Goldshell: the company, the BOX series and altcoin miners
ASIC manufacturers · Goldshell
Who Goldshell is and why quiet altcoin miners exist
20 models in our catalog, five algorithms, ten ASICs that draw under 500 watts. On paper these are the quietest and most compact miners we sell, and on energy use they lose to the competition by a factor of three to five. Who it fits and who it does not.
Goldshell in brief
Goldshell is a brand of compact miners for altcoins. It took a niche that large manufacturers avoided for a long time: ASICs you can put in a home or an office, and coins nobody builds data centers for. The format took shape in 2020 and 2021 and got the name BOX: a case the size of a thick book, weight from two to four kilograms, rated noise up to 35 dB.
Since February 2024 Goldshell is no longer an independent company. The brand belongs to the public Intchains Group Limited, which trades on Nasdaq under the ticker ICG. The deal closed on 28 February 2024, the brand assets cost the buyer 550 thousand dollars, and the seller was the Singaporean entity Goldshell Pte. Ltd. Intchains itself was founded in December 2017 in Shanghai and works on the fabless model, designing chips and outsourcing production.
First about the weak point, because it defines everything else. In power consumption Goldshell lags behind specialized competitors by a wide margin. On Scrypt the miners of this brand deliver from 0.53 to 1.27 J/MH, while the best models from ElphaPex and Bitmain fit into roughly 0.2. On Kaspa Goldshell has 339 and 375 J/TH, while the top Antminer of the KS series is around 77. That is a threefold and a fivefold gap.
In return you get what competitors do not offer, an ASIC from two kilograms that fits on a shelf and plugs into an ordinary room. You pay for that with efficiency. We look separately at the state of the coins the brand was built for. Some of them make no sense today, and the KD line was orphaned in October 2025.
Company card
| Parameter | Value |
|---|---|
| Brand | Goldshell |
| Brand owner | Intchains Group Limited, Nasdaq, ticker ICG |
| When the brand was bought | 28 February 2024, deal value 550 thousand dollars |
| Previous owner | Goldshell Pte. Ltd., Singapore |
| Parent company founded | December 2017, Shanghai |
| Market entry | The company claims 2017, and the first miner under the brand shipped in 2020 |
| Production model | Fabless: own chip design, third-party fabs |
| Platform process nodes | 22 nm since 2018, 12 nm since 2024 |
| Foundry | The name is not disclosed in filings |
| Algorithms in our catalog | Scrypt, kHeavyHash, Eaglesong, Blake2s, Blake3 |
| Bitcoin | The brand has no SHA-256 miners |
| Models in the asic.es catalog | 20 |
| Owner revenue for 2025 | About 31.6 million dollars, minus 21.6 percent year over year |
| Owner result for 2025 | Net loss of about 7.4 million dollars |
| Warranty | 180 days, though the starting point is described differently across the manufacturer's documents |
| Official site | goldshell.com, support through a ticket portal |
Goldshell history year by year
The history of the brand fits into less than ten years and splits in two, first an independent brand that released a few new miners a year, then an asset of a public chip company.
- 2017Intchains, the future Intchains Group Limited, is founded in Shanghai. At that point it has nothing to do with miners: it is a chip developer.
- 2018First tape-out of its own ASIC on the 22 nm node.
- 2020HS1 Plus arrives, a Handshake miner at 115 watts weighing 450 grams with a rated 34 dB. That is where the format the brand is bought for started.
- 2021Seven new miners in one year. CK5 and KD5 in March, KD-BOX in May, Mini-DOGE and CK-BOX in July, LT5 Pro in August, CK6 in December. The BOX format finally takes shape as a series.
- 2021In December an independent researcher publishes a security review of the HS-BOX and KD-BOX firmware: a weak root password, a debug page without authorization and device requests to an external domain every three seconds. This is not a backdoor and not hidden mining but poor hygiene, typical for cheap ASICs of that period.
- 2022CK Lite, KD6, Mini-DOGE Pro, LT Lite, then CK6-SE and CK-BOX II. An intermediate Lite class appears: more power than BOX, noise of 55 dB instead of 80.
- 2023Mini-DOGE II. In April the company itself warns about fraudulent clones of its site, and there really are many of them in search results. In December Intchains announces the purchase of the brand.
- 2024On 28 February the deal closes and Goldshell moves to Intchains. KA-BOX for Kaspa arrives in March, AL-BOX for Alephium and KA-BOX Pro in May, AL-BOX II Plus in August, E-DG1M in September. The tape-out on the 12 nm node was completed in the first quarter of the same year.
- 2025In March BYTE is presented: a desktop unit with two slots and hot-swap cards for different coins. In October the Kadena foundation ceases operations, the coin collapses, and three models of the KD series lose their purpose.
- 2026The owner's report for 2025: revenue of 31.6 million dollars, minus 21.6 percent, a net loss of 7.4 million, gross margin of 7.2 percent. Cost of sales rose because of a write-down of altcoin miner inventory. The company has promised new models for the second half of 2026, the half year that is running now.
Goldshell miner lines
Goldshell lines are split not by generation but by algorithm, and in most groups there is a compact BOX version and a larger one, while under Blake3 and kHeavyHash there are only compact ones. The order of the groups is the same as in our catalog.
The liveliest line of the brand, and the reason is merged mining: one miner computes Litecoin and Dogecoin at the same time, with no extra power spent on the second coin. It includes the desktop Mini-DOGE at 220 and 233 watts, Mini-DOGE II at 400 watts, plus the larger LT Lite and LT5 Pro, and the E-DG1M at 1800 watts. E-DG1M is noticeably more efficient than the rest: 0.53 J/MH against 1.27 for LT5 Pro.
Who it suits: the only direction of the brand with two assets from the top thirty.
KA-BOX at 400 watts and KA-BOX Pro at 600. Both are compact and both are rated at 35 dB. Kaspa cuts its reward every month, roughly by half over a year, and by the middle of 2026 about 95 percent of the whole emission has been mined. Calculating payback from today's figures is especially dangerous here.
Who it suits: those who count on a horizon of no more than a year.
The widest line together with Scrypt, six miners in each: the compact CK-BOX and CK-BOX II, the intermediate CK Lite and the industrial CK5, CK6 and CK6-SE at 2400 and 3300 watts. The Nervos network is running, but the next halving is expected in November 2027 and will cut the reward in half. The brand has not released new miners for this algorithm in a long time.
Who it suits: only with very cheap power and with an eye on the halving in November 2027.
KD-BOX at 205 watts, KD5 and KD6 at 2250 and 2630. In joules per terahash the KD figures look like the best in the catalog, but they cannot be compared with the J/MH of the Scrypt miners. These are different units. In practice buying them for mining today is out of the question: in October 2025 the Kadena foundation announced that it was ceasing operations, the coin collapsed and exchanges started delisting it. Blake2s chips physically cannot switch to other coins.
Who it suits: nobody, for mining. Three models, from 205 to 2630 watts, and all of them under a dead network.
AL-BOX, AL-BOX II and AL-BOX II Plus, from 180 to 480 watts. The network is alive but very small: daily turnover of the coin is measured in hundreds of thousands of dollars, not millions. That means slippage when selling and a delisting risk.
Who it suits: a small volume and money you can afford to lose: three miners from 180 to 480 watts.
A common form factor that runs across every algorithm of the brand. The case is roughly 150 by 200 by 90 millimeters, weight from two to four kilograms, powered from 12 volts DC through an external supply and a six-pin connector. Rated noise up to 35 dB, up to 45 for AL-BOX II Plus.
Who it suits: an apartment or an office. From two kilograms, a shelf, a rated 35 dB: that is what people buy the brand for.
CK5, CK6, CK6-SE, KD5, KD6 and LT5 Pro in a case of 264 by 200 millimeters with a depth of 270 to 290, weighing from 8 to 11.5 kilograms. Noise of 80 dB, a level incompatible with a living space. The Lite class sits in the middle: 55 dB at 1200 and 1450 watts.
Who it suits: a separate room with extraction, the same as for any regular ASIC.
Catalog of Goldshell models on asic.es
The asic.es catalog lists 20 Goldshell models across five algorithms, and not a single one for bitcoin. Power draw runs from 180 to 3300 watts, and half of the ASICs, ten out of twenty, stay within 500 watts.
Comparing efficiency across algorithms is pointless. A joule per terahash on Kaspa is not the same quantity as a joule per terahash on Nervos. The only meaningful comparison is inside one group, so the table is sorted by algorithm and then by efficiency.
| Model | Algorithm | Hashrate | Power draw | Efficiency |
|---|---|---|---|---|
| Goldshell E-DG1M | Scrypt | 3400 MH/s | 1800 W | 0.53 J/MH |
| Goldshell LT Lite | Scrypt | 1620 MH/s | 1450 W | 0.9 J/MH |
| Goldshell Mini-DOGE II | Scrypt | 420 MH/s | 400 W | 0.95 J/MH |
| Goldshell Mini-DOGE Pro | Scrypt | 205 MH/s | 220 W | 1.07 J/MH |
| Goldshell Mini-DOGE | Scrypt | 185 MH/s | 233 W | 1.26 J/MH |
| Goldshell LT5 Pro | Scrypt | 2450 MH/s | 3100 W | 1.27 J/MH |
| Goldshell KA-BOX | kHeavyHash | 1.18 TH/s | 400 W | 338.98 J/TH |
| Goldshell KA-BOX Pro | kHeavyHash | 1.6 TH/s | 600 W | 375 J/TH |
| Goldshell CK6 | Eaglesong | 19.3 TH/s | 3300 W | 170.98 J/TH |
| Goldshell CK Lite | Eaglesong | 6.3 TH/s | 1200 W | 190.48 J/TH |
| Goldshell CK-BOX II | Eaglesong | 2.1 TH/s | 400 W | 190.48 J/TH |
| Goldshell CK6-SE | Eaglesong | 17 TH/s | 3300 W | 194.12 J/TH |
| Goldshell CK5 | Eaglesong | 12 TH/s | 2400 W | 200 J/TH |
| Goldshell CK-BOX | Eaglesong | 1.05 TH/s | 215 W | 204.76 J/TH |
| Goldshell KD6 | Blake2s | 29.2 TH/s | 2630 W | 90.07 J/TH |
| Goldshell KD5 | Blake2s | 18 TH/s | 2250 W | 125 J/TH |
| Goldshell KD-BOX | Blake2s | 1.6 TH/s | 205 W | 128.12 J/TH |
| Goldshell AL-BOX II Plus | Blake3 | 1 TH/s | 480 W | 480 J/TH |
| Goldshell AL-BOX II | Blake3 | 0.72 TH/s | 360 W | 500 J/TH |
| Goldshell AL-BOX | Blake3 | 0.36 TH/s | 180 W | 500 J/TH |
Which algorithms Goldshell supports
Five algorithms across twenty models. Bitmain has eleven algorithms, but almost its whole catalog is SHA-256, so among the brands without bitcoin the coverage of Goldshell in our shop is the widest. The flip side is that each miner handles one algorithm and you cannot move it to a coin from another network.
| Algorithm | Main coins | Models in catalog | Comment |
|---|---|---|---|
| Scrypt | LTC, DOGE | 6 | The liveliest direction: merged mining yields two coins for one job |
| kHeavyHash | KAS | 2 | The network is running, but the reward drops every month, roughly by half over a year |
| Eaglesong | CKB | 6 | The network is running, and the next halving is expected in November 2027 |
| Blake2s | KDA | 3 | The Kadena foundation ceased operations in October 2025 and the direction is dead |
| Blake3 | ALPH | 3 | The network is alive but very small, and daily coin turnover is in the hundreds of thousands of dollars |
What you can mine on this hardware
Not a single Goldshell miner mines bitcoin, and that is the main difference from Bitmain, MicroBT, Canaan and Bitdeer. The whole risk profile follows from it. Instead of the largest network you are betting on one specific small project.
| Coin | Ticker | Algorithm | Goldshell models |
|---|---|---|---|
| Litecoin | LTC | Scrypt | 6 models of the Mini-DOGE, LT and E series |
| Dogecoin | DOGE | Scrypt | The same 6 models, mined at the same time as LTC |
| Kaspa | KAS | kHeavyHash | KA-BOX and KA-BOX Pro |
| Nervos | CKB | Eaglesong | 6 models of the CK series |
| Kadena | KDA | Blake2s | 3 models of the KD series, a network with no core team since October 2025 |
| Alephium | ALPH | Blake3 | 3 models of the AL-BOX series |
Technology: chips, boards, power
Less is known about Goldshell internals than about any other brand in our catalog. Official product pages simply contain no chip row: no die marking, no process node, no chip count per board.
What is known reliably concerns the brand owner. Intchains works on a fabless model. It designs the ASIC itself and orders wafers from an outside manufacturer whose name it does not disclose even in its exchange filings. Two process nodes of the platform have been named publicly: 22 nm since 2018 and 12 nm since the first quarter of 2024. The only die marking that has reached the open market is the ICT560 from the LT5 Pro.
| Chip | Year | Process node | Where it is used |
|---|---|---|---|
ICT560 | not disclosed | not disclosed | LT5 Pro. Found in the retail range of repair suppliers |
not disclosed | 2018 | 22 nm | The first process node of the brand owner's platform |
not disclosed | 2024 | 12 nm | The second process node, tape-out in the first quarter |
not disclosed | not disclosed | not disclosed | The KD, CK, LT, KA and AL series: the mapping of specific dies to models is not publicly described |
Control boards
| Board type | Where it is used | Comment |
|---|---|---|
| Control board | KD-BOX, KA-BOX, AL-BOX, Mini-DOGE, CK Lite, CK5, CK6, LT5 Pro | From 130 to 140 dollars on the secondary market. For the cheapest miners of the BOX series this is more than half the price of the ASIC itself |
| Hash board | CK5, CK6, LT5 Pro | Sold by specialized suppliers, but usually without warranty and without returns after shipment |
| Power supply | The top series and the BOX format | From 180 to 220 dollars. The manufacturer sells it itself, the only repair item in its store |
| Fans | 120 and 80 millimeters | The cheapest consumable, 3 or 4 dollars apiece |
| BOX format hash boards and chips | All BOX models | These are not available on the open market, neither the boards nor the chips |
Power on the BOX series is arranged differently from an ordinary ASIC. The miner takes 12 volts of direct current through a six-pin connector, so it does not plug into a wall socket directly. It needs an external power supply, sold separately or in a bundle. The manufacturer specifically requires a cable no thinner than 16 AWG and warns that a thinner one overheats.
There are also models with an ordinary alternating current input from 100 to 240 volts, the E-DG1M for example. For the Spanish 230 volt grid this works without adapters.
Official operating conditions for the BOX series are stricter than for industrial miners of other brands: air temperature from 10 to 35 degrees according to the manufacturer's manual, while seller listings show a range from 0 to 35 and humidity no higher than 65 percent. For coastal Spain it is the humidity that gets in the way, since 65 percent is exceeded there regularly.
Case formats and noise
Goldshell has no hydro and no immersion. All miners are air cooled. But within air cooling the brand keeps three different classes, and the E-DG1M with an ordinary alternating current input stands apart from them. The difference between the classes decides where the ASIC can be placed.
A case of roughly 150 by 200 by 90 millimeters, weight from two to four kilograms. Rated noise up to 35 dB on most and up to 45 on the AL-BOX II Plus. Nobody has published instrument measurements of the noise, and third-party catalogs disagree about the same KA-BOX, 35 dB in one listing and 55 in another. The rated 35 is reached at low fan speed, and the fan in these cases can spin up to 4500 revolutions per minute, at which point it is noticeably louder.
An intermediate option: the case is already industrial, but noise is rated at 55 dB. That is louder than a refrigerator and quieter than a vacuum cleaner, still a lot for a living room, acceptable for a separate room.
CK5, CK6, CK6-SE, KD5, KD6 and LT5 Pro. A case of roughly 264 by 200 millimeters with a depth of 270 to 290, weight from 8 to 11.5 kilograms, noise 80 dB. Here Goldshell is no different from an ordinary ASIC of any other brand and needs a separate room with proper extraction.
Firmware for Goldshell
Goldshell publishes firmware openly in a public repository, but only as binary files, with no source code and no formal description of the changes. The web interface opens at the address of the ASIC on the local network, and the brand's own discovery service helps to find that address. The factory password is the same on all miners and widely known. Change it at the first power on.
For a fleet there is a cloud platform from the manufacturer: group pool setup, scheduling of power modes, firmware updates, offline notifications and an automatic restart if hashrate stays below a tenth of nominal for longer than half an hour. There is also a mobile app and a desktop monitoring utility.
Two particulars show up only after the purchase. First, the firmware uses its own dialect of the pool protocol, and when connected to a pool that does not implement it the miner just sits there doing nothing. The dashboard shows zeros or a message that the pool is not ready, while giving no errors. Choose a pool from those confirmed to work with the brand, and recheck it after every firmware update.
Second, a firmware update on these miners carries a real risk. Owners describe a wave of failures in the bug tracker after a run of builds, especially when updating over a wireless network, and the whole BOX series was affected. Recovery is done by writing an image to a memory card, but the manufacturer does not publish the images themselves. They are requested by email to support, and the reply arrives within one to three days. Update only over cable and only when there is a reason, not for the sake of the version number.
Spare parts and components
The spare parts picture is mixed. Among repair items the manufacturer itself publicly sells only power supplies and has no catalog of boards, fans or chips. The secondary market meanwhile covers control boards, including for the BOX format, from 130 to 140 dollars. But BOX hash boards and chips of almost all series are not on the open market.
| Component | When it is replaced | Comment |
|---|---|---|
| Control board | No network, the miner does not respond, no address | From 130 to 140 dollars on the secondary market, available for the BOX format too. On the cheapest miners of the brand this is more than half the price of the ASIC itself |
| Hash board | A rise in hardware errors, a drop in hashrate | Available from specialized suppliers for the top series, usually without warranty and without the right of return after shipment. There are no hash boards on sale for the BOX format |
| Power supply | The miner does not start | The only repair item the manufacturer sells itself, from 180 to 220 dollars. On the BOX series the unit is external, and that is a plus: it is replaced separately, like an ordinary computer one |
| Fans | Hum, vibration, rising temperature | The cheapest and most frequent consumable, 3 or 4 dollars apiece from repair suppliers. In a compact case the fan runs at high speed and wears out faster |
| Power connector | A melted six-pin connector | A distinct weak spot of the brand. The cause is almost always the same: a connector that is not fully seated or a cable that is too thin |
| Chips | A single faulty die | Not sold for the BOX series and for almost every other series. In practice a single faulty die means replacing the whole board |
Repair: what fails and what to do about it
Repairability on Goldshell is limited, but the reason is not only technical. A control board costs from 130 to 140 dollars, while miners of the brand in the manufacturer's store run from 234 dollars for the simplest BOX up to fifteen hundred for the industrial ones. On a cheap ASIC the board eats up more than half its price, and with shipping both ways the repair loses its point. On a miner at seven hundred dollars and above the repair makes sense. For comparison, new Bitmain and MicroBT ASICs cost from a thousand dollars upwards, and repairs on those almost always pay off.
| Fault | How it shows up | What to do |
|---|---|---|
| Power connector and cable | Heating, melting, the miner goes into reboot | Excellent repairability and the most frequent cause. Seat the connector all the way in, use a cable no thinner than 16 AWG. A third-party power supply voids the warranty |
| Fan | Hum, then rising temperature and a stop | A simple swap, 3 or 4 dollars |
| Overheating | Temperature above 80 degrees by owner reports, emergency stop | Cured not by repair but by placement: free space around the case and the room temperature. Excess power density is not compensated by external airflow |
| Failure after a firmware update | The miner blinks its indicator and does not take an address | Recovered by writing an image to a memory card, but the image has to be requested from the manufacturer by email. It is not published publicly |
| Control board | No network, no response at the address | Technically an easy swap, and the unit is sold on the secondary market. The only question is the price of the miner: on the cheapest models of the BOX series the repair is not economically justified |
| Hash board | Hardware errors above five percent, a drop in hashrate | Low repairability. Third-party workshops do list the brand among those they service, but chips of almost all series are not on the open market and schematics are not published, so in practice it all comes down to replacing the whole unit |
Reliability, service life and warranty
There are no failure statistics for Goldshell: no return rate, no mean time between failures. The brand owner does not publish them, and neither does any other miner manufacturer. What does exist is an open firmware bug tracker, where real owner reports are visible, and a handful of user measurements.
Hashrate in those measurements is confirmed. In a user review the Mini-DOGE showed around 187 MH/s against the rated 185, while consumption came out at 222 watts against the rated 233. On the KA-BOX Pro consumption goes the other way, higher than rated, around 625 watts against 600. All three deviations fall inside the plus or minus five percent tolerance the manufacturer allows.
The weak spot is a different one, temperature. An independent review of the KA-BOX Pro records 75 degrees at 25 degrees in the room, and the manufacturer's open bug tracker holds separate reports of overheating on this same miner and on the AL-BOX II. The manufacturer's official hardware error threshold is five percent, and cases of exceeding this threshold also appear in the reports.
On noise, the very thing the brand is mostly bought for, there are no instrument measurements in the public domain. The manufacturer rates them at up to 35 dB, and third-party catalogs list 35 dB for the KA-BOX in one place and 55 in another. We quote the rated figures, but there is no independent check on them.
| Parameter | Value |
|---|---|
| Official term | 180 days. The starting point is described differently across the manufacturer's materials: from the shipping date in the main terms, from the receipt date in the service policy |
| What is covered | Repair or replacement of a unit with a defect in material or workmanship. Parts may be refurbished |
| Third-party power supply | Voids the warranty |
| Third-party firmware and overclocking software | Voids the warranty |
| Disassembly and modification on your own | Voids the warranty |
| Moisture, dust, corrosion | Not covered |
| Miners with boards swapped between units | Not covered |
| Claim procedure | A request through the support portal, remote diagnostics, then shipping the miner |
| Shipping | To the service center at the buyer's expense, back under warranty at the manufacturer's expense |
| Deadline for providing a tracking number | Seven business days, otherwise the request is closed |
| Damage in transit | To be reported within 48 hours after receipt |
| Service centers | The manufacturer does not publish physical addresses, and all communication goes through the request portal |
| Terms for the European Union | There are no separate provisions in the manufacturer's documents |
Energy efficiency by series
Efficiency cannot be compared across algorithms, so the ranges are given inside each group. The comparison with competitors goes in a separate section, and for this brand it is not flattering.
Inside Scrypt the spread is wide. Between the best miner of the group and the worst the difference is almost two and a half times. Model age matters more here than the case format.
| Series | Years | Efficiency | Comment |
|---|---|---|---|
| Scrypt, Mini-DOGE series | 2021 to 2023 | from 0.95 to 1.26 J/MH | Desktop miners from 220 to 400 W |
| Scrypt, LT series | 2021 to 2022 | from 0.9 to 1.27 J/MH | Industrial case: the spread between the two miners is almost one and a half times |
| Scrypt, E series | 2024 | 0.53 J/MH | E-DG1M, the best Scrypt miner of the brand |
| Blake2s, KD series | 2021 to 2022 | from 90 to 128 J/TH | The best J/TH figures in the catalog, not comparable with J/MH, and the network has been orphaned |
| Eaglesong, CK series | 2021 to 2022 | from 171 to 205 J/TH | Six miners, from desktop units up to 3300 W |
| kHeavyHash, KA series | 2024 | from 339 to 375 J/TH | Two compact ASICs for Kaspa |
| Blake3, AL series | 2024 | from 480 to 500 J/TH | Three compact miners for Alephium |
On Scrypt the best Goldshell miner gives 0.53 J/MH, while the best competitor miners are around 0.2, a gap of roughly three times. On Kaspa Goldshell has 339 and 375 J/TH against about 77 for the top Bitmain ASIC, almost five times.
If electricity is expensive for you and space is available, Goldshell will almost certainly lose on money. If there is no space and noise is critical, the brand has no direct competitors, but you will pay for that in efficiency.
Comparison with competitors
We compare against three competitors with different profiles: Bitmain as the all-rounder, ElphaPex as the Scrypt specialist, IceRiver as the Kaspa specialist.
| Criterion | Goldshell | Bitmain (Antminer) | ElphaPex and IceRiver |
|---|---|---|---|
| Scrypt efficiency | From 0.53 to 1.27 J/MH | About 0.21 J/MH on the best miner | About 0.19 J/MH at ElphaPex |
| Kaspa efficiency | From 339 to 375 J/TH | About 77 J/TH on the top ASIC of the KS series | About 283 J/TH on the IceRiver KS5L |
| Noise | The BOX format is rated up to 35 dB, industrial models 80 | About 75 dB, home models practically do not exist | About 75 dB on the larger miners |
| Suitability for the home | The strong point of the brand: from two to four kilograms, a shelf, an ordinary room | Not intended for it | Not intended for it, though ElphaPex has one compromise entry model |
| Range of algorithms | Five in our catalog, and not one of them for bitcoin | Eleven algorithms, but the focus is on SHA-256 | ElphaPex has one, IceRiver three |
| Bitcoin | Not one | The core specialization | No miners |
| Spare parts | The manufacturer sells power supplies only and the secondary market has control boards, but there are no hash boards for BOX units and no chips | The best secondary market in the industry | Weak |
| Economics of repair | On cheap miners a control board eats more than half the price, while on the larger models repair is justified | A new miner costs more, so repairs almost always pay off | Closer to Goldshell than to Bitmain |
| Third-party firmware | Practically none | A mature ecosystem: our AsicBoost covers 28 models | IceRiver has amateur builds, and they void the warranty |
| Warranty | 180 days, with the starting point described two ways in the documents, and no service addresses | Usually from 180 to 365 days | IceRiver also gives 180 days and also publishes no service addresses |
Goldshell wins on two points, suitability for the home and coverage of altcoin algorithms. It loses on efficiency, spare parts, repair, firmware and warranty. The brand is niche. Buy it with open eyes.
If the task is to earn money per kilowatt, look at the competitors. If the task is to put a miner where an ordinary ASIC cannot go, and you accept that you pay for it in consumption, Goldshell does what nobody else does.
Who Goldshell hardware is for
The main scenario for this brand. The BOX format means from two to four kilograms, a case the size of a thick book, a rated 35 dB and consumption from 180 to 600 watts. Neither Bitmain, nor MicroBT, nor Bitdeer has anything comparable in this format, and the home Avalon miners from Canaan only mine bitcoin.
What to take: Scrypt miners as the liveliest direction.
The barrier to entry is low: the miner costs hundreds rather than thousands of dollars, an ordinary socket is enough, and no separate room is needed. It is a sensible way to work out how pools, firmware and telemetry fit together without putting in serious money.
What to take: the cheapest miner in whichever direction interests you.
A 400 watt miner heats like a 400 watt heater. In winter that is a plus, and unlike an industrial ASIC it can be placed right in the room. But remember the other side: in a Spanish summer the same heat has to be pushed out with air conditioning, and that quietly adds about a third to the electricity bill.
What to take: an ASIC for seasonal use, not year-round.
If you believe in a particular project and want to hold it mined rather than bought, a compact miner for it is the cheapest way. Just be honest about the horizon: on Kaspa the reward falls every month, and on Nervos the halving is expected in November 2027.
What to take: only for a coin you genuinely follow.
From ten miners up. This is where the industrial models of the brand, from 2250 to 3300 watts with 80 dB of noise, come into play, and all the advantages of the BOX format disappear. At this scale it is simpler to compare with competitors on efficiency, and the comparison will not favor Goldshell.
What to take: first do the math on competitors for the same algorithm.
Wrong brand for that. The brand has no hydro cooling, no immersion, no megawatt class ASICs and no service network: the most powerful model is 3300 watts at 80 dB.
What to take: look at other brands.
Common problems and how to prevent them
The coin matters more than the miner. The main risk of this brand is not in the hardware. In October 2025 the Kadena foundation ceased operations, the coin collapsed, and three models of the KD series turned into furniture: Blake2s chips physically cannot switch to other coins. Before buying, look at the trading volume of the coin and at the state of the network. The specifications of the ASIC matter less here.
Coin liquidity. The price of a coin means nothing if there is nobody to sell it to. Handshake daily volume is measured in thousands of dollars, Alephium in hundreds of thousands. You can mine it, but you cannot cash out without losses.
Falling reward. The Kaspa block reward drops every month, roughly by half over a year, and by the middle of 2026 about 95 percent of the whole emission has been mined. On Nervos the halving is expected in November 2027. Payback has to be calculated with the decline in mind, not on today's figures.
Overheating in a compact case. Power density in the BOX format is high and the case volume is small. Owner reports in the bug tracker mention temperatures above 80 degrees at an ordinary room temperature. Do not put the miner in a cabinet, an alcove or behind furniture. Leave a gap of at least twenty centimeters.
The power connector. The most common mechanical failure. A six-pin connector that is not fully seated, or a cable thinner than 16 AWG, heats up and melts. Check that it seats with a click.
Firmware updates. Updating these miners carries a real risk of failure, especially over a wireless network. The recovery image is not published publicly and has to be requested by email. Update over a cable and only when there is a reason.
Pool incompatibility. The firmware uses its own dialect of the protocol, and with an unsupported pool the miner just sits there doing nothing: no errors, no hashrate either. Pick a pool known to work with the brand and check again after an update.
Humidity. The official limit is 65 percent. On the Spanish coast it is exceeded regularly, so the constraint is a real one. Condensation and corrosion are not covered by the warranty.
Operating tips
Half of these points are specific to this brand. Miners like this most often go into an apartment, so the advice is written for that.
- Work out the profitability before you buy, not afterAt a typical Spanish tariff most altcoin miners run at a loss in 2026, across every brand at once. Do the math on your own tariff and on a conservative coin price, and take a horizon of no more than a year.
- Check the state of the coin's networkLook at trading volume, at a living development team and at the schedule of reward reduction. Kadena in October 2025 showed how that ends: an ASIC for a dead coin is a zero.
- Put the miner in an open placeNot in a cabinet, not in an alcove, not behind furniture. At least twenty centimeters of free space in front of the intake and behind the exhaust. An external fan will not solve the power density problem.
- Keep the room temperature within limitsThe manufacturer's manual states from 10 to 35 degrees, and seller listings give from 0 to 35. At 35 degrees in the room the ASIC runs at its very limit and may go into emergency shutdown. In a Spanish summer 35 degrees in a room is easy to reach, so it is either air conditioning or a miner standing idle.
- Watch the humidityThe official limit is 65 percent. In coastal areas it is exceeded regularly, and then you need a dehumidifier or a different room. Otherwise the miner runs with condensation inside.
- Do not save on the power cableOnly the standard or a suitable power supply, a cable no thinner than 16 AWG, the connector pushed all the way in. A third-party unit voids the warranty, and a poor contact melts the connector.
- Change the factory password straight awayIt is the same on every ASIC of the brand, and anyone can look it up. After that, keep the miner in a separate segment of your home network. And never forward its port to the outside.
- Connect by cable, not over a wireless networkEspecially when updating firmware: a wireless connection is named as a typical trigger of failure. And do not connect the miner through a repeater or a powerline adapter instead of the router: that is a separate common cause of lost connectivity.
- Monitor two numbersThe share of hardware errors and the temperature. The norm for errors is up to five percent, and the hashrate should stay within five percent of the rated figure. Rising errors on an unchanged pool are an early sign of board degradation.
- Clean it every one to three monthsIn an apartment, dust, pet hair and pollen clog small heatsinks faster than large ones. Only on a powered down miner and only with compressed air.
Frequent questions about Goldshell
What is Goldshell and who owns it?
Goldshell is a brand of compact altcoin miners. Since 28 February 2024 the brand has belonged to the public company Intchains Group Limited, listed on Nasdaq under the ticker ICG. Before that it was owned by the Singapore entity Goldshell Pte. Ltd., and the deal was worth 550 thousand dollars.
Is it true that Goldshell does not build bitcoin miners?
Yes. The brand has no SHA-256 miner, and that is a fundamental difference from Bitmain, MicroBT, Canaan and Bitdeer. All twenty models in our catalog mine altcoins: Litecoin together with Dogecoin, Kaspa, Nervos, Kadena and Alephium.
How many Goldshell models are in the asic.es catalog?
Twenty, across five algorithms. Six miners on Scrypt, six on Eaglesong, three each on Blake2s and Blake3, two on kHeavyHash. Ten of the twenty draw under 500 watts.
What is the BOX series?
It is a form factor rather than a product line: it runs through every algorithm the brand covers. The case measures roughly 150 by 200 by 90 millimeters, weight ranges from two to four kilograms, power is 12 volts DC from an external supply through a six-pin connector, and the rated noise level is up to 35 dB.
Is Goldshell really quieter than a normal miner?
On paper yes, and by a wide margin: up to 35 dB for the BOX format against 75 to 80 for a standard industrial ASIC, the industrial models of Goldshell itself included. You cannot check that independently: there are no instrument measurements of noise in the public domain, and third-party catalogs list 35 dB for the KA-BOX in one place and 55 in another. The rated 35 is reached at low fan speed, and in hot weather the fan spins up and the ASIC gets louder.
Can an ASIC like this be installed in an apartment?
That is exactly what the format was made for. But a 400 watt miner heats the room like a 400 watt heater, the official humidity limit is 65 percent, and even 35 dB is a constant tone that becomes a nuisance in a bedroom.
How does Goldshell efficiency compare with the competition?
It is worse, and clearly so. On Scrypt the brand's best miner delivers 0.53 J/MH against roughly 0.2 for ElphaPex and Bitmain, a gap of almost three times. On Kaspa Goldshell offers 339 and 375 J/TH against roughly 77 for Bitmain's top ASIC, a gap of almost five times. People buy this miner for the quiet. Efficiency is not what it offers.
What happened to the KD line and Kadena?
In October 2025 the Kadena foundation announced that it was shutting down, the coin collapsed and exchanges began delisting it. The network formally keeps running on community effort, but there is no economic sense in mining it. The worst part is that Blake2s chips physically cannot be switched to other coins, so the KD5, KD6 and KD-BOX cannot be pointed at anything else.
Are the other coins still alive?
It varies. The Litecoin and Dogecoin pair is the most stable: two assets from the top thirty plus merged mining. Kaspa and Nervos are running, but the first has its reward cut every month and the second has a halving expected in November 2027. Alephium is alive, but trading volume is very thin. Handshake, the coin the brand built its first miners for, has almost no market today.
What is merged mining of Litecoin and Dogecoin?
Since 2014 Dogecoin has used merged mining. The ASIC hashes once and the solution counts in both networks at the same time. No extra energy is spent on the second coin and the revenue adds up. That is why the profitability of a Scrypt miner has to be calculated on both coins together, while a Litecoin-only calculation understates the result. The pool has to support this mode explicitly.
Which chips are used in Goldshell miners?
The manufacturer does not publish this: the official product pages simply have no chip field. All that is known concerns the brand owner: it designs its ASICs in house and outsources production, with 22 nm named publicly since 2018 and 12 nm since 2024. The only marking that ever reached open sale is the ICT560 from the LT5 Pro.
Can your AsicBoost firmware be installed on Goldshell?
No. AsicBoost is built only for Antminer. That is 28 models: the S19 and S21 series plus T9+, T19, T21 and several others.
Is there any third-party firmware for Goldshell?
Practically none. The manufacturer ships its firmware as binary files only, with no source code. Hobby builds for older models do turn up, but they are not supported solutions, and using them voids the warranty just like overclocking software does.
Is updating the firmware risky?
Caution is needed. Public reports describe a wave of failures after several builds, especially when updating over wireless, and the whole BOX series was affected. Recovery is done by writing an image to a memory card, but the manufacturer does not publish that image openly, so you request it by email from support. Update over a cable and only when there is a reason to.
Why does the miner not work with my pool?
Goldshell firmware uses its own dialect of the pool protocol. If the pool does not implement it, the ASIC accepts the connection, receives jobs and silently discards them: no errors, and no hashrate either. Pick a pool from those proven with this brand and check it again after every firmware update.
What is the factory password for the web interface?
It is the same on every ASIC of the brand and it is common knowledge, so the first thing to do after powering up is to change it. Keep the miner on a separate network segment and never forward its port to the internet.
Have there been backdoors or hidden mining in Goldshell firmware?
There are no publicly documented cases of hidden mining or a deliberate backdoor. What did happen was a 2021 security review of two models: a weak root password, a debug page with no authorization and regular calls from the device to an external domain. The author described the observation neutrally and did not claim malicious behavior. Whether these specific issues were fixed in recent builds cannot be verified from open sources: the manufacturer publishes no changelog.
What warranty does Goldshell offer?
180 days, but the starting point is described differently across the manufacturer's own materials: the general terms say the shipping date, the service policy says the date of receipt. For delivery to Europe that is a difference of two or three weeks, so the condition is worth confirming with the seller. The warranty is voided by a third-party power supply, third-party firmware, overclocking software, self-disassembly, and also by moisture, dust and corrosion. Claims are filed through the support portal, the buyer pays shipping to the service center and the manufacturer pays the return leg on a valid warranty case.
Where is the service center in Europe?
The manufacturer publishes no physical addresses, and all communication goes through the ticket portal. Its documents contain no separate provisions for the European Union either. That is why, with this brand, buying from a seller inside the European Union beats ordering direct: with a local seller the Spanish rule on liability for non-conforming goods applies for three years.
Is it worth repairing a miner that has failed?
Usually not, and it comes down to arithmetic. A control board costs from 130 to 140 dollars, while miners of the brand start at 234 dollars for the simplest BOX and reach fifteen hundred for the industrial ones. On a cheap ASIC the board eats more than half the price, and once shipping both ways is added the repair stops making sense. On a miner of seven hundred dollars and up it does make sense. The only parts that are cheap and sensible to replace are the fan, the cable and the power supply.
How solid is the manufacturer itself?
Not brilliant, and that affects support and spare parts. For 2025 the brand owner reported revenue of about 31.6 million dollars, down 21.6 percent year over year, and a net loss of about 7.4 million on a gross margin of 7.2 percent. Costs rose because of a write-down of altcoin miner inventory. The company has promised new models for the second half of 2026, the half year that is running now.
So should you buy Goldshell or not?
It depends on what for. If the goal is to earn on every kilowatt, then almost certainly no: on efficiency the brand is three to five times behind specialized competitors. If the goal is to run a miner where a normal ASIC simply cannot go, or to learn the subject cheaply, Goldshell has no direct competitors. Just run our calculator at your own tariff and look at the trading volume of the coin before you pay.
Technical statistics of the brand
Figures from our catalog and from the brand owner's public reporting.
| Parameter | Value |
|---|---|
| Market entry | The company claims 2017, and the first miner of the brand came out in 2020 |
| Brand owner | Intchains Group Limited, Nasdaq, ICG, since February 2024 |
| Models in the asic.es catalog | 20 |
| Supported algorithms | 5, Scrypt, kHeavyHash, Eaglesong, Blake2s, Blake3 |
| Coins available to mine | 6 main ones, LTC, DOGE, KAS, CKB, KDA and ALPH, plus small Scrypt forks |
| Miners for bitcoin | Not a single one |
| Models up to 500 W | 10 out of 20 |
| Power range across the catalog | From 180 to 3300 W |
| Rated noise of the BOX format | Up to 35 dB, up to 45 on one model |
| Rated noise of the industrial models | 80 dB |
| Scrypt efficiency | From 0.53 to 1.27 J/MH |
| Blake2s efficiency | From 90 to 128 J/TH |
| Eaglesong efficiency | From 171 to 205 J/TH |
| kHeavyHash efficiency | From 339 to 375 J/TH |
| Blake3 efficiency | From 480 to 500 J/TH |
| Models with our AsicBoost firmware | None: it is only for Antminer |
| Warranty | 180 days from shipment or from receipt: the manufacturer's documents differ |
| Owner revenue for 2025 | About 31.6 million dollars with a net loss of about 7.4 million |
Company achievements and what they gave the buyer
- 2020HS1 Plus: a 115 watt miner weighing 450 grams with a rated 34 dB. For the buyer this was the first case where an ASIC could simply be put on a desk.
- 2021The BOX format takes shape as a series and runs across several algorithms at once: seven new miners in one year, from the KD-BOX at 205 watts to the industrial CK6 at 3300. For the buyer that meant a choice of coin inside one case.
- 2022The intermediate Lite class appears: 1200 and 1450 watts at 55 dB instead of 80. Between a shelf in the room and a separate workshop a third option appeared.
- 2024The brand passes to a chip company. For the buyer this means public reporting from the owner: finances and risks became visible from outside, and most miner manufacturers do not offer that.
- 2024KA-BOX for Kaspa and AL-BOX for Alephium. The brand was one of the first to give a compact home format for these networks. On Kaspa it was beaten by IceRiver with the KS0 model, which came out back in 2023.
- 2024E-DG1M: 0.53 J/MH against 1.27 for the previous industrial miner of the brand on the same algorithm. In a single generation the gap to competitors on Scrypt narrowed by almost two and a half times.
- 2025BYTE: a desktop unit with two slots and hot-swap cards for different coins. You change the card without changing the case or the power supply. It is the only answer the brand has to chips that cannot be switched, the thing that killed the KD line.
- 2026Five algorithms in one catalog. No other brand in our shop covers five altcoin algorithms, and for those who mine something other than bitcoin there is often simply no choice.