Tilera

Over at Accelerated Times, an article was posted about the Tilera. Now I haven’t heard much about Tilera, other than pre-releases. [update: look at the comment here] The author focuses on several important aspects. The business model, the money raise, are they are where they say they are.

What strikes me is that if they raised a B-round, this usually … usually happens post initial revenue, when you start to see interest and traction. I am not agreeing/disagreeing with the author over his comments. Just pointing out that it is possible to be further along than press releases indicate. That said, for any new architecture/ISA, programming is a bear. Whether or not people like it, its the applications. It is always the applications. It is never the chip technology. The reason why Opteron did so well initially, was you didn’t have to recompile code for it. It ran existing code better/faster. You got a free performance boost by recompiling, but you did not have to recompile. The reason why Itanium has been on life support for the past decade has been that you have to recompile code. The reason why RISC is effectively over (there are a few pushes to keep it alive, but the CBA just isn’t worth it in the end with the economies of scale for IA type machines) is that each RISC is different, and that you can get about the same or better performance on IA, with fewer ports, lower cost of support (Linux vs Unix flavor of the day). And so on. This is what Tilera and every single other proprietary architecture chip shop has to deal with. It is not pretty. Sadly, for technologists, it is not about the technology. Its about how much better you can do with this one 20+ year old ISA. Well, AMD managed to extend the ISA a bit, and now we have a new ISA, which Intel is managing to win deals with. There is nothing wrong with competition. The problem is market momentum. Anyone peddling a chip/system built on a proprietary non-common ISA (e.g. non IA32/x86_64) is going to struggle commercially. Unless of course, they are selling to people whom have no interest in commercial apps, and have no problem rebuilding code. I was under the impression that Tilera was focused on the embedded market. There are other large core count chips out there. Cradle comes to mind. That is a somewhat different market. If they have traction in it, great. There IA is not so important. Ease of programming is.