Jasminer: the Sunlune company and Etchash miners

Jasminer is a trademark of the Chinese company Sunlune Technology (Beijing) Co., Ltd., listed in Chinese documents as 中科声龙. The company is older than the brand itself: by its own account it has operated since 2009, while miners under the Jasminer name have been sold since 2022. The specialization is narrow: one algorithm, Etchash: Ethereum Classic and its younger relative ETHW.

What sets Jasminer apart from almost every other brand in our section is that here the chip carries both the compute and the memory. Etchash requires a large amount of fast memory, and the company's entire engineering is built around placing memory next to the logic. An independent lab that disassembled the first generation chip found something atypical for the industry there: the logic die and the memory die are joined by direct copper bonding at the wafer level. As far as can be judged from published chip teardowns, no other examples of this technology in mining have been described.

The company is private and does not disclose its financials. In September 2022 it closed an investment round, and Intel's investment fund was named among the participants. The legal structure is complex: the operating company is in Beijing, the head office is in Singapore, and the holding company is in the Cayman Islands. The process node of the second and third generation chips has not been named anywhere.

We will say it right away, before any tables. Jasminer is the efficiency leader for Etchash, but that leadership comes from the X44 series miners at 0.10 J/MH, which are not in our catalog. The two models that are available, the X16-P and X16-Q, belong to the mid segment of the lineup and consume about 0.32 J/MH, three times more for the same hash. At today's Ethereum Classic price, neither of them turns a profit under Spanish electricity rates.

A separate note on the block reward. On July 22, 2026, at block 25 million, the Ethereum Classic reward dropped from 2.048 to 1.6384 ETC. This happened less than a month ago, and some public profitability calculators have not yet caught up with the change: they calculate using rewards from earlier years, 2.56 and even 3.2 ETC, and overstate income by one and a half to two times. All figures on this page are calculated using the new one.

The history splits in two: for thirteen years the company made chips not meant for mining, and the Jasminer brand has existed for only the last four years. The second stretch is dense, containing three generations of dies.

• 2009According to the company itself, Sunlune Technology was founded this year. There is no mining in its work: the company handles custom chips.
• 2018A pivot toward high performance computing chips with an emphasis on memory bandwidth. This is the very pivot that, four years later, would produce miners for Etchash.
• 2022The Jasminer brand and the first X4 series go on sale. The first generation processor itself is dated 2021 in the company's release, while sales began in 2022. The miners are built around a chip with five gigabytes of memory on die. The lineup includes several form factors right away: desktop, quiet three slot and single slot server.
• 2022In September the company closes an investment round, and Intel's investment fund was named among the participants. This is rare for a mining hardware manufacturer: major funds almost never enter this niche.
• 2022The year of the Merge. On September 15, Ethereum moves off proof of work, and the entire worldwide fleet of GPUs and Ethash miners is left without its main coin. Part of the hashpower flows into Ethereum Classic, part into ETHW. For Jasminer this is both a catastrophe and a chance: the market shrank, but competitors began leaving it, while Jasminer stayed.
• 2023In spring the X16-Q launches, the first ASIC on the second generation chip. Eight gigabytes of memory instead of five, a quiet case, claimed noise of about forty decibels. In August it is joined by the X16-P, an industrial variant of the same generation at 5800 MH/s. Both miners are in our catalog.
• 2024Expansion of the second generation: the X16-Q Pro in April, the X16-QE in May with six gigabytes of memory instead of eight. The QE is a cost reduced version, and what was cut is exactly the memory, and with it the miner's service life.
• 2024The third generation begins, the X44 series. Claimed efficiency of about 0.10 J/MH, roughly three times better than the X16 series and better than anything competitors have for Etchash.
• 2025Third generation shipments: the X44-Q at 7800 MH/s and the X44-P at 23400 MH/s. The brand secures first place for efficiency in its niche.
• 2026On July 22 the Ethereum Classic block reward drops by a fifth, from 2.048 to 1.6384 ETC. No new chip generation after the X44 has been publicly announced, and we found no separate company press release for 2026. Sales and support continue in the meantime.

Three chip generations and four approaches to the case. We go through the whole lineup, including what is not in our catalog: without the full picture it is unclear where our two models stand within it.

The asic.es catalog carries two Jasminer models, both running Etchash, both from the second generation of chips.

We do not show an algorithm column: there is only one algorithm for both miners.

There is one algorithm, Etchash. The ASICs compute regular Ethash as well, and that is what ETHW is mined on: the only difference between the algorithms is epoch length. The brand has never had miners for SHA-256, Scrypt, or anything else.

There are two coins, and the second is much worse than the first.

This is where Jasminer has the most interesting engineering of any brand in this section. Etchash cannot be sped up by logic alone: the calculation keeps referencing a large file in memory, and the ASIC is bottlenecked not by computation but by memory access speed. So the whole design is built around getting the memory as close to the chip as possible.

In the first generation chip, an independent reverse engineering lab found a technology that is rare for the industry: the logic die and the memory die are joined by direct copper to copper wafer level bonding, with a bump pitch under five microns. The memory itself is not conventional, but embedded stacked DRAM. We found no other documented examples of this technology in mining chips. For the second and third generations, the company discloses neither the process node nor the die architecture, but the same approach of memory on chip has been kept: for the X16 series that is eight gigabytes with a claimed bandwidth of about one terabyte per second.

Control boards

Power supply. On both of our models the power supply is built in, and this is a key point: you cannot swap it for a separate unit the way you can on Bitmain miners. Input is only 200 to 240 volts, both spec sheets confirm this. Claims from some sellers about running on 110 volts have no official confirmation, and given the power draw involved, they should be treated with skepticism.

The manufacturer does not specify the input connector type anywhere. Before ordering, ask the seller for a photo of the back panel of the specific batch: it will save you one order of adapters.

The brand publishes no chip designation for any generation, and the process node is known only for the first die and only from an independent teardown, which is why those cells in the table above are empty. The number of hash boards per model is also not published. The only known figure is the total chip count for the X16-P, nine chips, and even that comes from supplier data, not the factory. For a brand selling ASICs worth several thousand dollars, this is notably less transparency than Bitmain or MicroBT offer.

The brand has four case types, but the main division runs along noise rather than power. This is reflected directly in the name: the letter Q is the quiet version and the letter P is the industrial one.

The firmware is proprietary, Linux based, and there are no other options. No third party firmware exists for Jasminer: none of the known teams that make alternative firmware for Bitmain have taken on this brand. Our AsicBoost firmware also does not install on Jasminer: it is built for Bitmain hardware.

Default access: login root, password root. This is stated directly in the manual, and it is the first thing to change after powering on. There is an access point mode for initial setup: the ASIC raises its own wireless network with the password 12345678 and responds at address 83.85.78.1. That address comes from the manufacturer's manual and looks odd for a local network, so check it against the label on the miner itself. The access point password is also the same standard one on every ASIC of the brand, and the manufacturer itself has publicly warned users about this. The wireless module works only in the 2.4 gigahertz band.

There are four operating modes: economy, balanced, performance and maximum. The last one only appears after a firmware update, before a certain version it simply is not in the interface. These are preset profiles, not an automatic frequency and voltage tuning for a specific chip: there is no auto tuning here in the sense that Bitmain has it.

The miner reboots on its own two or three times a week, and that is normal: it is how the DAG file gets updated. Loading the DAG after startup normally takes up to forty minutes, and for that whole time hashrate is either absent or reduced. If loading takes longer than that, the cause is usually not the miner but the pool connection. Firmware updates are distributed only officially: through the brand's website and its messenger community.

The parts market for Jasminer is noticeably thinner than for Bitmain or MicroBT. Some things can be found, but the selection is narrow.

The brand's repairability is average, and it is distributed unevenly: what is on the outside can be fixed reasonably well, what is inside almost cannot.

The manufacturer does not publish separate failure statistics, and no one else has them for this brand either: the installed base is small and there is little forum material. We have to judge based on our partner's service manual and parts listings, which is indirect data.

Since the control board is sold separately by several suppliers, it fails often enough for a market to exist for it. Fans are a normal wear item, no surprises there. There is nothing to judge about hash boards precisely because there is no market for them.

Of firmware issues, the manufacturer has officially acknowledged two and fixed both: hashrate display dropping out and reboots tied to networking. The routine reboots two or three times a week are not a defect but a DAG file update, and many owners mistake them for a malfunction.

Service life. Both catalog miners were released in 2023, so as of August 2026 they are about three years old by model date, although a specific unit could still be new. The main limiting factor on service life here is not wear but memory: eight gigabytes will be enough until roughly 2038 for Ethereum Classic. On this metric the ASICs will outlast their own bearings, and most likely the owner's interest as well.

Both of our models are close in power draw, 0.32 and 0.33 J/MH, but they trail the brand's flagships by a factor of three.

Figures for our catalog and public data for models we do not carry.

Where it stands in the industry. As of August 2026, the efficiency ranking of Etchash miners looks like this: Jasminer X44-Q at 0.10 J/MH, X44-P at 0.11, iPollo V2H and V2X at 0.14, iPollo V2 at 0.15, three Bombax EZ100 variants from 0.18 to 0.20, Bitmain's Antminer E11 at 0.26, and only then our X16-Q at 0.32 and X16-P at 0.33. Older Bitmain E9 series ASICs are worse still, from 0.60 to 0.80. So the brand holds first place in the niche, but these two specific models sit in tenth and eleventh place.

Now for the money. Take the miner's efficiency in joules per megahash and divide 0.0153 by it: the result is the maximum electricity price in euros per kilowatt hour at which it breaks even. For the X16-P that is about 4.6 cents, for the X16-Q about 4.8 cents. For the flagship X44-Q, the same formula would give about fifteen cents.

We run the numbers at Spanish electricity rates. The X16-P at 1900 watts generates roughly 2.13 euros a day in gross revenue at an Ethereum Classic price of about six euros, while consuming 45.6 kilowatt hours. At a rate of ten cents that is minus 2.43 euros a day, at fifteen cents minus 4.71, at twenty cents minus 6.99. The X16-Q generates about 0.72 euros and consumes 14.9 kilowatt hours: minus 0.77, minus 1.51 and minus 2.26 euros a day respectively. This is without pool fees, cooling, or hardware depreciation.

Calculate profitability for a specific model →

Comparing with three brands that make miners for the same algorithm: Bitmain as the generalist with its newest E series ASIC, iPollo as the closest chaser on efficiency, Bombax as the new player in the niche.

Looking at the brand as a whole, Jasminer is the best engineering in its niche. Bitmain, iPollo and newcomer Bombax also stayed in the space after the Merge. But it's Jasminer that has shipped two new chip generations since then and held onto first place on power efficiency: the X44 flagship draws about forty percent less than its closest rival and 2.6 times less than Bitmain's newest ASIC. Weak points are not hard to find. Closed technical information, no hash boards sold separately, no service of its own in Europe, and a murky history of websites posing as representative offices.

Now our two models. The X16-P and X16-Q are mid range models from a lineup that is three years old. They have eight gigabytes of memory and a very long lifespan on the algorithm, the X16-Q has a quiet enclosure, and both are cheaper than the flagships. But on power draw they are three times worse than the X44, and at today's Ethereum Classic price neither one turns a profit at Spanish electricity rates. Buying them today only makes sense with very cheap electricity or as a bet on the price rising.

Profitability calculators are lying right now. On July 22, 2026 the Ethereum Classic block reward dropped from 2.048 to 1.6384 ETC. As of our check in late July 2026, some popular calculators were still using 2.56, and one major one was using 3.2, a value from four years ago. That inflates income by a factor of one and a half to two. Check what reward a calculator is using before you trust it.

Calculate profitability before buying. Both miners have a break even point of about 4.6 to 4.8 cents per kilowatt hour. The formula is simple: divide 0.0153 by the efficiency in J/MH to get the maximum electricity price in euros per kilowatt hour. The coefficient is worked out at an Ethereum Classic price of about six euros. At standard Spanish rates, both ASICs run at a loss.

The next reward reduction. The reward drops by a fifth every five million blocks. The next reduction, to 1.31072 ETC, is expected at block 30 million, roughly in 2028 or 2029. An ASIC bought today needs to pay for itself before that point.

ETHW will die for these miners before ETC does. Sellers often list both coins in the description, ETC and ETHW. But the ETHW DAG file grows twice as fast, and eight gigabyte ASICs will stop being able to mine it sometime between mid 2027 and early 2028. There's no point buying these miners for ETHW.

Half the network in one pool. As of August 2026, the largest pool controls about 52 percent of Ethereum Classic's hashrate, more than half the network. It has no comparable rival, and the rest is spread across dozens of small pools. This is a structural risk: cheap rented hashrate is exactly what powered the attacks on this network in 2019 and 2020. When choosing a pool, it's worth deliberately avoiding the biggest one.

Restarts two or three times a week are normal. That's how the DAG file gets updated. Owners often mistake this for a fault and start swapping hardware. Check the log before doing anything.

No hash boards for sale. If a hash board fails, there is only one standard path: return to the factory. From Europe that means shipping and customs clearance at your expense one way, plus turnaround time. Factor that into your decision.

Websites posing as representative offices. The brand has only one official website, and the company has publicly stated this itself, naming the fake domains. Steer clear of sites with additions to the brand name and prices several times above market.

Half of these points are specific to Jasminer and to the state the Ethereum Classic network is in today.

1. Change the password immediatelyThe default login and password are root and root, the access point password is 12345678, and they're the same on every miner. The manufacturer itself has publicly warned about this. First thing to do on first boot.
2. Check the memory size on your invoiceThe brand has variants with five, six, and eight gigabytes. This determines whether the miner will last until 2028, 2032, or 2038. The invoice should say eight.
3. Confirm the variant by its letterQ is the quiet version and P is the industrial one. Pro and QE have different specs despite similar names. The noise difference between the X16-Q and X16-P is almost double, forty decibels against seventy five.
4. Size the circuit by rated current, not consumptionThe X16-P draws 12.5 amps and the X16-Q 3.6 amps, both at 200 to 240 volts. The X16-Q's built in power supply is rated for 1200 watts while the ASIC consumes 620, so there's headroom, but still size the circuit to the rated current. On the X16-P the stated supply rating matches the draw, so that headroom is not there, and the figure is worth confirming separately.
5. Don't count on 110 voltsOnly the 200 to 240 volt range is officially specified. Claims by individual sellers about a universal input are unconfirmed.
6. Ask for a photo of the back panelThe manufacturer does not specify the input connector type anywhere, so one photo before ordering saves you an extra order of adapters afterward.
7. Use a wired connectionThe wireless module only supports 2.4 gigahertz, and most complaints about slow DAG loading and pool disconnects are resolved by switching to a cable.
8. Keep the room coolThe spec allows up to forty degrees, but the protection trips at a chip temperature of around seventy, and the service recommendation is no warmer than twenty five in the room. For Valencia in summer that's a challenge of its own.
9. Blow it out once a yearThis is the official recommendation. Clogged grilles and worn fans are the most common cause of performance drops.
10. Give the miner forty minutes after startupLoading the DAG file normally takes up to forty minutes, and during that time the hashrate is reduced or absent. Don't draw conclusions about the miner in its first hour.

Figures from our catalog and public data about the brand.

• 2018The company's pivot toward high memory bandwidth chips. They made that decision four years before the first miner. Everything the brand does today grew out of it.
• 2022Memory on the chip bonded directly to the logic die via wafer level copper to copper bonding, with a contact pad pitch under five microns. An independent lab that took the first generation chip apart found this. Published teardowns show nothing like it in mining chips.
• 2022An investment round in September 2022, with Intel's fund named among the participants. An investor like that is rare in the mining niche.
• 2022The company did not abandon Etchash after the Merge. When Ethereum switched to proof of stake and the niche shrank by several times over, most manufacturers wound the direction down. Jasminer has since shipped two more generations of its own chips, and the later of them, the X44 series, is still the most efficient in the niche.
• 2023Second generation: eight gigabytes of memory instead of five. Memory size on its own does not change hashrate, but the service life on Ethereum Classic moves from December 2028 to 2038.
• 2023A quiet enclosure at forty decibels, 1950 MH/s and 620 watts, with no clock cut to buy the silence. Before this, mining Ethereum Classic at home meant either GPUs or the noise of an industrial miner. The X16-P of the same generation runs at 75 decibels.
• 2024Third generation X44 at 0.10 J/MH. Three times better than its own second generation and about forty percent better than the closest competitor.
• 2025X44-P shipments at 23400 MH/s. As of August 2026, this is the most powerful ASIC for this algorithm on the market.