Originally posted by Chagi
I've been thinking a bit recently, and I wonder if a Transmeta CPU for mobile Mac purposes is a possibility.
Interesting that you bring Transmeta up.
The reason I mention this is that the TMTA chips emulate x86 in real-time on their VLIW architecture, and their watt/performance ratio blows both AMD and Intel out of the water. If they can do it with x86, why not do the same for Power architecture?
Well while their watt/performance figure is good there overall performance figure isn't all that great. Since Transmeta is going the way of the Dodo the best thing that could happen is if Frescale where to buy them. Take that technology and incoroprate it into a processor that can run PPC code at low power. Freescale shouldn't be underestimated with respect to power usage though, the trick is to get the low power and performance in the same package.
The saddest thing about Transmeta and probally one of the reasons that they were never well accepted is that there is little access to that VLIW engine and it is next to impossible to get a OS to run on that engine. I would have thought that they would have learned from the poor response their CPU's got in the market and upgraded their VLIW design into a real compute platform but they didn't so I don't see ahuge future for them. In any event I suspect that it would be very difficult to get a VLIW machine to emulate any of Freescales processors and be under their watt/performance figures.
Add to that the variability in the performance of a Tranmeta based device and I fail to see a happy ending. Its sad to see Transmeta going the way they are at the moment but it is often a fact of life that companies bcome so focused on their technology that the fail to adapt to new realities.
Ironically, the biggest reason I can see for this not to happen is not technical, but instead legal, since Transmeta isn't part of the PowerPC triad (Apple, IBM, Freescale), not sure what would be involved licensing-wise.
That is rather funny as I see the major problems beign technical not legal. Apple was part of AIM there is no reason why they couldn't sponsor another processor endeavor. We all remember that other company that wanted to go the bipolar route.
The problem is performance. Maybe the latest from Transmeta are performing better, but I find it hard to believe that they can beat a RISC system using equivalent processes. That and the fact that sooner or later high integration devices will be the norm, just reduces any advantages that the current Transmeta chips may have. At some point, for portables anyways, it will be a big advantage to go to high integration devices.