Intel Building New 7nm Chip.... to mine Bitcoin

legcramp

[H]F Junkie
Joined
Aug 16, 2004
Messages
12,401
https://www.techpowerup.com/291142/...tomer-a-usd-3-3-billion-crypto-mining-startup

Just a few days ago, we reported that Intel is preparing to unveil the company's first application-specific integrated circuit (ASIC) dedicated to mining cryptocurrency. To be more specific, Intel plans to show off its "Bonanza Mine" ASIC at the 2022 ISSCC Conference, describing the chip as "ultra low-voltage energy-efficient Bitcoin mining ASIC." We have yet to see how this competes with other industry-made ASICs like the ones from Bitmain. However, it seems like the startup company GRIID, valued at around $3.3 billion, thinks that the Bonanza Mine ASIC is the right choice and has entered a definitive supply agreement with Intel.

According to the S-4 filing, GRIID has "entered into a definitive supply contract with Intel to provide ASICs that we expect to fuel our growth. The initial order will supply units to be delivered in 2022 and GRIID will have access to a significant share of Intel's future production volumes." There are a few other mentions of Intel in the document, and you can see another exciting tidbit below.
UVIMHn8lgMNPxbp4_thm.jpg
GRIID S-4 FilingOn September 8, 2021, GRIID entered to a supply agreement (the "Intel Supply Agreement") pursuant to which GRIID may purchase Intel-designed BZM2 ASICs. The Intel Supply Agreement is for an initial four-year term and will automatically renew thereafter for one period unless either party provides at least 90 days' notice prior to the end of the initial four-year term. The Intel Supply Agreement provides GRIID with fixed pricing for the BZM2 ASICs for all orders placed prior to May 2023. In addition, subject to certain conditions, GRIID will be entitled to purchase from Intel at least 25% of all qualified Intel-designed ASICs through approximately May 2025.
 
Price these right and pump them out in abundance, Intel will make billions while solving the global GPU shortage (and price gouging).
Hope this works out!
Not sure gouging would be right, they just need to provide a price/performance ratio better than AMD and NVidia’s AIB’s have set. Technically Intel is positioning itself as the bargain alternative here. If it forces AIB’s to drop pricing to match then I’m not sure who’s loosing here. But Intel will have an advantage here as they don’t have to worry about the 25% tariff as their 7nm facilities are in US, and Germany so none of it will touch China. Could be a very big advantage for them especially now that most coin farms have moved state side.
 
Price these right and pump them out in abundance, Intel will make billions while solving the global GPU shortage (and price gouging).
Hope this works out!

Since nobody mines bitcoin with GPU's how exactly will this impact the GPU shortage in any way? Were Intel to start pumping out low cost ETH asics this would be a thing.
 
Not sure gouging would be right, they just need to provide a price/performance ratio better than AMD and NVidia’s AIB’s have set. Technically Intel is positioning itself as the bargain alternative here. If it forces AIB’s to drop pricing to match then I’m not sure who’s loosing here. But Intel will have an advantage here as they don’t have to worry about the 25% tariff as their 7nm facilities are in US, and Germany so none of it will touch China. Could be a very big advantage for them especially now that most coin farms have moved state side.

There's definitely gouging going on.
I got lucky and was able to snag an MSI 3080 Ventus at MSRP $750 shortly after launch.
Today, that same SKU now has an inflated MSRP of almost $1100.
 
There's definitely gouging going on.
I got lucky and was able to snag an MSI 3080 Ventus at MSRP $750 shortly after launch.
Today, that same SKU now has an inflated MSRP of almost $1100.
Yeah the AIB’s are all posting record numbers. They’ve jumped from like 8% to the low 20’s. But the tariffs did add $187 to your $750 card, tack on freight costs which have more than doubled that means the AIB has tacked on between $50-$75 to their selling cost before it moves up channel.

It’s just a shit show right now.
 
Bitcoin uses a CPU based algorithm, using a GPU to mine for Bitcoin is terribly inefficient. People use GPUs to mine various alt coins that they sell and trade up into bitcoins.

Ah, I understand now...thank you! Always thought best BTC mining results was with the strongest GPUs.
 
The main take that I am getting from this is these new Intel chips are more efficient then the current market solutions.
This means if they price them right and have availability they will sell well.
 
Bitcoin uses a CPU based algorithm, using a GPU to mine for Bitcoin is terribly inefficient. People use GPUs to mine various alt coins that they sell and trade up into bitcoins.
Btc is mined better on gpus then cpus. Its a relatively simple algo so Asics have been around on it for a long time (and drastically beat gpus)
 
Ah, I understand now...thank you! Always thought best BTC mining results was with the strongest GPUs.

This was true back in 2012-2013 when ASICs hadn't taken off yet and the difficulty was low. If you try to mine Bitcoin with a GPU now, you'd just be wasting money. If you were on a mining pool, you'd maybe see a couple fractions of a couple cents a month at best.

Bitcoin and other SHA256 clones are entirely in the realm of hideously expensive ASICs now. GPUs are used to mine Etherium and other non-SHA256 coins, and have been for years now.
 
The point is if you have a stack of GPUs and want to turn a profit with them then Bitcoin is not going to be where you are looking with them.
Definitely correct, most gpus are on eth directly or not. Just figured I would clarify as the op is directly talking about a specific type of asic made by intel on 7nm to mine btc. Which honestly is a questionable venture as asic companies have historically been fairly scummy companies using that to profit off it.

Asic mining has been a race to who has the lowest power and operational costs on the largest mine. Its not like eth mining where you just have to get in at the right time with decent GPU.
 
This was true back in 2012-2013 when ASICs hadn't taken off yet and the difficulty was low. If you try to mine Bitcoin with a GPU now, you'd just be wasting money. If you were on a mining pool, you'd maybe see a couple fractions of a couple cents a month at best.

Bitcoin and other SHA256 clones are entirely in the realm of hideously expensive ASICs now. GPUs are used to mine Etherium and other non-SHA256 coins, and have been for years now.
I'm making around $100 a month after electricity running a single 3080 on NiceHash. Not sure where you're getting that info.
 
He said "you'd maybe see a couple fractions of a couple cents a month if you were on a mining pool"

That's false.
Not if that pool is mining Bitcoin directly, so your both correct depending on how you want to look at it.

Regardless Intel coming to market with a high end ASIC unit I’m sure will make all sorts of waves. Bitcoin gets headlines but communications controllers are where Intel is going to make their big dollars with these.

I’m hoping part of these new ASIC chips means a new line of FPGA boards to develop for them. The existing Cyclone series chips are all 5+ year old designs and an update would be more than welcome.
 
Last edited:
He said "you'd maybe see a couple fractions of a couple cents a month if you were on a mining pool"

That's false.
It's not false if you're trying to mine bitcoin directly with a GPU. You would spend more in electricity than you would get back in crypto even with a pool which is what GiGaBiTe is talking about. If you look at mining calculator pages these days you can't even find their stats for SHA256 because its pointless.
 
He said "you'd maybe see a couple fractions of a couple cents a month if you were on a mining pool"

That's false.

I was referring to mining Bitcoin directly with a GPU in a pool. Yes, you will only get a few fractions of a cent per month. Even the most expensive pimped out GPUs cannot even come close to one ASIC box from even years ago.

Back in 2013/2014 when I was mining with my HD5870, A single block erupter USB stick could match it in performance. Higher end ASICs would curb stomp it, and the distance has only gotten farther as ASICs have gotten better.
 
If the Eth and other mining pools are paying out in BTC, and these ASICs make it enticing to mine BTC directly, rather than mining Eth or other alt-coins to convert them to BTC, wouldn't this be a more streamlined approach to mining? Is that enough to move people from massive GPU farms mining alt-coins to massive ASIC farms mining BTC?
 
Not if that pool is mining Bitcoin directly, so your both correct depending on how you want to look at it.

Regardless Intel coming to market with a high end ASIC unit I’m sure will make all sorts of waves. Bitcoin gets headlines but communications controllers are where Intel is going to make their big dollars with these.

I’m hoping part of these new ASIC chips means a new line of FPGA boards to develop for them. The existing Cyclone series chips are all 5+ year old designs and an update would be more than welcome.
I've been wanting to hear more about FPGA for such a long time. I feel like someone out there is using them to mine every algo out there.
Incidentally, I am off the eth train now too due to difficulty rises from what I assume are ASICs. There are plenty of ASIC resistant algos still out there to be had. Profit might be slightly lower now, but you know how crypto works...

I mined over half a million doge back in the day in 3 months with a couple 6870s.
 
Going to be some interesting taxes for those who think they're mining btc directly/pool mining with nicehash and not realizing they are selling their hash power to the highest bidder and being paid in BTC. And you're being paid in BTC for two reasons only, it has super low transaction fees and it's stableish. If you were paid in eth, unless you're pulling in $40-50 a day you'd be eating almost as much as taxes (income and capital gains) will with gas fees.
 
I've been wanting to hear more about FPGA for such a long time. I feel like someone out there is using them to mine every algo out there.
Incidentally, I am off the eth train now too due to difficulty rises from what I assume are ASICs. There are plenty of ASIC resistant algos still out there to be had. Profit might be slightly lower now, but you know how crypto works...

I mined over half a million doge back in the day in 3 months with a couple 6870s.
Im pretty sure the first move past GPU bitcoin mining was custom FPGA's before people started making ASIC's.
 
I’m hoping part of these new ASIC chips means a new line of FPGA boards to develop for them. The existing Cyclone series chips are all 5+ year old designs and an update would be more than welcome.
What do you mean? Why would new ASICs imply new line of FPGAs? You can take any FPGA you want (assuming it's big enough) and implement your RTL. I suppose they do release new FPGAs every once in a while for a new node, but that's not directly tied to any particular ASIC right? Cyclone is their budget FPGA's? Budget FPGAs usually dont hop onto new nodes, they're usually using older nodes anyways.
 
I'm making around $100 a month after electricity running a single 3080 on NiceHash. Not sure where you're getting that info.
You are not mining Bitcoin your are likely mining an alt coin that someone is paying nicehash bitcoin to mine.
 
  • Like
Reactions: travm
like this
What do you mean? Why would new ASICs imply new line of FPGAs? You can take any FPGA you want (assuming it's big enough) and implement your RTL. I suppose they do release new FPGAs every once in a while for a new node, but that's not directly tied to any particular ASIC right? Cyclone is their budget FPGA's? Budget FPGAs usually dont hop onto new nodes, they're usually using older nodes anyways.
It's just that they have more or less kept their Cyclone and eASIC releases somewhat inline, and yes both are currently using either a 45nm or 28nm process depending on which side of the sku you are ordering from. But these ASIC chips being built out on a new modern process doesn't really have an appropriate FPGA (from Intel) to test your development from, the gap in capabilities would be too wide to realistically test any design. I find it hard to believe that Intel would sell an updated ASIC chip but then require you to reach out to a 3'rd party to order the necessary hardware to test your Intel designs on.
 
Not sure if you've seen this, but they already have a buyer.

https://bitcoinmagazine.com/business/intel-closes-sale-for-its-upcoming-bitcoin-mining-chip

What I'm not sure about is if they are only doing the chip, or the entire unit? They have the resources to either or.

As someone in a startup for a btc mining company, I'm very curious about these. Price, hashrate and the joules per terahash is of concern.

Bitmain and MicroBT needs some American competition!
 
I hope this works out. Miners would get dedicted system and GPU prices would normalize again.
Your logic is somewhat flawed, Bitcoin miners and other ASIC centric coins will get access to newer faster lower power draw chips. The other 50 GPU based coins that are currently out there will continue doing what they do completely independent of this.
 
But these ASIC chips being built out on a new modern process doesn't really have an appropriate FPGA (from Intel) to test your development from, the gap in capabilities would be too wide to realistically test any design. I find it hard to believe that Intel would sell an updated ASIC chip but then require you to reach out to a 3'rd party to order the necessary hardware to test your Intel designs on.
Hmm... I think I'm not understanding something about crypto ASIC chips. Why would you EVER have an FPGA chip "appropriate" to the asic come out? For example, Intel releases CPU's, there is no FPGA that coorelates to that CPU. Intel releases network chips and various other chipsets. None of these have FPGAs that coorelate to those chips. They might share the same node process technology, but thats about it.
 
Hmm... I think I'm not understanding something about crypto ASIC chips. Why would you EVER have an FPGA chip "appropriate" to the asic come out? For example, Intel releases CPU's, there is no FPGA that coorelates to that CPU. Intel releases network chips and various other chipsets. None of these have FPGAs that coorelate to those chips. They might share the same node process technology, but thats about it.
Especially when a btc asic is MUCH simpler then making a top of the line fpga
 
Hmm... I think I'm not understanding something about crypto ASIC chips. Why would you EVER have an FPGA chip "appropriate" to the asic come out? For example, Intel releases CPU's, there is no FPGA that coorelates to that CPU. Intel releases network chips and various other chipsets. None of these have FPGAs that coorelate to those chips. They might share the same node process technology, but thats about it.
ASIC chips are generally printed as single-purpose chips, so a customer will design the ASIC unit and submit their design to the manufacturer for printing the most common ASIC chips have simple predefined layers, and how they are customized is by connecting or breaking connections between the layers and the transistors on said layers during the metallization stage of manufacturing and by joining and breaking certain switch connections you get the desired functions. So they are fully custom designs made to order, but to design an ASIC chip you generally want to be able to test it first and that is where FPGA's come in so you can build out your circuit on the FPGA to test and validate the design before you start submitting orders.

When Intel launched the eASIC N3X, and N5X chips they launched the Cyclone 5 and 10 FPGA's as part of the development suite.

So here Intel has announced the BZM1 ASIC on their 7nm process which they have built in-house to sell to bitcoin miners hoping to rival the S19j or M30S++ chips that currently lead the market, but that means Intel has developed a new ASIC silicon on their 7nm process, their existing ASIC solutions are either 45nm or 28nm and when they launched they were launched alongside a similar FPGA being the Cyclone 5 and 10 series. So here they have a potentially new member of the eASIC lineup, but no FPGA to do testing and validation on.

So as it currently stands Intel doesn't need a new FPGA to go alongside those BZM1 and later the BZM2 chips, but if they want to add this ASIC silicon to the eASIC lineup they will have to launch a development kit for it and that means a new FPGA.

Currently, there is a rapidly growing need for ASIC chips for cellular towers, networking equipment, and other communication centric devices as general-purpose chips are getting too hot and requiring too much juice, so the big L1 providers are looking towards custom silicon to cut down on power and heat requirements as throughput requirements rapidly increase. Intel knows this and I doubt they plan on leaving that market alone, given that 2 of the new facilities they have under construction are slated for 3'rd party fabrication it would make sense that their ASIC lineup is part of those facilities product list, so again if they are going to offer a new ASIC lineup they will need to offer a new FPGA as part of its development suite.

If they aren't going to offer up a new ASIC lineup then they are just leaving billions on the table for somebody else to gobble up.
 
So as it currently stands Intel doesn't need a new FPGA to go alongside those BZM1 and later the BZM2 chips, but if they want to add this ASIC silicon to the eASIC lineup they will have to launch a development kit for it and that means a new FPGA.
I think I dont understand this. What does their eASIC lineup have to do with this announcement? (I'm also not familiar with eASIC btw, something between an FPGA and standard cell design? So a sea of gates and you burn the routing you want in? Same tools as FPGA? HDL input -> synthesized/mapped/routed bitstream output?).

How would they add this to the eASIC line? As HDL source code available for customers to implement on the eASIC? As hardened logic inside the eASIC? As some sort of black-box encrypted module that designers can interface with? I thought they were just creating a bitcoin chip that they would sell to customers and thats it.
 
I think I dont understand this. What does their eASIC lineup have to do with this announcement? (I'm also not familiar with eASIC btw, something between an FPGA and standard cell design? So a sea of gates and you burn the routing you want in? Same tools as FPGA? HDL input -> synthesized/mapped/routed bitstream output?).

How would they add this to the eASIC line? As HDL source code available for customers to implement on the eASIC? As hardened logic inside the eASIC? As some sort of black-box encrypted module that designers can interface with? I thought they were just creating a bitcoin chip that they would sell to customers and thats it.
The eASIC lineup is Intels ASIC computer units. They sell a lot of them to car manufacturers, cellular providers, L3 network switches. Lots of industrial applications.

How it relates to this announcement is to have developed these Bitcoin mining ASIC units they would have had to develop a new line of ASIC compute units. So this is very likely a precursor to an announcement later this year about a new lineup of ASICs, which pairs into their recent announcement for their two new fabs that will be taking customer orders for custom silicon on their latest design nodes when those go live in 2023.

My interest comes into play with the hope that this means a refresh to the CycloneV FPGA’s which they launched in 2013 and are the industry standard for FPGA development. I’m working on a few things that it’s just under powered for.
 
Last edited:
The eASIC lineup is Intels ASIC computer units. They sell a lot of them to car manufacturers, cellular providers, L3 network switches. Lots of industrial applications.

How it relates to this announcement is to have developed these Bitcoin mining ASIC units they would have had to develop a new line of ASIC compute units. So this is very likely a precursor to an announceme
lol im so lost i give up.

I've mostly worked on Lattice Semiconductor Mach XO3 and XO2 small FPGAs. Did some work on Xilinx Spartan 6 and Xilinz Zynq 7010. Have never used Altera/Intel.
 
FPGA's are under the radar for good reasons, but there's tens of thousands out there mining away. I have some mining eth currently. Others are not designed with high memory bandwidth in mind, and are much more profitable on other core-intensive shitcoins.

Eth hashrate growth has been completely linear (sans the china drop off) for the last year. Asics are nothing new, and there's no sudden inrush of asics hitting the network. It's still a mix of gpu's and asics. I personally know of someone currently deploying ~1.2TH/s of gpu's, there's plenty of others doing the same Im sure.
View attachment 437450
We have cheap subsidized power and a large solar array. Accounting has been weighing the options on this for a while.
 
Back
Top