Topic: Intel Hardware Accelerator for Mining (Read 765 times)

ASIC engineer, not the most helpful as he very low on totem pole.  He is physical engineer which is a part of the puzzle I needed to fill and he wants to help.  His thought was that if I can make a case to Intel Foundries for 10nm chips then that should be my route in.  Bitmain was a huge customer of TSMC in 1st quarter so this may be getting their attention now.

I will ask him if he can reach out to the gentleman at Intel Labs and see what their reason was for submitting the patent in 2016.

Well, of course speculation isn't everything. The main influence on price of shares is through the actions of the company's management; what  they do and what they don't do. Apparently "what they don't do" is the major opening for shareholder litigation. So the smart lawyers convert that to "we did it, but we didn't see the promised return on investment".

Well how about that. Yet another good reason to add to my list of "never sell shares or take investments". Why would a lawsuit in any way solve that problem? After the initial offering, isn't share price determined entirely by market speculation?

Also, 21e6 really seemed, from the beginning, to have very little idea what they were doing (or how to do it right). I believe at one time they mined a significant share of the network, but their chips were lackluster and the one "product" was basically a single large-scale ASIC duct-taped to a Pi with about a $300 price tag and no real purpose other than to sound cool to people who didn't know any better. If I remember right, they were one group pushing for Bitcoin blockchain-as-a-service microtransactions (in an already cluttered network) doing entirely-non-currency things, which is pretty counter to reason.

Yeah, the story is as old as shareholder litigation in the USA. It is a long term defensive strategy. IANAL, but the way it was explained to me is as follows:

Large company listed on US stock exchanges is bound to be sued nearly anytime anybody is unhappy with the price of their shares. Those lawsuits are quite expensive to defend against. The best long term strategy turns out to be investing small sums in scatter-shot way in all kinds of businesses related to their main business. Even if that joint venture is a scam, not a real business, the large corporation for a period of time gains inside knowledge of the operation. That knowledge is being obtained cheaply on a friendly terms, not in a very expensive way of discovery during a litigation.

In other words the large corporation obtains inside knowledge to successfully defend itself against future lawsuits. That defensive information is obtained extremely cheaply (compared to the actual adversarial litigation) that even investing small sums in hundreds of businesses is cheaper than a single completely unanticipated and unprepared defense.

Color me skeptical.

A couple of years ago, a company called 21 Inc (became 21.co --> earn.com) got a great deal of press in terms of Venture capital from big names. They also supposedly had some kind of "in" with Intel for chip fabrication. If the buzz from them had panned out, they would be a major player in terms of Bitcoin mining hardware.

Instead their one hardware product has been completely orphaned, and their current business model has no mention of Bitcoin mining chips nor Intel. Maybe all that they claimed was real, but Intel is a huge company and can "squander" tens of millions of dollars, and think very little of it. The last big thing with them could completely evaporate in a year. Intel knows where their bread and butter comes from, and as long as these little side adventures don't imperil the X86 business, they can make these "odd" investments without seriously impacting their business. Not so much with the companies Intel "partners" with.

Just my $.02 on where this might be going (or not).

Have a call with Intel ASIC Engineer today  to try to get my foot in the door.  Baby steps guys.  Where there is a will there is a way.

My original attempt was to reach out to the principal author of the patent at Intel Labs.  He didn’t respond to the LinkedIn.  So I put it on the back burner.

Coincidently, while doing searches for capital providers I stumbled across an Engineer at Intel that was in directory focussed on Blockchain.  Reached out to him yesterday, he asked for further information about projects I am working and he was wanting a call this weekend.  Hopefully the call is fruitful.

Ja. There is no doubt the models can be reproduced but the 'recipe' of what the fab steps are such as the exact chemistry of the photo resists, exposure intensity/time, supplier of equipment used in critical areas, etc.is a whole 'nother story. I highly doubt Intel would release that other than in very general terms.

In our case with our main customer in Taiwan,  the process we developed for them in 2007 is still unmatched in the world. They and the few other much much smaller 'boutique chip' companies that use our systems to make ceramic heat spreaders/interposers hold tight to how we do the voodoo that we do so well. Initially they went as far as covering all machine make/model graphics for when they give dog and pony shows to their clients.

A year after our 1st customer's exclusive use clause expired we published a short bit about it here. While our patents on it of course describe what we do, they *do not* give any info other than chemical family names on what the secret sauce that only we provide is nor any laser parameters used. That very particular bit is what Phillips LumiLEDs needed most to create a 2nd source vendor for their components. Now just past the halfway point of the patents life AFAIK no one has yet replicated it though I'd think that any good  chem lab could easily break it down if they got a sample of it...

As for NODEhaven, ja unless they have connections in the semi biz that can put in a good word for them with Intel they are in for an uphill battle...

edit: hmm. That short blurb and the more detailed PR description really need to be updated.... We've progressed far beyond that...

I don't doubt what you wrote, but I just think that it needs to be way more carefully explained, with many more technical and legal details.

Here's a short snippet about Intel's collaboration with Mentor Graphics from the above link:


Which basically means that Intel had shared very detailed electrical and mechanical information with Mentor Graphics to build a detailed analog model of the actual circuits manufactured. The analog SPICE models are electronic, but many electrical parameters have direct equivalence with the mechanical and thermal parameters of the process. In turn anyone licensing Mentor Graphic's toolchain would easily reverse-engineer those.

So when you say "customer demands to know specific process parameters and chemistries": they will get the final specifications of the finished products, but won't get the access (or just very rough access) to the intermediate steps of the fabrication.

My guess is that when NODEheaven "reached to Intel", his query got the same handling that e.g. "reaching to American Airlines about chemtrails" would have gotten. Any semiconductor-related business must be getting so many public inquiries about "coin mining ASIC" from crackpots that one has to really stand out to get any other handling that simply trashing.

^^ Well I stand corrected. At least I did preface with 'As Far As I Know'...

Is also a safe bet that Intel would not bow to customer demands to know specific process parameters and chemistries.

In our case Phillips-LumiLEDs Lighting tried to do that a few years back to a customer of ours by saying that they needed all of that info "for their in-house QA records as part of six-sigma". Both we and our customer flat-out refused. Process yields and production records dealing with their product - fine but no way in hell do they need specific 'how-to-info"...

Kicker is -- a few months later when Phillips was in a meeting with that customer during a presentation one of the reps had a brain fart putting it together and somehow an email mentioning their efforts to get process IP to develop a 2nd source popped up on the projector screen for all of management there to see. Needless to say they were pissssssed!

Then what is "Intel Custom Foundry"?

https://www.intel.com/content/www/us/en/foundry/overview.html

Even somebody on this forum tried to manufacture Bitcoin ASIC using the https://en.wikipedia.org/wiki/Achronix technology, which is another fabless company. And according to the Wikipedia they fabricate at Intel and TSMC. That was when Intel's 22nm process was the state-of-the art.

IIRC that Bitcoin ASIC on Achronix was something obviously stupid, like they wasted chip space on testing logic despite SHA256D being self-testing and PoW mining being a lottery.

Personally, I think the American-headquartered companies with fabs (like Intel, etc...) are just much better at enforcing the legal contracts with their partners about mutual-nondisparagement (or whatever is that legal concept called) in their fab-for-hire business. So there's much less opportunity for some dumb rumors to spread, like the recent "Samsung enters Bitcoin mining business."

I really don't feel like digging back through the history in this forum. But if somebody is interested and they want to verify what I just wrote, then you are free to dig for the actual references to the Achronix Speedster and how quickly the rumors about Intel's involvement in coin mining had gotten quashed.

AFAIK Intel has never contracted out their fabs to anyone outside of the Intel family.
I can understand why - doing so inevitably leads to process IP leakage. One way it happens is from a customer pushing to get enough information so they can leak to another fab under the guise of "creating a 2nd source 'just in case'..."

I am reaching out to Intel labs.  I will report back if I find out anything interesting.  I don’t currently know of anyone using their foundry for BTC ASICs.

Edit: haven't heard back from them at this point

Good question and I don't think was ever answered. Remember part of the kerfuffle was Bitmain being granted a Chinese Patent on their rendition of AB. I do remember a lot of noise about just how that happened considering Timo's prior Patent on it.

One possibly could be that the examiners used Timo's as Prior Art but only as what is refered to as a Basic Patent which only outlined the concept of AB. Don't think Timo ever worked with anyone to have it implemented into a chip. Bitmain could then be free to patent their hardware design to actually use AB.

Another possibility could be that that because Timo did not register the Patent in China, at the time the Chinese examiners simply adhered to the letter of their law and ruled it was fair game for a Chinese company to claim as theirs. That particular point is why my company refuses to sell our equipment in China. Yes we did go through the process and have not only US but also EU and Chinese Patents but we are still not about to let them get their hands on actual systems for them to then dissect and copy.

Or they just said screw Timo and stole it...

Either way  (barring theft) as Prior Art Timo nonetheless could/should enforce royalty payments if AB was ever actually activated by Bitmain.

Patents are only as strong as the financial muscle of the legal teams used to defend them.

Was Bitmain's version a license of Timo's patent, or a clever workaround, or just stealing?

The only thing I am happy about is that the patent does not follow the onerous Halong version that to me is strictly their attempt to get hold of other companies tech for free. Yes Intel will probably require a license/royalty fee but at least everybody can use it without giving away tech *they* developed.

Same applied to the original Timro's(?) AB patent until Bitmain stepped in with their version. That is when all hell broke loose.

Right right, because patenting and licensing improvements is definitely supporting a thing. How irate was the crypto world with AsicBoost first showed up as a patented licensed improvement? How is this different, that we're happy about it?

Sounds like the beginning of a major shakedown for BTC mining! And I'd say Hell! its about time.

More to the point, what must be the beginning of an informed discussion and definitely more useful than a news article, Intel's patent application

God reading pat apps always makes my head hurt!
Anywho, sure sounds like how Halong's AB works. Glad to see someone giving an alternative to it and Halon's onerous 'you can use ours ONLY if you give us access to all of yours' Patent.

Came across this article today and thought you guys might be interested:

https://www.coindesk.com/intel-wants-patent-bitcoin-mining-hardware-accelerator/

Intel say that it would be possible to reduce energy consumption by up to 35%.

That should help your cost mining ratio but do you think it will affect how your setup runs?