February 2022
Miner's Newsletter
Hey there and thanks for taking the time to read my first newsletter. This is all new to me so I welcome any and all feedback. That being said, let’s get into the mix.
Tons of things came out for the Bitcoin mining industry this year already! Things I am going to focus on this news letter are
The New Intel ASIC coming
Blockstream’s ASIC
Satoshi Roundtable Brief
And I will try to always leave a footnote on every newsletter that is a quick tip of sorts based on questions I got the month before. This month I am going to discuss why we don’t put multiple ASICs on the same workername for the pool.
INTEL ASIC INCOMING!!!
So first things first on this front, calm yourself. To my knowledge they have not even begun tapeout yet so this is likely 18 months out still minimum.
The intel chip is VERY interesting for a few reasons. Firstly, it is being done in conjunction with some long time friends of mine over at GRIID. If you don’t know the folks at GRIID are mostly based in Austin, Tx. Mike Hamilton is a good friend of mine and definitely has been mining for a long time. He has done alot of work in the space and I am definitely interested to see what he can do with Intel at his back. The cited specs are somewhat impressive (should edge out the Bitmain s19’s) but those are likely all based on simulation results. Wafer yields and therefore overall miner pricing are going to be a big deal here. GRIID will be taking 25% of the entire allotment so that leaves 75% to us plebs. I can say that the INTEL chip certainly will be robust and likely better from a reliability standpoint over the longer term. However, with the new s19XP coming out, and them being 18 months out from production, gives me pause for concern when it comes to will it be good enough. My gut reaction is, its probably going to take 2 or 3 iterations on this design to really dunk it.
Specs aside, there are many companies who will buy this device even if it is not completely blowing the latest and greatest chips out of the water. Many public companies are interested in buying only from American hardware manufacturers for a multitude of reasons that I won’t get into here. But if it is good enough, they will definitely sell them no doubt.
Blockstream’s ASIC
As you know, Blockstream acquired Spondoolies late last year in the anticipation of working on their own chip. If you are not and old salty dog to the mining industry, you might not recognize the name Spondoolies. So let me start there.
Spondoolies was a team out of Israel. CEO is still a friend of mine, named Guy Corem. Genius hardware developer. The CFO was a guy named Gadi. Both gentlemen are standup folks. The hardware was truly great. Their first really successful device imho was called an SP30. A 4U rackmount server. Pictured here
These things were tanks. Never needed repairs. Usually arrived on time in pristine condition. Also, they had a great UI to boot.
They were manufactured by FlexTronics in Israel to the best of my knowledge. Truly great hardware.
Now on the surface this makes the acquisition seem like a slam dunk. Here is the problem I personally see with it. The Spondoolies designs were all centered around 24nm process nodes (if i remember correctly). Nowadays we are well beyond that design paradigm using 3D gates and all kinds of insane tricks. So while Blockstream started with some great PCB design etc, it will be no easy task to turn that into something competitive by today’s standards. That isn’t to say that they can’t get there. However, this is likely a much longer play requiring a few iterations to get into something that is very competitive. But I am very excited to see the Spondoolies tech revived and hopefully Blockstream can make something that once again is very stout!
Satoshi Roundtable Brief
This past weekend was the Satoshi Roundtable #8. I have been fortunate enough to attend every roundtable to date. For the unaccustomed, this conference is an invite only event where industry leaders get together and problem solve/brainstorm new ideas for a weekend.
This year there were MANY miners and hardware producers. It was really nice. The topics of discussion revolved around ESG, Intel, public listings, and new sites coming online.
It was very bullish to me to see everyone still so happy about everything despite the price sliding the difficulty increasing. We have an industry of truly legendary miners now and it is a very good thing for the network!
Quick Tip : VARDIFF & WorkerNames
So this past week the homie @RoninMiner on twitter posted this tweet
That got me started down a path of explaining why this is NOT an optimal set up. You can find some of the response here.
But to explain this more in detail here is why this is.
How do machines talk to the pool? So technically a pool does NOT know your hashrate. The pool kind of “derives” and I use that term loosely, your hashrate by a calculation that looks something like this.
The pool then has a hashrate that should closely resemble the hashrate of the machine. Now how the pool negotiates the share difficulty is more interesting.
Right when your miner connects to the pool, the miner exchanges the first TCP packet. We call this the Extranonce. That tells the pool things like your workername etc. Right after that the pool starts sending a ton of low difficulty shares that the miner will be able to crunch these shares quickly and effectively get TONS of jobs within the first minute or two of connecting. So your mine is kind of DDOS’ing the pool for that short time. During this time the pool is trying to figure out the optimal share difficulty to send to that specific miner, based on its interpretation of hashrate. So it will throttle the difficulty of the shares based on the hash power of that device. This whole methodology is called VARDIFF.
Now the reason we don’t use two miners on one workername is basically because now the pool thinks there is a badass machine that has 200TH/s instead of 100TH/s and it will send more difficult shares to BOTH machines. If the machines aren’t able to solve this share (usually within 20 seconds or so) then it will be a stale share. The pool might also call it a rejected share. This means your miners are wasting hash power.
**DISCLAIMER: Pools might use different metrics than what I described above. Some may prefer 25s share computation etc.
**EDIT**: After Nick H from Luxor pinged me, this seems to be a somewhat outdated explanation of VARDIFF. In the original pool days, this is indeed how it works. However, now it seems that pools actually are logging each individual TCP Socket connection and using VARDIFF to average the hashrate across all sockets that share the same workername. It is still not ideal as one miner may be hashing slower than the others and therefore your larger devices might get sub optimal share difficulty so the logic is still sound. Thanks to @hash_bender and @ausenhus on twitter for pointing that out!
Thanks for taking the time to read my first Newsletter. Hopefully it didn’t put you to sleep! If you folks like it, I will keep trying to do them.
nice overview.