VISC - How to throw your money away 101

See:
http://www.anandtech.com/show/10025/examining-soft-machines-architecture-visc-ipc


Here is an idea that is sure to bankrupt a lot of people (while making some millionaires too).

The idea itself is great (see link above), if it could be made to work.


See:
http://www.anandtech.com/show/10025/examining-soft-machines-architecture-visc-ipc/5

The assumptions, twisting of data and plain lies of the processor reality as it is today (Feb 2016) is astonishing though. See the 9 points in the link above why the proposed 'benefits' VISC offers is so full of holes that even BS can't stick to it.


See:
http://www.anandtech.com/show/10025/examining-soft-machines-architecture-visc-ipc/4

Yeah; AMD and Samsung. Intel nowhere to be found. Best hint to steer clear of these jokers.

AMD is still into promising the moon when they can't find the light to their back porch.

Samsung is doing what it does best: obfuscating (hard) facts with graphs manipulated, tortured and beaten into showing whatever they're backing to look like it came from a superior alien race from the future (and yeah; some will believe...).

Just like the conversations we had in these forums not too long ago about comparing whole platforms to whole platforms and drawing conclusions on real life possibilities (instead of artificially limiting platforms to identical clock speeds and then trying to interpret what that inaccurate IPC difference might/should/could give us in real world workflows)...

what VISC is presently trying to sell here is the idea that if time were to stop here/today and everyone else's platforms were brought to the low, low level of their working silicon shown at it's best with almost every other platform shown at it's worst, they would be worth looking at.


I believe that the tech itself may be viable given enough time. After all, a rocket scientist I am not.

What I find hard to believe though is that Intel hasn't already gone down this road already and have found it a dead end vs. what they're doing today and have planned for the foreseeable future.

It would be great for VISC to put a fire under Intel and get them to give us future architecture faster. No one would want this more than me (even if I still would stick to Intel anyways, for one reason or another).

But when the players behind the company are made known (especially the Samsung connection...) and the facts as presented so far are made up fairy tales (Skylake is represented as an i5-6200U? Really? Worse: as offered in a Dell notebook (overheating/throttling pieces of junk)?) all I can say is this is the companies your fathers warned you about.

Luke, listen to your father...

 

From the beginning of the article, it seems Anandtech was showing the potential benefit to older systems. Intel has clearly chosen to use its turbo feature to accelerate single-core performance in lightly threaded applications, and as such is their approach, they're not going to be interested in this alternate method. Samsung & Global Foundries are consistently looking for ways to jump ahead of Intel, which makes their involvement logical.

The concept of VISC does seem like it could be more efficient, but my guess would be that the reason they didn't show the raw data in their chart was because they were trying to show that their process can be both more efficient AND more powerful, when it probably isn't both. Seems like Soft Machines adjusted the data to get that extra wow factor and got called on it.

Regardless, they will have to provide data on their actual processors to gain the contracts that will allow them to be in any production devices.

 

If this approach was viable (tech and cost-wise), Intel would have served it up already. Combining what VISC promises with what Intel is already capable of would only set them a further decade or two ahead of anyone else.

Intel 'jumps' ahead by delivering actual value for their products to their customers. Samsung just keeps trying to bamboozle theirs' anyway they can.

As long as products are assessed on their own merits (without the Samsung 'cheat all you can while you can get away with it' angle...), the best products (and companies) will rise to the top and stay there, solidly, in the long run.

Even in the article, it is mentioned that GF and Samsung were noticeably absent in the actual production of the chip produced. Timing was mentioned as a possible issue, but I see this more as Samsung having a clear line to the exit door instead...

 

The world doesn't work like you think it does. Intel have failed many times to 1) adjust to new markets 2) predict the future of computing 3) predict the future of logic processes (they still are leading though).
Intel isn't using magic and can't know how the future will be. What they have is a lot of money and many smart people, but that doesn't guarantee anything.

Right. This isn't related with the real world and what happens there, but right - in your mind that's so. No need to back it up with facts.

There are a lot of other reasons why that may be, but why consider things logically when you can lash out with any factual support?

BTW. I agree with your thread title - VISC have burned a lot of cash already and with very little to show. That computer companies (including Intel) give some cash to far out projects isn't uncommon, that doesn't mean they necessarily think it will ever result in anything.

 

You do not negate any facts I've presented. At least not with any such facts as you claim I lack... to support what you say...

The world does work as I think. Intel has failed to adjust immediately just a few times over it's history, small missteps in the overall journey actually of where they are now.

And yeah; some of the things I state are my educated opinions. So?

And that is the point too; Intel doesn't need magic nor predict the future 100%. They deliver faithfully on what they know they can and what customers want.

VISC and it's backers are not even in the same ballpark, but try to strut and preen like they are. If anyone thinks that Intel is going to be 'surprised' by anyone else (even by some former employees) with this line of thinking is not paying attention to what is happening here.

Intel is not at the pinnacle of the industry by fluke. Hard work and exploring multiple scenarios simultaneously is what keeps it ahead. Especially as each further move forward gets harder and harder to do lately for everyone.

 

Like they weren't surprised when Itanium took a beating with AMD64? There's always room for surprises

Problem is when following instructions depend on the results of previous instructions. For instance if A=B+C you have to wait until A is worked out before it can be used in any further calculations.

 

I'm all for some potentially disruptive technology even if, ultimately, it fails to deliver all that is claimed. I've got a Transmeta Crusoe powered notebook (Fujitsu P2020). It certainly didn't win any performance awards but the battery life caused Intel to do a hasty rethink about where they were going with mobile CPUs. And we have to thank AMD for putting enough pressure on Intel that their CPU prices became a little more affordable.

To me, VISC is something to be aware of and watch with interest.

John

 

Mine is, does it perform as expected? Can it do my job?