/The NX range → Generation 5 → Extended

NX5/X

The fold engine, unrestricted. Twenty-four physical blocks presenting as 41.3. The widest fold table we ship, the whole eviction predictor, and 41 dB(A) of fan when you ask it for everything at once.

11,200 credits · deck code NX5-X
Sustained
1,180MB/s
Thirty minutes unbroken, enclosure closed.
Effective blocks
41.3
1.72× on 24 physical blocks.
ICE traversal
14ms
Median, Barrier class. 63 ms at the 99th.
Trace floor
1.6/min
At idle. The Mk V measures 1.4. Theirs is better.

01What it is

Bookkeeping, at speed

The /X is the NX5 with nothing held back. Twenty-four physical blocks — the same count the NX4 shipped with, and four more than the /T’s twenty — but the fold table runs to 4,096 entries instead of 1,024, and the eviction predictor gets its full ninety seconds of history rather than thirty.

That is the whole difference. There is no extra memory on this board. The gap between 24 and 41.3 is the fold engine finding pages that two residents were both carrying, and keeping one of them.

What it is bad at

It is loud. 41 dB(A) at one metre under sustained load, and the fan curve steps audibly at 71 °C rather than ramping — that is issue NS-2029-051, it is open, and the safe fix is louder rather than quieter. The Mk V is 21 dB(A) doing similar work. If you share a room, buy the /T.

It thrashes past twenty-two residents. A wide fold table costs lookup time, and beyond roughly twenty-two concurrent residents the incumbent’s flatter design overtakes us. The crossover is charted below rather than left for you to discover.

Twelve months of warranty. Against thirty-six. We priced the deck accordingly; we did not price the argument.

Against the /T

Credits
11,200 vs 7,400
Physical blocks
24 vs 20
Effective blocks
41.3 vs 32.4
Sustained MB/s
1,180 vs 690
Noise dB(A)
41 vs 31
Draw W
118 vs 72

The /X costs 51% more credits for 71% more throughput, and ten more decibels. That is a good trade on paper and a bad one in a shared room. We would rather you bought the right deck twice than the wrong deck once.

The NX5/T

02Benchmarks

Where it wins and where it stops

Throughput held against resident count — NITROBENCH 4, NX5/X against a retail Mk V
  • NX5/X
  • Mk V
03006009001,20026101418222630WAREZ RESIDENT AT ONCECROSSOVER250505MB/s

Both curves are ours, run on the same day with the same profile. We lead by 310 MB/s at a typical set and we lose the lead at 22 residents. If your working set is routinely wider than that, the Mk V is the better deck and we would rather say it here than in a ticket.

Figures
Throughput held against resident count
ResidentNX5/X (MB/s)Mk V (MB/s)
21,180870
61,176868
101,168862
141,140848
181,010812
22720744
26430640
30250505
Sustained cache throughput — the field as we measure it
  • Nitrosoft, generation 5
  • Cybrscape, measured by us
  • Nitrosoft, generation 4
NX5/X1,180 MB/s
Mk V870 MB/s
NX5/T690 MB/s
NX4/T520 MB/s
Mk IV430 MB/s
NX5/S410 MB/s

Cybrscape units measured 14 June 2029: two retail Mk V, two retail Mk IV, unmodified firmware. These are our measurements, not their specification.

Figures
Sustained cache throughput
UnitMB/s
NX5/X1,180
Mk V870
NX5/T690
NX4/T520
Mk IV430
NX5/S410
Physical against effective — the /X in the line it came from
  • 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

G4 to G5 added no memory at all. The whole generation is in the ratio.

Figures
Physical against effective
UnitPhysicalEffectiveRatio
G1 NX11213.41.12x
G2 NX21619.01.19x
G3 NX32026.61.33x
G4 NX42435.01.46x
G5 NX52441.31.72x

03Specification

Everything we measure

Memory and residency
Physical blocks24
Effective blocks, mixed set41.3
Fold ratio, measured1.72×
Fold table width4,096 entries
Page granularity64 B
Residents, typical mixed set19
Residents, hard ceiling26
Eviction predictor history90 s rolling
Throughput
Sustained cache throughput1,180 MB/s
Peak burst, under 2 s1,760 MB/s
Sustained at 22 residents720 MB/s
Sustained at 30 residents250 MB/s
Cold jack to first resident2.1 s
Warm re-jack0.4 s
On the wire
ICE traversal, median, Barrier class14 ms
ICE traversal, 99th percentile63 ms
Trace floor at idle1.6 counts/min
Trace floor, full resident set9.8 counts/min
Link interfaces2 × duplex link, 1 × service bus
Physical
Draw, sustained load118 W
Draw, idle19 W
Noise at one metre, sustained load41 dB(A)
Noise at idle24 dB(A)
Thermal ceiling before throttle84 °C
Mass3.1 kg
Enclosure312 × 208 × 74 mm, folded steel, no window
Mountingdesk, or two-unit rack tray (sold separately)
Commercial
Deck codeNX5-X
Credits11,200 cr
Credits per 100 MB/s sustained949 cr
Warranty12 months, parts and labour
Firmware support toApril 2034
RuntimeHotfloor 4.1

In the box: the deck, a service bus cable, a fold-table reference card, and the twelve-month warranty statement. No link cables — yours will be better than ours. Runtime is Hotfloor 4.1, updated over the service bus, changelog published the day it ships.

04Ordering

Deck code
NX5-X
Quote this to your yard’s supply desk with a quantity.
Credits
11,200
Settled on the yard account on despatch. No card details are taken on this site and there is no checkout.
Lead time
11
Working days, median, August 2029. Worst case in that month was 24. We publish the worst case because that is the one that ruins a schedule.