CPU, GPU makers gussie up their wares for Hot Chips
This is still Silicon Valley
The annual Hot Chips symposium on high-performance chips will be hosted by the IEEE at Stanford University this August 21-23, and heavyweights in the microprocessor and other chip rackets will be on hand to talk up their latest innovations. As usual.
To get everybody fired up about the event, Hot Chips has posted the program schedule for the event, with just enough information about what chips will be discussed to be a tease. (Even techies need to do marketing, and this is the kind of free publicity that chipheads and their corporate sponsors revel in.)
The first day of Hot Chips 22 will be filled with tutorials, and the first day of the conference focuses on high-performance computing, systems-on-chip (SoC), and networking and data center processors. Nvidia is giving a presentation on its "Fermi" GF100 graphics processing engine, which has been making its way into discrete graphics cards and GPU coprocessors this year. IBM will be giving a session on how its eight-core Power7 chips have adaptive energy-consumption features that are useful for HPC clusters.
On the SoC front, Microsoft will be divulging details on its new Xbox 360 SoC, which is rumored to be a 45 nanometer SoC that will allow Microsoft to launch a smaller, revamped game console. Hot British chip designer ARM will be showing off extensions to the ARMv7-A architecture, and Mindspeed Technologies will be talking up how multicore baseband-processing SoCs will be able to make devices on 4G networks work better.
Researchers from the Massachusetts Institute of Technology and the University of California at San Diego will be spilling the details on GreenDroid, which they say is "a mobile application processor for a future of dark silicon." Which sounds suitably ominous to garner some attention. (By the way, Michael Taylor, who is one of the presenters for GreenDroid, wrote his PhD thesis at MIT on multicore SoC processors with mesh networks that was commercialized as the tile chips from Tilera.)
On the networking and data center front, IBM will be talking some more about its Power Edge of Network, or PowerEN, which is a 16-core, Power-derived processor that IBM also talked about at the International Solid State Circuits Conference in February. This chip, known as the Wire-Speed Processor back then, has 16 cores, each with four threads, derived from the Power7 architecture but with lots of stuff ripped out that's not necessary for an inline processor that chews on data and passes it on as it sits on the network.
Day two of Hot Chips 22 will see IBM trot out its next-generation mainframe processor, which everyone is calling the z11 but which probably has a different name. Advance Micro Devices will be showing off its "Bulldozer" cores for next' year's Opteron server and workstation processors as well as the "Bobcat" low-power x64 chips that will be used for other more power-conserving devices (and maybe more than a few servers if AMD plays its cards right.)
Intel will be talking about the "Westmere-EX" kicker to the current Xeon 7500 processors, which the company says will have 20 threads — that means it will have 10 cores compared to the current top-end Xeon 7500 parts, which have eight cores and sixteen threads. And the Chinese Academy of Sciences will be showing off its future Godson 3 GS464V (sometimes called Loongson-3), a low-power, MIPS-derived processor that will sport 512-bit vector extensions and that will be used in the next-generation, multi-petaflops super installed at CAS.
Oracle, which is supposed to be debuting its 16-core "Rainbow Falls" Sparc T3 processors right about now, gave out some details in February at the ISSCC event, but is not schedule to talk about microprocessors at the Hot Chips symposium. Oracle's techies will, however, be on hand to participate in a tutorial on optical interconnects the day before the conference begins. ®
Rainbow Falls is falling off a cliff
Oracle will stay in the hardware business, just not as a strategic part of the business. It's all about high cost appliances which drive software dollars and lock in customers. The only reason Oracle is keeping the T chips around is because they use tons of licenses. The real problems is all the talent has left and you cannot just high top chip talent off the street when the good people have gone to CISCO, HP and even Microsoft.
The M-Class systems will be gone from Oracle's portfolio even before the next speed bump comes out. The SPARC64 VIII has been canceled so SPARC64 will never see 45nm.
Oracle is also looking to OEM Dell for Exadata V3.