlier than that.
The only problem is guessing at what the new products will be. A smaller iPhone makes a lot of sense as does a larger. That is just dedicated iPhone's, the potential for other products is huge. Personally I'm all for the larger iPhone. As to a tablet device I'm still thinking they need to start at paperback book size and go from there.
Dave
The first generation may not have the SGX that is capable of multi-core, but it might.
Actually, calling it a Netbook would be misleading. Just imagine a multi-core Netbook capable of running hi-res games, running next-generation, Javascript-heavy web 2.0 apps, and running current generation iPhone apps. (...and imagine a slightly updated version of iPhone OS X on it)
Where is kormac when you need him?
Doubtful that the 54x series will make the nextgen iPhone.
Guesses are that we'll see a PowerVR SGX 53x series mated to a faster ARM core in the
next revision.
The 2010 iPhone will be a powerhouse. We could see a custom chip based on ARM Cortex processing and PowerVR SGX 54X chips.
Apple could also deliver a Netbook class processor in 2010 with a ARM Cortex MPA9 and a PowerVR SGX 54x running full on OS X. That would be a quick little device.
Excuse the lack of technical points here but I believe the PalmPre is running on OMAP3? The first phone to do so I think? Im sure it's only a matter of time before a certain Finnish manufacturer makes use of this also and at the same time support for OpenGL2.0.
Will the 54x chip be able to compete with that?