China plans national, unified CPU architecture

According to reports from various industry sources, the Chinese government has begun the process of picking a national computer chip instruction set architecture (ISA). This ISA would have to be used for any projects backed with government money — which, in a communist country such as China, is a fairly long list of public and private enterprises and institutions, including China Mobile, the largest wireless carrier in the world. The primary reason for this move is to lessen China’s reliance on western intellectual property.

There are at least five existing ISAs on the table for consideration — MIPS, Alpha, ARM, Power, and the homegrown UPU — but the Chinese leadership has also mooted the idea of defining an entirely new architecture. The first meeting to decide on a nationwide ISA, attended by government officials and representatives from academic groups and companies such as Huawei and ZTE, was held in March. According to MIPS vice president Robert Bismuth, a final decision will be made in “a matter of months.”

China has a long history with MIPS and Alpha. Loongson processors, which power millions of Chinese school computers, use MIPS — and the ShenWei processors (pictured right) found in China’s first homegrown supercomputer, theSunway Bluelight MPP, are based on the Alpha ISA. MIPS Technologies (the company) hasn’t been doing very well recently, and it’s rumored that the Sunnyvale-based company could be up for sale — a purchase I’m sure the Chinese government could afford.

According to EE Times, there are some 34 ARM licensees in China, but at $5 million for a single Cortex-A9 core license, it’s unlikely that ARM will be China’s choice. The Power ISA is cheaper, but lacks the software ecosystems that ARM and MIPS enjoy. ShenWei/Alpha is also a possibility, but again it cannot compete with MIPS’ installed base.

The other option, of course, is developing a brand new ISA — a daunting task, considering you have to create an entire software (compiler, developer, apps) and hardware (CPU, chipset, motherboard) ecosystem from scratch. But, there are benefits to building your own CPU architecture. China, for example, could design an ISA (or microarchicture) with silicon-level monitoring and censorship — and, of course, a ubiquitous, always-open backdoor that can be used by Chinese intelligence agencies. The Great Firewall of China is fairly easy to circumvent — but what if China built a DNS and IP address blacklist into the hardware itself?

Taking a leaf out of South Korea’s hardcore gaming scene, what if the Chinese government decided to implement a hardware-level 10pm curfew for video games? Or some code that automatically turns negative mentions of Hu Jintao (the Chinese president) into positives, and inserts a few honorifics at the same time. Or a latent botnet of hundreds of millions of computers that can be activated upon the commencement of World War III. Or, or, or…

The dream is dead: AMD gives up its share in GlobalFoundries

Three years ago today, AMD spun off its fab division, in a move the company claimed would allow it to more effectively leverage its assets, inject new capital into the foundry side of the business, and make it more competitive vis-à-vis Chipzilla. Today, that dream is dead. AMD announced today that it would give up its 8.8% equity stake in the company. When AMD created GlobalFoundries in 2009, the company held a 34.2% share in the foundry.

AMD always intended to allow its share of GF to shrink over time; the company’s antitrust lawsuit settlement with Intel significantly relaxed the constraints that previously prevented Sunnyvale from taking a smaller stake in GlobalFoundries from day one. Nevertheless, it’s surprising to see AMD agreeing to give up its share entirely, and it speaks to both Rory Read’s new direction and the size of the problems between AMD and its erstwhile manufacturing partner.

The original structure of the GloFo spinoff

According to the new agreement, AMD has negotiated a “take or pay” agreement for wafer prices in 2012, as well as a “framework” for pricing in 2013. A take-or-pay agreement is a contract in which the buyer (AMD) either accepts a product or pays the manufacturer a penalty. In 2011, AMD negotiated a wafer price arrangement with GF in which it only paid for “good” dies. Scuttlebutt indicated that GF was quite unhappy with this deal, as it left the company losing money on every Llano wafer it could build.

So what did AMD get? Manufacturing flexibility. Previously, Sunnyvale had agreed to manufacture 28nm APUs solely with GlobalFoundries. This new agreement voids that arrangement, freeing AMD to work with TSMC other foundries. It’s not an agreement that came cheap — not only is AMD giving up its 8.8% equity share of GF, it’s agreed to pay the manufacturer some $425 million by the end of Q1 2013. AMD will take a $703M charge against the transaction.

It’s a tad odd for AMD to be paying GF, given that AMD is giving up its valuable shares, but there’s a likely reason behind it. GlobalFoundries invested significant amounts of capital in order to meet AMD’s timeline for 32nm and 28nm parts. The foundry built a 28nm SOI variant — 28nm-SHP — specifically for AMD. With AMD now looking for other foundry partners, GF is left with production lines it can’t flip a switch and convert to bulk silicon. The $425 million probably accounts for some of that cost.

AMD may have regained some flexibility, but the company has paid dearly for the privilege. We’ve remarked on the cooling relationship between AMD and GF before and broke the company’s move to TSMC last fall, but this decision still comes as something of a surprise. GlobalFoundries has taken AMD to the cleaners with this one, securing a major cash infusion, far better per-wafer prices, and regaining full control of its own stock.

Whatever residual goodwill that existed from pre-spinoff days is dead and gone. Virtually all of AMD’s top positions have been replaced and GloFo sacked former CEO Doug Grose last year. We’re not going to claim that AMD will never build parts with GloFo — business is business, after all, and if AMD can make nice with Intel, it can make nice with anyone. Nevertheless, we don’t expect to see any new x86 parts rolling out of GlobalFoundries once Piledriver is completed. Jaguar, Steamroller, and Kabini/Kaveri will almost certainly debut at TSMC and be built on 28nm bulk silicon.

Is this a bad thing? That’s unclear. We know AMD killed Krishna/Wichita due to manufacturing problems, Llano limped along for most of 2011, and GF’s problems at 32nm impacted AMD’s ability to sell 45nm chips into the channel. From a macroeconomic perspective, AMD is simply transferring its business to a foundry partner that’s more able to meet its needs. The irony is that when GF spun off in 2009, there was a great deal of chatter about how GF would capitalize on TSMC’s tremendous difficulty in ramping 40nm. Instead, we see TSMC stealing 28nm business away from GlobalFoundries.

One could argue that AMD’s decision to get out of the foundry business is a logical extension of Rory Read’s plan to de-emphasize cutting-edge silicon in favor of SoCs. An 8.8% equity stake isn’t very much; it may not have made much sense for AMD to continue holding shares when it owned such a small percentage of GlobalFoundries and accounted for a shrinking percentage of the company’s revenue. It’s even possible that this was AMD’s strategy from the beginning, provided it could renegotiate its x86 license with Intel. Either way, it’s a decision that comes as AMD faces major long-term challenges with a new executive team at the helm and a very different CEO calling the shots. Thus far, Read has demonstrated that he’s willing to take risks and do things differently, as evidenced by AMD’s purchase of server manufacturer SeaMicro just last week. Whether or not he’s taking the right strategy, we can’t say — but he’s definitely willing to explore new options.

Source