NX5/X
11,200 crThe fold engine, unrestricted.
- Blocks, physical
- 24
- Blocks, effective
- 41.3 · 1.72×
- Sustained
- 1,180 MB/s
Generation five · shipping nowTwenty-four blocks.
Twenty-four blocks.
Forty-one blocks of work.
Memory is the scarce resource in every deck ever built, and no supplier has ever changed that. What a supplier can change is how much of it goes to waste. Hotfloor 4.1 folds a resident set down to 1.72× on the NX5/X — the same 24 physical blocks the NX4 shipped with in 2028.
01The four numbers that matter
Sustained throughput
1,180MB/s
NX5/X, mixed-set profile, thirty minutes unbroken.
Fold ratio
1.72×
41.3 effective blocks out of 24 physical.
Credits per 100 MB/s
949cr
The incumbent Mk V measures 1,931 cr on the same profile.
Generations since 2026
5
37 months. Median 9 months between them.
02The argument
We did not give you
more memory.
The NX5 has exactly as many physical blocks as the NX4 did. We could have added four and charged for them. Instead we spent the generation on the fold table.
Block folding finds pages that are byte-identical across two resident warez and keeps one copy. Most working sets are more repetitive than their authors think: shared library pages, shared constant tables, three Barrier probes that all carry the same handshake. On a mixed set of nineteen residents the NX5/X folds 24 blocks into 41.3 blocks of usable residency.
The eviction predictor does the other half. It watches which residents you actually re-enter and holds those; it pages out the ones you loaded out of optimism. Nothing about this is exotic. It is bookkeeping, carried out faster than a person would bother to.
It has a ceiling, and we publish where the ceiling is. Past roughly twenty-two concurrent residents our fold table thrashes and the incumbent’s flatter design pulls ahead. That crossover is drawn on the benchmarks page, at full size, with the numbers under it.
- Physical blocks
- Effective blocks, folded
G1 · NX11.12×
1213.4
G2 · NX21.19×
1619.0
G3 · NX31.33×
2026.6
G4 · NX41.46×
2435.0
G5 · NX51.72×
2441.3
Physical blocks are the silicon. Effective blocks are what a mixed resident set actually fits, measured on NITROBENCH 4. The ratio is the only thing we have ever really sold.
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 |
03Throughput, ours and theirs
- Nitrosoft, generation 5
- Cybrscape, measured by us
- Nitrosoft, generation 4
Cybrscape figures are our own measurements of retail units on unmodified firmware, not manufacturer specification. Our method is published; theirs is their business.
Figures
| Unit | MB/s |
|---|---|
| NX5/X | 1,180 |
| Mk V | 870 |
| NX5/T | 690 |
| NX4/T | 520 |
| Mk IV | 430 |
| NX5/S | 410 |
Sustained, not peak
Peak burst is a marketing number and we print ours too — 1,760 MB/s on the NX5/X — but it lasts under two seconds and you should not buy a deck on it.
Every throughput figure on this site is a thirty-minute unbroken run at a controlled 24 °C with the case closed, averaged across the last twenty-five minutes. If a deck throttles, that is in the number. Ours throttles. It is in the number.
Read the method04Cadence
Five generations in 37 months
G1NX1MARCH 2026120 MB/s
G2NX2DECEMBER 2026215 MB/s
G3NX3SEPTEMBER 2027390 MB/s
G4NX4JULY 2028700 MB/s
G5NX5APRIL 20291,180 MB/s
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 |
Shipping this often has a cost and the cost lands on support. Our median first reply is fourteen working hours; the incumbent’s is two. We are not going to pretend that is a rounding error — it is on the support page with the rest of our service figures, including the targets we miss.
05The range
Five decks, one runtime
G5 · Extended
NX5/T
7,400 crThe one most people should buy.
- Blocks, physical
- 20
- Blocks, effective
- 32.4 · 1.62×
- Sustained
- 690 MB/s
NX5/S
4,900 crA first deck that is not a toy.
- Blocks, physical
- 16
- Blocks, effective
- 24.6 · 1.54×
- Sustained
- 410 MB/s
NX4/T
3,400 crLast generation, current price.
- Blocks, physical
- 20
- Blocks, effective
- 29.2 · 1.46×
- Sustained
- 520 MB/s
NXD-5
2,750 crAn /X with no case and no excuses.
- Blocks, physical
- 24
- Blocks, effective
- 41.3 · 1.72×
- Sustained
- 1,180 MB/s
Ordering
Quote through your yard’s supply desk with a deck code and a quantity. Credits settle on the yard account. We do not take card details and there is no checkout on this site.
Deck codes06For people writing warez
Hotfloor tells you why
A fold engine you cannot inspect is a fold engine you cannot trust. nitro fold --explain prints the decision for every page in your resident set: folded, refused, and the reason it was refused.
The SDK is free, the benchmark harness is the same one we publish our own figures with, and the block-budget linter will fail your build before a customer finds out the hard way.
$ nitro budget --target nx5-t --set ./ward.set reading 14 residents … 20 blocks available fold candidates 2,118 pages folded 1,404 pages (66.3%) resident cost 12.3 blocks of 20 PASS headroom 7.7 blocks $ nitro budget --target nx5-s --set ./ward.set resident cost 12.3 blocks of 16 PASS headroom 3.7 blocks — tight, see E-BLK-12