Nehalem-EX: Big Memory for Big Science

I was at SuperComputing’09 last week in Portland, Oregon. I talked with some brilliant people, and saw some fantastic stuff.

It was good timing on my part because last week Intel also announced that it would offer a 6-core, frequency-optimized version of its Nehalem-EX product due out next year. This part is intended for use in tackling some of the types of high performance computing (HPC) workloads prominently displayed at SC’09.

Most people know that the majority of HPC workloads today are based on clusters of relatively small-memory, 2-socket systems. That is because most HPC workloads may be broken into smaller, discrete units of work that can be efficiently processed using such clusters. For these workloads the primary hardware capability selection criterion is typically a balance of both memory bandwidth and compute FLOPs (floating point operations per second).

But there are other types of HPC workloads. Specifically, those that deal with very large datasets (some as large as a terabyte) or those that have to deal with non-sequential memory access. This means the workloads simply aren’t easily divisible--or it is inefficient to do so-- into the relatively small memory footprints used in traditional clustered 2-socket HPC solutions. Examples of these types of bigger memory applications can be found in a variety of fields such as weather prediction, manufacturing structure analysis, and financial services.

The high-speed processing requirements and size of these workloads put a greater premium on system memory capacity/bandwidth than on compute FLOPs.

If the larger dataset won’t fit into available memory, and dividing up the dataset to spread across multiple nodes cannot easily be done, then data has to be moved in and out memory to hard disk.  But using hard disk drives (which are many times slower than RAM memory) can drastically impair performance.

There are now two better alternatives to the use of hard drivers. One is SSDs and the other is having a larger memory footprint. Solid State Drives have fairly high data density vs RAM, but much faster access than hard-disk drives--albeit still markedly slower than RAM. Another solution is to simply have more capacity of the faster RAM. This last one is what the Nehalem-EX HPC part is aimed at.

Nehalem-EX is the Expandable Class of Nehalem. The Expandable Class brings all the goodness of the Nehalem architecture (Xeon 5500 product line) to the HPC market, but in the form of a “super node” that has greater: a) core/thread count, b) socket scaling (up to 256), c) I/O and memory capacity (up to 1 terabyte in a 4 socket system) and bandwidth at capacity, d) reliability features, e) and other features.

The 6-core frequency-optimized Nehalem-EX part has also been tuned to offer the highest core frequency possible for this chip.   In creating this part, Intel is meeting the needs of the HPC community that want higher scalar performance along with the benefit of large memory capacity and bandwidth per core.

Of course the 8-core version of NHM-EX is still an option for those HPC workloads that scale well with more cores while still looking for the high memory capacity of the expandable class.

By having both 8-core and frequency optimized 6-core versions of the NHM-EX class of processors means HPC researchers have greater choice in selecting the processor best suited for their specific workloads.

After talking with some of the researchers at SC’09 last week I’m really excited to see how the Nehalem-EX “super node” will deliver the necessary compute and memory capabilities to help those researchers solve some of their biggest challenges.

Three short years ago, this would have taken 32 Xeon 5100 (Woodcrest) servers, accounting for 64U of rack space... this pic is from the upcoming Xeon MP (Beckton) platform with Nehalem-EX processors that many of you have seen at IDF 2009.  This server only takes 3U of rack space... less than 5% of the space of what it could replace.

Sometimes you see a screenshot and it just makes your jaw drop...

Just to give a comparison of CPU density... here's a diagram showing the comparison of 3 year old technology compared to the upcoming Nehalem-EX.  If each of those 32 old servers burns 400W of power - that's 12.8 kilowatts - compared to one server, burning less than 1kW.

What's even more amazing, is that some design wins are based on a 1U server with the same cpu footprint - that's AWESOME!

What are your thoughts on these upcoming multi-core technology improvements?

Sun has recently published a whitepaper that discusses how the Solaris OS will take advantage of the next generation Intel Xeon processor (codename Nehalem-EX) for expandable servers (4 sockets & greater).  Sun with over 20 years of experience in larger socket, core & threading capabilities is working to have the Solaris OS be ready to take advantage of the features & new capabilities of “Nehalem-EX”.  The three areas of collaboration for Solaris & Nehalem-EX are around  scalable performance; advanced reliability and energy efficiency between the specific features in Solaris and the next generation Intel Xeon processor.  Read this recently published whitepaper

I finally finshed the content for the talk that I am scheduled to deliver next week at IDF on Sept 22 (TCIS001 10:15 am – Room 2004).  The content covers examples of optimizing for multi-core using our software tools to accelerate performance,  and more importantly the seamless use of the same software base with minimal or no changes in next-generation architectures (what we call scaling performance forward). Personally, I am excited about the potential of multi-core optimizations with today’s architectures. When I was a graduate student in parallel computing from 1988-1994, it was extremely difficult to take any algorithm and map it to the parallel architectures since most of the algorithms were not very efficient once you took the communication delay’s into account.  The key is to get total delivered performance at an application level, not at the kernel level. However, given the architectures of today which are better balanced, and the availability of multi-core, the memory bandwidth, software tools that work, and the faster interconnects, the number of algorithms that can be parallelized and that actually benefit with accelerated performance (Delivered total application time)  is huge, pretty much every industry vertical is taking advantage of multi-core architectures, software tools, and clusters.

John Gustafson, from our Intel Labs, an industry HPC veteran, is my co-author, and I am thrilled to have him speak about Balanced Computing. Wes Shimanek, a colleague of mine at Intel introduced me to John and after listening to his explanation of balanced computing, and his views on what works and what doesn’t, we immediately  knew that John’s expertise will be greatly valued by the IDF audience, and invited John to be part of the talk. John graciously accepted to participate, and I hope that folks interested in computing architectures, especially in the HPC world will make the time to come listen to John’s talk.

I will also be giving you a high level view on the challenges that drive our products and briefly introduce you to the various aspects of our strategy. I will be followed by 3 more talks that will cover the key aspects of what we do at Intel in HPC, Software Tools for Scaling Application Performance Forward (TCIS002), Delivering more to HPC than just Performance (TCIS003), and Intel® Cluster Ready (TCIS009).

I am looking forward to IDF next week. See you all at the developer forum

Nash Palaniswamy (Intel)

I wrote a while back about how the Xeon 7400(Dunnington) processor series compared to RISC. Since then I have shared information through other blog posts and sharing content about how Xeon 7400 and Xeon 5500 will compare to both SPARC and POWER.

Xeon 7400 and Xeon 5500 are the current products shipping into the marketplace today. I.M.H.O they offer a pretty compelling alternative from both a performance and TCO perspective Vs SPARC and POWER. But I will not try and repeat all the reasons here

What I wanted to share with you was some thoughts about what the next product to succeed Xeon 7400 will bring to the RISC party. Nehalem-EX is the code-name for our next generation of product designed to serve workloads currently serviced by Xeon 7400 today (i.e. Database, ERP,  BI etc). EX btw is what we all would traditionally call MP or multi processor servers

Don't stop reading now, here is why I'm EXCITED about what Nehalem-EX will bring to the RISC party.

My excitement is actually based on real customer discussions about what Nehalem-EX will do for them and why it delivers some new stuff (my code for features and benefits) which they see as a pre-requisite to make the move from RISC to Xeon. For some customers the TCO and performance of  products have been enough to convince them to move. For some other customers there are still some checkboxes remaining which I believe Nehalem-EX will address

Here is a snapshot of some of the cool new stuff which is actually convincing customers (from some real deals that I have worked)

1. Improved bandwidth. Up to 9 times memory bandwidth of previous generations
2. Introduction of Quickpath Interconnects to the EX systems
3. Add new RAS features previously seen on Itanium products to Xeon products
4. Significant improvement in performance vs previous generations e.g. Database 2.5xe
5. More scalable platforms through 8 OEMs offering >8S. These platforms are key to manage large databases and for large scale consolidation
6. Mainframe class availability in scalable platforms

For more information check out the press briefing from May. See more the details in the presentation

Nehalem-EX goes into production later this year and I am pretty excited about how it will change the game. What do you think?

A MONSTER CHIP IS COMING. The next generation of MP processor is targeted for production later this year, and by all accounts it is going to be a monster. Nehalem-EX is part of the Nehalem family of processors, but compared to its siblings it has the highest cores/threads count, largest shared cache, highest CPU-to-CPU bandwidth, highest I/O bandwidth, highest memory capacity, highest memory bandwidth, greatest scalability, and highest level of Reliability/Availability/Serviceability. It’s expected to bring a gargantuan, unprecedented leap in capabilities and performance--the biggest leap in all of Xeon product history.

IT’S TARGETED AT “BIG BOXES”. Big box servers are multiprocessor systems using the most capable processors and platform components. These systems are targeted at applications and usages that require the largest memory footprints, the highest amounts of single-box processing power (for workloads that don’t decompose well into lots of independent threads) and/or advanced levels of RAS. Such systems are typically the best choice for large databases, ERP apps, Business Intelligence apps, large-scale server consolidation and business-critical virtualization, mission critical applications and large scale high performance computing.

IT USES THE SAME PROCESSING TECHNOLOGY AS THE SUCCESSFUL XEON 5500, BUT MORE OF IT. Just like with Xeon 5500, the Nehalem micro-architecture brings improved single-threaded performance via IPC (Instructions per Clock) enhancements and Intel’s Hi-k 45nm manufacturing process. Greater multi-threaded performance comes via Hyper-Threading and more cores. But while the Xeon 5500 has up to 4 cores/16threads per socket, the Nehalem-EX monster doubles that to 8 cores/16 threads.

HAS A BEEFIER MEMORY AND INTERCHIP COMMUNICATION SUBSYSTEMS. Monster thread processing capabilities require monster size feeding to bring out the best performance. Nehalem-EX’s raw processing potential is made viable by a heavy duty memory subsystem and inter-chip communication system.

Nehalem-EX has 24MB of shared level 3 cache--that’s 50% more than the current Xeon 7400 and 200% more than Xeon 5500. The memory channel bandwidth was increased to 9-times that of Xeon 7400. And it’s all attached to up to 16 DIMM slots per socket (that’s 64DIMMs slots for 4 sockets)—double the current generation of Xeon 7400.

In a multi-socket system, processors need to communicate with each other in order to most efficiently coordinate their shared workload. They also need lots of I/O bandwidth. Nehalem-EX has four QuickPath Interconnects on every socket--double that of Xeon 5500. The four QPI links enable Nehalem-EX processors to be directly connected to each other in a 4 socket system. This offers significant performance advantage over a so-called ring architecture wherein some processor-to-processor communication must go through an intermediary processor. The extra QPIs also mean that there’s plenty of CPU to I/O bandwidth.

EXPECTED TO BRING THE GREATEST LEAP FORWARD IN XEON PERFORMANCE EVER. On key server performance benchmarks (e.g. SPEC_int_rate, SPEC_floating point_rate, TPC-C, etc) Xeon 5500 using Nehalem technology brought gains of over 100-200% greater than prior generation. Generational gains of this magnitude come along just about once a decade. Nehalem-EX’s generation-to-generation performance gains are expected to be substantially higher than those of Xeon 5500. We’ve already seen measured memory bandwidth of 9X vs. prior generation. That’s an early indication of the level by which new performance records will be set when this monster chip comes to market.

Related Topics:

NHM-EX Press Fact Sheet

NHM-EX May 26^th Press Briefing Video – condensed version

IBM 8Socket Demo Video

NHM-EX--A New Standard

Evolution happens.  In matters of nature, Charles Darwin is clearly the dominant expert and the Theory of Evolution has stood the test of time and much scrutiny by scientific experts for decades.  However, the pace of innovation and change that happens in nature is insufficient as a means to evolve technology.

Darwin's Theory of Evolution is a slow gradual process. Darwin wrote, "…Natural selection acts only by taking advantage of slight successive variations; she can never take a great and sudden leap, but must advance by short and sure, though slow steps”

Personally, I’m very happy that the Theory of Evolution does not govern the pace of innovation for computing.  In matters of technology evolution, Gordon Moore is the expert.  In 1965, Dr Moore observed a trend in silicon manufacturing that has subsequently driven the pace of innovation, revolutionalized an industry and quite possibly society.  Moore’s Law states that the number of transistors on a chip will double about every two years.  Read more about Moore’s Law

In March 2009, Intel introduced the Nehalem microarchitecture where the innovative use of these transistors is truly phenomenal.  Besides offering nearly double the performance for many 2 socket server applications since last year, these processors offer a FIVE - fold improvement in energy efficiency versus the first generation of intel xeon quad-core processors introduced only two years ago.

ð       5x lower idle processor power (10W vs 50W)

ð       5x the number of power states between full power and idle power operation

ð       5x faster transitions (lower latency) between these power states

Today, Intel’s Boyd Davis discussed the innovations coming soon for the expandable 4-socket and larger servers, codenamed Nehalem-EX processor.  For this segment of the server market the bandwidth gains are expected to be staggering - offering up to 9x the memory in bandwidth over the highly scalable six-core xeon 7400 based servers available today.  Learn more about Nehalem EX

The question for IT managers and business leaders is how fast are you evolving your compute infrastructure capability.  Older single-core processor infrastructure is consuming valuable resources (space, power/cooling, maintenance) while often running underutilized and consuming full power.  Replacing that infrastructure can deliver dramatic benefits in performance and operational efficiency.  The savings from server replacement can provide a rapid payback on investment – stimulating re-investment or improved business results.

The faster you move … the more competitive you can become.  Just like in nature, business is survival of the fittest.  How fit is your IT infrastructure? 

Take advantage of Moore’s Law – Evolve Faster.