Kadena (KDA): a guide to the coin, Chainweb and mining on Blake2s

Kadena is a proof of work blockchain with an unusual architecture: instead of a single chain it runs twenty parallel chains tied together by cross references. That construction is called Chainweb and was meant as a way to raise throughput without giving up proof of work. Mainnet went live in 2019. The mining algorithm is Blake2s, and over its whole history Kadena remained the only significant coin for which that algorithm was ever put into silicon.

The network was created by Stuart Popejoy and William Martino, who had previously worked at JPMorgan Chase and at the SEC. At JPMorgan they took part in building the predecessor of the bank enterprise blockchain platform known today as Kinexys. The second notable part of the project is Pact, its own smart contract language: human readable, with formal verification and without arbitrary loops, that is, deliberately built so that a contract is hard to write unsafely.

The main thing to know about the project today. On October 21, 2025 the Kadena organisation announced that it was ceasing operations and active support of the blockchain. Since then kadena.io has carried a short notice saying that Kadena LLC ended its operations on that date. The announcement itself said that the proof of work network would carry on existing for as long as miners and the people maintaining it remain.

On November 8, 2025 at 04:00 UTC the community launched a fork. Node version 3.0, released on November 6, 2025 in the unofficial kda-community repository, carried a direct reference to that date and time of separation from the former mainnet. Since then the node has been developed by the community rather than by the original team. As of August 1, 2026 the network produces blocks, and public pools and calculators return live difficulty and hashrate. This page is written for someone choosing hardware, so from here on we look not at the prospects of the coin but at what all of that means for an ASIC owner.

• 2019Kadena mainnet goes live. Chainweb architecture with parallel chains and Blake2s as the mining algorithm. The choice of algorithm was explained by the fact that it is well studied, lends itself to mining on graphics cards and still allows purpose built hardware to appear later.
• 2020The network runs on graphics cards and software miners. An ecosystem grows around the Pact language, with the first contracts and developer tools. The number of chains in Chainweb is brought up to twenty.
• 2021The year of peak interest in the project. On November 11, 2021 the all time price high of 27.64 dollars was recorded. By 2026 the coin had lost more than 99.9 percent from that mark. In the same period Goldshell enters the niche with the KD line and pushes graphics cards off the algorithm.
• 2022The year the hardware market filled up. In April the Goldshell KD6 arrives at 29.2 TH/s, in September Bitmain releases the Antminer KA3 at 166 TH/s and 3,154 W, that is 19 J/TH. The KA3 becomes, and remains, the most efficient machine on the network: the next model behind it loses by roughly four and a half times.
• 2023The Blake2s hardware market stops. Neither Bitmain nor Goldshell release any new models. Everything available from here on is the secondary market.
• 2024Peak network hashrate. Industry publications of the time quoted figures around 666.88 PH/s dated 28 January 2024. We repeat that number with a direct caveat: the primary source could not be confirmed, because the kadena.io blog those publications pointed to is currently unavailable. It should not be relied on as exact.
• 2024The year of organisational contraction. Timelines for the Ethereum compatible environment slip, part of the engineering team leaves the project, and the price of the coin falls by an order of magnitude against the values seen at the start of the year.
• 2025In July the Chainweb EVM testnet, an environment with Solidity compatibility, is shown at the EthCC conference. That work never reached mainnet, and this matters, because some reference sites still describe it as a completed launch.
• 2025On October 21 the Kadena organisation announced that it was ceasing operations and active support of the blockchain. The statement said that market conditions were the reason and that the company could not go on promoting and supporting the project. It noted separately that the proof of work network itself would carry on existing while miners and maintainers remain.
• 2025Late October: exchanges begin winding down trading. Binance.US delists KDA on October 28. OKX halts spot on October 26 and removes the trading pairs on October 29, leaving withdrawals open until January 22, 2026. Bybit ends perpetual contracts from October 24.
• 2025On November 6 node version 3.0 is published in the unofficial kda-community repository. The release notes state that separation from the former mainnet will happen on November 8, 2025 at 04:00 UTC. The same version moves the key sets under community control and replaces the bootstrap nodes with public ones.
• 2025On November 12 KDA is delisted from Binance, with the withdrawal window closing on January 12, 2026. A major pool publishes a notice about moving to the community version and states plainly that it cannot guarantee the stability and technical reliability of a node maintained by someone other than the original team, and that it accepts no liability for mining losses.
• 2025In December node version 3.1 arrives. It removes the forced shutdown by date and introduces miner voting on forks: every 14,400 blocks, roughly once every five days, with a threshold of two thirds of the hashrate. Every version below 3.1 stopped working on January 7, 2026.
• 2026By August the network is still producing blocks, the node is developed by the community, and liquidity sits at a few thousand dollars a day. Part of the infrastructure has fallen away: the former explorer on the explorer.chainweb.com domain does not respond and one of the large mining aggregators dropped KDA from its coin list. Third party explorers work.

Chainweb is not one chain but twenty parallel chains, each with its own blocks, its own history and its own set of transactions. Every chain references not only its own previous block but also the latest blocks of several neighbouring chains. Neighbours are fixed by a graph in which each chain has exactly three of them. Thanks to those cross references, work invested in one chain confirms the others as well: rewriting the history of one chain would mean rewriting the chains linked to it too.

The target block time in each individual chain is thirty seconds. Since there are twenty chains and they run at the same time, across the network as a whole a block shows up roughly once every second and a half. The daily volume comes from the same arithmetic: 20 chains times 2,880 blocks gives 57,600 blocks per day. For a miner that means frequent small credits instead of rare large ones, which is convenient for smoothing variance on a pool.

The hardware does not pick the chain. Mining work is assembled by the node: it is the node that writes the chain id into the header, gathers the neighbour references and hands the machine a ready block of data. The miner can change only the block creation time and the nonce, every other field of the header is fixed for it. That is a point of principle, and it is why the hashrate of the whole network is counted as the sum across all twenty chains rather than across one.

Total hashrate is not measured directly, it is derived from the combined difficulty of all twenty chains over a chosen interval. Different trackers take different averaging windows and treat chains that happen to be lagging differently, so their estimates diverge more visibly than for single chain coins. On the build date of this page the spread between sources was around 1.36 times, and that is a normal property of the architecture rather than an error on any particular site.

1. The node assembles the headerThe node builds a block header of 318 bytes for the graph in which each chain has three neighbours. The header holds the flags, the creation time, the parent block hash, the references to neighbouring chains, the difficulty target, the payload hash, the chain id, the accumulated weight, the height, the version, the epoch start, the nonce and the hash of the block itself.
2. References to neighboursA separate 110 bytes of the header, at offsets 48 through 157, are given over to references to neighbouring chains. Inside, that is two bytes of length and three records of 36 bytes, each holding a four byte chain id and a thirty two byte hash of that chain latest block. This is the field that stitches twenty chains into a single construction.
3. The work is handed outThe node hands work to a machine or a pool. The chain id in the header is already set by the node, the hardware neither chooses it nor can change it. Out of the whole header the miner is allowed to change only two fields: the block creation time and the nonce.
4. Grinding the nonceThe nonce occupies eight bytes at offsets 278 through 285, that is, it sits at the very end of the hashed part of the header. The machine grinds those eight bytes and on every attempt computes one Blake2s hash over the first 286 bytes of the header. The pass is single: there is no second round of hashing here of the kind SHA-256 uses in Bitcoin.
5. Comparison with the targetThe resulting value is compared with the target, which sits in the header at offsets 158 through 189. If the hash does not exceed the target, the block counts as found. One important detail: the proof of work hash and the hash of the block itself are different values, and the last 32 bytes of the header are not part of what is hashed.
6. The nonce space runs outEight bytes of nonce do not last long for a modern machine. Developer discussions noted plainly that an ASIC of around 1 TH/s exhausts the entire nonce space in less than a second against a thirty second block. Pools therefore have to feed fresh work constantly and to use shifts of the block creation time as an extra source of variation.
7. Propagation and confirmation by neighboursThe block that was found spreads across the network and nodes verify it by recomputing the same single Blake2s. From there the block starts being confirmed not only by its own chain but by the neighbouring ones: as soon as the neighbouring chains reference it, the cost of reversing it grows by the amount of their work.

Blake2s is a cryptographic hash function from the Blake2 family described in RFC 7693. Unlike many mining algorithms it was not designed for mining at all: it is an ordinary fast general purpose hash function, which the Kadena developers picked because it is well studied and well tested. Blake2s has no memory requirements, no auxiliary tables and no second pass. That is exactly why a dedicated chip for it comes out simple and very dense, and why the gap between ASIC and graphics card turned out to be enormous.

The key point for a hardware buyer: Blake2s and Blake2b are different algorithms, not variants of one. Under RFC 7693, Blake2b has a 64 bit word, a 128 byte block and 12 rounds, while Blake2s has a 32 bit word, a 64 byte block and 10 rounds. Maximum hash length is 64 bytes for the first and 32 for the second. Even the rotation constants in the mixing function differ: 32, 24, 16 and 63 for Blake2b against 16, 12, 8 and 7 for Blake2s. These are physically different chips, and no reflash will move a machine from one to the other.

From this follows the main practical limitation. A Kadena ASIC cannot be moved to any other coin. Kadena was the only significant project mined on Blake2s with purpose built hardware. Verge formally uses Blake2s as well, but it has a different header structure and different plumbing, it is mined on graphics cards, and Kadena machines do not apply to it. The remaining projects on this algorithm are either long closed or without liquidity. There is no fallback use for the KA3 or for the KD line.

Graphics cards on Blake2s in Kadena make no sense today. In the first year of the network mining ran precisely on them, but once purpose built machines appeared the efficiency gap became impossible to close: the KA3 delivers 166 TH/s at 3,154 W, which no graphics card rig can match either in throughput or in energy per unit of work. GPU software miners still exist, but they are of interest only as a teaching or testing tool.

The strength of the algorithm is that it is completely transparent and trivial to verify: one call to a well documented function over 286 bytes. The weakness is that the simplicity of the algorithm gives the coin nothing if the network has few users: a narrowly specialised chip turns the entire hardware fleet into a hostage of one project. Kadena is a textbook example of that dependency, and it is worth keeping in mind when buying any hardware for a niche algorithm.

Full description of the Blake2s algorithm →

Kadena has no halvings. The block reward does not drop in a step once every few years, it declines smoothly along an exponential curve: each following period brings a little less than the one before, with no discontinuities. For a miner that is more comfortable than the Bitcoin model, because there is no date on which income suddenly halves. The flip side is that the decline is continuous and easy to miss, and reference sites update this figure rarely.

The current reward as of August 1, 2026 is 0.9119135 KDA per block. That is the value shown by the coin page of a major pool and by specialised calculators. And here lies the most common error in third party sources: on the date we checked, one of the well known aggregators was still calculating with a reward of 0.960 KDA. The difference looks small, but it is about five percent, and it runs through every derived quantity: daily emission, income per terahash and final profitability.

Daily emission is simple to work out. At 57,600 blocks a day and a reward of 0.9119135 KDA the network issues about 52,500 KDA per day. Against a circulating supply of 338,586,083 KDA that gives annual inflation of about 5.7 percent. Maximum supply is capped at one billion coins, roughly 33.9 percent of that is in circulation now, and the issuance curve itself is designed for a very long horizon, reaching into the twenty second century.

The network has no burn mechanism with any noticeable effect on supply. Gas fees in Kadena are small and next to invisible against emission: miner income consists almost entirely of the block reward rather than of fees. That sets Kadena apart from networks where the fee share gradually replaces emission, and it means that as the reward keeps falling there will be nothing to compensate for it.

The practical conclusion on economics is single and unpleasant. The break even threshold on the electricity bill alone is about 0.034 dollars per kilowatt hour for the Antminer KA3 and about 0.008 dollars for the Goldshell KD6. The average price of electricity in Spain on August 1, 2026 under the regulated tariff was around 0.135 euro per kilowatt hour. The gap here is not a matter of percent but of multiples, and no amount of tuning closes it.

Fewer than ten models were ever built for Blake2s, all of them more than three years old, and not one of them can compute anything but Kadena.

Browse miners in the catalogue →

The ranking follows efficiency from the asic.es catalogue strictly, and the gap between first place and the rest is wider here than on any other algorithm.

Straight to the point, with no illusions: there is no firmware for Blake2s. The HashCore Toolkit support lists contain no machines on this algorithm, and we neither build nor plan AsicBoost releases for the Antminer KA3 or for the Goldshell KD line. Our tooling is aimed at SHA-256 and Scrypt, where the hardware fleet runs into hundreds of thousands of machines and development pays for itself. Four catalogue models for Kadena is a fundamentally different scale, and it is more honest to say so plainly than to sell a capability that does not exist.

Separately about the built in hidden overclocking menu on Goldshell hardware, around which there is a lot of confusion. According to user reports on specialist forums, tapping a menu item repeatedly in the web interface of the machine opened hidden frequency and voltage settings. On the KD5 at a frequency around 900 MHz users described hashrate rising from roughly 17.8 to 18.6 TH/s. The price of that rise was the share of rejected shares climbing to about 2 to 3 percent and, most importantly, payouts themselves did not change. From firmware 2.1.6 onward access to those settings was restricted, and some users rolled back to 2.1.5.

Two caveats are needed here. First: this is a built in function of the manufacturer itself, not third party firmware and not something we developed. Listings that present the hidden menu as a "firmware for overclocking Goldshell" are misleading. Second: the source of these observations is user reports on a forum, not manufacturer documentation, and they date from the period when the KD5 was in active use. Checking how a particular machine behaves is something you will have to do yourself.

The practical bottom line for a Kadena hardware owner: there is no lever here that reads install firmware and earn more. The only things that genuinely affect the result are the price of electricity and the price of the machine itself on the secondary market. If someone offers you firmware that supposedly raises profitability on Blake2s, that is a reason to be wary, not a reason to pay.

Getting started is technically simple, because the scheme is no different from any other ASIC mining. You need a Blake2s machine, a wallet to receive payouts, an account or an address on a pool, a stable internet connection and a suitable room. The machine is connected over the network, the pool address, worker and password are entered in the web interface, and after that the hardware starts receiving work. Nothing about the twenty chains has to be configured: the chain is picked by the node, not by the miner.

What really does need checking before you buy is pool availability. That is where the main practical risk for Kadena sits today: after the autumn of 2025 some operators paused payouts while the move to the community version was under way, and some reference sites still list pools whose working state we could not confirm. Test the stratum and live credits on a small amount of hashrate before you commit to a fleet.

On noise and heat the picture is the usual one for industrial machines and very uneven within the line. The Antminer KA3 means more than three kilowatts of heat and a noise level around eighty decibels, and the Goldshell KD6 and KD5 are barely quieter. None of those three models is suitable for a living space, they need organised intake of cold air and removal of hot air, plus a tolerance for constant drone. There is one exception: the KD-BOX at 205 W, which is comparable to an ordinary desktop computer in noise and heat.

The home scenario and the industrial one differ here not in degree but in kind. At home only the KD-BOX is realistic, and even then as a teaching bench rather than a source of income: at current network parameters it earns cents a day against a household tariff. The industrial scenario assumes your own generation or a tariff at the level of a few cents, and even there practically only one model still makes sense, the Antminer KA3.

Separately about the electrics. The Antminer KA3 at 3,154 W needs a dedicated circuit and a quality power supply with headroom, and so does the Goldshell KD6 at 2,630 W. Calculate from rated power plus a margin of around twenty percent, not from the figure in the listing. Remember that the meter runs continuously: a 3 kilowatt machine is about 75 kilowatt hours a day, and on a Spanish tariff that is a noticeable sum regardless of what the coin price does.

Miner income on Kadena is the product of three factors: the share of your hardware in the total network hashrate, daily emission, and the price of the coin. The first depends on how many machines are running in the network, the second is fixed at about 52,500 KDA per day, the third is set by the market. Network fees can be left out of the calculation: they are small and practically invisible against the block reward.

The distinctive feature of this particular coin is that the calculation depends heavily on which data source you took. On the build date of this page the price differed between venues by about 1.45 times and the network hashrate estimate by 1.36 times, while one aggregator was on top of that calculating with an outdated reward of 0.960 KDA instead of 0.9119135. Multiply those divergences together and the spread in the final profitability figure comes out twofold. That is why any Kadena calculator screenshot today is worth cross checking against at least two sources.

The break even threshold is a steadier number than profitability itself, because it is tied to the efficiency of the machine, and that you know exactly. For the Antminer KA3 at 19 J/TH it sits around 0.034 dollars per kilowatt hour on the build date of this page, and for the Goldshell KD6 at 90 J/TH around 0.008 dollars. Allowing for the spread between sources, the honest range is wider: roughly 0.029 to 0.057 dollars for the KA3 and 0.006 to 0.012 for the KD6. Even the upper bound is below any Spanish tariff.

From which follows the direct answer we owe you instead of promises. At any realistic Spanish tariff both main machines run at a loss on the electricity bill alone, before the cost of the hardware, the premises and maintenance is even counted. The average price under the regulated tariff on August 1, 2026 was around 0.135 euro per kilowatt hour, and the cheap hour of the day dropped to roughly 0.005 euro, but a few cheap hours do not pull the daily balance up. A payback period is not calculated under those conditions, because there is none.

Open the profitability calculator →

The main rule for a coin in this situation is simple: keep KDA in a wallet whose keys only you hold. In the autumn of 2025 several venues closed trading, and for some of them the withdrawal window was limited in time. Anyone who held the coin on the exchange itself and missed the announcement ended up in a noticeably worse position than someone whose funds sat in their own wallet.

A separate warning about maintenance. The wallets listed above work, but after the organisation wound down the ecosystem has no central maintainer, and some of the tools are developed by the community or not developed at all. Before moving any meaningful amount, check that the app you chose was updated recently, and be sure to store the recovery phrase off the device. We make no claims about hardware wallet support: its current state could not be confirmed from a reliable source on the build date of this page.

Blake2s hardware is bought on the secondary market only: there are no new machines from the factory at either Bitmain or Goldshell, and the last models of the line were released in 2022. That means any machine offered to you has been running for at least several years. Look at the state of the fans and the hashboards, at the operating history, and at whether the seller shows you the machine under load rather than just the interface loading.

Separately about how these machines are sold. Product listings promising payback on Kadena hardware are not supported by any calculation today. The break even threshold of the best machine on the network is around three and a half cents per kilowatt hour, which means any payback calculation at a European tariff is built either on an outdated coin price, or on an outdated block reward, or on data from before October 2025. If a listing quotes a payback period, ask to be shown which price and which reward it was calculated from.

Two frequent substitutions in listings worth knowing about in advance. First: the Goldshell KD MAX at 40.2 TH/s is sometimes presented as a mode or a firmware for the KD6. It is a separate model with a different case and different power draw, and buying a KD6 in the hope of turning it into a KD MAX is pointless. Second: the hidden overclocking menu on Goldshell hardware is also sometimes called firmware, although it is a built in function of the manufacturer itself which, according to user reports, raised hashrate but did not raise payouts.

And the general advice on transaction safety, which does not depend on the coin. Check the seller, do not pay into personal accounts, insist on a purchase document, and if the delivery comes from outside the EU work out duties and VAT in advance. For Kadena machines this matters especially, because in a disputed situation the residual value of the hardware is very low: it can only be resold to another miner on the same network, since the machine cannot be moved to any other coin.

The original purpose of the network was a smart contract platform on proof of work. The bet rested on three things together: scaling through parallel chains, the Pact language with formal verification, and an expectation of enterprise use that owed a lot to the background of the founders at JPMorgan. A separate trait of Pact is that it has no arbitrary loops, while contracts can be upgraded, which is an unusual decision for smart contract platforms.

What is left of the ecosystem as of August 2026. The community list still names decentralised exchanges, a bridge to Ethereum, several applied projects and infrastructure services. The lists are not marked up by activity, however, so you cannot judge the real state of each project from such a listing. All we record is that the list itself exists and is maintained.

Separately about Chainweb EVM, the environment with Solidity compatibility, around which there is a lot of confusion. It was shown as a testnet in July 2025 at the EthCC conference. We found no launch on mainnet in any source. That matters, because some aggregators still list the launch of this environment among the events of 2026, although there is no confirmation.

The practical conclusion for anyone looking at the coin from the hardware side: the applied load on the network is small today, fees are barely visible in miner income, and that income consists almost entirely of the block reward. Which means the fate of Kadena mining depends on the price of the coin and the price of electricity, not on how many Pact contracts get deployed in the coming year.

Kadena security rests on chains confirming one another. To rewrite the history of one chain an attacker would have to redo the work of the neighbouring chains that referenced it, and through them the next ones. Formally that raises the cost of an attack above that of a single chain network with the same hashrate. In practice, though, the absolute level of protection is set by the absolute hashrate, and that is orders of magnitude below the historical values of this network today.

The concentration risk has to be discussed honestly. At a network hashrate on the order of ten petahashes, the cost of renting or deploying comparable power is incomparably lower than at the peak, and on an algorithm where in effect one large operator holds a noticeable share, the distribution of power is uneven by definition. We found no publicly confirmed successful double spend attacks on Kadena, and we will not claim there were any. But the safety margin of the network is now substantially smaller than it was, and exchanges would do well to account for that in their confirmation counts.

A separate layer of risk in Kadena sits not in consensus but in maintenance of the node. Since October 21, 2025 the code has lived in an unofficial community repository, and a major pool warned plainly in its notice: it cannot guarantee the stability and technical reliability of a version maintained by someone other than the original team, it pointed to the possibility of synchronisation errors, further forks and vulnerabilities, and it disclaimed liability for mining losses. That is not an accusation aimed at anyone, it is a record of how the operator itself described its risks.

There is another side that should be noted honestly too. Version 3.1 removed the forced shutdown of nodes by date and introduced miner voting on forks with a threshold of two thirds of the hashrate and a period of about five days. That is, decisions about further protocol changes are now taken by the people who produce blocks rather than by the owner of a repository. For a network without an owning organisation that is a sensible construction, and it has been working for more than half a year.

A block reward of 0.960 KDA. The costliest error on the build date of this page. The current reward is 0.9119135 KDA, shown by the coin page of a major pool and by specialised calculators. One well known aggregator meanwhile keeps calculating with 0.960 and therefore reports daily emission of about 55,300 KDA instead of roughly 52,500. An error of about five percent runs through every derived calculation. If someone shows you a Kadena calculation, the first thing to look at is which reward it uses.

A single price for the coin. There simply is no single price today. On August 1, 2026 a major market aggregator showed about 0.0058 dollars, while mining calculators and a pool page on the same day showed about 0.0040 to 0.0042. That is a spread of about 1.45 times. Any profitability figure calculated without naming the price source hangs in the air, and the difference between the upper and lower estimate exceeds the entire profit.

A single network hashrate estimate. The same story. Different trackers on one date gave from 11.80 to 16.02 PH/s, a divergence of about 1.36 times, and correspondingly a difficulty from 17.7 to 24.0 P. The reason is architectural: total hashrate is derived from the combined difficulty of twenty chains, and different sites take different averaging windows. This is not an error on any particular site, but a single source cannot be treated as the truth.

The project is under active development. A significant number of reference sites and coin cards describe Kadena as if nothing had happened, down to phrases about the only truly scalable proof of work blockchain. The factual position is different: the organisation announced that it was ceasing operations on October 21, 2025, its site now carries only a notice about that, and since November 8, 2025 the node has been maintained by the community in an unofficial repository.

A machine generated roadmap for 2026. A separate case worth attention. One of the largest aggregators publishes an automatically generated summary of Kadena events for 2026, including the launch of the Ethereum compatible environment on mainnet, a move to DAO governance and a relaunch of the bridge. We found confirmation of none of those events in any source. The only thing confirmed about the Chainweb EVM environment is the testnet shown in July 2025. Such summaries are easy to mistake for facts, because they are published on an authoritative domain.

Internal contradictions in hardware reference sites. One well known ASIC reference site showed, for one and the same date, daily revenue of about 2.8 dollars for the Antminer KA3 and about 0.17 dollars for the Goldshell KD6. At their hashrate ratio of 166 to 29.2 the second figure should be around 0.49 dollars. Which means one of the two cards was calculated from an outdated data snapshot. It is worth checking even the internal consistency of a source, not just the presence of a number.

Product listings promising payback. Any promise of a payback period for the Antminer KA3, Goldshell KD6, KD5 or KD-BOX on Kadena is not supported by calculation today. The break even threshold of the best machine on the network is around 0.034 dollars per kilowatt hour, against an average Spanish price of about 0.135 euro. If a listing quotes a payback period, it was calculated either from an outdated coin price, or from an outdated reward, or from data predating October 2025.

Blake2s is the same thing as Blake2b. This turns up both in hardware descriptions and in algorithm roundups. Under RFC 7693 they are different functions: a word of 32 against 64 bits, a block of 64 against 128 bytes, 10 rounds against 12, a different maximum hash length and even different rotation constants. These are different chips, and a Kadena machine will not compute Blake2b under any reflash.

The hidden Goldshell menu is overclocking firmware. A substitution of terms that appears in listings. The hidden frequency and voltage settings in the Goldshell web interface are a built in function of the manufacturer itself, not a third party release. According to user reports on the KD5 hashrate rose from roughly 17.8 to 18.6 TH/s, the share of rejected shares climbed to 2 to 3 percent, and payouts did not change. Access to those settings has been restricted since firmware version 2.1.6.

The KD MAX is a firmware or a mode for the KD6. It is not. The Goldshell KD MAX at 40.2 TH/s and 3,350 W is a model in its own right with its own case, its own board and its own power draw. Turning a KD6 into a KD MAX through a software update is impossible. The KD MAX is absent from the asic.es catalogue, so it takes no part in the ranking or the comparison table on this page.

Links to an explorer that no longer exists. Many reference sites and coin cards still point to the former chain explorer on the explorer.chainweb.com domain. On August 1, 2026 that domain does not respond. Third party explorers work, while one of the large mining aggregators additionally dropped KDA from its coin list, so links to its Kadena page have stopped working too.

A Kadena ASIC can be moved to Verge or another Blake2s coin. It cannot. Verge formally uses Blake2s, but it has a different header structure and different plumbing, and it is mined on graphics cards. Other projects on this algorithm are either closed or without liquidity. Kadena was the only significant coin mined on Blake2s with purpose built hardware, and the residual value of the machines is tied to it completely.