Intel is once again wasting electrons and oxygen with a new low in briefings, this time ostensibly about the Knights Landing Xeon Phi. There was a lot of empty hollow grand strokes but precious little if anything worth listening to.
It is hard to decide if this was actually a new low or just plain worthless and awful, the Intel briefings of late have been so pointless and hollow picking an absolute is hard. That said the entire briefing contained nothing actually new other than a few marketing names. We are sorry to waste your time with it but we promise this article will be less pain than the actual brief.
Of the 15 slides there were two that contained anything worth calling information. OK there were three if you count the fact that the NERSC Cori system contains 9300 Knights Landing nodes is news, We don’t but either way it wasn’t worth a full slide. That said this slide was far more scintillating and on point than 12 of the 15, someone please shoot me.
The first of the 15 that had anything new was the Knights Landing Phi slide. Knights Landing is old news but the fact that it is DDR4 with stacked memory was sort of new. Actually our own Thomas Ryan told you about in 2011 even if he was a generation early. The memory is co-developed with Micron but no details were actually given other than up to 16GB per card is the maximum.
To be fair there were some BS technobabble details like stacked memory uses 1/3rd the space, 1/5th the power, and has 5x the bandwidth of DDR4 but these are obvious to anyone who has a clue about the tech. This slide is mostly the PR equivalent of saying a CPU is made on silicon and rectangular, then claiming that is breaking news. In short Intel wouldn’t say anything worth a damn but wanted headlines so they put out obvious technical specs as new. Sadly most of the ‘technical’ press was too illiterate to understand the difference and wrote it up. *SIGH*
The cores are Silvermont based, something that we told you about years ago, this is the ‘converged’ architecture we have been babbling about. We’ll save you the rest of the bullet points, they are not worth the IQ loss to repeat. In short Intel said nothing but is really proud of it for some reason, why do they repeatedly waste our time like that?
Slide two was on Omni Scale, the next generation of Intel True Scale Infiniband-ish fabrics. In typical Intel quasi-honest fashion they claim it is coming in 2015 and have some silicon photonics bullet points. Other than that there is nothing other than marketing BS, no real information. What were you expecting them to say, “Our approach to silicon photonics is so backward and broken that we are well over a year late delivering products but still can’t get yields up to production levels.“? That would be both honest and contain information so it wasn’t said, instead there were some vague hints engineered so that journalists wrongly extrapolate something positive. Most did. They shouldn’t have.
This is what scares Intel silly internally
The rest of the briefing was both vacuous and repetitious, nothing worth talking about other than the single slide above. Intel has been big on the Phi vs GPU debate and try to play up the programming headaches of a GPU vs the mostly x86 Phi. There is a lot of truth to that argument but the problem is that ARM servers are perfectly compatible, easier to program than a Phi, and most importantly, immensely cheaper than a Xeon, Phi or vanilla. ARM servers are currently steamroller-ing Intel in key high-margin areas but for some reason the company is pretending they don’t exist. Not sure why here but feel free to reach your own conclusions.
In the end the Knights Landing Xeon brief was the better part of an hour that I will never get back. Other than being painfully worthless the only other thing we can say is that it was once again purposefully made to lead journalists down the wrong path so they would write things Intel didn’t say. It worked, go read it for yourself. If you concluded that Intel has a chance in HPC against the new wave of competition or that their other programs are on track, that is what they wanted. Unfortunately it is wrong.S|A
Have you signed up for our newsletter yet?
Did you know that you can access all our past subscription-only articles with a simple Student Membership for 100 USD per year? If you want in-depth analysis and exclusive exclusives, we don’t make the news, we just report it so there is no guarantee when exclusives are added to the Professional level but that’s where you’ll find the deep dive analysis.
Latest posts by Charlie Demerjian (see all)
- AMD shrinks GPUs to 12nm with the Radeon RX590 - Nov 15, 2018
- Coolit water cools Cascade-AP CPUs - Nov 14, 2018
- Intel tries to pretend they have 5G silicon with the XMM 8160 - Nov 12, 2018
- AMD’s Rome is indeed a monster - Nov 9, 2018
- Intel announces Cascade Lake-AP MCM - Nov 5, 2018