the amount of memory has been more than doubled from 144 MB to 300 MB based on some set of experimental parameters though the reference implementation uses over 500 MB.
One of the core benefits of the Equihash algorithm, is its memory hard property for increased ASIC resistance… In order to increase Aion PoW algorithm’s resistance to future ASIC designs… The implementation of 210,9 as the parameters set for Aion’s Equihash2109, doubles the amount of memory required. Aion Blog (Feb 2018)
ASICs will be apart of the long term vision… 4 months to release the Equihash design and at the end of 6 months, have several solutions on how to keep hardware decentralized so the community can decide how to proceed.
So in summary: Aion selected equihash and tweaked the parameters for improved Asic resistance, then partnered with a chip company to prototype an FPGA miner, and the chip company now seeks funding from the Zcash foundation to develop an ASIC design for equihash. Meanwhile Aion will replace its PoW with a delegated proof of stake system.
It will be really interesting to see how this develops. I think coin devs working with hardware devs is the future. I cant wait to see the layout inside this thing. I will be exceptionally interested in getting my paws on one of these.
But it does imply that equihash really is not the algo for asic resistance. I’m surprised this was not caught earlier by the foundation. Yes I think this is more the foundations responsibility than zookos. It really seems Zooko had the best intentions and knew his limits so passed oversight of this sort of thing to the foundation… This is just my opinion though. The structure is still being worked out.
Also if anyone gets an ASIC, when it stops being profitable for you, hit me up I would like to take it apart, I will buy it from you obviously. I have a feeling an oscilloscope and some differential power analysis could explain a lot about the inner workings. This is what I do for a day job. Red team hardware testing for a defence company - FIPS 140-3 mainly, and crypto software RCE/exploit dev. [mistfpga is not my day job.]
If you come across any more info like this please post it here.
Hurm, seeing as reverse engineering hardware is legal in the UK if you do it for compatibility reasons, could I get a grant for this? heh. (jk)
edit: The more I read up on this the more it seems these people would know the most and up-to-date information about the interaction between equihash and hardware… It is probably worth reaching out to them…
The ePIC Aion partnership will result in the first open source implementation of Equihash on an FPGA (Field-programmable gate array), producing a 10x efficiency gain over a Graphic Processing Unit (GPU), resulting in a more secure, decentralized, and scalable processing network.
I’am curious. How is this achieved in direct competiton/compare to GPU’s? More Secure, decentralized, more scalalable processing? I admit i do not know much about FPGAs, so i’am always eager to learn something or hear an expert explaining things.
It says it is a 10% efficiency gain, not a 10x hashrate gain + 10x less power. We will have to wait an see, but it probably does 1000 sols for 25 watts. but costs 10x as much. This is normally how FPGA’s play out. We will have to see though. I think volta does 1000 for 100 watts. I get ~1000 for ~250 watts on my 1070’s
Some of the advantages of fpgas over asics.
FPGA’s can be bought by anyone from a number of resellers worldwide.
There are a lot of different makes of FPGA.
Bitstreams are the same as gpu miners like claymore.
Any number of companies can attach a fpga to a board, or you could try it yourself with a skillet (although the cheapest method is to do a night course in electronics and use their machine)
non-programmed FPGA’s face no import or export restrictions(like gpus). Cryptographic hardware is one of the most restricted import/export. it is an easy point of failure for asics.
FPGA’s can be reprogrammed to a much greater extent and cope with some protocol changes. (although not all changes, and some changes might make your fpga not work for the algo)
cpus and gpus are generally more versatile. by this I mean a fpga is more likely not to have redundant hardware, like 8gb of ram, or the fpga has ddr3 ram rather than hbm2 or gddr5x, etc. I am pretty sure the idea with zcash is to work on as much commodity hardware as possible. FPGA’s are on the line of commodity hardware.
once the designs are out, anyone can make one or get one made for them.
This probably wont obsolete GPU’s, bitcoin is a really good example of this. FPGA’s and GPU’s mined side by side until asics hit. ASIC’s pushed them both out. but equihash is a different algo, so who knows.
Thx for the explainations. I didn’t want to have it compared to asics to avoid whatever non technical discussion again. I’am fine with it of course, just hoping that this topic stays technical and on topic.
I heave readed of course here and there something about FPGAs, not that much to be honest as it always sounded a bit too far in future or not accessable to a guy like me, having just basic technical knowledge.
However, like everything new, it’s interesting of course.
Just to ask again some things about it:
The last time i readed a bit more about them, some experts at bitcointalk had discussed some private FPGAs projects and i think to remember that for a private person “only 3 or 4” producers are selling low quantities and to private consumers FPGAs. Is this correct or do i remember that wrong?
Personally like everything related when it comes to mining hardware i think the road would be the same as with GPU’s. They would be pretty fast sold out. Do you think there would be enough supply to have it indeed available to everybody wanting some?
I hope you do not mind if i ask some other things as well, but it seems right now you have good knowledge and can explain some things.
Where does the Bitmain Sophon Processor fits in? https://sophon.ai/product/introduce/bm1680.html
(Tensor Computing Processor, BM1680Sophon BM1680 is customized AI ASIC for tensor computing acceleration. It is designed for both deep learning training and inference of neural networks like CNN, RNN, DNN etc) Peak performance 2TFlops, Data precision FP32, Internal memory size 32MB, Average power consumption 25W.
I have somehow the feeling this is not a pure Asic chip or? Some kind of hyprid or what is this Sophon chip exactly? From what i can see on their webpage they use these Chip as well in their Deep Learning Accelerating Card SC1+
Thx in advance for everybody that can technically easy explain something.
I forgot to add this information to my previous post:
PASCAL A1 ASIC,
Claims are possible Algos to mine with this miner: SHA256, Scrypt, X11, Qubit, Quark, Ethash, Blake2b,
Equihash, Cryptonite, Lyra, Neoscript, Groestl, Pascal, Lbry, Black14r
Now my first thought when i saw this was, this is a scam, but on reading furthr i found 2 things interesting:
first the Algos mentioned above, exactly Algos we have now new released Asics: Ethash, Equihash, Cryptonite Pascal, Lbry, Blake*,
2nd thing, the mentioned performance: Hash Rate (±5%) ETH:520MH/s, Zcash:10000 Sol/s. Ok the hashrate for ETH is a bit off compared to E3, but the Zcash one fits perfectly that of the Z9.
3rd thing, Installed Memory 32GB
I have no idea if this project is real, a scam, or just that a project. However, the 10.000 Sol/s catched my attention as we now know it’s possible. Could it be that these new Asics use new chips like that Sophon Chip, something similar that makes them multi algo miners? Or are they even hybrids btw an Asic and a FPGA?
P.S.: If it doesn’t fit the topic, sorry, just move it away, just postd it hear as i forgot it to mention with the Sophon BM1680 chip above.
That being said, the purpose of the ZCash Grant would encompass an Equihash (200,9) design and for research into how we can incorporate ASICs into the ecosystem while maintaining decentralization without compromise of privacy. Our team has brainstormed a few ideas this past week which will require further research and analysis with the end goal to present the solutions to the community. We strongly believe ASICs offer performance and efficiency over GPUs but also understand the risk of centralization. The ZCash foundation ultimately has control over our research and if the community comes to agreement on one of our findings, we hope ePIC would be apart of creating this ASIC for the foundation.
I thought the proposal/offer/whatever was for FPGA, why i read Asics? Maybe my English is not good enough, but i can’t find the word FPGA anywhere or is FPGA nowadays the same as Asic? What do i miss?
Start off with providing a base Equihash digital design suitable for an FPGA and/or ASIC.
But they also mention that they
We have decades of experience in ASIC development and digital design and our team has already completed several designs and is in the late stages of releasing our first ASIC.
So I’m not exactly sure what they are proposing, ASIC hardware or FPGA?
Ok, so at least i’am not the only only one not being sure what the proposal is exactly about. Thx for confirming. Was allready afraid my English has some serious lacking
AFAIK many softwares for fpga synthesis can easily export their outputs in order to convert the design from fpga to an Asic. At least this was the way that Altera Quartus II was working up to 3/4 years ago.
So, having a working fpga project means that the same asic implementation is not far.
Why not just use SHA256d POW if you are going ASIC? they are readily available and it makes no difference if used for Zcash POW mining. Sorry to be mean but it appears to me as in most companies the place you where you knowledge ends i.e. moderator, security guard etc. ZcashCo needs to clarify their position if they aspire to be ASIC resistant or not so investors can join or move on.
As I have stated before I am against ASIC’s as they lead to control via regulation (specialized hardware can be regulated, general purpose hardware cannot).
However, FPGA’s are a completely different story from ASIC’s. Anyone can buy an FPGA, they are widespread and general purpose. They would be impossible to control via regulations as they are even more multipurpose than a GPU. I would be in favor of ASIC resistant and FPGA friendly Zcash.
I am sure you and the leadership team of the Zcash community are giving ASIC mining serious thought and discussion. Decentralizing hash power is important to maintaining censorship resistance.
As Zcash holder (tiny) I certainly believe in the mission of free expression in the digital realm. A key element of why we started ePIC Blockchain is to work directly with protocols like Zcash to make them inclusive partners to the design and deployment of the hardware that will run blockchains. We see the vision of Trustware that Andreas Antonopoulos explained just over a year ago; a blockchain is where hardware and software design merge into Trustware.
Too many protocols are making a passive decision on hardware design, essentially letting Bitmain make an ASIC or using AMD/Nvidia GPUs, this is leaving immense value un-examined. It is also creating moments like now when the community’s energy is sucked into an infrastructure debate and potentially held hostage by miners who have no interest other than profits.
ePIC is the leading Trustware hardware developer, our mission is to work with protocols like Zcash to design, develop, and deploy best-in-class hardware to produce secure, decentralized, and scalable protocols. How do we best engage with you, the team and Zcash community to determine how to collaborate?
ePIC submitted a Zcash Foundation grant proposal today, we would love any and all feedback.
