AMD is holding a contest, looking for the best answer to “What Would You Do With 48 Cores?” I must admit it: I prefer AMD CPUs. I have been an AMD customer for years, starting with a 486 overdrive CPU based on an AMD DX5-133 CPU. Friends could not believe I had a 486 and not a Pentium 75! For the last 10 years my home has been Intel free except for an old laptop. My children all have AMD based PCs, and whenever family or friends ask me what type of home PC to buy I make sure to add “…and look for an AMD CPU because you will get more performance for your money. “ And I have been promoting Opteron CPUs for servers and workstations since that line came out. I will be very proud to win this contest, and I can assure you that, if I do, you will get to read all about it right here, in full detail. Now, here is what I will do with 48 cores:
Short answer – I will improve the free and open source software (FOSS) community by using the 48 core machine to help test FOSS projects and advance FOSS speech compression.
Full answer – First, I will provide my time and machine time to test free and open source code projects to see how they scale and insure they operate properly. Many-core systems are the way of the future, and I can think of no better way than to stress current software packages with this ultimate AMD 48-core machine to insure they are ready as many-core systems replace our current multi-core systems. For example, I will test operating systems and multi-threaded applications for scalability and race/deadlock conditions. I will do this by soliciting requests from the maintainers of OS projects for disc images preloaded with test suites ready to boot and test. I will also solicit applications developers, and provide various OS images ready to load their applications for testing. Depending on their needs, I will provide remote shell or desktop access so they can directly interact and monitor the machine and applications.
Second, I will use this machine to develop new, and refine current, high quality FOSS speech compression algorithms. Speech compression is one of the last desert wastelands of FOSS, with few projects totally free of the threat of patents. There are a number of techniques which are unencumbered, such as wavelets, linear predictive coding (LPC), and code-book excited linear prediction (CELP). However, it takes a lot of computing time to run each new algorithm through hours of speech samples – recorded in multiple languages, from men/women/children, and with different amounts and types of background noises – in order to compare the speech quality and performance. All my advances in speech compression will be provided as FOSS.
In conclusion, I have the right mix of varied experience, writing skills, and vision to fully utilize 48 cores and to promote FOSS and the AMD brand.