TR Forums

It seems VIA's subsidary CenTaur has been cooking up something new in the form of an Isaiah II architecture. Right now their company page links to a teaser site with a count down, set to run out on 31st Aug.

This has all come as a bit of a surprise as I haven't seen any new VIA CPUs for a long time, well since they released the VIA Nano (formerly code-named VIA Isaiah) in 2008. At the time the Nano was a competitive with Intel's Atom performing between 15 - 20 percent faster but not matching it in power efficiency. Linky Unfortunately Via doesn't have the resources of Intel and wasn't able to make it a widely available success.

What's even more intriguing is that Extremetech suggest the Isaiah II might be a hybrid x86/ARM processor. Chew on that.

I'd say roughly 8 years ago, I was intent on building a car-puter. Since all the other options at the time were incredibly anemic, I wanted to use VIA's newly announced Nano processor; which sounded perfect for the job. All I can tell you is that you should add at least 2-3 years to the predicted launch date, and then expect ongoing supply issues. I eventually gave up on the idea entirely.

Hi Egg I know exactly what you mean. I followed their development pretty closely at the time and was disappointed by the long wait and eventual release of products that had by then been outclassed. I hope things might be different this time around. I mean, VIA must know exactly what they are up against with the proliferation of power efficient x86 CPUs, small form factor PCs like Intel's NUC, as well as ARM development boards like the Raspberry Pi and Beagle Bone Black becoming quite popular.

Who's VIA?

*badum-psssh*

This is an interesting development. About as interesting as AMD's upcoming ARM and x86 cores. Some thoughts:

1. Considering the cost of developing CPU cores, existing product sales will have to be strong enough to fund the next project. But given VIA's tiny market presence, I wonder where VIA is getting funding. They're part of Taiwan's Formosa Plastics Group but still, unless Formosa Plastics just wants to burn some cash it seems difficult to continue developing x86 cores. Considering VIA's cores are quite a bit simpler than Intel's or even AMD's, R&D costs are probably lower, but still, VIA CPU revenues are too small.

2. How is VIA's x86 license doing? Do they renew it every few years the way AMD does?

3. Good to see how AMD's SkyBridge isn't the only game in town. In fact, this actually looks even more interesting from a technical perspective (never mind performance) because both ISAs can be serviced without swapping chips. That is, unless SkyBridge is actually a hybrid x86/ARM part as well.

4. I really hope VIA can target higher clocks. A similar attempt at hybrid ISA CPUs was done by IBM back in the 90's in the form of the PowerPC 615, but I think it ultimately failed because Microsoft didn't want to support it and performance wasn't groundbreaking (don't quote me on this though). Here's hoping VIA can pull it off better than IBM could, but don't hold your breath.

With the rough times VIA and AMD are going through I wonder if they shouldn't team up.

Losergamer04 wrote:

With the rough times VIA and AMD are going through I wonder if they shouldn't team up.

Because failing company + failing company = successful company?

They're probably both better off focusing on niche markets. AMD on GPUs, APUs for consoles, and (hopefully) their new ARM initiative; and VIA on... well, whatever it is that VIA focuses on these days. VIA could give Realtek a run for their money in the onboard audio chip market (VIA's sound better IMO), but I imagine audio chips are a really low margin business these days.

I disagree with the notion that two failing companies would not work. I think, in this case at least, that it's a good idea. There is little overlap in customers but much in the skills and tallent. They could even use their existing IP and do what Intel has done, a low-power/mainstream CPU and a powerful/server CPU.

OK, let me rephrase that... two failing companies that historically have both had issues executing on their business plans. I suppose if you could find an outside management team + venture capital firm willing to acquire them both and make some difficult decisions, you might be able to find some synergies there and turn the resulting company into something successful. I don't think a simple merger of the two companies (with their respective management teams retaining key roles) would end well. The value in both companies lies in their IP, and any key engineers who haven't already quit or been let go.

I wonder if Centaur's new core will have the option to use the instruction set of the underlying RISC core like their old Nehemiah did. I always thought that was so interesting: set the ALTINST bit in the FCR, execute your special jump instruction and from then on you use AIS instructions. Then when you're ready to come back to standard x86 execution mode execute your AIS special branch instruction and carry on executing x86 instructions. The AIS instructions were actually fixed length 32bit instructions that were embedded inside an lea instruction (as the displacement). We need more crazy stuff like that in the world

So the countdown has hit naught, and Centaur's site just sits there with a link to their old homepage looking like it is out of the 90s. This launch doesn't seem to be terribly well stage managed and yet I am still very interested.

Edit: Looks like they have reset the counter for another month.

Yeah, I noticed that too. So much for the supposed big reveal on September 1st...

Like you say though, still interested. Clearly I just can't help myself