By: mas (mas769.delete@this.hotmail.com), August 22, 2013 3:17 am
Room: Moderated Discussions
Michael S (already5chosen.delete@this.yahoo.com) on August 22, 2013 2:47 am wrote:
> Ronald Maas (rmaas.delete@this.wiwo.nl) on August 21, 2013 11:54 pm wrote:
> > mas (mas769.delete@this.hotmail.com) on August 21, 2013 6:10 pm wrote:
> > > Single thread performance has primarily determined PC sales
> > > since the beginning and that will not change going
> > > forward. Top bin Haswell has about 6x better performance than
> > > top bin A15/A57. AMD is much much closer in performance
> > > and cheaper yet it has done nothing but hang on in a money
> > > losing fashion to Intel's coattails. The primary threat
> > > of ARM is not direct competition but indirectly by delaying/canceling
> > > a potential PC purchase when a smartphone/tablet
> > > is bought. Intel has finally whipped Atom into competitive
> > > shape so that it can be in that tablet or smartphone.
> > > This is more a form factor revolution than an architectural/business model one.
> >
> > After years of hard work and God knows how many billions of dollars spent, Intel finally achieved parity on
> > performance per Watt with some existing ARM SOCs. Wow. Great. Intel definitely earned a pat on their back.
> > So now they are finally able to compete with companies like
> > Rockchip and other Chinese design houses that are
> > able to complete their designs in 6 week time with a just
> > a bunch of engineers. And these companies are able
> > to make comfortable profits when they charge just a couple of dollars for their SOCs.
>
>
> "Couple of dollars" sounds exaggerated. I'd think that for likes
> of RK3066 "comfortable profits" zone starts above 15 dollars.
>
I have seen a figure of $7 quoted for one of these Chinese vendors A7 chips, competition is getting tough amongst each other and Qualcomm and ARM did say this in their last CC ...
http://seekingalpha.com/article/1566712-arm-holdings-plc-armh-adr-ceo-discusses-q2-2013-results-earnings-call-transcript?page=5
'If we look at entry-level phones, which today are being based designed around Cortex-A7, I mean really Intel don't have a products that are suitable for that market. Clover Trail Plus is the smallest thing they have but it is enormous in comparison to Cortex-A7. Cortex-A7 delivers roughly the same performance but at considerably less power and crucially in a really small die size.
The benefit of that die size is that every thing else that you need to put into, effectively a single chip application process with modem can be integrated onto one device at low cost. So the modem, the connectivity, the GPU, the memory can all be integrated on to one SoC. That build on a foundry process can sell for as low as $5 and our customer who is doing that today, these devices are shipping today.
When you look in the mid range, which again, devices are shipping today based on ARM, based on Cortex-A9. Cortex-A9 outperforms Clover Trail Plus. Again, its lower power. Its considerably smaller. So again, it can be built into an SoC with everything else you need and build on a foundry process and sell for around $10. Very low cost. Very low power. Great performance. Shipping Now.
In the premium end, as I mentioned in the beginning, we are seeing big.LITTLE devices shipping now that adds significant performance in comparison to Clover Trail Plus but at the same time, allows power to be lower because of the big.LITTLE architecture. Again, relative size wise, the cluster configuration is still very small. So build out on a foundry process you can build a whole SoC and sell it for $15. So you can deliver the performance at a very low power, and we can hit these very low price points that require these large volume markets.'
Back in the real world Clovertrail+ is easily competitive with A9 SoCs like Tegra3 but you get the idea about prices.
> How much does Intel need
> > to get to earn back their investments for Atom / Silvermont? Probably a whole lot more than that.
> >
>
> In case of phone CPU, it's unlikely that Intel is thinking in terms of return of investments.
> The goal is probably much more moderate - to sell chips above the cost of manufacturing.
>
> Ronald Maas (rmaas.delete@this.wiwo.nl) on August 21, 2013 11:54 pm wrote:
> > mas (mas769.delete@this.hotmail.com) on August 21, 2013 6:10 pm wrote:
> > > Single thread performance has primarily determined PC sales
> > > since the beginning and that will not change going
> > > forward. Top bin Haswell has about 6x better performance than
> > > top bin A15/A57. AMD is much much closer in performance
> > > and cheaper yet it has done nothing but hang on in a money
> > > losing fashion to Intel's coattails. The primary threat
> > > of ARM is not direct competition but indirectly by delaying/canceling
> > > a potential PC purchase when a smartphone/tablet
> > > is bought. Intel has finally whipped Atom into competitive
> > > shape so that it can be in that tablet or smartphone.
> > > This is more a form factor revolution than an architectural/business model one.
> >
> > After years of hard work and God knows how many billions of dollars spent, Intel finally achieved parity on
> > performance per Watt with some existing ARM SOCs. Wow. Great. Intel definitely earned a pat on their back.
> > So now they are finally able to compete with companies like
> > Rockchip and other Chinese design houses that are
> > able to complete their designs in 6 week time with a just
> > a bunch of engineers. And these companies are able
> > to make comfortable profits when they charge just a couple of dollars for their SOCs.
>
>
> "Couple of dollars" sounds exaggerated. I'd think that for likes
> of RK3066 "comfortable profits" zone starts above 15 dollars.
>
I have seen a figure of $7 quoted for one of these Chinese vendors A7 chips, competition is getting tough amongst each other and Qualcomm and ARM did say this in their last CC ...
http://seekingalpha.com/article/1566712-arm-holdings-plc-armh-adr-ceo-discusses-q2-2013-results-earnings-call-transcript?page=5
'If we look at entry-level phones, which today are being based designed around Cortex-A7, I mean really Intel don't have a products that are suitable for that market. Clover Trail Plus is the smallest thing they have but it is enormous in comparison to Cortex-A7. Cortex-A7 delivers roughly the same performance but at considerably less power and crucially in a really small die size.
The benefit of that die size is that every thing else that you need to put into, effectively a single chip application process with modem can be integrated onto one device at low cost. So the modem, the connectivity, the GPU, the memory can all be integrated on to one SoC. That build on a foundry process can sell for as low as $5 and our customer who is doing that today, these devices are shipping today.
When you look in the mid range, which again, devices are shipping today based on ARM, based on Cortex-A9. Cortex-A9 outperforms Clover Trail Plus. Again, its lower power. Its considerably smaller. So again, it can be built into an SoC with everything else you need and build on a foundry process and sell for around $10. Very low cost. Very low power. Great performance. Shipping Now.
In the premium end, as I mentioned in the beginning, we are seeing big.LITTLE devices shipping now that adds significant performance in comparison to Clover Trail Plus but at the same time, allows power to be lower because of the big.LITTLE architecture. Again, relative size wise, the cluster configuration is still very small. So build out on a foundry process you can build a whole SoC and sell it for $15. So you can deliver the performance at a very low power, and we can hit these very low price points that require these large volume markets.'
Back in the real world Clovertrail+ is easily competitive with A9 SoCs like Tegra3 but you get the idea about prices.
> How much does Intel need
> > to get to earn back their investments for Atom / Silvermont? Probably a whole lot more than that.
> >
>
> In case of phone CPU, it's unlikely that Intel is thinking in terms of return of investments.
> The goal is probably much more moderate - to sell chips above the cost of manufacturing.
>