Intel has a big problem and today’s Knights Something and Infiniband announcement is a glaring reminder of it. What is this problem? Intel or to be more specific, Intel’s technology and messaging.
Lets start out by telling you a bit about todays ‘news’, and we use the word very lightly. Ostensibly Intel is talking about the new Knight’s Landing CPU to be unveiled today along with a new variant of their Omni Scale not-Infiniband interconnect This does indeed sound like an interesting brief, high-end products that supposedly push new boundaries in an area where Intel is leading the field. What could go wrong?
That is the problem, the tech is probably very good but you would never know it from the call. Why? Intel flat-out refuses to say anything about the products or the tech that they bring to the table. No we aren’t joking, it was as bad as the last time they tried this BS, no actually it was worse. Why? The ‘disclosures’.
What did the Intel talk reveal about Knights Landing? Absolutely nothing. No we are not joking, they said it uses stacked DDR4 from Micron but but that was already disclosed last time they wasted our precious oxygen. But fear not gentle reader, there was a real disclosure, the next generation called Knights Hill, is built on the next gen 10nm process, and has a next generation not-Infiniband built-in.
If you don’t see how this leaps from sad to surreal, think about this. Intel is purportedly talking about Knights Landing but the only disclosure was the name of the successor! Seriously? They are so desperate to grab headlines and so unwilling to do the barest minimum disclosures that they skipped over what they were talking about entirely. It has long been known that there will be one generation of Knights* per process node so that part is a no brainer, and the next gen not-Infiniband is pretty much a given too. So we got nothing but a name. Which was widely known everywhere. For years. Pleural. SemiAccurate wrote it up in January of 2012, see here.
And then there is the new version of True Scale/Omni Scale fabric. The biggest news of this new ‘disclosure’ is that some marketing jenius didn’t like the name so it is now Omni-Path. Sets the heart aflutter with anticipation, eh? Not to slight this major reveal but there was an equally stunning technical side of it too. What is that you say? Intel will have 48-port switches!!!!! No really, stop the presses, the entire technical side of the Omni-waitforittochangeagainagain briefing was a 48-port switch is coming! Yup, that is it on the technical side.
Now we started out this rant about the briefing [my editor won’t let me use the appropriate term] by saying that it pointed out that Intel has a problem. That isn’t actually true, Intel has two problems that are both highlighted by the rather pathetic state of their messaging. Those problems are internal buck-passing and external perception.
On the internal side, every time we point out that Intel’s messaging has gone from class of the field to something between non-existent and a painful joke, the obvious thing happens. Yes the executive who made the decision to not say anything worth a damn chews the [synonym for posterior] off of the messenger, you know the person who didn’t have a say in the problem but had to carry it out. Yes the people making these decisions to not say squat all shoot the messenger and then make things worse the next time for reasons that defy logic. Other than server tech which has historically been a high point of Intel messaging, I can’t see how things will get worse. That said I have faith that they will surprise me with their ingenuity, they have done so several times in the past.
The other problem is their public perception or alternately the internal views of their public perception. Intel has long been clueless about why no one seems to care about their products anymore and why all the mindshare, momentum, and excitement around tech is no longer x86 based. This was exemplified with their ostensible desktop revival talk at the last GDC where you could tell that they really wanted to turn things around. Unfortunately they didn’t do anything to actually address the problems, just put out more of the same stuff that people have turned their backs on Intel for. In short they tried to grab headlines to message that desktops were back in the hopes it would change things but it was a hollow message. It fell flat, badly flat.
With Omni-Whatsitsname and Knights* they are doing the same, banging a drum that no one wants to listen to any more while stonewalling the few that still do. Worse yet they have made it very clear that to listen to Intel’s messaging is a waste of your time but for some reason they expect positive headlines from it anyway. A nearly three-year old name and a port count while not talking about the really good tech they have is not a way to keep people from turning down the next briefing or 12 much less keeping the public interested.
How bad was it? Last time they ‘revealed’ Knights* information they said there would be optical interconnects integrated. This time they didn’t even say that but begrudgingly admitted it when directly asked. When asked about the Knights Landing chip itself, ostensibly why we were all on the call, they flat-out refused. Worse than that they went so far as to ‘explain’ the Knights Hill name release by saying it was there for potential customers at the Supercomputing show.
Stop and think about that, supercomputer and HPC customers who will buy hundreds of nodes of an accelerator, port their code over, and optimize for a new architecture, interconnect, and paradigm will be swayed by a name release but are OK with being stonewalled on the barest minimum technical details of the stuff that is on sale now. Does Intel really think we are this stupid? Don’t answer that until you see how many other sites put that explanation in their stories, I fear for humanity now.
And that is the root of the perception problem, no one in the press really cares about Intel any more because their stuff is boring. The company then replies with how it isn’t boring and there is tons of innovation under the hood while absolutely refusing to talk about any of it. Most of the press has moved on to the SoC/ARM/phone/tablet world because there is a lot of actual innovation there and plenty of the vendors actually want you to know about the cool stuff they are doing. For some odd reason, the press listens to this and not the empty headline grabs from Intel. It isn’t rocket science but for some reason, Intel seems hell-bent on digging their hole deeper. In this aspect they are wildly successful, hi down there!
So in the end we have another pathetic ‘brief’ about new ‘tech’ from Intel. Oops, should have put single quotes around new there too, sorry. There is no excitement around their products because Intel is actively stifling anything that would bring excitement to the mix, then spinning around in circles until they pass out trying to figure out what is going on. Meanwhile every tech enthusiast has moved on to ARM based systems and will be a hard sell to even listen to Intel’s meager attempts to be relevant there. In the mean time we get a ~3 year old name, a port count, and a few echoing gunshots when the decision makers read this story. Welcome to the new Intel, does anyone wonder why no one cares?S|A
Latest posts by Charlie Demerjian (see all)
- More on Intel’s 10nm process problems - Sep 17, 2018
- Intel puts out another 14nm 2020 server platform - Sep 11, 2018
- Why Can’t Intel Supply Enough 14nm Xeons? - Sep 10, 2018
- Intel can’t supply 14nm Xeons, HPE directly recommends AMD Epyc - Sep 7, 2018
- AMD reintroduces the Athlon name with two CPUs - Sep 6, 2018