The Future of Computing Is Here

[Radar]

Originally Posted by tw

IOW they are still distributing prototypes to selective and hopefully future developers and customers.

It may be a breakthrough. Such technology has been discussed and proposed even 20 years ago. In ballpark numbers, it has maybe a one in ten chance of ever making it to the market.

It's already in the market. There are several companies already using the Zii chip. If you go to CES, you're in the market.

[tw]

Originally Posted by Radar

It's already in the market. There are several companies already using the Zii chip. If you go to CES, you're in the market.

Sometimes. Generally, major companies with existing products do not demonstrate in shows (ie CES) until ready for production. Small upstarts, instead, will use CES to test the market or maybe encourage entrepreneurial investors. I am currently thinking of a Microsoft product involving turtles (name and purpose long forgotten) that was to be some kind leadership product. It got unveiled. Then it completely disappeared. Different companies unveil for different purposes.

GM has so few new products that GM unveiled their Volt in shows how many years before it will ever be on the market? It is rather difficult to know (from what I have seen) if that CPU is still in prototype or Alpha testing. I seriously doubt there will be a retail product this year. They are a new company. So they need the exposure long before actually selling product.

Sigma Designs once unveiled a revolutionary camera many years ago. Good luck finding that camera in the marketplace. They unveiled. Could never find it except in ‘new product’ publications.

Meanwhile, what new software makes it possible to program? That too has been a complication for every revolutionary CPU products including the original CMOS CPU sold by RCA. No complete solution means no retail sales. Its difficult to predict any kind of sucess from a produt that, if it does as prmoted, may be wildly successful.

[mbpark]

Creative Labs is involved. They're well known for PR gaffes (think of their latest driver fiasco and their not-so-good financial shape).

They're also known as being quite hostile to developers (closed SDKs), while NVidia, ATi, and IBM are working on OpenCL support.

Don't get me wrong, 3DLabs is a great company, but this isn't exactly groundbreaking. This is roughly the same technology that NVidia and AMD are already working on (CUDA for NVidia and ATi Stream), and a variation of what IBM has done with PowerPC/Altivec/Cell and their Blue Gene (which run on very low power PowerPC chips) processors. It also reminds me a bit of the AXE engine on the Freescale PowerPC based MPC5200 series of chips, but with dual ARM cores instead of one PPC core.

SGI has done a lot with this technology too, esp. with their RASC (Reconfigurable Application Specific Computing) technologies, which they've worked on extensively with Intel for full integration with Linux.

The difference is that NVidia has been open with CUDA, with many design wins. SGI has a ton of design wins out there, and has integration with Intel's compilers, which are widely considered to be some of the best in the industry. IBM has published what you need to develop for Cell to the Linux kernel already and can use multiple Linux distributions to code for it (Cell has 6-8 SPEs depending on the platform). The supercomputer used to cause the MD5 hash collision was based on Sony PS3s with Cell chips.

The point is that Creative is "creative", so to say, with their marketing. This sounds a lot like CUDA, RASC, Freescale AXE, Cell, or ATi Stream to me.

TW, funny that you mention the CMOS CPU worked on at RCA. My uncle was on that team .

[tw]

Originally Posted by mbpark

This is roughly the same technology that NVidia and AMD are already working on (CUDA for NVidia and ATi Stream), and a variation of what IBM has done with PowerPC/Altivec/Cell and their Blue Gene (which run on very low power PowerPC chips) processors.

If I remember correctly, CUDA is about letting the programmer use the 60 some CPUs inside the video processor to execute complex, multiprocessor programs.

TW, funny that you mention the CMOS CPU worked on at RCA. My uncle was on that team.

It was called COSMAC, if I remember. Was he in Somerville NJ?

Back then, RCA was doing breakthrough technologies including CMOS chips that are now standard and required for all computers today. Chips that were necessary for 1970s spacecraft when most semiconductor manufacturers will still using bipolar transistors (ie Texax Instruments TTL). Even Intel eventually gave up on NMOS to convert to CMOS (with help from Harris Semiconductor).

Once designed a car alarm using those RCA cmos integrated circuits. Amazing how much could be done with almost no power.

So when was that CPU team first started? Why did RCA not create a programming development language for it?

[mbpark]

TW,

My uncle worked out of both Phoenixville and Somerville, I believe. He started there in the late 60's, and told me he started working on the 4 and 8 element CMOS chips around '68.

They grabbed him right out the US Army, literally. Since the first applications were for defense, and he just got back from Europe, that makes perfect sense.

If you PM me, I will give you the name.

BTW, CUDA is about using the 240+ stream processors, aka SPEs, within a standard video card as programmable units that can be programmed using C. You can gang up 3 of these cards in a standard PC using SLI and really crank out the numbers.

The second they have a SAS interface I have a customer who will buy 6 cards (three for him, and three for his developer).

[mbpark]

TW,

Chips were slow back then. High-level languages just added to the complexity. ASM and ML was the way to go.

[tw]

Originally Posted by mbpark

Chips were slow back then. High-level languages just added to the complexity. ASM and ML was the way to go.

I was told that the CPU was sold without even an ASM development language. Never could confirm that. Never got to use one.

However it was the first computer chip used in communication satellites where I worked. Its function was to maintain the spacecraft only if communication had been lost for a given time period. All commercial satellites back then typically had no intelligence and would be quickly lost if not in constant control from the ground. Satellites once were kept as dumb as possible for reliability purposes meaning that processor was a breakthrough.

By "4 and 8 element CMOS chips around '68", do you mean 4 bit and 8 bit versions? That confirms what I had originally learned. Those processors were available in '68. Others insisted they were not available until about 1972.

I think Intel finally used CMOS around mid 1980s - to put perspective on how innovative some parts of RCA still were back then.