Overt AsicBoost Released today?

[PeaMine]

asicboost.com
I am seeing a few news sites picking this up, but nothing I personally trust.
Anyone have more information on it?

From what I can tell, covert asicboost is now invalid, but overt asicboost works, but only for companies that give away their patents.
Not sure if it's related to the huge price drop, or if that's from the Whales dumping in Japan and at the Auction

[score1more]

When can I get the patch for my Bitmain miners?  LOL

[ccgllc]

When can I get the patch for my Bitmain miners?  LOL

I had the same thought... one firmware update release, and we are all back on a level playing field again (at least all of us Bitmain owners).  More likely, Bitmain will just create a new S9, using old stock with new firmware, and sell it as their V11 - advertising the 20% improvement in hashrate/watt.  Only trick would be to keep all S9 miners from reflashing using the V11 firmware...

[wedatin]

When can I get the patch for my Bitmain miners?  LOL

I had the same thought... one firmware update release, and we are all back on a level playing field again (at least all of us Bitmain owners).  More likely, Bitmain will just create a new S9, using old stock with new firmware, and sell it as their V11 - advertising the 20% improvement in hashrate/watt.  Only trick would be to keep all S9 miners from reflashing using the V11 firmware...

If the Bitmain does that, does it have to sign away the patent rights?

[NotFuzzyWarm]

ref my post https://bitcointalk.org/index.php?topic=3218166.msg33465262#msg33465262
IF Bitmain were to revise the physical logic inside the chips to use the Halongitosis's overt vs BM's covert AB protocols then yes they would be required to open all of their IP to the public. Don't see that happening.

[PeaMine]

ref my post https://bitcointalk.org/index.php?topic=3218166.msg33465262#msg33465262
IF Bitmain were to revise the physical logic inside the chips to use the Halongitosis's overt vs BM's covert AB protocols then yes they would be required to open all of their IP to the public. Don't see that happening.

Thanks NotFuzzyWarm! I always like your posts on here.
The thread seems to be off limits to me now?

[NotFuzzyWarm]

Just looked and that message usually means that the post or even entire thread thread was deleted by mods. It was probably in the Halong Dragon miner thread and the mods there have been rather free about removing disparaging posts and specifically any with the word 'Bitmain" in it....

Beyond that, both Biteme and Halonitosis's versions of AB rely on physical circuits inside of the ASIC chips. The chips themselves need to be changed. It is not just a change in firmware.

[2112]

Beyond that, both Biteme and Halonitosis's versions of AB rely on physical circuits inside of the ASIC chips. The chips themselves need to be changed. It is not just a change in firmware.

This is both true and false.

It is all true if you really believe that the particular chip really implements the AB.

On the other hand is it possible to write a firmware that drives the regular non-AB mining chips in such a way that they appear to produce results that look like made with the AB.

But the only way to actually verify the existence and support for AB is via reverse-engineering of the firmware or attaching the logic analyzer to the signals sent to the mining chips.

Edit: As far as reverse-engineering the firmware, the cost and complexity depends on the MCU. I believe the recent hardware has been using Xilinx Zynq chips which is a dual core ARM chip paired with the Artix FPGA logic slices. Reverse-engineering the FPGA programming bit-streams is nothing like disassembling the microprocessor's instruction ROMs, it is several orders of magnitude more cumbersome and expensive. In particular there are no tools available publicly for that task.

On the other hand writing a firmware that fakes the AB (provided that you have the source for non-AB firmware) is fairly trivial exercise.

[sidehack]

If I ever get ASIC samples and data from those guys (it's been discussed a bit), I'll drop you a line and figure out what to look for.

[NotFuzzyWarm]

@2112

This is both true and false.
It is all true if you really believe that the particular chip really implements the AB.
On the other hand is it possible to write a firmware that drives the regular non-AB mining chips in such a way that they appear to produce results that look like made with the AB.

Interesting...

Regarding BM's AB, everything I'd read is/was saying that it was a hardware implementation in the chips but able to be switched on/off. Considering it was never put to use, the ability to simulate (fake) AB by reprogramming the FPGA makes more sense as oppose to wasting die space on unused circuitry.

Intel's patent application definitely is a hardware implementation probably destined to become a Foundry IP block for use by whoever wants to pay the royalty.

As for Halong's  'open' version...
A) To me it is a blatant attempt to hijack other mfgr's IP by saying, "Sure. You can use our 1 patent for 'free' -- provided you open access to all of your IP and lift all vendor NDA's as well.

B) Their miners only work with a pool that implements the stratum AB switches. While writing/testing the T1's driver -ck has been emphatic on that. When he forced the firmware to connect to a non-AB pool only 1/4 of the hash's produced are valid. Now if they manage to convince most of the pools to support the new stratum code to use AB then yes the Halongitosis patent does become a rather large carrot to dangle even if it is a very malodorous one-- assuming the T1 miners eventually live up to their promised specs. So far the jury remains out on that point.

C) My feeling is that not only Intel but also Bitmain, Canaan, Bitfury et al will all pursue their own implementation of overt AB. Hell, with Samsung getting into the ASIC miner chip biz it would even make sense for them to also come up with a Foundry IP block for AB to drop into whoever wants it in their chips.

[2112]

Regarding BM's AB, everything I'd read is/was saying that it was a hardware implementation in the chips but able to be switched on/off. Considering it was never put to use, the ability to simulate (fake) AB by reprogramming the FPGA makes more sense as oppose to wasting die space on unused circuitry.

I read some reports from people reverse-engineering Bitmain's firmware. According to them the AB is not only switchable on/off but also allows variable levels of boosting.

Intel's patent application definitely is a hardware implementation probably destined to become a Foundry IP block for use by whoever wants to pay the royalty.

I wouldn't make that inference for any patent in the USA, probably even for most of the world. In my understanding the use of "device" abstraction is a method of strengthening the patent against counter-claims, not anything technical or technological.

As for Halong's  'open' version...
A) To me it is a blatant attempt to hijack other mfgr's IP by saying, "Sure. You can use our 1 patent for 'free' -- provided you open access to all of your IP and lift all vendor NDA's as well.

B) Their miners only work with a pool that implements the stratum AB switches. While writing/testing the T1's driver -ck has been emphatic on that. When he forced the firmware to connect to a non-AB pool only 1/4 of the hash's produced are valid. Now if they manage to convince most of the pools to support the new stratum code to use AB then yes the Halongitosis patent does become a rather large carrot to dangle even if it is a very malodorous one-- assuming the T1 miners eventually live up to their promised specs. So far the jury remains out on that point.

C) My feeling is that not only Intel but also Bitmain, Canaan, Bitfury et al will all pursue their own implementation of covert AB. Hell, with Samsung getting into the ASIC miner chip biz it would even make sense for them to also come up with a Foundry IP block for AB to drop into whoever wants it in their chips.

I think you are needlessly conflating two separate things:

1) legalistic maneuvering related to patents and their licensing

2) deception and disinformation tactics like "covert AB"

They are only very loosely related and untangling the two is the key to understanding the situation.

It seems like a lot of bitcoiners had fallen into believing their own bullshit and now have hard time returning back to reality.

I think the company 21e6 (now 21.co ?) was a good early example of what happens when you believe your own propaganda. Their "Bitcoin Computer" was utterly non-competitive because it included some complicated DRM logic to guarantee coinbase payment address. If you think that Intel is somehow above making such a strategically stupid mistake, all I have to do is to remind you about Itanium and their completely unfounded in reality claims about advances in compilation technology.

[NotFuzzyWarm]

I *do* find it odd that Intel makes specific mention in point 15 and later in the filing go into great detail about BTC mining. Possibly as a cover-all-bases approach and response to Halong?

I would think they are also targeting the myrid of other blockchain applications such as ledgers of Public Records. Don't forget that their FPGA division (formerly Altera?) is pushing their tech like crazy for blockchain processing.

re the Intel patent, in USPO filings I've done they are rather picky about differentiating between a physical device or assembly vs what would better be described as a Method or a Process.. Given heavy mention is made of it being a SoC along with:

Point-16. A system, comprising: a circuit board; a processor disposed in a first location of the circuit board; an off-chip logic device operatively coupled to the processor, disposed in a second location of the circuit board, wherein the off-chip logic device comprises: <description of the logic blocks and such>

Sounds like a description of a physical hardware/chip/SOC or such to me.

Back to the main point here, aside from the nature of Halong's patent my major problem with their approach is that their gear MUST be pointed at an AB-enabled pool or it just does not even connect. That is most definitely something that should have been clearly stated when they started pre-order sales.

Ya know. The not-so minor omission that unlike every other previous BTC miner made, you cannot connect them to any mining pool you want or even Core unless the pool/Core are patched to work with AB.

edit: wording change to par.1 to reflect further reading of the patent.

[sidehack]

Ya know. The not-so minor omission that unlike every other previous BTC miner made, you cannot connect them to any mining pool you want or even Core unless the pool/Core are patched to work with AB.

Yeah that part's kinda crap. Hopefully some more decent pools implement compatibility, now that CK and such have done most of the legwork.

[2112]

re the Intel patent, in USPO filings I've done they are rather picky about differentiating between a physical device or assembly vs what would better be described as a Method or a Process.. Given heavy mention is made of it being a SoC along with:

Point-16. A system, comprising: a circuit board; a processor disposed in a first location of the circuit board; an off-chip logic device operatively coupled to the processor, disposed in a second location of the circuit board, wherein the off-chip logic device comprises: <description of the logic blocks and such>

Sounds like a description of a physical hardware/chip/SOC or such to me.

IANAL in general, moreover IANA patent L, but I used to sit next one at the table during conferences and lunches.

The way they explained it to me that the history of US jurisprudence and legal precedents make it easier to defend a "device patent" against any sort of infringing "emulation" or alternative "embodiment" than vice-versa or any other combination in-between. Basically, if you think that your understanding of English grammar and Boolean algebra is sufficient to understand the patent, you are sorely mistaken. If your experience of patenting is only as far as getting it granted by the USPTO then you have less than half of the required knowledge.

Patent lawyers are so well paid, because they known the best combination of claims that's easiest to get approved and hardest to overturn in litigation.

[itzazkrit]

Can complain all we want about them not advertising that it works on certain pools, but don't you guys realize that the only reason they didn't was to hide AB implementation, which was obviously withheld for legal reasons?  Kinda hard for them to disclose that info ahead of time without letting cat out of the bag.  In due time, I would bet all the pools will support AB and this whole point will be moot.  How would you guys have handled it differently?

[sidehack]

Probably mention, without specifics, that the control code uses a non-standard stratum implementation that's currently only available at one pool but they're working with other pool operators to patch in support.

[NotFuzzyWarm]

Can complain all we want about them not advertising that it works on certain pools, but don't you guys realize that the only reason they didn't was to hide AB implementation, which was obviously withheld for legal reasons?  Kinda hard for them to disclose that info ahead of time without letting cat out of the bag.  In due time, I would bet all the pools will support AB and this whole point will be moot.  How would you guys have handled it differently?

By not following almost every page from "Scamming for Dummies" and being an honest and open company.
1) NOT doing pre-order sales at all
2) providing verifiable business contacts and company information
3) NOT implementing their version of AB as an all or nothing feature which would let the market decide if AB is a good thing or not. So far the T1 performance compared to the s9 and A841 is -- meh.

[itzazkrit]

By not following almost every page from "Scamming for Dummies" and being an honest and open company.
1) NOT doing pre-order sales at all
2) providing verifiable business contacts and company information
3) NOT implementing their version of AB as an all or nothing feature which would let the market decide if AB is a good thing or not. So far the T1 performance compared to the s9 and A841 is -- meh.

Are you saying that if you were in their shoes, had spent the so called millions on development, had working units in hand and a pretty good feeling about success rate of the design based on experience, that you wouldn't have tried to do presale?  As an Engineer, I hate pre-sale stuff as much as the next guy because in reality, at that point it isn't ready. And then you have a time limit your final R&D phases have to be restricted by.  I could probably argue both ways, sure business sense wise and the optics was not good, however, I can't say that if I were in their shoes that I wouldn't be one of the first to want to hawk my 'wares ASAP.  In the end its about the miner and not about the launch PR, or lack thereof.  I'd much rather them focus on getting the hardware right than trying to quiet down the peanut gallery.

[sidehack]

As both an engineer and a miner manufacturer, I have refused to take a single dollar in presales without having first distributed several working prototypes to well-established people for public critique.

Never ask for money without working units in hand, and if you have working units in hand never ask for money until they've been verified by independent parties.

Never send money to someone who hasn't had independent parties verify their working units.

[NotFuzzyWarm]

By not following almost every page from "Scamming for Dummies" and being an honest and open company.
1) NOT doing pre-order sales at all
2) providing verifiable business contacts and company information
3) NOT implementing their version of AB as an all or nothing feature which would let the market decide if AB is a good thing or not. So far the T1 performance compared to the s9 and A841 is -- meh.

Are you saying that if you were in their shoes, had spent the so called millions on development, had working units in hand and a pretty good feeling about success rate of the design based on experience, that you wouldn't have tried to do presale?  As an Engineer, I hate pre-sale stuff as much as the next guy because in reality, at that point it isn't ready. And then you have a time limit your final R&D phases have to be restricted by.  I could probably argue both ways, sure business sense wise and the optics was not good, however, I can't say that if I were in their shoes that I wouldn't be one of the first to want to hawk my 'wares ASAP.  In the end its about the miner and not about the launch PR, or lack thereof.  I'd much rather them focus on getting the hardware right than trying to quiet down the peanut gallery.

That is exactly what I am saying.
As you said, doing public pre-sales sets a huge milestone that MUST be met regardless of how the technology co-operates. ref the fallout from BFL's Monarchs and Bitmine.ch/AMT A1 Coincraft miners. Unfortunately input from the Engineering side of many companies - not only those involved in crypto - is ignored once the Marketing group is unleashed. In the crypto mining business pre-sales translates to anything from 'we have this idea...' (ala the GMO vaporware) to Halong's 'we have some engineering test rigs built and running' (but still a work in progress and not ready for Prime Time).

Much to their credit Canaan and yes even the 'evil' Bitmain never did that. They stay quiet while developing and once factory production trials start only then announce upcoming products.

In the beginning of their thread Halong specifically stated that they were fully funded all the way through to at least pre-production trial runs and "we have no need of further funding via pre-orders". Then went on with their spiel never to this day saying *why* they did pre-orders.