You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
"We stated at our Financial Analyst Day that we're already investing heavily in 7-nanometer. The 7-nanometer will be key for us on both the CPU and the GPU side. And I would say that development is progressing well. We're working with multiple foundries on that."
I'm assuming that 'multiple' means mainly GF and TSMC, right? Does anyone have an idea of how these will be divided between the CPU, GPU, APU and semi-custom parts?
I'm assuming that 'multiple' means mainly GF and TSMC, right? Does anyone have an idea of how these will be divided between the CPU, GPU, APU and semi-custom parts?
Currently they use GF 14nm for CPU and GPUs and TSMC 16nm for consoles (CPU/GPU) combo. It is my understanding that TSMC has much better margins on 16nm and 7nm so the low margin console business is a good fit. The first TSMC 7nm is SoC/cost focused (Apple) which is also a good fit for the console business.
The first GF 7nm is high performance and is custom tailored to AMD's Ryzen architecture. This is the first time AMD has had a custom process since they went fabless. AMD is really going to give Intel 10nm parts a run for their money on this one. At least I hope so!
The first GF 7nm is high performance and is custom tailored to AMD's Ryzen architecture. This is the first time AMD has had a custom process since they went fabless. AMD is really going to give Intel 10nm parts a run for their money on this one. At least I hope so!
So, for the first time ever AMD will have a process that is both custom-tailored and essentially on par with Intel's. Very good.
The only question that remains is timing. There are rumors that Intel 10nm has been delayed to 2018. GF 7nm will enter risk production in the first half of 2018, is there any hope of seeing 7nm Ryzen (Zen 2) by year end? Or is it gonna be 1H-19?
So, for the first time ever AMD will have a process that is both custom-tailored and essentially on par with Intel's. Very good.
The only question that remains is timing. There are rumors that Intel 10nm has been delayed to 2018. GF 7nm will enter risk production in the first half of 2018, is there any hope of seeing 7nm Ryzen (Zen 2) by year end? Or is it gonna be 1H-19?
Not a huge lead but GF will insert EUV into 7nm to increase their lead in 2019. I don't expect Intel to have 7nm EUV until 2021 so AMD really has an opportunity here, absolutely.
However, I have understood that there are many variations of a given process -- GF may support 6T cells at 7nm for low-power mobile, but may have e.g. 9T cells for high-performance chips (lower capacitance/resistance). The de facto value for Intel at 10nm is 7.56T, don't know if they will have more configurations beyond that.
Currently they use GF 14nm for CPU and GPUs and TSMC 16nm for consoles (CPU/GPU) combo. It is my understanding that TSMC has much better margins on 16nm and 7nm so the low margin console business is a good fit. The first TSMC 7nm is SoC/cost focused (Apple) which is also a good fit for the console business.
The first GF 7nm is high performance and is custom tailored to AMD's Ryzen architecture. This is the first time AMD has had a custom process since they went fabless. AMD is really going to give Intel 10nm parts a run for their money on this one. At least I hope so!
Well said Daniel. For the first time after AMD went fabless , with GF 7LP they will be building CPUs on an extremely competitive high performance process node. I have high expectations from Zen on 14nm+ in H1 2018 and Zen 2 on GF 7LP in H1 2019.
Not a huge lead but GF will insert EUV into 7nm to increase their lead in 2019. I don't expect Intel to have 7nm EUV until 2021 so AMD really has an opportunity here, absolutely.
True. GF has already confirmed multiple generations with EUV at 7nm. Zen 2 will use 7nm DUV while Zen 3 will most likely use the 7nm EUV node with performance, area and power improvements.
Whats more impressive is IBM has already demonstrated test wafers using silicon nanosheets and Globalfoundries is likely to use that for its next major node whatever its name is. The foundries have well and truly caught up with Intel. Intel has not confirmed a new device architecture for their 7nm node. If Intel does not step up their pace they are at risk of falling behind foundries. The current FINFET device architecture is likely to have serious difficulty when contacted gate pitch gets down to 40 - 42nm.
The big decision comes when the gate-pitch approaches 40nm. Based on simulations from Imec, the finFET begins to teeter at a 42nm gate-pitch. “The nanowire will scale below that and still have good electrostatic control,” said An Steegen, executive vice president of semiconductor technology and systems at Imec. The nanowire FET, according to Imec, has demonstrated good electrostatic control at a 36nm gate pitch. Imec has also devised a nanowire down to 9nm in diameter.
However, I have understood that there are many variations of a given process -- GF may support 6T cells at 7nm for low-power mobile, but may have e.g. 9T cells for high-performance chips (lower capacitance/resistance). The de facto value for Intel at 10nm is 7.56T, don't know if they will have more configurations beyond that.
For high performance Intel 10nm should have the lead in density and a significant one if the library used is 7.56T as GF 7LP uses 9T for high hperformance.
