ProductTechnologySolutionsValidationCasesCustomersIPCompanyInvestorsNews Contact 中文

ZK-Storage WS5000

The WS-HBMM5000 all-flash accelerated storage appliance: storage upgraded from supporting act to compute amplifier.

Quick answer

What are the key specs of ZK-Storage WS5000?

Product
WS5000 disaggregated all-flash accelerated storage appliance
Aggregate bandwidth
300 GB/s (vendor spec S9)
Random IOPS
~50M (S9)
Access latency
~20 µs (S9)
Domestic-GPU coverage
90%+ (S9)
Deployment
~48-72 hours, no framework change (S9)
WS5000 · ALL-FLASH EBOF
CORE SPECS

Core specifications

The hard numbers at a glance. Figures are vendor-disclosed (S9); independent results are on the Validation page.

300 GB/s
Aggregate bandwidth
Line-rate data path
50M
Random IOPS
Friendly to high-concurrency small files
20 µs
Access latency
Microsecond response
90%+
GPU coverage
Broad mainstream accelerator support
48-72 h
Fast deployment
Turnkey, live within a day
-40%
Total cost
vs. mainstream 3-yr TCO
-60%
Scale-out cost
Elastic, on-demand
2-3×
GPU utilization gain
High-switch / long-context
PORTFOLIO

Product portfolio

Two product lines — WS5000 and WS7000 — across multiple delivery models for different customers.

Product / ServiceFormCustomerCore value
ZK-Storage WS5000 applianceHardwareNew AI clustersHigh-bandwidth all-flash, turnkey
ZK-Storage WS7000 applianceHardwareAI compute centersExtreme-performance disaggregated acceleration (70M IOPS)
ZK-Storage storage softwareSubscriptionExisting hardwareDisaggregation, continuous updates
Brownfield retrofitSolution + serviceExisting data centersSpeed-up without downtime
Accelerated storage serviceCapacity / computeSMB / cloudOn-demand, low barrier
SECOND LINE

WS7000 · for the AI compute center

A disaggregated all-flash storage acceleration platform for AI centers / AI factories. Figures below are vendor spec, blueprinted from the GP7000 white paper.

70M
Random IOPS
Vendor spec
300 GB/s
Aggregate throughput
Vendor spec
2.4 Tbps
Network bandwidth
Vendor spec
20 µs
Access latency
Vendor spec

WS7000 adopts a disaggregated architecture in the NVIDIA G3 ICMS lineage with end-to-end NVMe, GPU-direct acceleration and Active-Active HA — for large-model training, long-context inference and agent workflows.

Explore AI compute center × WS7000

WHY STORAGE

Why storage?

More GPUs deliver diminishing returns. The real bottleneck is data supply: model loading, checkpoint I/O and KV-cache scheduling.

  • Cut KV-cache-related cost by ~74% (measured)
  • No changes to your training / inference framework
  • Independently scale compute and capacity; pool and share
  • Deeply adapted to domestic accelerators, self-controlled

Go deeper into the tech

WS5000 · ALL-FLASH EBOF
CERTAINTY

Four certainties

From concept to maturity — certainty grounded in verifiable facts.

Validated
Technology
Third-party test by Beijing Information Science and Technology University
Finalized
Product
WS5000 in mass production
1,000/mo
Manufacturing
Luxshare Precision foundry
In test
Ecosystem
AMD / xFusion in progress

Benchmark it on your own workload

2 live demo units are ready for immediate PoC. Let the data do the talking.

Last updated: