Posted on 12/22/2003 12:15:15 AM PST by Swordmaker
Blacksburg, VA — Dr. Srinidhi Varadarajan knew that he wanted to build a world-class supercomputer. Also, he wanted to solidify the position of Virginia Tech’s world-class computer science program. But with only a fraction of most supercomputing budgets to spend, it seemed like a hopeless dream. He crunched numbers, solicited every likely vendor, examined and ultimately discarded all possible options using other platforms and chips. Then in June of 2003, news of Apple’s Power Mac G5 hit the airwaves. At last, Varadarajan realized, he’d have as much 64-bit processing as he needed to power his dream, without overtaxing his budget.
Soon after the announcement, Varadarajan took delivery of his very first PowerBook laptop running Mac OS X. Within days, he placed an order for the 1100 dual processor, 2.0 GHz Power Mac G5 computers that now drive Virginia Tech’s new supercomputer. Smart choice: In November of 2003 the giant system — named System X — became the third fastest supercomputer in the world.
System X is radically different from traditional, high-performance supercomputers. Unlike most, it is based on a “supercluster” of Power Mac G5 computers, each of which has 4GB of main memory, and 160GB of serial ATA storage. Not only is System X the world’s fastest, most powerful “home-built” supercomputer, it quite possibly has the cheapest price/performance of any supercomputer on the TOP500 list.
Select an area to explore:
". . . for Lockhart, the key to the group’s success is simple: “I’ve always said that people will have to pry my Mac away from my cold, dead hands,” he laughs. “But it’s true — Apple clearly has the right tools for the job. And I say that as someone who has very broad experience with every platform and operating system out there. This project never would have been possible at this price, while getting this performance, with any other [platform],” Lockhart finishes.
Jason Lockhart, director of the College of Engineering’s High-Performance Computing and Technology Innovation group at Virginia Tech
"“I’d never used a Mac before in my life,” admits Varadarajan. “I was a Linux/UNIX kind of guy, and used Windows to run the Office productivity suite. But I [bought] a 17-inch PowerBook laptop, and found that it was quite easy to do everything I’d been doing under the other platforms. I could compile the software application I’d been developing; it just worked under Mac OS X. Plus, I could use the productivity applications I needed, without having to reboot or launch an emulator. Within three days I’d made up my mind that Mac OS X and the new Power Mac G5 were the tools I wanted to use.”
Dr. Srinidhi Varadarajan, Assistant Professor of Computer Science at Virginia Polytechnical University, Director of Virginia Tech's Terascale Computing Facility
Record-Breaking Processing Speed
Lockhart says several features of the new Power Mac G5 computers were especially attractive for Virginia Tech. “AMD’s Opteron can only execute two double-precision floating-point instructions per clock cycle, which kind of limits system performance,” he notes. The G5 processor has two floating-point units. Thus, unlike the Opteron, it can perform fused multiply-adds, giving four operations per clock cycle and achieving a theoretical limit of 8 gigaflops from a 2 GHz processor. All this is in addition to the G5 processor’s Velocity Engine floating point and integer units.
Says Lockhart, “We would have had to buy more nodes to equal the amount of performance needed to hit our ten-teraflop goal. But the hypertransport memory architecture of the new Macs gives us lots of [speed]. And, having the PCI-X bus in the system gives us the bandwidth we need for our communications fabric.”
“AMD’s Opteron can only execute two double-precision floating-point instructions per clock cycle, which kind of limits system performance . . . The G5 processor has two floating-point units. Thus, unlike the Opteron, it can perform fused multiply-adds, giving four operations per clock cycle and achieving a theoretical limit of 8 gigaflops from a 2 GHz processor.”
Actually, this is FALSE. If you use the SSE2 instructions on the Opteron, you can perform 2 DP fused multiply-adds per clock, just like the G5. As long as your compilers properly supports SSE2, the G5 does not offer better per-clock performance than the Opteron.
not supports
The designers of the supercomputer stated "unlike the Opteron, it can perform fused multiply-adds, giving four operations per clock cycle and achieving a theoretical limit of 8 gigaflops from a 2 GHz processor."
That was a direct quote from someone who should know. Seems to me that the architects of the Supercomputer are a little more informed than you on the capabilities of the two processors.
The article says the G5's IBM 970 can perform FOUR per clock tick. That is twice the capacity of the Opteron.
For most purposes, including extreme purposes, computing speed is not the issue. Programming and programming languages are. Right now, if I wanted to, I could put together a system for three to four thousand dollars that would monitor the progress of a PETN explosion in real time down to the inch or less. PETN reacts at a rate of about 27,000 fps.
FWIW - Beta was (supposedly) better than VHS too, but I don't see too many beta machines around...
prisoner6
Just exactly how is a 66 MHz processor going to do that on a 33Mhz clock??
Gee, RLK, why don't you build your computer and show the world what you can do. Perhaps the world has just overlooked this extremely capable and cheap computer?
Where are the Beowulf configured 486's on the list of the world's fastest super computers??? Not even there.
This computer is capable of monitoring much faster and much more complex processes... say like nuclear explosions that propagate much faster than your PETN.
There are Beowuld architecture supercomputers on the Top500 list. But I garantee they are not 80486s! From the Beowulf site:
Top500.org's 8th Fastest Supercomputer in the world - 2nd Fastest Intel-based supercomputer ever built (2002). This is part of a 768-node dual 2.2 GHz Xeon Beowulf cluster which operates at 3.3 Teraflops.
The Macintosh Supercomputer produces 10.6 teraflops... for less money.
I wondered when the first ad hominem attack would take place... They seem to be required on any thread about Macintosh computers.
Thanks for getting it out of the way so quickly!
That's the problem with Mac folks...they're nice enough but too focused on their machines...sigh.
Kinda like the folks who drove Chevettes...Except for a few older folks, everybody else drove them at 10/tenths ALL THE TIME. They weren't sports cars or muscle cars but lots of Chevette owners thought they were and never tired of telling you so or trying - and failing - to prove it.
Ricers are in the same boat...so to speak.
prisoner6
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.