ChipCrunch.com

It seems that at times we go through dry spells with relatively little news relating to semiconductor or chip startups, but then when things do happen they seem to happen all at once. With that said, a lot of news is not necessarily good news, as is the case with Ambric Inc. According to this eetimes.com article, the company failed to close on a $15 million round of series C funding, and had no choice but to close down. We wrote about Ambric last December, and found their software approach as well as the proposed Am2045 processor quite elegant. The combination of Java for its object-orientation, the Eclipse development framework, and a compiler which translates the code into a native machine language that maps simple objects to single cores and complex objects to multiple cores, seemed particularly promising. Further, according to the eetimes.com article, the company had quite a few design wins particularly in the video processing and medical fields. The additional funding was likely required because the company failed to successfully commercialize the Kerstel prototype chip which they first announced back in 2006. It is very disappointing that the company needed to obtain additional cash to stay afloat in this less than startup friendly environment, yet at the same time surprising that VCs decided to pull the plug on what at least externally seemed like a very promising approach.  Hopefully, some more well-funded company or a wealthy individual out there will find Amric’s ideas interesting enough to buy them and carry them forward – Who knows, Transmeta, a company that a lot of people have written off for a long time now, just got acquired by startup Novafora the other day.

Back in August, we wrote about Plurality and the additional funding that the company obtained for their HyperCore Architecture Line (HAL) of chips. At that point, we had several questions for the company. For example, we wondered what happened to the 64-core 90nm part that was supposed to ship in Q3 of 2007? We speculated that the company either encountered some problem with the initial architecture or that what the company attempted was not feasible in 90nm. We also had concerns about the efficiency of the task map programming model that the company was pursuing. Well, recently Alan Singer, Plurality’s VP of Sales and Marketing, commented on our article to explain the product delay. Additionally, just a couple of days ago Plurality announced at the Multicore Expo in Japan the beta release of a complete set of tools for the HAL processors line.

You can read Alan’s complete comment regarding our post here (below the article). In summary, there were no issues with the initial architecture, however, due to price, performance, size, and power requirements expressed by potential customers the company opted to use the 65nm processes instead of the 90nm process for the 64-core chip. So we were partially right about 90nm not being the right process node for what the company was attempting. Plurality also intends to use the 65nm process for the 256-core part. Unlike what management sometimes believes, taking a design from one process node and porting it to the next one is not exactly a walk in the park, as many an engineer will attest.  As such, the new target dates for these processors are the second half of 2009 and 2010, respectively.  An FPGA-based evaluation board featuring 32 cores should be available in Q2 of 2009.

In addition to working on the processors themselves, Plurality has been busy optimizing their development tools. As mentioned above, the company has just released a beta version of their development tools which consists of a cycle-accurate simulator, a GCC cross-compiler and GNU Binutils, a cross-debugger which is compatible with the Eclipse development environment, and finally an emulator. The tools can be immediately downloaded free of charge over here, although a brief registration is mandatory.  Additionally, Plurality is also developing acceleration boards featuring the HAL chips which can connect to the main CPU either via the PCI Express interface, or specifically to AMD processors, through a HyperTransport link via AMD’s Torrenza initiative.  Good to see that Pluarity is making progress and given some spare time it would be interesting to take the development tools for a spin.

First things first: Why would you pick a name which sounds so similar to nVidia? I’ve had friends who’s startups had names that differed by more than a mere two letters from other brand names, but still got letters from brand lawyers from similarly names companies within days. Just strange, but I suppose investigating naming choices is beyond the scope of this blog.

Movidia is a fabless semiconductor startup based out of Dublin, Ireland, and has just secured $14 million in Series A funding led by Celtic House Venture Partners and Enterprise Ireland. The company intends to enable users to be able to edit videos directly on their mobile devices. Especially, they are targeting the always on-the-go users, who contribute a lot of User Generated Content (UGC), but rarely have time to sit down and edit their content on a PC. There really is not much information available regarding the actual processors that the company intends to develop, other than the claim that they are supposed to be low-power. Since the company targets a lot of video processing, where a lot of parallelization can be exploited by working on different parts of the image simultaneously, one could expect either a multi-core processor or something along a Single Instruction Multiple Data (SIMD) vector processor.

There are a couple of issues that seem to bother me about the company’s strategy, although they are simply guesses without knowing what the company is actually developing. First, it is unlikely that mobile device developers will want to integrate an extra chip into their devices. They are having enough trouble with battery life on the current crop of smart-phones as it is. As such, it would seem the company would have to develop a chip that implements all the other basic cell-phone functionality as well. Which leads directly to the second problem - they would have to go up against the mobile processor giants such as Qualcomm, TI, and even possibly Intel depending on the final power envelope. All of these companies have mobile chips and sophisticated development tools to go along with them – which would make it difficult for Movidia to compete. Then again, Sean Mitchell has been around the semiconductor industry for a long time, and has lead Parthus Technologies through an IPO in 2000, so he just might have a few aces up his sleeve. Until Movidia reveals more about what exactly they are developing, enjoy the video below that the company created as an introduction.

We wrote about XMOS Semiconductor about a year ago when the company first emerged from stealth mode and introduced its programmable semiconductor technology called Software-Defined Silicon (SDS). Essentially, the idea was to configure the underlying hardware to handle the software tasks at hand via a high-level description language. Obviously, there are many ways of solving parallel programming problems for multicore processor, however, according to Professor David May who is the CTO of XMOS, many approaches to date will simply not work, as he recently revealed in a column by David Manners over at ElectronicsWeekly.com.

Professor May is mostly critical of Intel and Microsoft, arguing that the basic definition of their problem is flawed in that they are trying to take current sequential applications and enable them to run better on multicore processors – which is "virtually impossible." He further points out that the shared memory approach is flawed as well since every time more cores are added, they will end up competing to gain access to the same resource, essentially making an already complex problem even more difficult. He is also not a big fan of people who put their faith into compilers or abstraction layers for legacy software, arguing that the former take too long to develop and optimize while the latter are simply inefficient. One of the more interesting quotes from the article is with regards to computer engineer in general and a paradigm shift that David thinks is necessary: "A universal computer is an infinite array of finite processors, not a finite array of infinite processors." - Now you have something to contemplate over you next coffee.