Linked by Thom Holwerda on Mon 16th Apr 2007 15:47 UTC, submitted by george
Permalink for comment 231366
To read all comments associated with this story, please click here.
To read all comments associated with this story, please click here.




Member since:
2005-07-08
My only comment on the dynamic core-to-core power shifting is how they worked around the issue of power density. It's been shown that the IHS doesn't really spread the heat effectively at all, it primarily distributes the pressure of the heatsink. Doubling the power dissipation on one side of the die would double the maximum power density and impact the leakage current. I bet that these processors will have lower performance per watt than its competition when running single-threaded workloads.
I question the theory behind flash caches. The storage hierarchy works because the various levels are each over an order of magnitude faster/slower than their neighbors in the worst case. Flash is an odd medium for caching hard disks because while it's about an order of magnitude faster than the disk for random access, it's comparable or even slower than today's commodity SATA drives in sequential access. Managing a flash cache is tricky because you not only need to worry about evicting least-recently-used filesystem pages, but you also want to bypass the cache during sequential I/O and evict pages that are usually accessed sequentially. I wonder if they even bother implementing these considerations.
Now that the world is being thrust toward Vista, all three major platforms effectively use main memory as a disk cache, and DRAM wipes the floor with any mass storage medium, including an enterprise-class SAN. It's about 20x as expensive as flash, but typical working sets for desktop applications are likely to fit in the memory left over for the disk cache. There's no substitute for plenty of DRAM. At least not until MRAM or PRAM hits the market. Flash is a great removable storage medium, but it's being shoe-horned into a new role as a cache, and it just isn't a good fit. Wear-leveling will make it feasible, but nothing will make it a strong performer.
As for the tablet device, this doesn't surprise me at all. These days, you should expect almost any new computing device (not marketed by Apple or Microsoft) to support Linux. How do you think they test the hardware during development? Linux is the bring-up platform of choice for device manufacturers. All of IBM's new hardware is debugged on Linux before AIX, i5/OS, z/OS, or Windows enters the picture. It's easier, more flexible, and cost-effective. It just makes more sense. As the "PC" evolves to take on new shapes and sizes, free software platforms will have a pragmatic advantage.