Stage 2: Datacenter

Intel’s Datacenter Group SVP, Navin Shenoy, also took to the stage at CES in order to discuss some new products in Intel’s portfolio, as well as to deliver updates on ones that were disclosed last year. Back in 2018, Intel held its Datacenter Summit in August, where it lifted the lid on Cascade Lake, Cooper Lake, and 10nm Ice Lake. Along with this, we saw new instruction support for AI and security as the top two areas of discussion.

Cascade Lake: Get Yours Today

Intel’s first generation of Xeon Scalable processors, Skylake-SP, was launched over 18 months ago. We’ve been hearing about the update to that family, Cascade Lake-SP, for a while now, along with its brother Cascade Lake-AP and how it will tackle the market. The announcement today from Intel is that the company is now shipping Cascade Lake for revenue.

This means, to be crystal clear, that select customers are now purchasing production-quality processors. What this doesn’t mean is retail availability. These select customers are part of Intel’s early sampling program, and have likely been working with engineering samples for several months. These customers are likely the big cloud providers, the AWS / Google / Azure / Baidus of the world.

It’s worth pointing out that at Intel’s Datacenter Summit, they said that half of all of its Xeons sold were ‘custom’ processor configurations that were not sold though its distributors – these parts are often described as ‘off roadmap’. It is likely that when Intel says Cascade Lake-SP is shipping for revenue to select customers that they are likely to be purchasing these off-roadmap processors. They might be running at a higher TDP than Intel expects for the commercial parts, or have different core/cache/frequency/memory configurations as and when they are needed.

One of the big draws for Cascade Lake is Intel’s Optane DC Persistent Memory support, which will enable several terabytes of memory per socket, but also the hardware security patches for Spectre v2. Businesses who want to be sure their hardware is patched can guarantee security if it's in the hardware, rather than relying on a firmware/software stack. So this might be part of why Intel’s demand for 14nm CPUs is at an all-time high and outstripping supply – if a company wants to be 100% sure it is protected, they need the hardware with baked-in security.

The full retail launch of Cascade Lake is expected in 2019. Based on what we saw at Supercomputing in November, given by a rolling slide deck at the booth of one of Intel’s OEM partners, that time frame looks to be somewhere from March to May.

Nervana for Inference: NNP-I coming in 2019

To date, when Intel has discussed the Nervana family of processors, we have only known about them in the context of large-scale neural network acceleration. The idea is that these big pieces of silicon are designed to accelerate the types of compute commonly found in neural network training, at performance and power efficiency levels above and beyond what CPUs and GPUs can do. It has been disclosed that Intel is working on that family of parts, NNP-L, for a while now, and we are still waiting on a formal launch. But in the meantime, Intel is announcing today that it is working on a part that's optimized for inference as well.

There are two parts to implementing machine learning with neural networks: making the network learn (training), and then using the trained network on new information to do its job (inference). The algorithms are often designed such that the more you can train a network, the more accurate it is and sometimes the less computationally intensive it is to apply it to an external problem. The more resources you put into training, the better. But the scale of compute between training and inference is several orders of magnitude: you need a big processor for training, but don’t need a big processor for inference. This is where Intel’s announcement comes in.

The NNP-I is set to be a smaller version of the NNP-L and built specifically for inference, with Intel stating that it will be coming in 2019. Exact details are not being disclosed at this time, so we don't have any information on the interface (likely PCIe), power consumption, die size, architecture, etc. However, we can draw some parallels from Intel’s competition. NVIDIA has big Tesla V100 GPUs with HBM2 for training that can draw 300-350W each, with up to eight of them in a system at once. However for inference it has the Tesla P4, which is a small chip below 75W, and we’ve seen systems designed to hold 20 of NVIDIA's various inference processors at once. It is likely that this new NNP-I design is along the same lines.

Snow Ridge on 10nm: An SoC for Networking and 5G (Next-Gen Xeon-D?)

The Data Center Group will be making two specific announcements around 10nm. The first is disclosing the Snow Ridge family of processors, focused on networking and specifically targeting the wide array of 5G deployments coming up over the next decade. The purpose of Snow Ridge is to enable wireless access base stations and deployments, as well as functions required at the edge of the network, such as compute, virtualization, and potentially things like artificial intelligence.

Intel gave no other details, however going back in my mind, I realise that we’ve heard this before with Intel. They already have processors on their roadmap focused specifically on networking, with 40 GbE support and features like QuickAssist Technology to accelerate networking cryptography: the Xeon-D line of processors. This makes me believe that Snow Ridge will be the name for the next generation of Xeon D, either the Xeon D-2500 or Xeon D-3100, depending on the power envelope Intel is going for.

Given this assumption, and the fact that Intel has said that this is a 10nm processor, I suspect we’re looking at a multi-core Sunny Cove enterprise design with integrated networking MACs and support for lots of storage and lots of ECC memory. There’s an outside chance that it might support Optane, allowing for bigger memory deployments, although I wouldn’t put money on it at this stage.

Ice Lake Xeon Scalable on 10nm

To finish up Intel’s announcements, Nevin also talked about Ice Lake Xeon Scalable. At Intel’s Architecture Day, a processor was shown at the event that was described as Ice Lake Xeon, so this is just Intel repeating the fact that they now have working silicon in the labs. There is still no word as to how Intel is progressing here, with question marks over the yields of the smaller dies, let alone the larger Xeon ones. Working silicon in this case is just a functional test to make sure it works – what comes now is the tuning for frequency, power, performance, and optimizing the silicon layout to get all three. I’m hoping that Intel keeps us apprised of its progress here.



What Happened at CES 2018, and why CES 2019 is Different

A memory that will stick in my mind is Intel’s CES 2018 announcements. At the heart of the show, we wanted to know about the state of Intel’s 10nm process, and details were not readily available. 10nm wasn’t mentioned in the keynote, and when I tried to ask then-CEO Brian Krzanich about it, another Intel employee hastily cut in to the conversation saying that nothing more would be said. In the end we got a single sentence from Gregory Bryant at an early morning presentation the day after the keynote, and that sentence was only after 10 minutes of saying how well Intel was executing. That single sentence was to say that Intel was shipping 10nm parts in 2017, although so far only two consumer products (in limited quantities, and region specific) have ever been seen.

This year, coming off the back of Intel’s Architecture Day last month, shows that Intel is becoming more open to discussing future products and roadmaps. A lot of us in the press and analyst community are actively encouraging this trend to continue, and the contrast between CES 2018 and CES 2019 is clear to see. Companies tend to hide or obfuscate details when product execution isn’t going to plan; now that Intel is starting to open up with details, the outlook is clearly returning to one with more optimism.

Stage 1: Consumer
Comments Locked


View All Comments

  • HStewart - Tuesday, January 8, 2019 - link

    One more - I predict in 2020 (maybe even in late 2019) AMD will be seriously going back to drawing boards
  • TheJian - Tuesday, January 8, 2019 - link

    Who here believes everything revolves around desktops? I want one soon on 7nm, but I think servers/hedt are what matters, along with jacked up laptops (well, ok, everything above $300 matters…ROFL, where 80% of the NET INCOME comes from that is). HIGH margin is what I've been telling AMD to chase for ages. I never said wccftech was reliable...In fact I've said the opposite, but that doesn't mean they get everything wrong. However my 1st sentence after the quote is I think their article is BS...Did you read the post or get hung up on a few caps? LOL. I said the same thing about wccftech article on AMD pricing in the Navi post (I quoted them in 2060 or AMD CES news, then called the article quoted BS). Even they don't believe their own rumors on that one...LOL. Not sure what caps have to do with reliability. GET over it.

    Is Intel 10nm out in volume? NOPE, why would I compare something that won't even arrive until xmas vs. 7nm AMD coming before it? As I stated, I don't believe they can make anything else but a small chip on 10nm due to LOW volume they use. NO different than AMD/NVDA doing the same with mobile etc gpu first then reticle limit (for NV anyway) after. YIELDS matter. Anandtech all but says it here too. Yay, we agree on something anandtech ;)

    I know the tech matters, which is why Intel stumbled at 10nm. They shot too high (like HBM vs. GDDR5x, hard and new tools vs. simple with same tools), and failed for 2yrs now, 3 if xmas. So they shot low with 7nm to speed the process up and avoid bumps in the road. If AMD has 7nm or shortly, EVERY review will be vs. Intel 14nm chips until xmas, and even then as noted (including anandtech noting it), it's a low volume part and puny, meaning easy to produce even on a crap bad process with defect city. Also, 7nm TSMC vs. Intel 14 IS better, even if TSMC process is not as good per gen. Intel's 10nm will merely bring things to an even playing field vs. TSMC 7nm. Intel will be possibly slightly better (not looking that way now), but it won't be squashing AMD or TSMC any time soon. That will take over a year (likely 18mo+), and even then 5nm will be here for TSMC, so Intel 7nm (like TSMC 7nm as they took the easy way out, unlike usually aggressive due to 10nm failure) won't win any races really. They admitted they cut corners to speed it up. TSMC is actually making Intel chips due to 14nm shortages due to 10nm being screwed so long. You don't see the humor in this point? ;) But yeah, nm, density, gates, etc all mean a lot. But I get the tech, so point wasted on me, unless you have something that says Intel 10nm is great...WTH are you talking about?

    No desktops will never be replaced for power users. I guess you haven't been around for the last 30+ years of this crap, as it is all ~80-120w on desktops, and usually ~85-95w. Let me know when your laptop can take that heat/watts. What cpu on top is using less power on desktops or servers? I see them all aimed at ~85w-120w pretty much at launch for every gen.

    I said AMD only had 12-18 months in my post to make hay while the sun shines, IE, before Intel comes back if they can. But TSMC is on a roll lately, w/tapeouts, risk production etc, 5nm on track, 7nm out the door massively for apple etc. If Intel 7nm hits and tsmc 5nm very close time wise, the bleeding won't stop for Intel for a while past 12-18mo. In an even race in fabs (I'm admitting it takes a smaller process here for TSMC to match/beat Intel process in any case), the cheaper chip will win, and AMD always starts price wars (stupidly IMHO) at launches when they should be charging as much as they can get at launch until SOMEONE else starts that war (INTC/NVDA I mean). You seem confused about what I actually posted. I was very clear however with dates, launches, net income, pricing mistakes by AMD, R&D money, die sizes, what I’m planning on buying etc. How much data do I have to give to make the point clear? I feel like I’m repeating myself at this point when people should just READ the post (again? If you didn't get it) ;) You asked me to consider stuff, I already covered, albeit without mass granular detail that would bore most to death anyway :) What about my first post said to you that I'm ignorant here? If nothing is HVM basically until 2020 (xmas puny chips don't count IMHO, as it's a low volume product chosen), I'm right.

    I'm not confusing the discrete gpu coming (supposedly) 2020 and don't think that will be worth much anyway. Hardware is not enough, and I've seen multiple failures in gpu stuff from Intel over the years (I sold i740 for a while). You need great drivers all year too (game ready for big AAA launches etc), and this I think will be lacking for Intel based on MUCH previous experience with their crap. They can't even take down plucky little AMD apu's and they've had YEARS to get that done. Always losing to a company with 1B R&D...LOL, never mind NV R&D and wads of cash in the bank to fight on gpu fronts.
  • HStewart - Tuesday, January 8, 2019 - link

    You are the one that stated if 10nm is not in desktops, wasting time. To be honest - your comment seems like if it is Intel then you wasting time - so be a kind person and leave you comments to AMD articles. Just for information, I notice in other articles that AMD 7nm will not be ready in 2nd half of 2019.

    nm size is not everything, 10nm is only small part of Intel change coming in 2nd half of 2019 - likely 4th quarter - holiday season does not mean Dec 25th, it means before that like Fall 2019.

    But with Ice Lake - it not just the stupid process - but architxture which is more signficant than the process - adding more cache and excution units.

    By this article, Intel is not going low productions - not only notebooks and surely desktop - because that is where AMD fans like to fight - but also in servers and low end and next year will attack in GPU - likely ready for CES 2020. Intel is preparing to do not a Tick or a Tock - but this time they are combining Tick and Tock - mean both Process and Architexture change in a single releases.

    For me, my XPS 15 2in1 will last me to late 2019 or 2020, we are going to see signifcant change in performance with this new breed. To me Sunny Cove looks like Intel's most signficant Architexture change possible even surpassing the original i series from old Pentium 4 and Core 2 days

    More information on Sunny Cove can be found in this article
  • TheJian - Wednesday, January 9, 2019 - link
    NEW INFO First and foremost, 7nm announced today on stage, but GPU (331mm) Radeon VII. So TSMC working, we’ll see about cpu, but they can clearly do a cpu too as this GPU is 331mm^2 and that would be a fairly LARGE cpu die from AMD. But this does mean 7nm DESKTOP cpu WELL before XMAS, stating Q2/Q3 (back to school). MATISSE (mid2019 stated). The GPU is here early I guess since they just re-used MI60 chips for quick release (yield good, or left over?)
    So, adding this info above as I’ve been reading on/off all day (CES crap). So again, AMD 7nm CPU READY Mid2019.

    We are talking HVM, which I called BS until a HVM product is out the door, such as a desktop for instance as noted. Or heck, "insert any HVM larger die product" you want which would mean HVM is a go! Intel has claimed (apparently) that they have ANY HVM product capability, which again, I see NONE and a SMALL product in pretty low volume doesn't scream "success, we've done it for ALL chips, even the larger ones"...Uh, no, or start producing some!

    I said what I WANTED in a desktop from Intel (10nm! which again is high volume) or stated I'd buy AMD and if you're waiting on Intel, might as well take AMD 12core etc right (12/16 core will require 7nm to be effective from tsmc surely)? I gave an argument based on perf/process of the alternative, which to me seems very reasonable (it’s the only other choice in town). We see how far Intel is getting with 14nm (8 core finally mainstream? Not really at $478!), and 10nm MAY enable Intel to do a decent 12 core, but we’ll see vs. 7nm AMD from TSMC with it seems, up to 16 core on mainstream boards (room on ryzen2 (3rd gen) for another 8core chiplet) and If AMD IPC claims are true, a decent upgrade from the current story playing out. Assumed leaks say 12/16 core just moved down from HEDT to mainstream boards with AMD at some point. It will be fun to see Intel 10nm response (cancel a chipset at release again? LOL, my 6 core+Board was dead the day I bought it), just like it was fun watching pathetic 9900k/9700k etc. They had to turn HT off just to cherry pick enough for 8core/16 thread model IMHO. IE, even though they turned off ½ the threads, cache, etc, 9700k uses the same watts? How bad are your yields if you have to drop ½ threads & slow it down to hit 95w still (anandtech discussed this issue of yields and LVM too)? Is it the same core too, so you disabled 4MB of cache maybe to get yields up? No reason to leave it out as 9700k, slower, ½ threads is no competition for the chip above it and again, 95w with all this downgrading? Smells like yield issues even here at 14nm, never mind you moved production to SERVER, because you can’t keep up with demand already (thus hiring TSMC for some 14nm help…Not good either).
    Intel execs have now admitted they “bit off a little too much” on 10nm.
    “But regardless, we’re now talking about whether Intel can match its competitors on feature sizes, not whether it’ll continue to lead the industry.” Says it all doesn’t it? My point exactly. They are not “ahead” any more. Ziff Davis here.

    Pentium 4 is a great chip now, and you were excited by it?
    Crash landing, failure says Anandtech 2005 (many others they wrote too, like everyone).
    Netburst arch, REALLY was that BAD…I could go on, as every review site said this at some point. You imply this is a tech that was some kind of amazing achievement. I disagree, as does the whole web. “ridiculously huge pipeline”. You’ve never heard of PRESSHOT? Cooking with Intel? ROFL.

    Intel has been promising 10nm for over 2yrs (late 2016, as noted in above article from extremetech) and now seems 3yrs late if xmas 2019 for even a SMALL chip, not really HVM as Anandtech even noted. Now you think they'll get a HVM gpu out the door for CES in Jan 2020? You’re aware they’ll be 3yrs late with 10nm at xmas right (I said it before and here again)? They failed with i740, they failed with larrabee, drivers have NEVER been good on Intel (I know I sold i740 as a PC biz I had for a decade) discrete or integrated, etc. What in Intel's history says "this time it will work" and the design time is super-fast here? I think H2 at best for discrete from Intel gpu dept based on no cpu in 2019 for HVM lager dies.
    List of companies that have attempted to take on NV/AMD (er, ATI?) and failed, which BTW have mostly been gobbled up by AMD/NV. Please can you provide a link to info that says Intel will get it right where EVERYONE has failed before for 20yrs no matter the size of their budget, including INTEL? What has change THIS time?
    Clearly states Intel discrete gpu will be 10nm (Intel said it), and 2020 while I don’t think they meant Jan. OK, so you think Jan 2020 CES for a LARGE chip that will compete with AMD/NV cards at, spitballing here, 300mm^2+? It will be competing with faster chips than current 12nm 445mm^2 dies @10.8B transistors (never mind ridiculous reticle limit 2080ti chips, titans etc). Both NV/AMD will have gpus at 7nm, so you have to perform even faster than these right? Intel can’t put out a 200mm^2 die at 10nm and think it will catch much larger dies on TSMC 7nm. It would appear NV will wait on 7nm until 2020 (no competition in high end this whole year, AMD going low die size THIS year), but that is to put out another reticle limit top end chip again like 2080ti and raising prices after a cut surely when AMD hits 7nm. I mean, 754mm^2 is really pushing limits on yields here, 18.6B transistors and that would even be BIG at 7nm if you just shrunk this (~400-500mm^2?).
    AMD 7nm MI60 gpu is 331m @13.2B, so major reduction from NV 18.6B, so again guessing Intel will be competing with NV ~450mm^2 or so (mid of my guess? Though NV probably does better than AMD here, so ~400? Whatever it’s big) if they go 18.6B or more based on AMD 331 7nm from the same process and I’ll be shocked if NV’s 3080ti (guessed name, whatever) will be less than 20B transistors and I’d guess much higher. I’d be shocked if AMD doesn’t get navi 10 out by July, and may hit early Q2 as MI60 is Q1 for purchase. They said MI60 was a pipe cleaner (smaller 331 die, but on a low volume HIGH margin product in PRO stuff), signaling a LARGER die size on at least some model for desktop later in the works right (NAVI10 is small, so not the chip discussed here)? You don’t have to pipeclean for a small desktop gpu chip, you do it for a LARGE model coming soon. AMD/Intel/NV routinely put out smaller laptops first, pipecleaning for future larger models on new processes. You are living in a fantasy world if Intel is putting out the first “supposedly” HVM 10nm small chips for xmas devices not back to school, which MIGHT make your point valid. It’s XMAS, so AFTER sept or you’d say back to school, and gpu has been said to be H2 2020. Intel isn’t dumb going after poor, there is no money in it, ask AMD (Navi10…LOL, puny). Intel will attempt to chase NV high end for margins, which means BIG, and that isn’t coming in Jan as a pipecleaner is not what is coming for xmas from Intel. It will likely be under 150mm^2, not 331 like AMD pipecleaner.
    Cannon lake, shown running CES2017, but delayed to 2018, then CES 2018, shipping would ramp 2018, ooops, April 2018, shipping LOW VOL 10nm & cannon would be delayed AGAIN to LATE Q2 2019. It’s in a nuc now, but we’re talking a 70mm^2 dual core chip here and NUC’s don’t sell volume anyway. How many people do you know that own HTPC’s? These are less than that and expensive for a puny weak dual core (~$540). As noted they’ve been fighting yields and costs due to them sucking. They only produced ONE model AFAIK.
    Architecture is important but no more so than shrinks and process upgrades (Intel 14, 14+, 14++ etc) as these allow massive watt drops, transistors, and shrinking die sizes to make more money. IE, from 4th to 5th gen on Intel, you got 35% more transistors and a 37% smaller die size (131 down to 82mm^2, and .96B to 1.3B transistors). Not sure how you can just discount this side of the equation and act like it’s just a “stupid process”…LOL. OK. You trolling me? Processes are stupid? C’mon man. Discrete gpu 2020 in January? Are you an Intel shareholder (I have been, not now) with wishful thinking or what? If Intel won’t even put out a “tentative” release date, it’s not JAN 2020. Why would you say something as non-specific as 2020, if you meant Jan, or at worst Q1? Heck they could have said H1 2020. You say 2020 when you have no idea yet as of CES 2019 speeches. NO date, just 2020. I’m sure they’re praying for Jan, but not unless they’re chasing poor people like AMD, and that is NOT Intel’s style. Also note, Intel was supposed to be on 10+ by now and 10++ this year.
    “A big part of keeping up the improvements to the system architecture remains the process technology.” AGREE, and the 10nm “stupid process” has held up the entire company product line right? But who needs this process crap…LOL.
    “Process technology details are almost always kept secret until released due to competitive reasons.” Well, if it is just “stupid process” junk, why the heck do you think Intel guards it so closely until RELEASE? GF process has literally killed a product or two. But no, not very important even though they can literally CRUSH your product launch. Just ask Intel. They are getting less out of process today than 10yrs ago, but it’s still worth a lot. We’re done here, as I’m realizing responding to your comment is kind of a joke at this point (perhaps on me for doing it…LOL), but for anyone buying these stocks, it’s relevant info anyway and I have to do the homework like it or not (or lose money) ;)
  • lashek37 - Tuesday, January 8, 2019 - link

    I’m more confused than than a homeless man on house arrest. your but must really hurt.
  • croc - Tuesday, January 8, 2019 - link

    What does "thejian" mean in Russian? I see one of the Russian Trump 'supporters', I call someone...
  • HStewart - Tuesday, January 8, 2019 - link

    Even though, I disagree with "theJian" technical, I would leave any type of politics out of technical discussions.
  • ASAPscotty - Wednesday, April 3, 2019 - link

    Blame everyone else for your problems, write a ton of shit no one cares about or will read, and then blame everyone for that too. You definitely aren't mentally unstable.
  • TheJian - Monday, January 7, 2019 - link

    "Intel also demonstrated how users will be able to play with its amped-up Gen II integrated graphics. Unfortunately, Intel didn't disclose any setting that they were playing the game at."

    Anandtech calls no roadmaps open? NO settings open? Did they say anything of SUBSTANCE today? I've been reading Intel crap for an hour and still don't see any data from CES 2019. I see 2018 repeat.
  • Jorgp2 - Monday, January 7, 2019 - link

    Dude, we know the EU counts.

    So we know around where it will perform.

Log in

Don't have an account? Sign up now