Benchmarks
Every figure below has a method and a signed run file, including the ones that go against us. NITROBENCH 4 is our own harness, which is a conflict of interest. We publish the harness, the profiles and the raw run files, and print the third-party index next to ours where the two disagree.
Bench 2 · where every figure below was taken24 °C ± 0.5, still air, enclosure closed30-minute unbroken run12 runs per figure, median reported
01The headline
Against 870 MB/s from a retail Mk V on the same profile on the same day. A 36% lead at a typical resident set, and a lead we lose entirely past 22 residents.
02Throughput
- Nitrosoft, generation 5
- Cybrscape, measured by us
- Nitrosoft, generation 4
Cybrscape units measured 14 June 2029: two retail Mk V, two retail Mk IV, unmodified firmware. Our measurements of retail hardware, not their published specification. If they measure differently we would like to see the profile.
Figures
| Unit | MB/s |
|---|---|
| NX5/X | 1,180 |
| Mk V | 870 |
| NX5/T | 690 |
| NX4/T | 520 |
| Mk IV | 430 |
| NX5/S | 410 |
- Nitrosoft, generation 5
- Cybrscape, measured by us
- Nitrosoft, generation 4
The NX4/T wins this outright because its tooling is paid for and no longer amortised into the price.
Figures
| Unit | cr (lower is better) |
|---|---|
| NX4/T | 654 |
| NX5/X | 949 |
| NX5/T | 1,072 |
| NX5/S | 1,195 |
| Mk V | 1,931 |
| Mk IV | 2,233 |
03The crossover
Where we stop winning
- NX5/X
- Mk V
A wide fold table costs lookup time. Up to about eighteen residents we are ahead by a wide margin; at 22 the curves cross; at thirty the Mk V holds 505 MB/s against our 250. If your working set is routinely wider than twenty residents, buy theirs.
Figures
| Resident | NX5/X (MB/s) | Mk V (MB/s) |
|---|---|---|
| 2 | 1,180 | 870 |
| 6 | 1,176 | 868 |
| 10 | 1,168 | 862 |
| 14 | 1,140 | 848 |
| 18 | 1,010 | 812 |
| 22 | 720 | 744 |
| 26 | 430 | 640 |
| 30 | 250 | 505 |
04Residency
- Physical blocks
- Effective blocks, folded
Physical blocks went from 12 to 24 over four generations and then stopped. Effective blocks kept rising.
Figures
| Unit | Physical | Effective | Ratio |
|---|---|---|---|
| G1 NX1 | 12 | 13.4 | 1.12x |
| G2 NX2 | 16 | 19.0 | 1.19x |
| G3 NX3 | 20 | 26.6 | 1.33x |
| G4 NX4 | 24 | 35.0 | 1.46x |
| G5 NX5 | 24 | 41.3 | 1.72x |
Why we quote the ratio
A block count describes the board. The fold ratio depends on your work: two decks with the same twenty-four blocks hold different amounts depending on how repetitive the resident set is.
Our figures use a fixed mixed set of nineteen residents (three Barrier probes, two Ward wrappers, a Sentry sweeper and thirteen utilities), specified page-for-page in the harness. A set of nineteen genuinely unique residents folds to about 1.08× on our hardware, which is barely any fold. We publish that case too; it is profile 34 and it is one of the nine we expect to lose.
The incumbent’s Mk V measures 1.08× on the same set with 28 physical blocks, giving 30.2 effective: four more physical blocks than the NX5/X and eleven fewer effective ones.
05Third party
The index we did not write
The Open Deck Index is a consortium figure. We submit hardware to it, we do not run it, and we do not get to choose the profiles. Index 100 is the median deck in their sample.
We lead on throughput and residency by a wide margin, lead narrowly on ICE traversal, and lose on trace discipline, acoustics and service turnaround. Two of those three are the same two things we lose on our own harness. The third, service turnaround at 64 against their 161, is the worst number on this site.
A 64 puts our service department well below the median deck supplier and a long way below Cybrscape’s. What we are doing about it is on the support page, with the targets we currently miss.
- NX5/X
- Mk V (measured)
Published by the consortium. We submit hardware; we do not run the tests and we do not see the profiles in advance.
Figures
| Metric | NX5/X | Mk V |
|---|---|---|
| Cache throughput | 168 | 124 |
| Resident set density | 181 | 112 |
| ICE traversal | 143 | 131 |
| Trace discipline | 129 | 152 |
| Acoustic | 71 | 148 |
| Service turnaround | 64 | 161 |
06Head to head
NX5/X against the incumbent Mk V
| Measure | NX5/X | Mk V | Ahead |
|---|---|---|---|
| Physical blocksThey have more memory than we do. | 24 | 28 | Cybrscape |
| Effective blocks (folded)And we hold more in it. | 41.3 | 30.2 | Nitrosoft |
| Sustained throughput | 1,180 MB/s | 870 MB/s | Nitrosoft |
| Credits per 100 MB/s | 949 | 1,931 | Nitrosoft |
| ICE traversal, median | 14 ms | 17 ms | Nitrosoft |
| Trace floor at idleTheirs is quieter on the wire. We are working on it. | 1.6 /min | 1.4 /min | Cybrscape |
| Noise under sustained loadTwenty decibels apart. Ours is a loud deck. | 41 dB(A) | 21 dB(A) | Cybrscape |
| Warranty | 12 months | 36 months | Cybrscape |
| Median first support replyTheir service department is better than ours. | 14 h | 2 h | Cybrscape |
| Generations shipped since 2026 | 5 | 2 | Nitrosoft |
Cybrscape builds a good deck and runs a better service department than ours. We beat the Mk V on throughput, residency and credits per unit of work. We expect to keep losing on warranty, noise and support until we fix them, and fixing them is engineering and hiring.
07Method
How every figure was taken
Conditions
- Ambient 24 °C ± 0.5, still air, enclosure closed and screwed shut.
- Thirty-minute unbroken run. The reported figure is the mean of the final twenty-five minutes, so warm-up and any throttle are inside it.
- Twelve runs per figure, median reported, full spread in the run file.
- Stock firmware, factory settings, no fold-table overrides, no thermal override.
- Mains supply through a regulator; draw measured at the wall.
- Noise at one metre, on-axis, in a room with a 17 dB(A) floor.
Profiles
Forty-one profiles in NITROBENCH 4. The headline mixed set is profile 12: nineteen residents, three Barrier probes, two Ward wrappers, one Sentry sweeper, thirteen utilities, specified page-for-page. Nine profiles are deliberate stress cases we expect to lose, including profile 34: nineteen genuinely unique residents, where our fold ratio collapses to 1.08×.
The competitor units
two retail Mk V, two retail Mk IV, unmodified firmware, bought at retail through an ordinary supply desk, measured 14 June 2029. Unmodified firmware, factory settings, same room, same day, same profile set. These are our measurements and we label them that way everywhere they appear; they are not Cybrscape’s specification and should not be quoted as such.
Conflict of interest
NITROBENCH 4 is ours. We wrote the harness, chose the profiles and ran the hardware. The harness ships in the SDK under nitro bench with the profile definitions, and the signed run files for every figure on this page are in the SDK archive. If you run it and get a different number, send us the run file.
What we will not publish
Figures for hardware we have not measured ourselves, extrapolations, peak-burst numbers without the sustained number beside them, and anything from a run we could not reproduce at least nine times out of twelve.
08Cadence
Figures
| Generation | Deck | Shipped | Blocks | Fold | MB/s |
|---|---|---|---|---|---|
| G1 | NX1 | March 2026 | 12 | 1.12x | 120 |
| G2 | NX2 | December 2026 | 16 | 1.19x | 215 |
| G3 | NX3 | September 2027 | 20 | 1.33x | 390 |
| G4 | NX4 | July 2028 | 24 | 1.46x | 700 |
| G5 | NX5 | April 2029 | 24 | 1.72x | 1,180 |
G1: Bench only. 40 units, all hand-finished, all replaced free. G2: First unit anyone paid for. Hotfloor 1.4 introduced block folding. G3: Eviction prediction lands. Support queue triples; see /support. G4: Still shipping as the NX4/T. Firmware supported to July 2033. G5: Same twenty-four blocks as G4. Fold ratio up from 1.46× to 1.72×.