My grandfather was born in the early 1900’s.  By all accounts he was a hardworking man with a strong degree of curiosity.  He passed away in his late 80’s and before he died I remember talking to him about my pursuit of an Electrical Engineering degree.  He nodded politely, asked a few questions and when I helped to fix the electrical outlet in his garage I got the sense that he thought I was heading down the path to be an electrician.  I believe that thought pleased him.  Several years ago I was explaining to my five year old daughter in layman’s terms what I did for a living and what my company made.  I said things like “We make tiny engines that run computers” or “I work with computers that run websites like Webkinz® and Disney®”.  She seemed impressed.  Months later when she was asked by a parent of her friend what her dad did for a living I was a combination of proud and surprised to hear that she replied “They make chips…”  (proud moment) “…and salsa!” (um OK.  I still have work to do).

_

Now the other day she walked up to me and said something like “Dad, I am having trouble getting the Slingbox to work on mom’s iPod Touch.  It is connected to the Internet but the remote does not seem to be changing the channel.  Can you help me?”  Clearly she has made some progress up the technology curve, but it also struck me how far she has come.  Kids these days are surrounded by technology.  In our house alone there are at least the following electronic devices; Oven, Microwave, AppleTV, refrigerator, smoke detector (3), carbon monoxide detector, programmable thermostat, furnace, radio, garage door opener (2), wireless speakers, televisions (3), set top boxes (3), ceiling fans with remotes (3), netbook, Slingbox, Clear wireless router, remote outlet, sprinkler control box, iPod Touch, desktop computer, Wii, iPod shuffle (2), alarm clocks (3), oven timer, electronic light dimmer, cordless phones (4), AV receiver, DVD players (3), VCR, iPod docking station, security system, motion sensor, camcorder, camera (2), USB hub, music keyboard, AV switch, computer keyboard, battery chargers (4), Wii remotes (4), Wii Fit Pad, Wii drums, copier/fax/scanner, computer monitor, AC, Power supplies (4), RFID credit cards (2), washer, dryer, noise canceling headphones, answering machine, internet modem, cell phones (2), handheld GPS, auto GPS and electronic battleship.

_

I am sure I have forgotten several things and I did not count cars or anything at my children’s school.  I am also sure each of the electronic devices in our house has either a processor, microcontroller, ASIC or multiple of each.  Admittedly, the silicon content in our house is probably above average given where I work and the personalities my wife and I have.  But when I think back to my grandfather he had none of these silicon laden items.  I am sure he didn’t care since it is hard to miss something you never knew.  Of the hundreds of pieces of silicon in our house about a dozen or so are smart enough to connect to each other or to “the cloud” in some way.  I put “the cloud” in quotes because it is not only the most over-hyped word of it’s time it is also the best way to articulate what I suspect my children and many others think of the services that they get when all of this stuff gets connected.

_

I can safely say two things are fact. First, my grandchildren will have in their house many more pieces of silicon than I do. Second, they will have more pieces of silicon that can connect to each other and communicate with “the cloud”.  There are many billions of devices connected to the Internet today and that number will grow.  At Intel we are building silicon, and increasingly software assets, that facilitate the processing and movement of data both on those devices and between them. Servers are increasingly becoming an important part of that over-hyped cloud word. My cable company has a cloud delivering me my on demand video content, A social media site allows me to upload pictures into their cloud to share with my friends, someone just used a cloud architecture to develop a perpetual motion machine.  OK, one of those things was false.

_

My grandfather thought a cloud was something in the sky.  My children think it streams video to their handheld device.  What will our great-grandchildren think?

At ISC09, the Top 500* results were announced: 399 out 400, nearly 80%, of the world’s supercomputers are using Intel processors.  The Top500 list is based upon one benchmark, Linpack.  While powering most of the world’s fastest computers is a great endorsement of the role Intel’s technology is playing to help solve the most complex high performance computing problems, no one buys a supercomputing machine just to run Linpack.  Linpack is a kernel that does not necessarily resemble any real application.  It’s just one evaluation vector among many. So, should you demand more?

Yes, look beyond the flops:  look at real application performance or benchmarks that might more closely resemble yours, look at the versatility, and look at ease of deployment of your solution.  

Today, Intel processors deliver more performance and throughput in less space and require less power than ever before.  The Intel® Xeon® 5500 Platform delivers up 3X performance over the previous generation Intel Xeon 5400 to decrease your time to discovery.   The Top500 list has 33 new entries based on the Xeon 5500, which launch only 3 months ago. Intel tools (compilers, libraries, and cluster kits) bring new levels of software versatility by enabling HPC users and ISVs write applications that extract peak performance and scale forward.  The Intel’s Cluster Ready program is easing cluster deployments, increasing reliability and lowering TCO by making it simpler to purchase, deploy and manage an HPC cluster.

So while providing flops is great, don’t forget (and demand) to look at real application performance, ask for software tools and technologies that maximize the value of your HPC system.

Jimmy

*Other names and brands may be claimed as the property of others

I haven’t been to see the new Terminator movie yet, but I certainly remember the Arnold movies of my youth, and the similar theme of machine vs. man in the geek lovin’ Matrix series.  Really, we all have a fear in a remote corner of our minds that someday the machines we all love will be smarter than us and somehow realize that we’re disposable.  Or useful as human batteries.  Which is why I love to work on the future of datacenter technology…after all, we’ll be the first to know the Top 500 list will mean something much more sinister.

Of course…I’m kidding.  The future of datacenters will bring great things to our planet from speeding the discovery in science, to making us much more efficient and lowering our collective carbon footprint.  And of course it’s datacenters that bring us Facebook, and who could really live without that?

The next transformation of the datacenter is almost within our grasp with the evolution of the enterprise cloud.  I wanted to shift focus from the nearer term technology innovation covered in our most recent podcasts to this broader technology movement, and to do so I recently chatted with three very smart people.  First, I talked with Dylan Larson about Intel’s view of the enterprise cloud and what technology trends he sees as critical to the creation of the architectural framework for the cloud.  I then spoke to Jim Greene about the future of security in the datacenter.  Finally, I visited with David Jenkins about our vision of instrumentation and why this technology is so important to the datacenter of the future. None of them mentioned anything about Christian Bale or Neo…but they did say a lot about where we’re going to create the next stage in datacenter computing. Take a listen at our Chip Chat Channel…

...and if you like what you hear subscribe to Chip Chat on Intel's RSS feed or on iTunes.

The debate on how to best increase system capacity to accommodate growing applications has raged on for years; “scale up” with more CPU, memory, and I/O, or “scale out” with loosely connected systems.    Scaling out by adding networked systems to increase capacity has been a good economical solution for many IT managers because it allows them to grow by using less expensive, industry standard building blocks.  However, there are some notable exceptions to this line of thought.  One is that the class of applications that require shared memory and large database support are much better suited to run on a single, expandable system that scales up.  These are typically transaction processing, business intelligence and ERP solutions.   Until now, IT managers running applications that require scale-up systems larger than 4 or 8 CPUs have had limited platform choices and most were proprietary and expensive RISC-based servers.

The other problem with the scale out approach is the people, facilities, software and overhead costs and complexity of managing very large numbers of servers, which can grow to a point where the costs outweigh the performance and system cost benefits.  The industry solution to achieving better ROI has been to consolidate multiple scale-out servers onto single industry standard scale-up servers with virtualization solutions.  This is a good solution, but is limited by the number of application loads the IT manager feels comfortable placing on a single server, given the need to maintain peak performance and availability for each application.

Well, it looks like the scale-up, scale-out debate is about to take another turn.  In the server product update Intel gave on May 26th, they talked about new levels of system scalability and choice supported by the upcoming Nehalem-EX processor.  This processor will support systems that scale up to 8 sockets natively (shared memory, without any additional silicon), and up to 16 sockets and higher with node controllers from system manufactures that allow single systems to share memory beyond 8 sockets.   So far there are over 15 different designs from 8 OEMs that offer 8 socket or higher scalability.  But of course, for the class of application where scaling is important, socket count doesn’t tell the whole story of what’s needed for scalable performance.  Thread support, key for transaction processing and virtualization, scales at the rate of 16 threads per socket with 8 cores and Hyper Threading (2 threads per core).  That would be 128 threads for an 8-socket system, and 256 threads for 16 sockets.   And in order to keep those threads fed with data close to the CPU, each processor supports up to 24 MB of shared cache (1.5X current generation Xeon), and an impressive 16 memory slots per socket or 128 DIMMs on an 8-socket system.  In addition, the Scalable Memory Interconnect gives these systems 9 times the memory bandwidth of today’s top Xeon processor.  Finally, four QuickPath interconnect links per socket allow for high-bandwidth sharing of data across the system.

So the net of it is that the industry is going to see a broad selection of highly scalable, next-generation servers that significantly extend the economic advantage of industry standard scale-up solutions for business-critical, large database, and high-end virtualization/consolidation deployments.     I would expect these systems to give IT managers a very cost-effective alternative to the much more expensive and proprietary RISC-based servers they use today.

What are your thoughts?  Mike

Related Topics:

• NHM-EX Press Fact Sheet
• NHM-EX May 26th Press Briefing Video – condensed version
• IBM 8Socket Demo Video
• NHM-EX--A New Standard

Looks like the Intel® Xeon® processor 5500 series is making lots of noise in HPC.  The QPI and integrated memory controller are really providing the boost necessary to make it an all around performance leader for HPC applications.  With all this performance why did Intel add a third memory channel?

The third memory channel enables the platform to support a boat load of memory.  Matter-of-fact, up to 192GB can be supported in a two socket configuration.  It wasn’t too long ago when only 32GB was supported in a dual socket configuration.  By having the ability to support so much memory you can now meet the needs of almost every HPC application.  The 5500 series is intended for all server markets, but let’s face it, with the design changes Intel made with the new architecture the server segment gaining the most benefit appears to be HPC. 

It seemed like yesterday when the only way to have access to large memory configurations was through expensive, proprietary SMP systems.  The HPC market for large SMP systems is still out there but it is shrinking…fast.  Today, we are clustering low cost solutions to create some of the most powerful systems in the world.  Standard components are leading to lower and lower system costs, delivering a price/performance advantage alternative solutions cannot meet.

Now that a single dual socket node can support up to 192GB’s it is important to understand how to get there.  First, to enable 192GB you need 16GB DIMMs x 12 memory slots.  There will be a premium for a 16GB DIMM.  Knowing the options and determining the best, most cost effective solution is going to be dependent upon your environment.  When a large memory node is required, do you purchase the 16GB DIMM’s or go up to a Multi-socket solution?  If I decide to scale back on the memory (use 4GB or 8GB DIMMs instead of 16GB DIMMs) what is the performance impact to my application?  If I am cost sensitive, will the lower cost outweigh the lack of performance?  Can I use SSD’s (Solid State Disk drives) to compensate for any performance loss due to lower memory capacity?  There are many questions to think about when deciding the right configuration for your application and environment and I certainly can’t answer them here.

Let’s not forget the third memory channel enables a different set of optimal memory configurations.  Think x3 when deciding on how much memory to install into your node; 12GB, 24GB, 48GB, etc.  What happens when you don’t use an optimal configuration?  Well it depends, in most cases the impact is minimal, but let me add a bit of context around minimal:

·         Low bandwidth sensitivity (more dependent upon the processor for performance)

–        E.g. Monte Carlo, Black-Scholes (financial modeling), BLAST (bioinformatics), AMBER (molecular dynamics)

–        Expect less than a 2% difference between memory configurations*

Ÿ  Medium bandwidth sensitivity (somewhat balanced between memory and CPU usage)

–        E.g. CFD, Explicit FEA, Implicit FEA (with robust I/O system)

–        Expect approx. 5% degradation for non-optimal symmetrical configurations*

Ÿ  High bandwidth sensitivity (high access to the system memory)

–        E.g. WRF (weather), POP (climate), MILC (physics), Reservoir Simulation

–        Expect approx. 10% degradation for non-optimal symmetrical configurations*

The results are interesting.  In all three cases above, the degraded performance is always better than the performance you would have with only two memory channels.

When you hear about performance impact of non-optimal memory you can see by the examples above, it is application dependent and will not have a severe impact on your overall system performance.   

The Intel Xeon processor 5500 series offers support for huge memory nodes with the addition of the third memory channel.  Memory configurations in multiples of three are ideal, but if you decide to stay with a power of two configuration the performance should still exceed that of a solution based upon only two memory channels.

*Based upon Intel internal measurements

Having bounced from Engineering to Sales to Marketing in my career I have found some unique interactions between those organizations along the way. But I have recently come across something for the first time that seems particularly noteworthy. I am finding that many of the internal discussions I am having about our upcoming products are largely void of the usual marketing fluff. You could argue that this blog and my previous statement is itself marketing, but oh well.  I am also not saying that I don’t still visit an end user who is having trouble picking out a server topology, an infrastructure to virtualize on or maybe they are having datacenter challenges or power constraints and we provide them with advanced product info.  All of that still happens regularly and I expect it will continue for a long time. Rather, I am referring to the solutions we are starting to propose for those problems.

I am sure everyone in marketing can remember some product that they were responsible for that kept them up nights. The feature set wasn’t quite right, the price was out of whack, competition was breathing down their necks or competition was the incumbent in a certain area. Those are tough days and you only hope that the future products in the hopper are leadership and there is balance to your present day effort. For a while I have seen segments where products are “unmarketable”. You can pretty much leave the marketing guys at the door when you walk in to a High Performance Computing account, Financial Services Account or Internet Portal Datacenter. They want hardware and you can take your PowerPoint slides and “shove them $#@^%.” That may be a direct quote J

Still, that was certain segments. They did their own benchmarking and they made their decisions based on the exact workloads and configurations they are running. Many Enterprises, Datacenters and Small/Medium Businesses rely on third party data, benchmarks or word of mouth to make their purchase decisions. We have been talking to them under non-disclosure lately about our next generation Nehalem based products and the responses have been rather unique. In short, Nehalem appears to be “unmarketable”. I find myself pretty much trying not to mess things up when talking about the product. There have been some early public discussions about the performance and the message boards seem to be taking a keen interest in how the platform looks. The launch will happen later in Q1 and I for one am looking forward to seeing what exciting new things companies are going to be doing with them.