Find me on Forbes!

After nearly six years of sporadic musings on Arm server topics, I’ve joined Moor Insights & strategy as Sr. Analyst covering AI and HPC topics.  I’ve enjoyed writing for this site, but I am now posting to Forbes where I can reach a broader audience. This article, for example, has over 13,000 views since Monday, and my article today about Google’s cool but  mysterious TPU has already reached over 3000 readers.

So, I will miss writing about cool ARM stuff here, but will not abandon the mission as I write under a new banner.  Hope to see you there!



Is ARM Nearing “Xeon-class” ? And Does It Matter?

I’ve often opined that ARM V8 server SOC’s would struggle to meet Intel Xeon head-on until a) appliedmicro_x_gene-1200x794they get a few micro-architecture revisions under their belts to improve per-core performance, and b) they can narrow the manufacturing gap to 14/16 nm, or perhaps even 10 nm.   Now,  Applied Micro is aiming to do just that with the X-Gene 3.  They have claimed this lofty goal before, with their 1st and then 2nd attempt at X-Gene’s custom cores. Now, it appears that their third generation custom-core SOC will (finally) deliver performance that is “well within range of Xeon E5 products”, according to a recently released report by the Linely Group, the widely read and respected publisher of the “The MicroProcessor Report.”

This announcement tees up 2 important topics.  1) Will X-Gene3 really deliver “Xeon-class” performance? and 2) does it matter?  i.e., do ARM server partners really need to chase Intel Xeon, or is there a market for less powerful “Atom-class” SOCs in the data center?   See my perspective on The Next Platform.


Has OpenPower Crushed ARM at Google?

Anyone who still doubted that large datacenters are serious about inducing  a competitive imagesCPU field certainly got a wakeup call last week when Rackspace and Google announced that they are teaming up to enable their use of OpenPower when IBM releases their POWER9 chip in late 2017.   Some have questioned whether this means ARM is now far behind,  and one does have to wonder how this could impact Qualcomm and Broadcom’s ARM Server SOC plans.   Let’s take a look at how this might play out.

First,  these pronouncements do not amount to an endorsement of OpenPower at the expense of other alternatives, namely ARM.   To the contrary,  they amount to an endorsement of HAVING CHOICE.  This is really great news,  showing that some of the industry’s titans are already preparing their software for alternative architectures and are willing to invest significant efforts to enable choice.

Second, one thing I learned at Calxeda and AMD was that large datacenters like Google, Facebook, Baidu, and even Microsoft are NOT fans, per se, of any particular architecture or supplier.  A “Fan” is someone who wants to see his team (chip) win, but these companies don’t give a whit who wins, so long as they can get more performance for less.  Its not that they don’t like Xeon, or ARM for that matter.  Xeon has been a unstoppable force in lowering datacenter costs for well over a decade.  But, everyone knows that competition spurs innovation and lowers costs. Google has repeatedly stated that they would adopt a new architecture if it delivered a 20% advantage over the incumbent (Intel Xeon).  But the only thing better than having a choice of 2 architectures for the datacenter is having 3,  and ARM is still the odd-on favorite to grab a share of the pie.

Finally, did ARM somehow blow their “lead”?  The fact is that OpenPower is ahead of ARM V8 64-bit implementations in terms of performance, and features.  However this comes at a significant cost and power premium over ARM implementations such as AMD, APMC, and CAVM.  its important to note that neither Google nor Rackspace are deploying the current POWER8 chips at this time in production workloads, presumably because of those premiums.  IBM clearly admitted that themselves by announcing a roadmap that has Scale Up and Scale Out (cheaper) versions of POWER9.  6965821-14600374310280457_origin

So where does this leave ARM and their partners?   Let’s be honest: ARM V8 implementations to date have been too wimpy in terms of per core performance,  and their energy efficiency for datacenter workloads has not kept up with Intel’s  Xeon D and Atom chips. We will see if Qualcomm can change that,  probably by the end of this year when many expect them to start sampling their 64 (!!)  core beast more widely.  However, going to 64 cores almost assures that this chip will not deliver the ~20 SpecINT per core performance threshold most hyper-scale and cloud users will demand, limiting its target market and SAM.  As for Broadcom, they have been strangely silent since their acquisition by Avago, causing many industry insiders to wonder whether the Vulcan chip will survive the acquisition budget and prioritization process. After all, the company had set expectations that Vulcan would ship by the end of LAST year, and I haven’t seen any evidence that this goal was met.

But lets face it, ARM does need to up their game with faster microarchitectures and more datacenter features, or OpenPower9 may carve out the lion’s share of the non-Xeon market in the 2019-20 timeframe. So, I  suspect we will be hearing a lot about “Ares”. ARM’s high end processor design for 10nm FF, this fall at TechCon,  if not sooner!  (They’d love to spoil Intel’s IDF party in August!)



Is Google Really Ready to Switch to ARM?

There have been cascades of repeated headlines and rumor-rich articles running over the last few days that assert that Google is “looking to replace Intel with Qualcomm ARM chips” for their data center servers.   Everyone loves a good fight, so its not surprising that the press is jumping up and down, yelling “Fight!  Fight!  Fight!”   But seriously folks,  do you really think that Google would announce such a strategy before they even see silicon?

First, I have not seen any reports claiming that Qualcomm is sampling silicon yet.  It takes a lot of time and effort to produce, optimize, and test an SOC, especially one capable of meeting the challenges of the datacenter.  And even if they are sampling, its only first silicon samples, which means it won’t be full performance (frequency) nor fully functional.  Which means Google cannot be certain that it meets their requirements.  Hey,  Google didn’t become the largest company in the word (at least for a few moments) by making brash decisions about their technology.

Second, Google, no doubt, is anxious to lower costs by evaluating alternative technologies, including ARM and POWER.  But having worked with them before,  I can tell you that they will be thorough, methodical,  and meticulous in their analysis of any Xeon alternative. They will need at least a year, I would suspect, before they put anything into pre-production.   And they will likely avoid a patchwork approach of workload specific architectures, which creates lifecycle and system/network management process nightmares.

Finally, Google is very very smart. They will happily talk about potential Intel replacement technologies, because this puts Intel on notice and strengthens their negotiating position.

So, I think its great that Google likes the PROMISE of Qualcom’s ARM SOC!   But that is a far cry from the predicted premature announcement that they will adopt it.  More likely, they will say they are working with Qualcomm, just like they are working with

Calxeda lives! (Well, at least the Fabric does!)

Last week, in conjunction with the AMD announcement that they are (finally!) shipping the Opteron A1100 series ARM Cortex A57,  a little-known company debuted 2 fabric interconnect solutions for Hyperscale data centers. As some of you may recall, AtGames Holdings purchase the Intellectual Property from Calxeda’s bankers a little over a year ago.  AtGames was lined up to be Calxeda’s largest initial customer, and were left in a jam when Calxeda suddenly folded. The SilverLining Systems subsidiary announced that they had taken the Calxeda technology and have repackaged and reimagined it to produce two new products aimed at the Rack-scale Fabric market. Given all the excitement this fabric generated, this was perhaps more newsworthy than the expected AMD announcement,  so I wanted to chime in with a few thoughts


The first product is a PCIe card with one Calxeda ECX-2000 ARM SOC,  supplying 4 XAUI ports  to interconnect other servers on eh fabric,  and one 10GbE SFP for uplinks to the Top Of Rack Switch.  It also supports up to 8 GB of optional DRAM to enable the four ARM Cortex A15 processors (normally turned off when the card is acting only as a fabric adapter) to execute jobs such as packet inspection and other offload tasks.   The second product, which will be available in Mid 2016, is an ASIC for custom server developers.  The Fabric Interconnect Chip (FIC) is basically the ECX-2000 without the cache and the A15 cores, providing the 80 GbE fabric switch, the PHY’s, and the ARM Cortex A7 to manage the routing tables and optimize traffic flow across the fabric.  These products will reduce in-rack networking costs by around 75%, according to Silver Lining Systems.

With this announcement, the promise of the Calxeda Fabric is decoupled from the ARM world and can now be used to interconnect standard x86 servers, or build custom dense servers with virtually any processor or SOC.  TPM has already covered the products in this article, but for those who knew and loved Calxeda,  I’d like to add a few comments.   (Full disclosure:  I have been recently consulting with Silver Lining Systems.)


The SLS Newport fabric interconnect adapter (FIA)

There is a lesson here for all you guys trying to provide innovative new technologies. One that is so obvious it can easily be overlooked in the zest for New and Improved!.   Your new product must provide unique value but it MUST BE EASILY CONSUMED.   Calxeda changed too many things at once:  “Here’s a new processor  with a new (for servers) instruction set, a new (fledgling) ecosystem, in a new form factor,  and with a radically different networking topology and management approach.”  “WHOA!” the customers all said.  “One change at at time, please!”

Now, had we simply focussed on ARM, we might have been more successful, but without a differentiator we wouldn’t have lasted long.  Had we just been a Fabric company, going up against Mellanox and CISCO and Intel and …,  we never would have gotten the funding. BUT, had we produced two dies, the Fabric customer could have kept his Intel processors, as SLS has done now, or just used the ARM SOC  as AMD has done,  or they could combine the two.  But we didn’t offer them that choice.  We realized the problem, but it was too late to fix it and we ran out of cash when our investors ran out of patience.

So, bottom line?  Change as little as possible and still produce value through innovation. OR produce a plug and play complete solution like an iPhone that fits within the larger infrastructure the customer is already comfortable with.

Three New Year Resolutions for ARM Server & SOC Vendors

As we enter 2016, its probably a good time to reflect on where the ARM Server movement stands.  There are now two vendors with production V8 parts in the market, Cavium and Applied Micro, and more are on the way from AMD, Huawei, Qualcomm, and others.   So, the future looks bright, with lots of promises from major vendors.  On the negative side, there remains no mainstream ARM-based servers in the market,  and zero production use cases we can examine.  Even more concerning, there still aren’t any published benchmarks of note. Let’s examine what needs to be done to restore some luster and credibility to this powerful vision of an alternative architectural standard that can compete with Intel Xeon for the server market.

2015 ARM TechCon and SC’15 recently provided fresh opportunities for the ARM Server community to talk up the latest chips and roadmap announcements.  The surprising news was: there was no new news!  Sure, AMCC announced intentions to produce a 3rd generation SOC.  Thats not news; of course they will do a 16nm part with lots of cores and goodies after they get their 28nm part out.  And Gigabyte announced a Cavium-based server design,  a year after Cavium supposedly was ready for prime time. But where were the use cases and success stories?

A case in point: During a well-attended HPC session sponsored by ARM Holdings at SC’15,  Dr. Olof Barring of CERN, lamented that the reality has fallen short of the claims made by the SOC vendors.  CERN, one of the world’s leading technology institutions and champions of power-efficient computing and storage,  has been unable to  acquire 64-bit ARM servers that vendors claim are supposedly “in production”.  Perhaps worse,  the 64-bit prototypes he has been able to get his hands on did not demonstrate performance per watt  advantages touted by their vendors. “Neither ARM nor POWER 8 has delivered the performance per watt we see today with our Intel servers”, he said.

The harsh reality is that the current (initial) batch of ARM V8  SOCs are still seeking their niches in the market,  and it will take time before we see competitive ARM SOCS for general purpose server workloads.   So, the perennial battle cry has been “ARM Servers are coming!”, and unfortunately this will remain the case for some time to come.  (See Timothy Prickett Morgan’s excellent article on why we are still waiting.) But eventually, ARM will succeed with advances in core microarchitectures, and by narrowing the gap afforded by Intel’s Fabulous Fabs.  With this in mind,  here’s my list of suggested New Year’s resolutions for the industry’s players to consider:

New Years Resolution #1:  Be Patient. The effort to bring forward a competitive SOC and software ecosystem will take years to materialize. We will need faster cores, advanced process nodes like 14nm FINFET,  and lots of work on the optimized software stack such as those begun by RedHat, ARMH, and Canonical..

New Years Resolution #2:  Tell the truth.   ARM Server SOC vendors have been fairly undisciplined with in communicating the facts in terms of  schedule, performance, and power consumption, resulting in the perception of “late”, “slow” chips that are not as power efficient as their Intel competition when it comes to number crunching. Vendors should set realistic expectations for schedules, performance, and power consumption and be very explicit about their SOC’s applicability for specific workloads.

New Years Resolution #3:  Be transparent.  A corollary of #2.  Your customers are very smart; they can handle the truth.  And they tire of hearing claims of superior performance per watt without any reproducible benchmarks or 3rd party measurements. So, give your gear to folks like Anandtech and let them measure the efficiency with real-world workloads. Yes, synthetic benchmarks will always give Intel an advantage, and they aren’t relevant to the sort of workloads you are targeting.  So tell us something that *is* relevant to your markets.

Have I grown skeptical of ARM’s potential in Servers?  No, I remain enthusiastic and optimistic about the future.  Lots of hard work remains.  Be vigilant, take the high road,  and trust that your customers are smart enough to tell the difference between facts and BS.

Note: the opinions expressed in this blog are solely the views of the author.


I’m Baaaaccckkkkk!!!! (And just in time, too!)

Last week,as followers on LinkedIN may know,  I left AMD after a 16 month stint leading the strategy and marketing efforts for Servers (ARM and X86),  and more recently as GM for HPC.  Now that I am unemployed, I am free (within reason) to speak out again on whatever pops into my bald head, so get ready too rumble.

Karl at CastAt AMD, the latter (HPC) gig was a blast, while the former,  well, not so much. The AMD HPC strategy was picked up by the media and will be a hot topic at SC’15.  In the realm of the server, AMD’s market share continues to plummet as they await both the production ARM Seattle part as well as the Xen-based next gen X86.  Meanwhile. there’s not much to sell and market!  More to come on both of these topics (although don’t expect any disclosures of a confidential nature!).

Meanwhile, ARM TechCon is only a few weeks away, followed by SC’15 in Austin (My Home Town!).  I will attend both,  so ping me at if you would like to get together at either event!  I’ve been known to frequent the fine dining and drinking establishments around Austin and can show you a good time.

So, if anyone is still out there looking for an alternative to Xeon for their server apps, keep checking in here.   I will restart my blogging here and hope someone is listening!!!

Ciao for Niao!

Karl , Chief Explorer

RUFN (Retired Until Further Notice)

A not-so-small step forward for ARM-kind



Two major milestones were reached this week for fans of ARM-based server gear.   First, HP and Applied Micro announced the 1st production ARM server this week with appropriate fanfare.   Here’s an analysis by Paul Teich and Gina Longoria (another Calxeda Alumni) of Moor Insights and Strategy.   Second, AMD showed off 2 OS’es (RedHat and SUSE), 2 JVMs (OpenJDK and Oracle),  and Hadoop running for the 1st time on an ARM A57 based server at JavaOne.   As Harish Jonnalagadda of BSN noted,  “Adding Hadoop functionality to its software ecosystem is a natural move for AMD as its target clients will be looking to use its server clusters to process large data sets. The low-power nature of the CPUs make them ideal for processing large chunks of information and undertaking high I/O tasks.”

[Read more…]

Haswell E5 and the iPhone6: Why would anyone ever need more?


Aiphone6s expected,  Intel® announced the Haswell E5 processor family for Servers and Workstations at IDF on September 9.  Coincidentally the event was just up the valley from Apple’s event announcing the (ARM-based) iPhone 6, 6Plus, and Apple Watch.  Between the two media-saturation blitzes, one could barely find coverage of misbehaving NFL stars or Russian would-be Czars in the day’s news headlines.  While few would connect these two events in any way,  to me there is  a common thread,  best summarized by my interpretation of their messaging:  “It’s a floor wax! It’s a desert topping!  Its everything you ever wanted,  and more!”   If you’ll allow me, …

[Read more…]

I was so impressed with AMD’s strategy, I joined the company!


As an analyst I was happy to attend AMD’s industry analyst summit back in May. I learned how they are leveraging their expertise and IP to implement an ambidextrous SOC strategy, including both custom and standard ARM cores. In fact, I am so bullish on the strategy and team, I’ve decided to join up and become the VP Marketing for server products.

Not sure what that means for this blog, but hopefully I can continue to post interesting stuff for y’all!