Hewlett Packard Enterprise To Acquire Supercomputer Maker Cray for $1.3 Billion (anandtech.com) 101
Hewlett Packard Enterprise will be buying the supercomputer maker Cray for roughly $1.3 billion, the companies said this morning. Intending to use Cray's knowledge and technology to bolster their own supercomputing and high-performance computing technologies, when the deal closes, HPE will become the world leader for supercomputing technology. From a report: Cray of course needs no introduction. The current leader in the supercomputing field and founder of supercomputing as we know it, Cray has been a part of the supercomputing landscape since the 1970s. Starting at the time with fully custom systems, in more recent years Cray has morphed into an integrator and scale-out specialist, combining processors from the likes of Intel, AMD, and NVIDIA into supercomputers, and applying their own software, I/O, and interconnect technologies. The timing of the acquisition announcement closely follows other major news from Cray: the company just landed a $600 million US Department of Energy contract to supply the Frontier supercomputer to Oak Ridge National Laboratory in 2021. Frontier is one of two exascale supercomputers Cray is involved in -- the other being a subcontractor for the 2021 Aurora system -- and in fact Cray is involved in the only two exascale systems ordered by the US Government thus far. So in both a historical and modern context, Cray was and is one of the biggest players in the supercomputing market.
Relevance? (Score:1)
Let me preface this by saying I've been a fan of Seymour Cray's hardware for years - back when you could sit on them (without causing panic).
Is Cray really that relevant nowadays? They started out as a system integrator - of their own hardware and software - but nowadays, they deploy Linux and build systems based on x86 processors with NVIDIA compute cards. Yes, they are good at making it all work with high-bandwidth, low-latency networking... but anyone (including HPE) could do the same over time.
What doe
Re: Relevance? (Score:4, Informative)
No, that was spun off into its own separate company named Agilent. The peripherals from the previous post probably refer to the printer (inkjet and laser) business that they still have a hand in.
Re: (Score:2)
And the test equipment division was spun off from Agilent a few years ago to form Keysight.
These days Agilent primarily does hardware for bioscience.
Re: (Score:2)
..and Keysight sell re-badged Rigol...
And I love my Rigol gear. Cheap and good. I use the super expensive test equipment at work and they are generally a pain in the neck to use. If I didn't need 10GHz+ I'd bring in my own scope.
Re: Relevance? (Score:1)
(Not same AC) I use rigol at home, but at my work we had issues with cross channel coupling on rigol scopes. To be fair, lower tiers of other major scope makers have similar problems because they are all cutting corners on the front end. These days the front end is all that seperates a good scope from just an ADC, considering how much data is just processed on a computer using a scope in a headless manner. I am still surprised how many use cases there still are where a scope is cheaper than a display-less d
Re: (Score:2)
I prefer the permanence of a scope or similar bit of lab gear, over a headless thing with a UI on a PC. I've seen countless headless bits of test equipment laying in the corner of the lab, which require some unknown bit of lost software to run on Windows 3.1 and talk to it through a parallel port. Meanwhile a scope of the same age is still functional and useful.
Re: (Score:1)
Don't want your tools to become obsolete?
Demand and only BUY products with Linux or U*ix drive/display and published source code.
Spend a few bucks on some half-way decent Iron Box programmers
RESELL the constantly upgraded software.
Parallel port? Buy one of these and quit bitching [walmart.com]
Re: (Score:2)
Dunno about that.
The screens on the old tek scopes I used are still there. Never 'reproduced'. The screens on my laptops and PCs have been replaced many times over the same period.
Re: Relevance? (Score:4, Informative)
The old Hewlett-Packard has split into quite a few companies. As far as I can tell
HP make printers, PCs etc.
HPE make servers, network switches and other enterprisy hardware. They also have a few service offerings left.
DXC Technology* does the enterprise service stuff.
Agilient make life sciences lab equipment.
Keysight** make electronic test and measurement gear.
Broadcom*** make the optoelectronics stuff.
* Formed by the merger of CSC with the former enterprise services division of HPE in 2016
** Keysight split from agilent in 2014, who in turn split from HP in 1999.
*** Avago split from agilent in 2005 who in turn split from HP in 1999. In 2016 avago bought broadcom and renamed themselves to broadcom.
Re: (Score:1)
Well..... yes and no. Don't forget that HP Enterprise Services (HPE) is the combination of Electronic Data Systems (EDS) and Computer Sciences Corporation (CSC), two big system integrators/service providers by themselves. They are more the "How about we throw in some free laptops and printers with this $1B multi-year contract" and less the "For only $5 more, we'll throw in a free ream of paper!" types.
Re: (Score:2)
HP is back to hardware, they still deliver good servers and switches on top of the laptops and such most people see. I've installed a couple HP High Performance compute clusters and they work well and handle integration nicely.
Re: (Score:2)
Re: (Score:2, Informative)
In particular the Gemini product line from Cray provides high speed memory access and clustering up to 50 peta-flop capabilities.
Beyond that they have expertise in system design and job scheduling that will become commodity products as more companies jump on the personalized genetic medicine band wagon
Unfortunately, we have also seen HP kill the DEC ALPHA processor (while supporting Itanium and hiding ALPHA benchmarks that crushed the Intel product), so there is a fair chance that Cray will suffer a similar
Re: (Score:2)
Unfortunately, we have also seen HP kill the DEC ALPHA processor (while supporting Itanium and hiding ALPHA benchmarks that crushed the Intel product), so there is a fair chance that Cray will suffer a similar fate
Alpha was killed by the clock rates of x86, and its coffin was screwed shut by amd64. Once AMD adopted NUMA and HyperTransport, it became simple to build PC-based computers with very large memory and high core count. Intel followed suit some years later, and now there is no real benefit to doing anything other than throwing more x86_64 cores at your problem.
Re: (Score:2)
Then please help me understand why an ALPHA derivative is the core of the #1 Top500 machine (since the Chinese brought ALPHA back to life on newer fab processes)
They threw money at it. That doesn't make it the best option.
ALPHA could have reached much higher clock rates if moved to newer fab processes
Sure, but some of the best ideas of Alpha made it into AMD processors anyway, and we went another direction — one with greater compatibility.
AMD would have never gotten their hands on HyperTransport if HP had not killed the project and sent their engineers (and technology) to other companies.
So, I'm glad?
Re: (Score:2)
Sure, using risc to simulate x86 microcode is the "best" solution, you get an atta-boy for compromise, but at what cost?
Well, let's talk about that cost. The x86 decoder is a tiny, itsy-bitsy part of a modern CPU. It was a big deal, granted, but now it isn't.
bu, bu, bu Intel does it too doesn't make it the best solution, x86 processors are pretty much steaming piles of security leaks at this point as a result of trying to push for performance while maintaining backwards compatibility for installed software base
POWER 7-9 and the latest UltraSPARC were both vulnerable to both MELTDOWN and SPECTRE. The idea that x86 is especially insecure is utterly false. AMD, on the other hand, made processors not vulnerable to MELTDOWN, and where SPECTRE-type attacks are less serious and cheaper to mitigate. Granted, Intel failed pretty badly at security, but they're hardly alone in that.
Re: (Score:2)
an ALPHA derivative is the core of the #1 Top500 machine
You information is out of date. [top500.org] The top two machines are IBM PowerPC.
Re: (Score:2)
Alpha was killed by Compaq - before HP bought that company. Besides, HP itself had already decided w/ Intel that Itanium would replace both PA-RISC and x86. While the latter failed, primarily due to AMD coming out w/ AMD64, the former was achieved pretty easily. It didn't make sense for Compaq - having aborted its own RISC platform PA-RISC - to now bat for Alpha, which had worse yields and worse margins
Re:Relevance? (Score:5, Insightful)
Basically, if US government is going to build a top-10 system, Cray or IBM is simply going to be the one they go with.
Part of that is that it is simply not as easy as it sounds. You don't just slap linux, some cpus, and some gpus together and call it a day. There's a bit more to it.
Another is that they are doing their own interconnect. The only other companies in the world that are on par with Cray's is Mellanox (to be acquired by nVidia) and Intel (OmniPath). Those interconnects are absolutely required to deliver best performance. They are however niche, as they require very carefully designed topology to actually deliver value, and can't provide a significant chunk of capabilities that are typical in ethernet networks.
Cray also has software that is only paired with Cray solutions, so a Cray software shop will buy Cray.
Of course, their brand value and relationships with certain government entities effectively seal the deal. With this HPE basically bought a guarantee of most future national labs business. Now they really only have to compete with IBM for that business, since they had already bought SGI (which is a prime example of a company who at the end had no significant technology advantage, but brand strength and government habits basically handed them a lot of US federal business as well).
Now a few years ago, this probably would not have been worth it, as the US government super computing budget went anemic (US or US allies were the super computing leads, so hard to justify big purchases to reinforce that). However China overwhelming the top slot triggered the US govt to inject lots of funding to win the pissing contest.
For Cray this is also an easy call. They have had a tight grip on *gigantic* US super computing deals, but pretty much no viable business anywhere else. It is exceptionally difficult to pay for the dedicated R&D to keep that up on that low volume of business. Additionally with the reduced competition, I would expect winning bids to come in a bit higher and with some richer margins.
At the end, HPE will become 2nd place Top500 Vendor by number of entries behind Lenovo, and edge out Lenovo for top performance share (the performance share a better rough indication of the revenue involved). When the two current Cray national labs wins come online, that performance gap will almost certainly widen.
Of course the Top500 benchmark and logistics are pretty flawed, but it's still a coveted marketing win.
Re: (Score:2)
Re: (Score:2)
Wow, somebody actually remembers what made Cray special. Brings a tear to my eye. And no, ECL (or its derivatives) is still the fastest. But as MOS devices shrank and various capacitances went down, the tech got pushed not too far behind ECL. I think that once EUV runs out of room to shrink things and all these GAAFET fads will play out, we will be back to building things out of some nano version of BJTs. I also hope to live to see the day when nanofluidics on a chip with superfluid helium as coolant are st
Re: (Score:2)
Is Cray really that relevant nowadays?
And the other big question is, how is HP still in existence?
I dismember Momma . . . (Score:2)
Back in the late 80's . . . HP was a cool place to work for doing things "The HP Way".
. . . and DEC was a cool place to work, because we all logged in in terminal room to a VAX.
. . . and SUN was über cool because of their workstations.
IBM was totally out, because they did mainframes.
So the joke is . . . which one of the companies listed above still exists today . . . ?
Re: (Score:2)
You need to take a look at Slingshot.
Sigh (Score:5, Interesting)
Another once great company to get pushed through the HP meat grinder. Is there anything that HP hasn't destroyed? Their lab equipment division was lucky enough to spin off as Agilent (Keysight), but other than that has anything ever been touched by HP and survived to tell the sad tale?
Re:Sigh (Score:5, Interesting)
So this is the Cray that Seymour Cray left, that merged with SGI, then was sold to Tera Computer Company, which then renamed itself Cray, right?
As opposed to the Cray that Seymour Cray founded after leaving Cray, which then went bankrupt, or SRC that he founded after that.
Smart man he was... start a company, then stop being CEO and become a consultant.
Re:Sigh (Score:4, Informative)
The original Cray did very well for a while, but the main reason for their prosperity was government spending on supercomputers for weapons simulation. Towards the end of the cold war this became increasingly difficult to justify, and Cray began its decline. Cray (the man) was pressured out of the company since he wanted to pursue some pretty outlandish and risky designs that his board couldn't see as justifiable, but he left on good terms more or less. "The Supermen" is a great book if you're interested.
Re: (Score:1)
WTF?
Sure, and that is why China just took an open-sourced ALPHA design, upgraded it to modern fab specs and produced 2016's top super computer [pcworld.com]
Where is Itanium in the TOP500 list?
DEC ALPHA had pre-planned generations out to 2030, which HP scrapped.
Face it, sometime chasing quarterly numbers and pursuing profits over any long-term planning results in FAILURE
Re:Sigh (Score:5, Funny)
Isn't that an old joke in there?
Q. How do you know when a tech company is dead / past its prime? /zing!
A. When HP buys it!
Re: Next up. (Score:1)
I want HP to buy Apple, using borrowed money, then using Apple's piggy bank to pay off the loan. It seems win-win to me. HP laptops, with actual working keyboards, would be a step up from Macbooks.
Like this? (Score:2)
Engine Block Shredder
https://www.youtube.com/watch?... [youtube.com]
Cray SGI HP (Score:2)
Re: (Score:2)
This acquisition is about putting a stake through it's heart, as clearly SGI and HPE haven't been able to kill it.
Re: (Score:1)
This will be the advent of the super printer, but with insane ink prices.
Another historic company dies... (Score:4, Funny)
m
Re: (Score:2)
You do realize that Taco Bell and Pizza Hut are owned by the same company right? They own KFC too...
Re: (Score:3)
The joke was that in the US cut of "Demolition Man", the last restaurant in existence was Taco Bell, but Taco Bell was not considered to be well-known enough in overseas markets, so for cuts in most other countries the last surviving restaurant shown in the film is instead Pizza Hut.
Since the deal for last-surviving-restaurant was with Yum Brands, of course the alternate cut for overseas viewers was also a Yum Brands product.
Re: (Score:3)
Get started learning how to use those three shells...
Re: (Score:1)
Can you please tell me how to use the shells?
Re: (Score:2)
Re: (Score:3)
HPE also bought whatever was calling itself SGI at the time a little while ago too.
History goes like this ... (Score:5, Insightful)
History goes like this ...
Cray makes good Supercomputers
SGI acquires Cray
Original Cray devs leave for HP and build their Cross-Bar architecture machines (starting with T-class, V-class, SuperDome, RP/RX series)
SGI start building computers with x86 CPUs
Cray spins off from SGI. SGI dies a slow death.
HP fires most CPU & X-Bar architecture teams to save $$$
HP releases "x86 SuperDome crap" which nobody wants to buy
HP acquires Cray hoping to get some smart people to make them a sell-able product
Future:
HP fires most Cray smart people to save $$$
HP releases overpriced x86 crap nobody wants to buy
HP acquires another company
Re: (Score:2)
Re: (Score:3)
Um... I'll add a few tidbits
History goes like this ...
Cray makes good Supercomputers, but there are few sales.
Cray goes bankrupt, and is purchased by SGI
some Cray devs leave for HP and build their Cross-Bar architecture machines (starting with T-class, V-class, SuperDome, RP/RX series)
SGI designs the Nintendo 64
some SGI engineers leave and form 3dfx, the first widely adopted 3D graphics accelerator.
3DFX goes bankrupt and is bought by Nvidia.
SGI's Cray Engineers design new machines (Altix) using Itanium proc
Re: (Score:1)
Don't forget Convex!
Super Computer Record (Score:1)
HP Enterprise and HP are two different companies (Score:4, Informative)
In HHGTTG parlance, HP contained the hairdressers, TV producers, insurance salesmen, personnel officers, security guards, public relations executives, and management consultants.. HP Enterprise contained the high achievers, the scientists, thinkers, artists, and important leaders.
Re: HP Enterprise and HP are two different compani (Score:2)
No, the good part of HP split off to become Agilent.
The IT crap is still the IT crap. The only thing programmable from HP that is worth anything are the Corvalis Division calculators.
Re: (Score:2)
This is effectively a private Cray bailout, which is very much desired by the US government, particularly right now.
Cray subsists on very sporadic business from a relatively small number of customers. If the US government tightens its belt for any reason or a handful of procurement cycles line up the wrong way, Cray tumbles into the red. For 2017 through 2018, they've lost over $200 million against a revenue of about $800 million. We aren't talking about a small loss compared to gigantic revenue, we are
Re: (Score:2)
For supercomputers, the giant in the room is Amazon, though I suspect Azure is headed in the same direction.
AWS provides top secret certified, click-to-provision supercomputing. Full infiniband interconnect, high-speed Lustre cluster filesystems, hordes of memory, dedicated CPU, GPU, and other coprocessors.
All with no facility, no infrastructure, no cooling, no utility cost, no need to manage hardware failures, and more.
It's very competitive to use AWS to provision out a supercomputer, run your job, and the
This is HPE, not HP (Score:1)
This acquisition was done by HPE, NOT HP. Two different companies at this point.
HP does the PCs and printers.
HPE does servers, storage, networking, software and services around that.
Hmm (Score:1)
Nice to see one's name in print.
history has been not kind to HP acquisitions (Score:2)
period
I'd by Cray too just for the name... (Score:2)
I'd by Cray too just for the name... ...if I had the money. I guess that's why I don't have the money.
RIP Cray, again. (Score:2)
Welp, there's a death if I ever saw one. Cray was an amazing company that was on the cutting edge, died, came back, died again, but this time, HPE will likely kill them for good.
Both HP and IBM love to kill other companies. That's pretty much all they do now. /me pours one out for Notes/Informix/Compaq/Tivoli/Storwise/Trusteer/Cleversafe/Red Hat/Snapfish/Voodoo PC/3Com/Palm/SGI
I expect Dell to be joining their ranks eventually. Some how EMC and VMWare are still alive.
So sad news... (Score:2)
RIP Cray (Score:2)
RIP Cray, it was nice to know you. So sad you ended up in the clutches of zombie HP. Safe assumption that this constitutes the end of innovation at Cray and there will soon be a mass exodus of talent. Countdown begins now: 3, 2, 1...