Mice · Keyboards · Headsets · Chairs
Every CEMMORONX input device reports at 8 kHz and every stage between your finger and the operating system is measured, named and printed on this page — including the stages we are not proud of. Nine peripherals, three chairs, and no adjective doing a number's job.
What we make
Mice, keyboards and headsets are what CEMMORONX is for. The chairs exist because the same people asked for them and because an eight-hour session is an engineering problem too — but they are a line of three, not half the company, and this page is built to show that ratio.
Four shapes from 47 g to 68 g, each named for what it weighs.
Magnetic analog boards with rapid trigger, plus one hot-swap mechanical.
One wireless flagship, one light wired pair. Both with a detachable mic path.
Three seats on one frame platform. No latency figure — a different number entirely.
Two cloth pads and one hard pad, cut to the sizes our own testers use.
Skates, cables, earpads, keycaps, castors and gas lifts, for as long as we make the product.
The organising idea
Nearly every peripheral brand quotes one of these four numbers and calls it latency. We publish all four, the method behind each, and the total — measured on our own rig with a 10,000 fps camera and a logic analyser, on 2.4 GHz wireless at 8 kHz.
0.05ms
The switch changes state and the microcontroller knows. Magnetic switches are read as an analog value, so this is a sampling interval rather than a debounce delay — there is no debounce, because nothing bounces.
0.125ms
The device is only allowed to speak on its own schedule. At 8 kHz that slot comes round every 0.125 ms, so on average you wait half of it. At 1 kHz you wait eight times longer — this is the single stage a polling rate improves.
0.62ms
Our 2.4 GHz link, median, in a room with forty competing devices. This is the largest stage and the one nobody likes to print. A wired product removes it entirely, which is why BASE 68 and ANVIL TKL still exist.
0.31ms
Your operating system's USB stack and input queue. We do not control this one and we measure it anyway, because leaving it out is how a 1.1 ms product becomes an advertised 0.8 ms product.
Platforms
Choosing a peripheral is really choosing a sensor, a switch and a radio. We use three platforms across nine products, so a spec here applies to every product that names it — and a change here is a change to the whole range rather than a quiet substitution in one model.
AXIS 54 · AXIS 47 · CURVE 61
Nobody plays at 30,000 DPI. It is on the sheet because the sensor supports it; the setting that matters is the 50 DPI step size, which is what lets you land on your actual sens instead of near it.
RAIL 60 HE · RAIL 80 HE
Contactless, so it does not wear a contact out and does not need debouncing. It also cannot be hot-swapped for a mechanical switch — a magnetic board is a magnetic board for life, which is the trade we made and you should know before you buy one.
Every wireless product
One dongle, one device, no shared receiver. It is less convenient than a universal receiver and it is the reason the 99th percentile figure above is 1.4 ms rather than something we would have to bury.
The range
AXIS 54 weighs 54 g. AXIS 47 weighs 47 g. Weighed without cable, with the standard skates on, on a 0.1 g scale. It is a constraint rather than a gimmick: a revision that gains weight gets a new name instead of a quietly edited spec sheet.
The symmetrical flagship. 124 × 63 mm, CX-30K sensor, optical main switches, 90 h at 1 kHz and 26 h at 8 kHz.
The same mouse for a smaller hand. 115 × 60 mm, shorter hump, identical internals. 80 h at 1 kHz, 23 h at 8 kHz.
Right-handed ergonomic shell with a thumb rest and two side buttons. 95 h at 1 kHz, 28 h at 8 kHz. No left-handed version, and none planned.
Wired, on a paracord that does not remember its packaging. CX-30K sensor, same 8 kHz firmware, no battery — which is why it is the lowest-latency mouse we sell.
60 % magnetic analog board. Actuation anywhere from 0.1 to 4.0 mm, rapid trigger in 0.1 mm steps, per-key. Wired 8 kHz or wireless on CX-LINK.
The same switch and firmware in tenkeyless, with a function row, a volume encoder and a gasket mount. The board most of our own testers use.
Hot-swap mechanical for people who want to choose their own switch. Optical linears in the box, five-pin sockets, wired only, same 8 kHz firmware.
50 mm drivers, CX-LINK wireless plus USB-C, 70 h, 296 g on the head. Detachable boom with a cardioid capsule and a hardware mute that cuts the capsule, not the software.
40 mm drivers, 238 g, wired over USB-C or 3.5 mm. No battery, no dongle, no pairing — and no wireless version of it, because the weight is the product.
No products in that category.
The seating line
Two of the three figures at the top of this page do not apply to a chair. There is no polling rate and no radio; there is a frame, a foam density and a number of hours before your back tells you the session is over. So the chairs are sold on their own number — and the five-year warranty is the one claim that does carry across the whole company.
Cold-cure foam at 55 kg/m³, 4D armrests, 165° recline, class-4 gas lift, steel frame. Rated to 150 kg. For 165–185 cm.
The same platform, wider seat base and taller backrest, on a 350 mm five-star aluminium base. Rated to 180 kg. For 180–200 cm.
Elastomeric mesh back instead of a padded shell — cooler, and without the racing bolsters, which some people want and some people cannot stand. Rated to 130 kg.
How a product happens
01
Grip length, palm width and finger span from 400 players across claw, palm and fingertip grips. The shape starts here, not at a render.
02
Printed, held, thrown away, printed again. Roughly thirty iterations before a shell is tooled, because tooling is the point of no return.
03
Sensor position and lens height set against the tracking rig, on six surfaces, before the internal layout is frozen around it.
04
Optical for mice, magnetic for the RAIL boards. Force curve measured, not specified — a 40 gf switch that measures 47 gf goes back.
05
Antenna tuned in the chamber with forty competing devices running, because an empty chamber flatters every radio ever made.
06
Onboard profiles, no background service, no account. Signed images, and every release stays downloadable after it is superseded.
07
1.8 m drop onto oak in 26 orientations, 100 M switch cycles, and for chairs the full BIFMA X5.1 cycle set.
08
Moulded pulp, no plastic tray, no foam block. Spare skates, spare keycaps and the hex key go in the same box.
Build
Nothing on a CEMMORONX product is glued shut to make it thinner. Skates, keycaps, earpads, cables, castors and gas lifts are all ordinary orderable parts, and the shells are uncoated on purpose — a soft-touch coating feels better in a shop and worse in year two.
Measured, not asserted
Foam does not fail, it sags — slowly, and then all at once. These are our own cycle-rig results, expressed as the percentage of original seat height retained. The rig runs a 100 kg impact head at BIFMA X5.1 rates; a heavy user reaches 100,000 impacts in roughly four years of full-time use.
CEMMORONX Tuner
Tuner is a single page that talks to the hardware over WebHID. There is no installer, no background service, no account and no telemetry. Everything you set is written to onboard memory, so a configured device behaves identically on a machine that has never heard of us — including the tournament PC that will not let you install anything.
Four onboard profiles per device: DPI steps, actuation points, rapid-trigger depth, debounce, lighting and macros. Close the page and nothing is lost, because nothing was stored on your computer in the first place.
Tuner never asks who you are and never sends anything anywhere. There is no usage data, no crash reporting and no profile sync, which means there is also no profile sync — a real cost, stated plainly rather than hidden.
Every firmware image we have ever shipped stays downloadable, signed, with its changelog. If a release makes your board worse, you are one file away from the one that did not.
No SOCD cleaning, no snap-tap, no null-bind resolution. Several tournament organisers now ban that behaviour, and firmware that is legal on one stage and disqualifying on the next has no business in your bag.
WebHID is not available in Safari or Firefox. That is a genuine limitation of the approach we chose, and the alternative — shipping a background service to every customer — was worse.
Quality
If a figure appears on this page it was measured here, not taken from a component datasheet. Where we quote a supplier's number we say so — the CX-30K's 750 IPS is the sensor vendor's figure, and it is the only one on this page that is.
A solenoid presses the switch while a 10,000 fps camera and a logic analyser watch both ends. This is where the four latency figures come from.
A linear stage sweeps the sensor across six surfaces at up to 6 m/s, checking for smoothing, angle snapping and jitter that a spec sheet will not show.
100 million actuations per magnetic switch, 80 million per optical, with force measured at intervals rather than only at the end.
1.8 m onto oak in 26 orientations, then 500 tumbles in a drum. A peripheral that survives the desk should survive the bag.
Forty competing 2.4 GHz devices in one room, running while we measure. The published latency figures are the congested ones.
BIFMA X5.1: 100,000 seat impacts, 300,000 tilt cycles, 120,000 armrest cycles, plus a static load test to twice the rated capacity.
Shipping & returns
A mouse weighs 54 grams and a chair weighs 25 kilograms, and pretending both cost the same to move would mean one of the two prices is a lie. Peripherals ship free and tracked everywhere; chairs ship free inside the markets listed at checkout and are quoted before payment everywhere else.
Questions
Contact
There is no checkout here. Send the product, the quantity and the destination, and we reply within one working day with a payment link and a shipping date.