The Six Five team discusses the delay of Intel’s Sapphire Rapids CPU chips.
Watch the clip here:
If you are interested in watching the full episode you can check it out here.
Disclaimer: The Six Five Webcast is for information and entertainment purposes only. Over the course of this podcast, we may talk about companies that are publicly traded and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors and we do not ask that you treat us as such.
Patrick Moorhead: Intel was in the news recently and not for what I think maybe they wanted to be in the news for. Sapphire Rapids had a push and also there was a reorg particularly across DCG. And I don’t know if they’re combined, but we just decided to put them in at the same time. So, I think in a way, the Sapphire Rapids push was a little bit, the reaction I was pretty surprised. Wall Street, what are they down like five points, Daniel, since that was announced?
Intel had committed to shipping Sapphire Rapids by the end of 2021 and then to have volume in the first half. Intel is still going to hit volume in the first half, albeit in the second quarter. But, they really had to do it, I think, because a new leader came in at DCG. And listen, this is my speculation and nothing that I got from Intel, but I think Sandra Rivera came in to run DCG. She saw the schedule. She saw what the outbound message was and said, “Listen, this is the new Intel and our new CEO has made some commitments on how we communicate and we’re more open.” And one thing led to another and they gave more details. Now, was that connected to the reorg that was last week? I have no idea, but it is incredibly ironic and essentially what happened and listen, I like Navin Shenoy a lot. Got to know him a lot better over the last couple of years. He is exiting stage left and we’ve got new leadership coming in with Sandra Rivera.
I’ve gotten to know her really well when she was running the networking group and building that. And she spent the last couple years running HR. Additionally, networking and IOT were pulled together, which to me there’s absolute synergy there. And those are really the highlights. And on the reorg, listen, the new CEO, Pat Gelsinger came in and said, hey, we’re going to bring technologists to the top and Sandra Rivera has a double E and Nevin didn’t. That could very well be the reason for this.
And also, I mean, Daniel, how many times have we seen a new CEO comes in, 100% of the time they make changes? And so here we are. So people shouldn’t over-index on the reorg like, “Oh, they’re going to be a complete failure.” I mean, my gosh DCG still has 90% market share of general purpose servers and is growing in storage and growing in networking and carrier and were always up in the right, but, we can’t underplay the commitment that Pat Gelsinger made to schedules and the commitment that he made to be more open. So we have what we have.
Daniel Newman: Yeah, absolutely, Pat. A couple of things I wanted to notate too. This slip into the first half of 2022, while it was sort of textbook Intel and what’s happened with Intel over the last few years. To your point, the more transparent open, upfront and early admittance of this was a solid move to show this is a new Pat Gelsinger-led Intel, and it’s not going to be mistake-free, not possible. But it’s going to be proactive in communication.
The second thing is the company did take the chance to make this update not just about being late, but also being better. And that kind of got a little bit lost, but there was some pretty significant updates and I pointed to these in some of my hot takes this week. There was new support for Advanced Matrix Extensions, which is important to the deal, Boost technology, which is important to the ML strategy of Intel and their general purpose CPUs. There were enhancements to its data streaming architecture. It is the first Intel chip that is going to include DSA.
And then the second thing is in upcoming Sapphire Rapids, they did announce or officially memorialize that they’re going to feature up to 64 gigs of HBM2 so made enhancements on the memory or confirm some of the capabilities. Well, the point is all of these things were really part of the plan from the get-go. And from talking to the team about this delay, they’re taking their time and making really what is a modest extension to the go to market to try to make sure that these enhancements that have been committed in Sapphire Rapids are truly delivered. Because I think waiting a quarter or a month, because actually I think it was like Supercomputing or Computex, they’d already sort of alluded to this thing being delayed into the beginning of ’22 so this isn’t even really the first time they said it. It’s just the first time they’ve officially set it, Pat.
And I guess to that point, getting it right, I still believe at this point is going to be more important than getting it fast. AMD arm, very strong competition right now. But let’s just say Pat, they’re going to be better served to get it right. And by the way, with Ice Lake being a little bit late to market too, this actually kind of one of the convenient benefits is giving a little bit more time gap between those two launches for customers to settle in on these new generation. So, definitely imperfect, not the ideal situation. But I do believe the company is doing a pretty good job of making the best of this difficult moment.
Patrick Moorhead: Yeah, those are really good adders Daniel. And I actually talked to an OEM that was actually thankful which you’re like thankful? They’re thankful because it’ll be Intel doing the incremental testing for three months as opposed to them finding bugs and submitting to Intel. So they were thankful that that Intel is not going to push their problems down the line to them. Another OEM’s like we’re ready, let’s go. Right? We’re okay with this. Our customers aren’t going to use these two instructions, which by the way, I had originally mistaken, they thought that it was just these two instructions that were causing the issue. It’s not.
Intel with Sapphire Rapids, it’s a brand new platform and it’s a brand new chip, right? And anytime you introduce a new memory architecture or … sorry, a new memory standard with DDR5, which they’re leading in, problems always happen 100% of the time. When I was at AMD, this was some of the hardest part to get. Not only, you can get one vendor going on memory, but getting multiple vendors going on new memory is fricking tough because you’re going back and forth with the memory vendors and Intel didn’t tell me this. I think it’s memory. I think it’s a DDR5 thing that they’re trying to work out.