Slashdot Mirror
Windows Breaks Into Supercomputer Top 10
yanx0016 writes "Wow, that's some news this week at SuperComputing 08. Apparently Microsoft Windows HPC Server 2008, with a Chinese hardware OEM (Dawning), made #10 on the Top500 list, edging out #11 by only 600 Gflops. Folks were shocked to see Microsoft getting so serious around HPC; I think we are only beginning to see a glimpse of Microsoft in the HPC field."
That is, programmers who are familiar with Windows more than other systems.
And Microsoft is also looking to roll out a new language that is supposed to make parallel programming much easier for those programmers.
If it works, there would be a LOT more apps that take advantage of these systems.
"With the release of HPC Server 2008 a few weeks ago, Microsoft also offered an academic version priced at $15 per node to generate interest. By comparison, a commercial license runs $450 per node"
Ok, just because I'm strange I had to go and figure it out.
A C64, according to this guy runs at about 320 flops.
So, it would take that C64 600*10^9 / 320 = 1,875,000,000 seconds. That's 59.46 years.
Wiki says there were 30 million C64 units ever made.
So that would be 1,875,000,000 seconds / 30,000,000 = 62.5 seconds.
It would take every single C64 ever made about a minute to make up the difference.
Wow.
Crap I'm old. =)
Weaselmancer
rediculous.
"Erris" and "right handed" (who replied to you) are just two of twitter's 14 sockpuppet accounts.
See this thread for a recent fun shill session.
Web2.0: I love when people Flickr my cuil and digg my boingboing until my google is reddit and I start to yahoo
What's missing in the article is that there are only a few windows-based systems in the top500 and there numbers have been declining over the years.
New things are always on the horizon
Keep in mind: you're talking about a processor which doesn't even have integer multiplication, let alone any floating point operations. And you have only three 8-bit registers to perform these operations in. Executing 1k instructions for even a basic FLOP is not inconceivable here.
FLOPS = FLoating Point Operations Per Second
The C64 has no floating point unit in hardware. All floating point math had to be done through software emulation. It's a lot of extra work.
The C64 may have "fast" bytewise integer math in comparison, but there's a lot of overhead when you try to do floating point in an integer system.
What is the business case for this for Microsoft? Anyone know? This seems like an area where an executive decided it was simply unacceptable for Windows not to play in this space.
That's exactly right. Bragging rights. They get to say their platform can scale, and that 1% of the gee-whiz propeller heads that really know their stuff enough to build the world's most powerful supercomputers recognize the advantages of their platform.
Given the targeted nature of their involvement, a critical eye might look to the methods used to influence those propeller heads. In HPC as in national government the motivation is not perfectly on Total Cost of Ownership and value for price, and even when it truly is there are a few places on Earth where it's permissible to skew that metric a bit by pricing your per-node price seriously into the negative numbers. China might even be such a place. Which would open questions about Microsoft subsidizing high technology in China. Which might invite other questions about what China is actually doing with a computer powerful enough to more perfectly model nuclear explosions. It might be politically sensitive to deny China such a device, but paying for it? That would be... unsavory.
And 1% bragging rights is better than none, right? I've been one of the guilty ones who've pointed out over and again that Windows isn't ready for heavy lifting, as evidenced by its absence from this list.
The amazing story here is that in June 1999 Unix owned 498 of the top500 and Linux is well on its way to hit near that mark in only 10 years. Linux added 12 systems to hit 439 and eating share from every other category. If it gains another 12 in June that's 452. There's a good chance it will do better, but it's not likely to hit 498 next year, or even the year after.
Help stamp out iliturcy.
For the same reason that people knock Linux on the desktop -- software, software, and software. The codes that exist for HPC have been developed over years and guess what they target as a platform? UNIX. Do they have a scheduler/queue system? Is it torque/moab? How about a parallel debugger like Totalview? Are the install and cluster control (startup shutdown etc) tools functional and mature? How's hardware fault debugging under windows when headless?
The real question is what % of cycles do they deliver on it, and what's their job payload look like. VT's Big Mac made a huge splash a few years back, but talking to the admin, they broke it up into 64-core chunks and gave each research group a chunk. Most sat idle of course. Also that meant they didn't have to deal with a scheduler/etc.
The HPC machine I run never got that high (mid 60s) on the top500 list -- but we do deliver around 85% of the theoretical possible cycles (penalizing us for maintenance windows, dead/crashed nodes, unscheduled nodes, our 10% of the cluster debug queue) over our four year lifespan.
Slashdot Patriotism: We Support our Dupes!