I think IBM needs to approach POWER similar to how ARM has built their ecosystem. ARM has managed to get their CPUs in the hands of developers in huge volumes, has put effort into open-source software support for through organisations like Linaro, and has many partners in many diverse fields.
On the flip side, IBM has largely gone at it alone for POWER; very few people have access to their systems, most of the people supporting POWER in the open source world are IBM employees, and outside of Freescale and a few others, there aren't that many people pushing POWER.
Hopefully this Open Power Consortium and their new-found focus on Linux will lead to them building a larger community around POWER and clawing their way back.
IBM is a strange company. They seem to fade from relevance and appear to teeter on the brink of becoming obsolete, but somehow manage to remain a power house through it all. I have no doubt they'll pull through, but I haven't got the darndest idea how.
Perhaps IBM pulls through at least in part because it has the pockets -- and the foresight -- to think ahead -- way ahead. As the NYTimes recently noted, "I.B.M. was granted more patents in 2012 than any other company, **the 20th consecutive year in the top spot**". Personally, I have the pleasure of writing about IBM on a regular basis -- at least the folks I'm in contact with -- those working on their chips and basic science stuff -- are insanely smart and innovative. And as a company, they always seem to have an ace in the hole. They understand Big Data as few do. Last stand? To borrow from Mr. Twain, I think that falls into the domain of great exaggeration.... IBM is, was and always will be an amazing company.
IBM is now living from the past. The managers and the rank and file are oblivious to the current state of the company. There is no sense of urgency to innovate. They are mainly focused on maintaining their PBC ratings above 2. The employees are just there to push the buttons and that is rewarded with promotions and commendations from the managers and executives who are also products of the inept system. Last time I checked our org chart under STG, we have more than 13 layers of managers who only supervises 3-5 direct reports at the bottom. We create new departments for the sake of creating first line managers from the direct reports to makefirst line managers a second line manager. These managers have nothing to do with innovation or production or support. They are just there to act like managers and nothing more. Another example of inefficiency from this company is the practice of oursourcing jobs to India or China. Many understand the disadvantage of US based employees when competing against the salary of engineers from these countries. BUt instead of replacing one engineer with another engineer from those low cost courties, IBM managers here will insist on hiring more engineers from those countries thinking that they can accomplish more if they add more engineers. They are using the salary of one engineer here in the US to hire 5 engineers abroad. Guess what, adding more engineers has not accomplished more and employees from both US and those countries art now competing for the little work. There is no vision on how to utilize all these employees. The question of how many IBMers does it take to screw a bulb applies to the current state of the company. I am just a few years from my retirement and I hope I survive it before IBM implodes. Wish me luck.
Rick, i get your point about the non-existent Power architecture in new generations of game consoles...but what about Power architecture embedded in base stations and power trains for automotive? Are they being successful, or are they too getting replaced by ARM?
The trend is embedded processors from Freescale, LSI sand others are generally in the process of migrating from PowerPC to ARM, certainly in networking anjd comms for basestation chips. I am less sure about automotive-specific chips.
I am guess they might be slower to change given slower design cycles in automotive.
When I got to the part where you say "In the coldest cut of all, Amazon recently won a deal to supply computer services to the CIA..." I thought to myself "If I went back in time 20 years, what would IBM have said to me if I'd told them a 'book-seller' woudl one day be out-selling them on the computer front?"
Then I thought "It's a funny old world, and no mistake"
it is too late for IBM to do an ARM now. Too much of a head start, may work if bare isa is licensed free. Besides new Isa's like UCB's riscv are a better and cleaner option, besides being free. In fact for our open source processor, we did a ppc implementation but later switched over to riscv. Patents are not an issue in India but if we wanted our processor to be adopted worldwide, they can be a problem.
In fact a whole bunch of universities are collaborating to get these open source processors out. That could potentially eclipse ARM.
The adjunct story to that is the Chisel hardware development language from UCB. It is based on Scala and is again completely free. Use off scala opens up the programmer base to a wider audience compared to say System Verilog. In fact the US govt is also keen to see this language succeed. So now all you need is a PC and a bent for processor design and you can churn out your own custom cpu. A basic fpga board will do for a sub 500 MHz processor. Our lab results show 200 MHz is trivially achievable. The point is that commercial grade cpu code will now be open source, so most folks just need to tweak it, not write a CPU from scratch. We also plan to open source IP like PCIe,DDR3/4, RapidIO, 10ge. Hopefully someone will do a gpu.
See riscv.org . UCB plans to finalise the ISA in Sept.
Blog Make a Frequency Plan Tom Burke 17 comments When designing a printed circuit board, you should develop a frequency plan, something that can be easily overlooked. A frequency plan should be one of your first steps ...