Couldn’t resist the headline, but Joe is on quite a roll.
One of the first criteria to catch my attention is to link to me. Even better if you praise this blog (just my attempt at Kawasaki-esque humor). But seriously, Joe had a series of posts today on what he is calling “accelerated computing” (AC). We tend to agree on things from time to time and today was one of those occasions. In Drinking the koolaid, by the megaliter, Joe introduces the AC term and talks about the need for AC and the situations that drive usage and adoption. He also uses the term personal supercomputer. Now that is something I have used before and frankly, I like accelerated computing better. Sounds less grandiose for starters. Today, there are a number of choices (that too many people don’t choose to use them is another matter). One can choose between “supercomputers” like IBM’s Blue Gene, or clusters, or blades, some kind of grid or some form of accelerator system, like an FPGA, and I am sure that I have missed something. Each has its purpose and is useful in a different scenario. Identifying the need and the appropriate solutions is something that too many IT types and scientists ignore. Biases do play a part (I have more than my share), but in the end the good IT manager or computational scientist is one who can overcome those.
Anyway, to get back to the topic at hand, one thing that should change scientific computing is what Joe calls AC. When it becomes possible to get TFlops performance out of something that fits under my desk or in a little corner, does not need special facilities, etc, the productivity of the scientific community, from those who crunch numbers for a living to those almost unaware of the resources that help them run Blast again and again and again, will change dramatically. The question remains, why isn’t adoption higher? Economies of scale should be driving the cost of scientific computing down, but it seems to me that other than the cluster (a personal favorite), this is not happening. One should not need Koolaid to make the value of accelerated computing obvious. In the past, I have argued that computers are not fast enough. I stand by that argument, but on the flip side, I don’t think people are leveraging the resources available to them either.
Before I ramble too much, I strongly encourage you to read another post from Joe. I won’t try and dissect his logic, since it is so self-explanatory. It did bring back memories of a conversation I had while at an interview several years ago of building a cluster of PlayStations (the original one).
Technorati Tags: Scientific Computing, High Perfomance Computing, Accelerated Computing, Personal Supercomputing, FPGA, Cluster Computing, Scalability.org, Supercomputing, Grid Computing
powered by performancing firefox
4 Comments
Anyone who has tried to enocode home video on his PC knows that computers are quite far from being fast enough. It is hoever true that fast enough is relative..and I could solve a full protein structure on my 3 year old Athlon..but it crawls when it comes to encoding video..also who was it who said , software becomes slower faster than computers become fast.
Anyone who has tried to enocode home video on his PC knows that computers are quite far from being fast enough. It is hoever true that fast enough is relative..and I could solve a full protein structure on my 3 year old Athlon..but it crawls when it comes to encoding video..also who was it who said , software becomes slower faster than computers become fast.
Not just video. My 3yr old Athlon used to scream for recording music. Now as computers have become faster, algorithms have become more complex and some can easily bring it to its knees. There are actually DSP cards for effects, but they have some other issues.
Not just video. My 3yr old Athlon used to scream for recording music. Now as computers have become faster, algorithms have become more complex and some can easily bring it to its knees. There are actually DSP cards for effects, but they have some other issues.
4 Trackbacks
[...] By now I must sound like one of those old latin jazz records my father used to listen to. We have talked about accelerated computing before in these pages and I am still not sure what to make of news items like this one. How many FPGA applications of BLAST can the community support? I wish there was some consolidation here and a few clear cut winners. Otherwise all that I see is a wasteland of companies with some smart people and good ideas going under. Perhaps I am wrong, since I have never really done a complete business analysis of the field, but unless I had some technology that was a leap ahead of the rest of the community, I would NOT start a hardware acceleration company. The barrier to entry is too low, and the adoption rate to sustain the number of companies on the market is just not there. I would love to be proven wrong, cause the techie in me loves accelerators. [...]
[...] I am not going to be at SC ‘06 (How I’ve never made it to one of those conferences remains a mystery and a travesty). Wired is carrying an article about the next revolution in supercomputing, specifically using GPUs as accelerators, a topic that has been covered here and on other blogs before. Both ATI and Nvidia are jumping into the fray big time as the article suggests. One of the best parts of the article was that Folding@home was mentioned (its not all about gaming). The article suggests that, in general, the speed gains made by GPUs are far greater than those made by CPUs, however, the data from Folding@home and UNC suggests significant variance and algorithm dependence. In other words, certain applications will benefit greatly from GPUs (probably those with lots of floating point calculations, while others will not. I think given the push from ATI and NVidia, we will have an excellent idea of the applications where GPUs will provide the most benefit fairly soon. [...]
[...] Further reading: BLASTing FPGAs “Accelerated Computing†– The floodgates [...]
[...] Further reading: “Accelerated Computing” – The floodgates Blasting FPGAs [...]