2009/07/14
TheRegister: Nehalems make like elephants on HPC memory test
by Timothy Prickett Morgan
Intel's Nehalem EP chip has significantly out-peformed AMD's Istanbul on a set a memory-intensive benchmark tests.
The techies at supercomputer cluster maker Advanced Clustering Technologies are at it again, running their own benchmarks on single server nodes using popular high-performance computing tests normally used on entire clusters. This time around, ACT is putting the latest x64 chips into two-socket systems and running the Stream memory benchmark on the boxes.
By running various HPC tests on single servers, ACT is helping educate customers on the pros and cons of the new Intel quad-core 'Nehalem EP' Xeon 5500 and Advanced Micro Devices six-core 'Istanbul' Opteron 2400 processors.
A few weeks ago (http://www.theregister.co.uk/2009/06/25/act_super_duke_out/), ACT cluster engineer Shane Corder published a report (http://www.advancedclustering.com/company-blog/high-performance-linpack-on-xeon-5500-v-opteron-2400.html) after he slapped the Linpack Fortran benchmark test on two-socket servers using these new chips.
On that test, one of ACT's Pinnacle rack servers equipped with two quad-core 2.66 GHz Xeon X5550s with 12 GB of DDR3 main memory running at 1.33 GHz was able to deliver 74.03 gigaflops of sustained performance against a peak theoretical performance of 85.12 gigaflops. But a Pinnacle machine configured with two of the six-core Opteron 2435 processors running at 2.6 GHz and 16 GB of DDR2 main memory running at 800 MHz was able to deliver 99.38 gigaflops (against a peak theoretical performance of 124.8 gigaflops).
So, AMD won that one - especially when you consider that the Opteron-based Pinnacle HPC node from ACT cost $3,500 compared to the $3,800 price on the Xeon-based Pinnacle box.
Now, with the Stream benchmark, the test is not about flops so much as memory bandwidth, and given the higher clock speed of the DDR3 main memory compared to DDR2 memory, you'd expect the Nehalem EP server node to do better than it did on the Linpack test. And indeed it did.
Corder's home-done Stream benchmark tests (http://www.advancedclustering.com/company-blog/stream-benchmarking.html) were done on exactly the same iron as the Linpack tests, and for good measure, Corder tossed in some numbers for older quad-core Xeons and Opterons to show how much better the new chips are versus the old.
The Nehalem EPs really cleaned the Istanbul's clocks on this test. Using 1.33 GHz DDR3 memory, the server using the X5570 processors was able to 37,122 MB/sec of bandwidth on the Stream test, while the machine equipped with 1.07 GHz memory modules hit 32,770 MB/sec and one using 800 MHz memory could handle 25,490 MB/sec. A Pinnacle server equipped with the earlier "Harpertown" Xeon 5400s - quad-core chips using the old frontside bus architecture and 800 MHz DDR2 main memory - could only deliver 9,776 MB/sec of bandwidth on the Stream test, and dropping down to 667 MHz memory pushed performance down to 6,102 MB/sec.
By contrast - and this is a big contrast - the Istanbul-based Pinnacle server using 800 MHz DDR2 main memory - as fast as it gets - topped out at 20,534 MB/sec of memory bandwidth on the stream tests, which was actually a little bit lower than the results ACT saw with a Pinnacle server equipped with quad-core "Shanghai" Opterons, which came in at 20,687 MB/sec. A server using the older quad-core "Barcelona" Opterons and 667 MHz DDR2 main memory was able to deliver 16,965 MB/sec on Stream.
As Intel has promised, ACT confirms that the Nehalem EP chips and their new QuickPath Interconnect bus architecture delivers nearly four times the memory bandwidth as its Harpertown predecessors, and nearly double the memory performance of the current crop of AMD Opterons. And there is nothing AMD can do about it until it switches to DDR3 main memory early next year (http://www.theregister.co.uk/2009/04/22/amd_istanbul_forward/) with the "Magny-Cours" and "Lisbon" kickers to the Istanbuls.
AMD will be offering the G34 chipset with four DDR3 memory channels per socket (up to twelve DIMMs) and the C32 chipset with two channels per socket (up to four DIMMs). AMD's plan is to offer two different kinds of two-socket servers: one where memory bandwidth is key (that's the G34) and one where cheaper price and floating point or integer power are more important (that's the C32). AMD has the right idea. But it really needs this architecture to be here now to blunt Intel's considerable memory bandwidth advantage. ®
InfoWorld: Because Google says so
by Eric Knorr
I'm not the first to say this, but the idea behind Google's forthcoming Chrome OS reminds me of the Network Computer (NC), a driveless desktop unveiled by Larry Ellison in 1996. Back then, here's what I wrote about NC: "Do you really want to do without a floppy, hard, or CD-ROM drive? Be unable to compute – or even access your data – when the server goes down? Watch performance slow to a crawl during peak hours? An Internet appliance has everyman appeal at first glance; but on closer inspection, it's two steps back to those bad old mainframe days when Big Brother owned the computer, not you."
Now, 13 years later, Google has raised a similar proposition: an OS that pretty much dictates that you'll be living your computing life on the Internet and storing your data and preferences there, too. So let's break down that hoary old critique of mine and see if it still applies.
First of all, when I knocked the NC for lacking local storage, I was referring mainly to performance. At the time, 28.8bps modems were typical and putting personal storage at the end of such a slender connection seemed like a really bad idea. Now, some Chrome OS computers will have solid-state drives or hard disks, and some may only have a cache (who knows?), but it doesn't matter much. You'll be computing in the cloud. Broadband plus a fast JavaScript engine equals good enough performance, so score one for Google.
Now we come to the part about being unable to compute or access your data when the "server" goes down. (It could be the "server" or it could be the connection, but whatever.) Well, I imagine some implementation of Google Gears will be included, so you'll have some limited offline access to data. But more to the point, I can't remember the last time my work or home broadband connections went down and Google doesn't have outages. So Google gets another two points.
The reference to "peak hours" is a legacy of the days when mainframes or "online services" would choke on too many simultaneous users. But I believe in the magic of Google's hyperscalable server cloud, so I have to give 'em another one.
Which leads us to the final "Big Brother," point. Google is already the gatekeeper of the Internet; should it also be the keeper of your data? It does seem to be time to trot out the old cliche about absolute power corrupting absolutely. At the very least, I can't imagine enterprises ceding their data to Google (the SLAs on Google Apps, for example, aren't exactly business class).
But a near-zero-config thin client, with all my data and preference available from any Chrome OS device, is an awfully appealing idea. And if I had my choice, who would I want to play host? Oracle? Microsoft? IBM?
Well, I'm not ready to hand over the entire casket of family jewels to Google, either. But the technical hurdles to a modern-day NC have largely been vaulted, and HTML 5 and CSS 3 should enable desktop-class apps in the browser. The fact that I'm even considering what data I might or might not "give up" – and that a simple announcement implying no new groundbreaking technology has caused such an avalanche of speculation – is testament to the power of the Google brand. Should it be any more powerful than it is?
CNet: Smile at work-or the happiness detector will ding you
by Kelvin Low
Remember the gender recognition system we saw at the Singapore-based CommunicAsia trade show last month? Well, those zany Japanese have a more creative way of implementing a somewhat similar face recognition software.
Putting a new spin on the phrase "service with a smile," employees of Keihin Electric Express Railway will need to check their smiles in every morning. The software will determine the quality of their smile, and display visual alerts if they don't look happy enough.
According to an article in the Mainichi Daily News, the software assigns smile values to various parts of the face. It then adds those values up and determines a score.
The device recognizes eye movements, lip curves, and wrinkles. If an employee gets a low score, messages such as "You still look too serious" or "Lift up your mouth corners" will be displayed on the screen.
Maybe it seems cruel, but to us it's still a lot more humane than not having chairs at work. Or maybe Keihin Electric Express Railway was inspired by the ancient Chinese proverb "Hide your dagger behind a smile"? We wouldn't want to provoke them to find out.
