DragonMint T1 16TH/S halongmining.com

[philipma1957]

Still waiting for demo to review it.

I was looking for five units 1 then 4
Hoping for contact this week for demo


(Moderator's note: This post was edited by frodocooper to remove an unnecessary quote.)

[QuestionQuest]

Still waiting for demo to review it.

I was looking for five units 1 then 4
Hoping for contact this week for demo

Awaiting your details --- in 4-6 weeks the regular shipping should be done.

Anyways the final countdown

[HagssFIN]

After all these posts and weird things in this thread,
I will be so happy when people actually start receiving these units.

[NODEhaven]

So is the extra 2 TH/s in the DragonMint only due to the ASICBoost when compared to the S9?  I am interested in seeing tests of these machines when they come in.

[-ck]

So is the extra 2 TH/s in the DragonMint only due to the ASICBoost when compared to the S9?  I am interested in seeing tests of these machines when they come in.

Asicboost does not in itself increase hashrate, it decreases power usage. If you use that advantage by making your silicon fixed to only work with asicboost, it affords you extra space on the chip which means you can put more hashing units on it. Strictly speaking the answer is no, but yes

[NODEhaven]

So is the extra 2 TH/s in the DragonMint only due to the ASICBoost when compared to the S9?  I am interested in seeing tests of these machines when they come in.

Asicboost does not in itself increase hashrate, it decreases power usage. If you use that advantage by making your silicon fixed to only work with asicboost, it affords you extra space on the chip which means you can put more hashing units on it. Strictly speaking the answer is no, but yes

Thanks, that answers the whole $30 million debate.  They might have spent money elsewhere.  We are budgeted $16 million for 7nm BTC ASIC.  I do give Halong a lot of credit for making all of this happen.  If all the miners check out it is a big feat.

The patents are little weird, as the chip mask at the foundry is the patent.  It is a hard patent.

Edit:  
I looked into this BPDL license as it pertains to what I am working on.

This may be a big issue for any manufacturer wanting to use Halong version of Overt ASICBoost.

From: https://blog.bitmex.com/defensive-patent-licence/

"The second loophole is fixed by adding a new term to the licence. This is a requirement that specifies that the licence will be revoked if members use any patent licensed by a third party, when such a licence “is or is likely to be” enforced and restricted from use by another BDPL member. This should prevent a scenario such as the one illustrated above."

Correct me if I am wrong, that means that if I build a system that I have to rely on either the other BPDL license holders and their IP or open source.  That could be huge issue in the semiconductor space as proprietary IP is everywhere and means you can't use it if you don't get a license for everyone in the BPDL.

This may require everyone to come up a patent beating ASICBoost type solution or some alternate ways to optimize, which luckily there are.

[picominer]

Has Halong released their version of cgminer that supports overt asicboost? I'm curious to see how the finding the colliding work items has been implemented.

[-ck]

Has Halong released their version of cgminer that supports overt asicboost? I'm curious to see how the finding the colliding work items has been implemented.

Not yet, no. They'll have 30 days from delivering hardware to release their software under the terms of the GPL (provided someone asks for the code who has the hardware.) If they don't release it after that time, I will myself anyway since I've been working on their driver for the last 2 weeks.

[MyRig]

Thank you CK!  Yes, it will be released under GPL request, however, it is still being prepared for that.  A bit more final step announcements are being prepared by Halong Mining and once that is officially out, then you get to see the naked dragon's underbelly.

and phil, please email your shipping info to usual email or leave it in PM please.

Has Halong released their version of cgminer that supports overt asicboost? I'm curious to see how the finding the colliding work items has been implemented.

Not yet, no. They'll have 30 days from delivering hardware to release their software under the terms of the GPL (provided someone asks for the code who has the hardware.) If they don't release it after that time, I will myself anyway since I've been working on their driver for the last 2 weeks.

[philipma1957]

Thank you CK!  Yes, it will be released under GPL request, however, it is still being prepared for that.  A bit more final step announcements are being prepared by Halong Mining and once that is officially out, then you get to see the naked dragon's underbelly.

and phil, please email your shipping info to usual email or leave it in PM please.

Has Halong released their version of cgminer that supports overt asicboost? I'm curious to see how the finding the colliding work items has been implemented.

Not yet, no. They'll have 30 days from delivering hardware to release their software under the terms of the GPL (provided someone asks for the code who has the hardware.) If they don't release it after that time, I will myself anyway since I've been working on their driver for the last 2 weeks.

Thank you

pm sent

[picominer]

Thank you CK!  Yes, it will be released under GPL request, however, it is still being prepared for that.  A bit more final step announcements are being prepared by Halong Mining and once that is officially out, then you get to see the naked dragon's underbelly.

and phil, please email your shipping info to usual email or leave it in PM please.

Has Halong released their version of cgminer that supports overt asicboost? I'm curious to see how the finding the colliding work items has been implemented.

Not yet, no. They'll have 30 days from delivering hardware to release their software under the terms of the GPL (provided someone asks for the code who has the hardware.) If they don't release it after that time, I will myself anyway since I've been working on their driver for the last 2 weeks.

Thanks CK and MyRig!

[score1more]

what's the tl;dr answer to March shipping of these units?  Weren't the first batch promised to ship by March?

[-ck]

what's the tl;dr answer to March shipping of these units?  Weren't the first batch promised to ship by March?

Well I'm not privy to what's going on in the factory but I also never sign any NDAs so I can tell you there is a mad scramble to get hardware out ASAP and they've been leaning hard on me to try and make sure the first firmware it's being released with works as reliably and solidly as possible. There will definitely be firmware updates in the future but for now things seem to be pretty close. I got involved very late in the piece and I've worked very hard in the last couple of weeks but there are limits to my time and just what I can achieve to get working code out.

One thing I have managed to achieve is to add tuning to the driver. The driver has a 5 stage tuning process to try and set the lowest stable voltage possible for minimal power consumption and then it tunes each chip individually to the highest frequency it will tolerate to try and optimise the hashrate. It also has a PID type fan controller to try and maintain optimal temperature with minimal fanspeed. There are a few emergency things included too in case of fan failure, chain dropouts etc.

[philipma1957]

what's the tl;dr answer to March shipping of these units?  Weren't the first batch promised to ship by March?

Well I'm not privy to what's going on in the factory but I also never sign any NDAs so I can tell you there is a mad scramble to get hardware out ASAP and they've been leaning hard on me to try and make sure the first firmware it's being released with works as reliably and solidly as possible. There will definitely be firmware updates in the future but for now things seem to be pretty close. I got involved very late in the piece and I've worked very hard in the last couple of weeks but there are limits to my time and just what I can achieve to get working code out.

One thing I have managed to achieve is to add tuning to the driver. The driver has a 5 stage tuning process to try and set the lowest stable voltage possible for minimal power consumption and then it tunes each chip individually to the highest frequency it will tolerate to try and optimise the hashrate. It also has a PID type fan controller to try and maintain optimal temperature with minimal fanspeed. There are a few emergency things included too in case of fan failure, chain dropouts etc.

So auto tune  like the s-9  with no option to down clock  or over clock?

Or do I misunderstand you?

Btw in no way  am I  voting for or against auto tune

[-ck]

So auto tune  like the s-9  with no option to down clock  or over clock?

Or do I misunderstand you?

Btw in no way  am I  voting for or against auto tune

There is a plan to implement options for tuning frequency or for power or efficiency but not in the first firmware. From what I've seen of the web interface it will be pretty sparse to begin with. I also believe that first silicon is probably going to be less efficient while they finalise manufacturing.

[-ck]

By the way anyone mining on ckpool.org would have noticed a transient rise in hashrate there today. This is burn testing the first batch of hardware.

[HagssFIN]

Cool.
That's a nice thing from Halong supporting the ckpool.org by having it as a factory default for burn-in testing purpose,
similar to what Canaan is doing by having kano.is as the factory default in their burn-in test process.

[Shazam!!!]

Cool.
That's a nice thing from Halong supporting the ckpool.org by having it as a factory default for burn-in testing purpose,
similar to what Canaan is doing by having kano.is as the factory default in their burn-in test process.

Totally agree! Very cool indeed! Big props to Halong for supporting -ck's pool. It's truly Great to see these newer companies give
something back to the community.

[-ck]

Totally agree! Very cool indeed! Big props to Halong for supporting -ck's pool. It's truly Great to see these newer companies give
something back to the community.

It is indeed the community since there's no fee on my pool and only the miners on ckpool.org stand to benefit.  Not directing more hashpower at a large pool is great for decentralisation.

[Steamtyme]

It only makes sense for them to use a smaller pool. CKPOOL would be the obvious choice as they already have a working relationship.

I don't know if there is extra data they are looking for besides the burn in rates for their machines. If they are why not use the pool run by the person writing code for you.

It's not like their were going to go and point them at Antpool; and start mining for the competition