We Costed 646 Ice Machines. Half Won't Make Six Months on One Cartridge.
Every number we have published about filter sizing so far has come from one manufacturer's specification sheet at a time. That is fine for a single machine and useless for comparison, because two manufacturers do not necessarily test the same way.
So we went and got data that is comparable.
The U.S. EPA's ENERGY STAR programme certifies commercial ice machines, and it publishes the full certified list as open data — including, for every model, the harvest rate in pounds per day and the potable water use in gallons per 100 lb of ice, measured to a common test procedure.
646 models. 40 brands. We downloaded the lot and did the division on all of them.
Four things that came out of it
1 · Water use varies more than two-fold for the same ice.
The lowest figure in the dataset is 11.9 gallons per 100 lb. The highest is 25.0. That is a 2.1× spread, and it is the single biggest lever on your filtration running cost — bigger than brand, bigger than cartridge choice.
2 · Two machines making the same ice can differ 58% in filtration cost.
Twenty-three certified models make between 480 and 520 lb of ice a day. Across that group, water use runs from 12.0 to 19.0 gallons per 100 lb. Same ice, same day, and one of them puts 58% more water through your cartridge than the other.
3 · Half of all certified machines miss six months on a single cartridge.
At rated harvest rate, against an Everpure Insurice Single (9,000 gallons):
| Clears 180 days on a Single | On a Twin | On a Triple | |
|---|---|---|---|
| All 646 models | 50.3% | 84.8% | 92.6% |
| Cubers (batch) | 46.9% | 84.4% | 90.8% |
| Flakers and nugget (continuous) | 63.4% | 86.6% | 99.3% |
4 · Flake and nugget machines are substantially cheaper to filter than cubers.
Median potable water use: 12.0 gal/100 lb for continuous machines against 16.9 for batch cubers. That is 29% less water for the same weight of ice, and it is why 3M's own sizing sheet lets the same cartridge cover up to 1,200 lb/day on a flaker but only 750 on a cuber.
If you are specifying a new machine and the ice format is genuinely open — a nugget programme rather than a cocktail bar — that choice has a running cost attached that nobody mentions at the point of sale.
The number underneath all of this: 11.98
Once we had all 646 figures in one place, one thing jumped out. 126 of the 134 continuous machines report exactly 12.0 gallons per 100 lb — different brands, different sizes, 293 lb/day to 1,845 lb/day, all the same number to one decimal place.
That is not 126 independent measurements. It is a floor, and you can find the floor with a calculator.
One hundred pounds of ice is 11.98 gallons of water. A US gallon weighs 8.345 lb, so 100 ÷ 8.345 = 11.98. No machine can make 100 lb of ice from less.
So the certified figure tells you, directly, how much of your filtered water never becomes ice:
| Potable gal/100 lb | What it means | Down the drain |
|---|---|---|
| 12.0 | every certified flaker and nugget machine | essentially nothing |
| 12.1 | the three leanest certified cubers | 0.1 gal, essentially nothing |
| 13.9 | a lean mainstream cuber | 1.9 gal, about 14% of intake |
| 15.1 | the dataset median | 3.1 gal, about 21% |
| 16.9 | the median cuber | 4.9 gal, about 29% |
| 19.9 | a thirsty mainstream cuber | 7.9 gal, about 40% |
| 25.0 | the thirstiest certified machine | 13.0 gal, about 52% |
A batch cuber has to do this. It freezes clear cubes by freezing part of the water in the tray and flushing the rest, because that is where the minerals have concentrated. Purge more, less mineral stays on the evaporator, less scale. Purge more, more gallons through your cartridge. A continuous machine scrapes or extrudes instead, and freezes nearly everything it takes in.
One thing in that table we would not rely on. Three certified cubers — one ICETRO and two Noriota — report 12.1, which on this arithmetic means a batch machine purging essentially nothing. We cannot see how a batch cuber does that, so we would want to read those models' own specification sheets before sizing anything on the figure. We have left it in the table because it is what the certified list says, and flagged it because it looks wrong.
The floor is arithmetic from water density, not a published figure, and the reading of the gap above it as purge is ours rather than any manufacturer's. But it means a certified water figure is not an arbitrary number you have to look up — it is a direct measure of how much of what you paid to filter goes down the drain.
Why this comparison is legitimate and a brochure comparison is not
This matters enough to spell out, because we got it slightly wrong ourselves last week.
Manufacturers publish water figures from their own testing. The conditions are usually stated, but they are not necessarily identical between companies, which is why we have been careful elsewhere on this site not to declare winners from spec sheets.
ENERGY STAR figures are produced to a single test procedure for certification. That is what makes the table below worth reading at all.
Water use by brand, cubers only
Brands with twelve or more certified batch models, by median potable water use:
| Brand | Models | Median gal/100 lb | Range |
|---|---|---|---|
| Noriota | 12 | 12.4 | 12.1 – 12.6 |
| MAXX ICE | 38 | 12.8 | 12.5 – 16.7 |
| True Refrigeration | 24 | 14.5 | 13.8 – 25.0 |
| Summit | 12 | 14.8 | 12.8 – 18.2 |
| SAPPHIRE | 28 | 15.1 | 12.4 – 22.5 |
| ATOSA | 13 | 15.3 | 12.9 – 19.9 |
| Hoshizaki | 32 | 16.8 | 13.6 – 23.0 |
| EMPURA | 17 | 17.3 | 16.5 – 19.3 |
| Koolaire | 22 | 18.1 | 13.6 – 20.0 |
| Scotsman | 42 | 18.7 | 17.0 – 23.0 |
| Manitowoc | 103 | 19.0 | 13.5 – 20.0 |
| Ice-O-Matic | 58 | 19.0 | 17.0 – 22.5 |
Read the range column, not the median
This is the part that stops this being a league table.
The ranges overlap almost completely. Hoshizaki's certified models run 13.6 to 23.0. Scotsman's run 17.0 to 23.0. Manitowoc's run 13.5 to 20.0. The best Manitowoc is better than the median Hoshizaki, and the worst Hoshizaki is as thirsty as the worst Scotsman.
So "which brand uses less water" is close to the wrong question. The model matters far more than the badge, and a brand median tells you almost nothing about the specific machine in your back room.
Where we overstated it
Our Scotsman page said a Scotsman puts roughly 22% more water through a cartridge than a Hoshizaki. That number is correct for the two specific models it names — Scotsman's 18.0 against the KM-520MAJ's 14.7 — but the section it sits in reads as a statement about the two brands, and across the certified lineups the gap is 11% on medians with heavily overlapping ranges.
We have amended that page. The model-to-model comparison stands; the brand-level implication was further than the evidence goes.
One honest limitation
Everything here is an ENERGY STAR certified machine. Certification is voluntary, and a manufacturer submits the models it expects to do well.
That is a real selection effect, and it runs in one direction: the uncertified machines are more likely to be worse than this list than better. If your machine is not in the ENERGY STAR database, do not assume it sits in the middle of these figures. Assume it sits at the thirsty end until its own spec sheet says otherwise.
How to use this for your own machine
The arithmetic is the same one we use everywhere on this site.
- Find your machine's potable water figure — gallons per 100 lb of ice, on its spec sheet or in the ENERGY STAR database.
- Multiply by your real daily production ÷ 100. Not the rated figure. If the bin is emptied once a day and holds 300 lb, use 300.
- Divide your cartridge's rated capacity by that. Insurice Single is 9,000 gallons, Twin 18,000, Triple 36,000.
- Compare to 180 days and take whichever is shorter.
We have already done steps 1 to 4 for four full ranges: Hoshizaki KM, Scotsman Prodigy Plus, Manitowoc Indigo and Ice-O-Matic.
And one caveat that applies to every figure on this page: these are rated harvest rates. Most machines do not run at rated output. A machine at 60% duty uses 60% of the water and gets about 1.7× the cartridge life — which is why the honest answer to "do I need a Twin" is usually "tell me what it actually makes."
Send us the model and we will look it up
Tell us the make and model and roughly what it produces on a normal day. We will pull its figures, do the division, and tell you which system fits and what your real interval is.
If the answer is that the cheapest cartridge clears six months on your machine, we will tell you that — it is true for half the models in this dataset.
Related reading
- Hoshizaki KM water filter sizing — eleven models, worked through
- Scotsman Prodigy Plus water filter sizing — eight models, worked through
- Manitowoc Indigo water filter sizing — twenty-eight models, and a 41% split inside one lineup
- Ice-O-Matic water filter sizing — twenty-six models, cubes against nugget
- How often to replace a commercial water filter — capacity versus calendar
- The complete water guide for bars, cafés and restaurants — how it all fits together
Harvest rates and potable water figures are from the ENERGY STAR Certified Commercial Ice Machines dataset published by the U.S. Environmental Protection Agency at data.energystar.gov, downloaded 2 October 2026 — 646 records across 40 brands. Cartridge capacities are from Pentair Everpure's Insurice specification sheets. All cartridge-life figures are our arithmetic from those published inputs, calculated at rated harvest rate.