TI demos OMAP5 . . . not bad

[Bateluer]

http://www.engadget.com/2012/01/12/ti-omap-5-exclusive-demo-laptops-ultrabooks-ces-2012-video/

From the VP of OMAP.

"This is the greatest platform on Earth right now... way ahead of Apple, and it's the first Cortex-A15 (which runs 2x faster than the Cortex-A9) product on the market. When running two Cortex-A15 chips at 800MHz, it's more or less the same performance as running two Cortex-A9s at 1.5GHz. You'll see [commercially available products] ramping up with this stuff in late 2012 or early 2013.

Not too shabby. Keep in mind that inch thick brick is a demonstration model and won't actually be a commercial product.

 

[smartpatrol]

It's amazing how quickly mobile tech is advancing. Another year or two, our smartphones/tablets will be as powerful as Xbox 360.

 

[Red Storm]

It's amazing how quickly mobile tech is advancing. Another year or two, our smartphones/tablets will be as powerful as Xbox 360.

And yet with the same level (or more likely, worse) battery life. :|

 

[Mopetar]

The quote confuses me. He says it's the first Cortex A-15 product on the market, but as far as I know nothing uses it right now and there's no clear indication when the first product to use it will launch. It's true that it's faster than the A5, but I'm assuming that Apple will have a new chip in their next iPad, whenever that launches. Assuming that the next iPad comes out around March-April, even if this new OMAP is faster than whatever the A6 turns out to be, it will be the fastest chip on the market that's not actually available in any products for another 7 - 12 months.

I don't have any information about Apple's next chip, but some have speculated that it will also be a Cortex A-15 design so I can't see them being considered way ahead unless they've really customized the design. Historically, even the most heavily customized ARM chips have usually managed around 10% at best. Perhaps this can be further improved with the A-15, but it seems unlikely that TI is going to have something that can be qualified (or maybe even quantified) as "way ahead".

I suppose that one can't fault a person for hyping his own product up. I'd probably do the same in his position, but from my own position, it's hard to take his claims at face value.

 

[Mopetar]

It's amazing how quickly mobile tech is advancing. Another year or two, our smartphones/tablets will be as powerful as Xbox 360.

Considering that the CPU in the Xbox 360 was originally fabricated on a 90 nm process, and that we'll be having 22 nm parts shipping in a few months, which represents a 16x reduction in die size, it's not that out of line. This also doesn't take design improvements into account, which means that architectures become more efficient over time as well.

The Xbox 360 CPU had 165 million transistors and the GPU had 232 million transistors. Some of the ARM SoCs that I could find information on had 200+ million transistors, although all of those are not dedicated to the CPU or GPU, so it doesn't seem as though we're too much farther away from seeing that level of power in our devices.

 

[Bateluer]

And yet with the same level (or more likely, worse) battery life. :|

The Xbox 360 runs on batteries now? What'd I miss?

 

[Puddle Jumper]

The quote confuses me. He says it's the first Cortex A-15 product on the market, but as far as I know nothing uses it right now and there's no clear indication when the first product to use it will launch. It's true that it's faster than the A5, but I'm assuming that Apple will have a new chip in their next iPad, whenever that launches. Assuming that the next iPad comes out around March-April, even if this new OMAP is faster than whatever the A6 turns out to be, it will be the fastest chip on the market that's not actually available in any products for another 7 - 12 months.

I don't have any information about Apple's next chip, but some have speculated that it will also be a Cortex A-15 design so I can't see them being considered way ahead unless they've really customized the design. Historically, even the most heavily customized ARM chips have usually managed around 10% at best. Perhaps this can be further improved with the A-15, but it seems unlikely that TI is going to have something that can be qualified (or maybe even quantified) as "way ahead".

I suppose that one can't fault a person for hyping his own product up. I'd probably do the same in his position, but from my own position, it's hard to take his claims at face value.

TI has a license with ARM that allows them early acess to A15 while Apple does not afaik.

 

[lothar]

The quote confuses me. He says it's the first Cortex A-15 product on the market, but as far as I know nothing uses it right now and there's no clear indication when the first product to use it will launch. It's true that it's faster than the A5, but I'm assuming that Apple will have a new chip in their next iPad, whenever that launches. Assuming that the next iPad comes out around March-April, even if this new OMAP is faster than whatever the A6 turns out to be, it will be the fastest chip on the market that's not actually available in any products for another 7 - 12 months.

I don't have any information about Apple's next chip, but some have speculated that it will also be a Cortex A-15 design so I can't see them being considered way ahead unless they've really customized the design. Historically, even the most heavily customized ARM chips have usually managed around 10% at best. Perhaps this can be further improved with the A-15, but it seems unlikely that TI is going to have something that can be qualified (or maybe even quantified) as "way ahead".

I suppose that one can't fault a person for hyping his own product up. I'd probably do the same in his position, but from my own position, it's hard to take his claims at face value.

TI has an exclusive license with ARM on A-15.
They will be the first.

 

[Bateluer]

The major change on the iPad 3 will likely be the screen, not the SoC powering it. If it changes at all, it'll be a higher clocked A5. To make it more magical.

 

[MrX8503]

TI has an exclusive license with ARM on A-15.
They will be the first.

Apple has jumped ahead of the competition with the A4 and did it again with the A5. I suspect the same will happen again with the A6. If TI does get it out first, I don't think Apple would be far behind.

The major change on the iPad 3 will likely be the screen, not the SoC powering it. If it changes at all, it'll be a higher clocked A5. To make it more magical.

I guess the iPad 2's A5 demolishing the competition way ahead its time wasn't impressive enough?

 

[ChronoReverse]

I guess the iPad 2's A5 demolishing the competition way ahead its time wasn't impressive enough?

You know, it was the SGX543MP4 in the iPad2 that was smoking everything (and still is). The A5 itself wasn't nearly as impressive.

 

[Bateluer]

Apple has jumped ahead of the competition with the A4 and did it again with the A5. I suspect the same will happen again with the A6. If TI does get it out first, I don't think Apple would be far behind.

I guess the iPad 2's A5 demolishing the competition way ahead its time wasn't impressive enough?

The A4 wasn't a performance winner. The A5 wasn't either, as Chrono notes, it was the SGX543MP2 that gave the iPad2 is stellar graphics performance. Just a major let down with the low res 10x7 4:3 screen. If the iPad 3 gets the rumored 20x15 resolution, that will be impressive.

You know, it was the SGX543MP4 in the iPad2 that was smoking everything (and still is). The A5 itself wasn't nearly as impressive.

AFAIK, the A5 was on par with the other Cortex A9 chips. That GPU is pretty damn stellar though. Not sure what the delay with getting it to other devices is, some kind of exclusivity agreement?

 

[Mopetar]

The major change on the iPad 3 will likely be the screen, not the SoC powering it. If it changes at all, it'll be a higher clocked A5. To make it more magical.

They could probably get a way with a higher clocked CPU, but they'd definitely want to beef up the GPU as much as possible. Doubling the resolution means quadrupling the number of pixels that need to be pushed out. The GPU in the A5 was probably overkill, but they'll have their work cut out for them if they want to maintain the same level of performance.

 

[ITHURTSWHENIP]

Lead partners for Cortex A15 are: TI, Samsung, ST-Ericsson, Nvidia

They paid extra for early access. Its not exclusive to TI wich is why Samsung is releasing their SOC before TI

Apple usually prefers to not be named in these things, they get first access to IMG GPUs but never announce it. So its impossible to know if they are a lead partner or not

Would love to see OMAP 5 in the next Nexus phone

 

[Bateluer]

They could probably get a way with a higher clocked CPU, but they'd definitely want to beef up the GPU as much as possible. Doubling the resolution means quadrupling the number of pixels that need to be pushed out. The GPU in the A5 was probably overkill, but they'll have their work cut out for them if they want to maintain the same level of performance.

Agreed, the odds of them maintaining the same performance bump between the iPad 2 and 3 as they had between the 1 and 2 is slim to nil. One could argue that they don't really need to though. Even keeping the same A5 SoC in the iPad 3 combined with a 20x15 screen would still perform fine. And Apple users don't care about high performance or cutting edge tech anyway.

 

[ChronoReverse]

AFAIK, the A5 was on par with the other Cortex A9 chips. That GPU is pretty damn stellar though. Not sure what the delay with getting it to other devices is, some kind of exclusivity agreement?

No idea really. It was grotesquely overpowered though and only now is it finally trickling down to smartphones (in half size for the 4s). Looks like Medfield will get it too this year.

Perhaps Apple had intended to get a double resolution screen for the iPad2 in the first place but couldn't acquire enough until now? Who knows but it's certainly a nice GPU.

 

[ilkhan]

Would love to see OMAP 5 in the next Nexus phone

OMAP is in the gal nexus. I hope it won't be them for the next version as well.

 

[Mopetar]

And Apple users don't care about high performance or cutting edge tech anyway.

I would disagree. There are many ways to interpret 'high performance' but I think that Apple users in general (and even Apple itself) have fairly high expectations related to the performance and quality of their gear. To better illustrate my point, I would say that Apple users don't care about the results of a synthetic benchmark like SunSpider, but they do want the browser to be snappy and smooth.

I would also say that Apple likes cutting edge tech (e.g. high resolution displays, thunderbolt, etc.) but not at the expense of the overall user experience.

I was also thinking about their next chip more and it may make sense for them to simply find the bare minimum level of performance to maximize power efficiency. It's almost a certainty that the next iPhone will have LTE and that's going to require more power. Apple isn't going to want to sacrifice battery life so getting the power draw of their next SoC down to an absolute minimum may be their primary goal.

That said, we've probably derailed this thread enough.

 

[lothar]

Lead partners for Cortex A15 are: TI, Samsung, ST-Ericsson, Nvidia

They paid extra for early access. Its not exclusive to TI wich is why Samsung is releasing their SOC before TI

Apple usually prefers to not be named in these things, they get first access to IMG GPUs but never announce it. So its impossible to know if they are a lead partner or not

Would love to see OMAP 5 in the next Nexus phone

I've always thought the exclusive structure was more like this...:hmm:
A15 CPU: TI
A15(I'm not sure if ARM uses Axx nomenclature for their GPU parts) Mali(or whatever they want to call it.) GPU: Samsung

Apple may have access to the fastest GPU on the market, but they won't with A15 CPU part unless they release their iPhone * 6 months after everyone else does with A15. Of course Apple can always release a quad core A9 with the most advanced GPU on the market to counter a dual core A15 with inferior GPU or an octa core A9 with the most advanced GPU on the market to counter a quad core A15 with inferior GPU.

If you're a "lead" partner and everyone of your competitors have access to the product by becoming a "lead" partner themselves, then what is the benefit of becoming a "lead" partner again? So you don't get left behind in the dust? Rather than doing so to get the "exclusive" early performance advantage for a few 2-3 months or so?

 

[Bateluer]

OMAP is in the gal nexus. I hope it won't be them for the next version as well.

OMAP4 in the Nexus was paired with a weak GPU. Bumping to an OMAP5 with an appropriately modern GPU would address that short coming. Assuming TI gets the contract again.

 

[ITHURTSWHENIP]

I've always thought the exclusive structure was more like this...:hmm:
A15 CPU: TI
A15(I'm not sure if ARM uses Axx nomenclature for their GPU parts) Mali(or whatever they want to call it.) GPU: Samsung

Apple may have access to the fastest GPU on the market, but they won't with A15 CPU part unless they release their iPhone * 6 months after everyone else does with A15. Of course Apple can always release a quad core A9 with the most advanced GPU on the market to counter a dual core A15 with inferior GPU or an octa core A9 with the most advanced GPU on the market to counter a quad core A15 with inferior GPU.

If you're a "lead" partner and everyone of your competitors have access to the product by becoming a "lead" partner themselves, then what is the benefit of becoming a "lead" partner again? So you don't get left behind in the dust? Rather than doing so to get the "exclusive" early performance advantage for a few 2-3 months or so?

There is no A15 GPU. A15 is the name of their processor, the semiconductor companies themselves then choose the rest of the parts for their SOCs. Mali is their GPU, these are two different licenses and not related to each other

The benefit of being a lead partner is early access to the IP wich will give you an advantage in the market. Of course this comes at a premium price wich is why not everyone becomes a lead partner

ARM doesnt sell any processors, they just sell the rights to their IP, you are purchasing their blueprints to manufacture it.

My point about Apple is that even if they were a lead partner, they would never announce this. They never announce their GPU licenses either so until they announce the Ipad 3, we wont know what they will use

 

[lothar]

There is no A15 GPU. A15 is the name of their processor, the semiconductor companies themselves then choose the rest of the parts for their SOCs. Mali is their GPU, these are two different licenses and not related to each other

The benefit of being a lead partner is early access to the IP wich will give you an advantage in the market. Of course this comes at a premium price wich is why not everyone becomes a lead partner

ARM doesnt sell any processors, they just sell the rights to their IP, you are purchasing their blueprints to manufacture it.

My point about Apple is that even if they were a lead partner, they would never announce this. They never announce their GPU licenses either so until they announce the Ipad 3, we wont know what they will use

I know.

That's pretty much everyone already. Who else is missing?
LG doesn't make their own SoC's. Neither does Motorola. Samsung, Nvidia, TI, that's almost everyone already in the SoC market.
Qualcomm is the only one missing not listed that I could think off.
Isn't Krait an A15 core(if not, then what is it?)?

Who are you referring to? ARM or Apple? If Apple doesn't announce it themselves, it doesn't mean ARM can't make the announcement of their in their PR touting them as a lead partner. Every company has their own press release and they don't necessarily have to work in tandem.

Now, Apple could certainly have paid the ARM PR people some "hush hush" money...I'm not doubting that. That's "normal" business practice for most corporations.

 

[ChronoReverse]

Qualcomm doesn't do direct ARM cores. Instead they have their own designs that implement the ARMv7 instruction set. Usually, they're just about inline with the ARM designs of the time with maybe a miniscule performance advantage (although their GPU's have only been competitive at best, wholly inadequate at worst).

Krait should be on the level of the A15 though.

 

[ITHURTSWHENIP]

Qualcomm has an ARM architecture license but builds their own custom processors. Similar to AMD having an X86 license

Krait is supposedly not quite as powerful as A15 but much more power efficient and overall smaller chip

And yes the last paragraph is basically what i meant. Apple might be a lead partner for Cortex A15 but has asked ARM not to announce it. IMG usually announces when Samsung, TI, Intel buy a Power VR license but never Apple, yet Apple still has the latest hardware from them

 

[destrekor]

I know.

That's pretty much everyone already. Who else is missing?
LG doesn't make their own SoC's. Neither does Motorola. Samsung, Nvidia, TI, that's almost everyone already in the SoC market.
Qualcomm is the only one missing not listed that I could think off.
Isn't Krait an A15 core(if not, then what is it?)?

Who are you referring to? ARM or Apple? If Apple doesn't announce it themselves, it doesn't mean ARM can't make the announcement of their in their PR touting them as a lead partner. Every company has their own press release and they don't necessarily have to work in tandem.

Now, Apple could certainly have paid the ARM PR people some "hush hush" money...I'm not doubting that. That's "normal" business practice for most corporations.

This doesn't matter because unless Apple plans on actually getting into the chip design industry (they wouldn't be the fabricators, it would be one of the well-established fabs), they will not even be on ARM's radar; their only business is with chip designers. The big chip companies would be the one who holds business with Apple, and they would be the one from which Apple demands everything they've got for a few months.
It can basically be surmised Samsung will not be the provider of the "A6" SoC, so that leaves TI or Qualcomm. I highly, highly doubt Apple will source from nVidia, because a Tegra with Cortex A15 (Tegra 4?) will be in manufacturers' hands long after the rest of the industry has released product to consumers with an A15 SoC of some variety. And having what is essentially an outdated SoC, or at the least, considered last-gen half a year after you have it in a device, is not Apple's way.