Containerised mining is sold on density. You put a large amount of heat into a small steel box and then you have to get it out again, continuously, in whatever weather the site has. How well that works is decided less by the fans you bought than by whether hot air and cold air are kept apart.

The problem containment solves

A miner draws cool air in one face and pushes hot air out the other. In an open layout that hot air does not politely leave. It drifts, it finds the nearest low-pressure area, and a good deal of it goes straight back into the intake of the machine next to it, or back into the same machine.

That is recirculation, and it is the single most common reason a site with adequate cooling capacity still runs hot. The fans are moving plenty of air. They are moving the same air round in a loop.

The symptom is a temperature gradient that does not match the layout. Machines in the middle of a row running hotter than the ones at the ends. Units at the top of a rack hotter than the bottom. Chip temperatures that climb through the day and never fully recover overnight. Operators usually respond by adding fan speed, which raises power draw and noise and fixes very little, because the problem is not volume. It is that hot and cold are mixed.

Hot aisle versus cold aisle, and why the choice matters less than the seal

There are two ways to arrange this. Contain the cold side, so that cool intake air is sealed into a plenum and everything outside it is hot. Or contain the hot side, so exhaust is captured and ducted out while the rest of the container sits at intake temperature.

Hot aisle containment is generally the better fit for mining containers. The reason is practical rather than theoretical: the volume you are containing is smaller, the ducting runs are shorter, and the rest of the box stays at a temperature a technician can actually work in. Cold aisle containment leaves the working space at exhaust temperature, which in a mining container in summer is unpleasant at best.

What matters far more than which one you pick is how well it is sealed. A containment system with gaps is a containment system that leaks the hot air you just captured back into the intake. Blanking panels in empty rack positions, brush strips where cables pass through, and sealed joins at the top and ends of the aisle are not optional extras. They are the thing doing the work.

Where containment fails in practice

• Empty rack slots left open. A missing miner is a direct path from hot side to cold side. Blanking panels cost very little and are the highest-return item on this list.
• Cable penetrations. Power and network cables have to pass through the barrier somewhere, and an unsealed penetration is a hole.
• Doors propped open. Usually during maintenance, usually forgotten. A propped containment door defeats the system for as long as it is open.
• Mixed hardware in one aisle. Machines with different airflow rates or different depths create pressure imbalances that pull air the wrong way.
• Filters nobody changed. A loaded filter restricts intake, the fans work harder, and the pressure balance across the barrier shifts. Containment and filtration are the same problem viewed from two ends.

Getting the pressure right

A contained aisle works on a small pressure difference. The cold side should sit very slightly above the hot side, so that any leak pushes cool air into the hot aisle rather than pulling hot air back across.

Too little pressure difference and recirculation creeps back in. Too much and you are simply pushing conditioned air straight out of the container without it doing useful work, which is wasted fan power.

This is worth measuring rather than assuming. A differential pressure reading across the barrier tells you in seconds what a week of guessing at fan curves will not.

What it is reasonable to expect

Containment does not add cooling capacity. It makes the capacity already installed reach the machines rather than circulating past them. On a site that was recirculating badly the effect can be substantial, and on a site that was already well arranged it may be modest.

The honest way to find out is to measure intake temperature at several points in the rack before changing anything. If intake temperatures vary widely between positions, there is recirculation to recover. If they are already uniform and close to ambient, the cooling problem is somewhere else and containment will not fix it.

That distinction matters because containment is a real cost, and spending it on a site whose actual constraint is inlet air temperature or undersized ducting produces a tidier container and the same chip temperatures.

Containers versus buildings

In a purpose-built facility containment is a design decision made once. In containers it is partly made for you by the enclosure, which is why container manufacturers increasingly build the airflow path into the unit rather than leaving it to the operator.

If you are specifying containers, ask how the airflow path is arranged, whether blanking is supplied for partial fills, and what happens to the pressure balance when a row is half populated. A unit that only performs correctly at full occupancy is a unit that will underperform for the first months of every deployment.

Rax Mining builds natural gas modular data units with the airflow path designed in rather than added afterwards. If you are comparing containers and want to understand how each one actually handles heat at partial fill, talk to us.