Points from the Interview with Murthy

Key Points
Murthy's job is to help Intel to deliver leadership products in all markets it chooses to play in and that comes from aligning, Process, Architecture, Silicon engineering and executing on the roadmap

Apart from a few exceptions Murthy is head of all of Intel's Engineering and Technology Activities

Really excited to have Raja and Keller part of the tech leadership teams, key desire was to supplement a sports analogy "a great coach can make a great team even better" and thinks that the advantage of having Raja and Keller on his side is to bring deep technical leadership into Intel

Raja is focused on product architecture and product construction and driving roadmaps for biz units

Jim focused on driving core silicon engineering and execution, the boundaries between their roles are very diffuse, no overly constrained box between the teams, hopes the chemistry between the two drive Intel products to the next level

Jim has a pedigree in developing micro architectures and core cpu architectures

Raja comes from a background hes built up both engineering and biz scale in dGPU environment

Both will execute in those dimensions as well

Datacentric transformation stuff that we have all heard 10x over

Intel's "North Star Vision" - Imagine everyone had access to 10 Petaflops of compute and 10 Petabytes of storage within 10ms of access

Market is concerned that process gap is closing

Silicon Leadership is really important but is only one of the aspects required - Intel is a product company, not focused on being a merchant foundry

Need great arch capability, need good packaging and assembling tech but all much execute on a set timeline, bring all those together brings product leadership

Shipping 10nm in low volume

10nm was very aggressive, targeted 2.7x scaling factor, TMG had very ambitions goals on transistor scaling. More aggressive in hind sight than what was ideal and so they have had a greater challenge bringing 10 to market. 10nm problems are "not fundamental", "we know what to fix and are busy doing so"

In the mean time we found tremendous intra node capability in our 14nm process - delivered over 70% perf improvements due to intra node improvements

That has given them the ability to make sure 10nm yields are right before it goes into HVM

Intel is confident that 14nm will give Intel leadership products for the next 12-18 months as they bring up yields for 10nm

Is 10nm ramp in H1 or H2'19?

Products are shipping today, no specific timeline on HVM, when economic timeline is right and when cost on the yield graph is good enough, they will ramp when they can. Murthy is taking a "wait and see approach"

Will 7nm be pushed out? "No that would be a premature assumption"

Technical risk factors, 10 vs 7 are very different, 10nm focused on 2.7x density heavily, no help from EUV had to go to SAQP which is complex and time consuming.

As of today we are deep into product development on 10nm and are very very pleased with the progress

7nm is taking a slightly different balance point on density, performance and power and schedule predictability

 Have not disclosed 10nm server roadmap, very excited by product pipeline in our server roadmap, I am equally excited about the products we will be launch this year and next year on 14nm on our data center roadmap

Really excited by Ethernet, Silicon Photonics, 3DXP, Custom Accelerators and FPGAs products. Stellar execution on 14nm roadmap, will see a great deal of vibrancy in our DC Roadmap. 10nm products will come along at the appropriate time and believes 14nm will be able to carry their expectations as we transition to 10nm in 2019

Are you gonna have 14nm+++ and arch improvements? "Intel was all in one node then all out". 10nm delays has given then a real opener for the goodness of intra-nodes improvements, which was taken for granted at other foundries. Intra-nodes improvements were very significant - 70% gains from 1st gen 14nm(Didn't say no or yes to 14+++). Intra node improvements will continue, nodes will live longer.

Questions about EMIB and Stratix 10 This area is really exciting for Murthy, it is really going to transform the way they develop products, as part of their DC strategy, moving DC Processors to nearing of the beginning of new nodes. DC Products are very large exceeding 500-600mm^2 which is a challenge for new nodes. Large dies and new processes don't mix

Mix and Match Approach - Focusing on how they can target IP that benefits from density scaling, not all IP benefit from the scaling, like mix signal and high speed I/O IP

EMIB deconstructs silicon architectures, will be using EMIB and "other techniques that will become visible through the rest of this year and early next" to do this. It is beneficial to have IP on different nodes  but using contacts that provide virtually monolithic performance. Doing this on CPU and GPU cores to make smaller pieces of silicon that will yield higher on new process nodes, while the rest is left on already good yielding legacy nodes. Not constraint by moving all IP to new node. This is what you will see from Intel going forward

EMIB offers virtually monolithic performance in dis-aggregated silicon, it is not another version of MCP and  future technologies that are launching in the future will have much lower power, cost and much lower latency that traditional interposer techniques that others have spoken about

Reaffirms fixes in silicon fixes for Meltdown and Spectre for 14nm products like Cascade Lake and Whiskey lake in H2'18 this year.

What do you think about Intels 9th Gen Products vs your competitions? Vigilant about competition, peak performance is built on competition. Very pleased with product portfolio on 14nm. Across most major benchmarks Intel has demonstrated leadership performance. Coffee Lake 8700K vs R7 2700x is a good comparison point. Upbeat about product roadmap for the rest of this year and early 19. Build upon intra node improvements on 10nm. Excited where they are at on 8th Generation and 9th Generation. Very aware of the competitive environment. Must deliver the best to maintain leadership performance

Gaming Stuff is growing 21% growth YoY (Q1'17 to Q1'18) No reason for it to stop. Seeing more products from OEMs. Really exciting growth area. Esports is a growth area too. Which is driving much higher specs which is good for Intel.

Update on XMM7560 Modem: Really pleased with progress and strives in modems, you'll see the launch of the 7560  in H2'18. First Gigabit LTE product from Intel with GLOBAL CDMA SUPPORT(Emphasis his) that really drives feature parity across a broad set of benchmark metrics. First of many to come, very pleased with Modems. In HVM well before the end of the year


END