What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
There is no single independently verified best storage array for every 2026 workload. In MLPerf Storage v3.0, the reported leaders differ by task: YanRongTech leads the cited 3D U-Net training comparison, Everpure reports standout checkpoint and KV-cache results, and Azure Managed Lustre has a high checkpoint-write result from a large cloud deployment. Those figures are meaningful only alongside their workload and configuration; none is a universal ranking of storage arrays.
This comparison is strongest for AI and neocloud storage. It does not provide a complete ranking of mainstream enterprise block/file arrays or small-business products. For business-application random I/O, SPC-1 is a separate benchmark and should not be mixed into an AI throughput leaderboard.
As an Amazon Associate I earn from qualifying purchases.
Which storage system leads each AI workload?
MLCommons published MLPerf Storage v3.0 on September 1, 2026. The round covers training, checkpointing, vector database (VDB) indexing and querying, and large-language-model KV-cache reads and writes. It added VDB and KV-cache tests and S3 object access alongside POSIX. MLCommons reported 19 submitting organizations.
The following figures come from a comparison of v3.0 submissions. They are results for specific workloads and tested configurations, not head-to-head rankings across unlike tests.
#1 Best Overall
- Value NAS with RAID for centralized storage and backup for all your devices. Check out the LS 700 for enhanced features, cloud capabilities, macOS 26, and up to 7x faster performance than the LS 200.
- Connect the LinkStation to your router and enjoy shared network storage for your devices. The NAS is compatible with Windows and macOS*, and Buffalo's US-based support is on-hand 24/7 for installation walkthroughs. *Only for macOS 15 (Sequoia) and earlier. For macOS 26, check out our LS 700 series.
- Subscription-Free Personal Cloud – Store, back up, and manage all your videos, music, and photos and access them anytime without paying any monthly fees.
- Storage Purpose-Built for Data Security – A NAS designed to keep your data safe, the LS200 features a closed system to reduce vulnerabilities from 3rd party apps and SSL encryption for secure file transfers.
- Back Up Multiple Computers & Devices – NAS Navigator management utility and PC backup software included. NAS Navigator 2 for macOS 15 and earlier. You can set up automated backups of data on your computers.
| System | Reported result | What it indicates |
|---|---|---|
| YanRongTech F9000X | 543.9 GiB/s on 3D U-Net, feeding 99 simulated B200 accelerators. | Highest reported training throughput in the cited comparison. |
| TuringData F9200 | 541.5 GiB/s on 3D U-Net from three storage nodes. Its Checkpoint-70B result is 539.9 GiB/s read and 307.1 GiB/s write. | A strong reported result in both the cited training and checkpoint figures; the tests are distinct. |
| Everpure FlashBlade//EXA | At 30 data nodes: 877.5 GiB/s checkpoint write and 833 GiB/s checkpoint read; 1,623 GiB/s KV-cache read. | Leads the cited comparison’s checkpoint and KV-cache results; the KV-cache figure is a different workload from checkpointing. |
| Azure Managed Lustre | 642.2 GiB/s checkpoint write from a 4,096 TiB managed cloud deployment. | The cited comparison identifies it as the first hyperscale cloud service submitted. Its deployment scale and service model differ from on-premises systems. |
| NVIDIA AIStore | 20 submissions over S3 object storage across OCI, AWS, and GCP. | This describes the number and cloud spread of submissions, not a single throughput result or top-ranked score. |
MLCommons reported median and maximum efficiency figures for on-premises submissions: 14 GB/s per watt median and 201 GB/s per watt maximum for checkpoint writes; 34 GB/s per watt median and 277 GB/s per watt maximum for UNet3D reads. These are submission statistics for their respective workloads, not guaranteed efficiency for a particular installation.
How to compare the benchmark results fairly
MLPerf Storage is designed to test storage feeding simulated accelerators, saving and restoring checkpoints, and supporting specified AI data-access workloads. MLCommons describes the benchmark as architecture-neutral, representative, and reproducible. A valid comparison still depends on matching the test conditions and reading the full system configuration.
Rank #2
- Value NAS with RAID for centralized storage and backup for all your devices. Check out the LS 700 for enhanced features, cloud capabilities, macOS 26, and up to 7x faster performance than the LS 200.
- Connect the LinkStation to your router and enjoy shared network storage for your devices. The NAS is compatible with Windows and macOS*, and Buffalo's US-based support is on-hand 24/7 for installation walkthroughs. *Only for macOS 15 (Sequoia) and earlier. For macOS 26, check out our LS 700 series.
- Subscription-Free Personal Cloud – Store, back up, and manage all your videos, music, and photos and access them anytime without paying any monthly fees.
- Storage Purpose-Built for Data Security – A NAS designed to keep your data safe, the LS200 features a closed system to reduce vulnerabilities from 3rd party apps and SSL encryption for secure file transfers.
- Back Up Multiple Computers & Devices – NAS Navigator management utility and PC backup software included. NAS Navigator 2 for macOS 15 and earlier. You can set up automated backups of data on your computers.
Training throughput is not total training time
In the training benchmark, simulated accelerators read real data through PyTorch. The test skips accelerator arithmetic by sleeping for a measured batch-computation interval, and results must satisfy workload-specific accelerator-utilization thresholds. Listed training figures average five consecutive measured runs. The storage-to-client-DRAM data path is real, but the test does not measure end-to-end model training time, model quality, or GPU compute performance; Microsoft makes the same distinction in its explanation of the benchmark.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteMatch workload and accelerator type
- Compare results within the same workload and accelerator type. MLCommons says results are comparable within a workload, not across different workloads.
- Read the tested system type and configuration with the headline number. The reported throughput is aggregate storage-system throughput; Microsoft warns that client-node count is a load-generator setting, not a valid basis for normalizing or ranking storage performance.
- Do not compare checkpoint writes with training reads, KV-cache reads, or vector database results as if they were interchangeable measures.
- Where data is reported, useful additional axes include usable-capacity-normalized bandwidth, rack-unit density, and throughput per watt. Protocol support, such as POSIX or S3, may matter to the intended application.
Check the evidence label
A precise number is not automatically an audited benchmark. Keep these evidence classes distinct:
Rank #3
- Value NAS with RAID for centralized storage and backup for all your devices. Check out the LS 700 for enhanced features, cloud capabilities, macOS 26, and up to 7x faster performance than the LS 200.
- Connect the LinkStation to your router and enjoy shared network storage for your devices. The NAS is compatible with Windows and macOS*, and Buffalo's US-based support is on-hand 24/7 for installation walkthroughs. *Only for macOS 15 (Sequoia) and earlier. For macOS 26, check out our LS 700 series.
- Subscription-Free Personal Cloud – Store, back up, and manage all your videos, music, and photos and access them anytime without paying any monthly fees.
- Storage Purpose-Built for Data Security – A NAS designed to keep your data safe, the LS200 features a closed system to reduce vulnerabilities from 3rd party apps and SSL encryption for secure file transfers.
- Back Up Multiple Computers & Devices – NAS Navigator management utility and PC backup software included. NAS Navigator 2 for macOS 15 and earlier. You can set up automated backups of data on your computers.
- Peer-reviewed benchmark submission: a result submitted under a published benchmark’s rules. Check the official submission configuration and supplemental information before drawing a comparison.
- Vendor or partner claim: useful context about a claimed capability, but not equivalent to an audited result. Large rack-scale claims cited in current coverage—including Dell and WEKA throughput figures and NetApp’s projected pNFS namespace rate—are identified as claims or projections, not measurements at the claimed scale.
- Independent lab measurement: a test conducted by an independent lab. Do not conflate it with a vendor’s internal testing or a benchmark submission.
Deployment examples such as CoreWeave, xAI Colossus, and Nebius are evidence of adoption, not proof of a benchmark lead. Current coverage describes them as multi-vendor or named deployments and reports DDN, Hammerspace, WEKA, and VAST among AI-native specialists, alongside Dell and other incumbent vendors with AI product lines.
Why SPC-1 IOPS does not answer the AI-storage question
SPC-1 measures predominantly random I/O for business-critical applications such as OLTP, databases, and mail servers. That workload family differs from AI data feeding, checkpointing, vector database, and KV-cache tests. A high SPC-1 IOPS score therefore cannot be used to rank a system against an MLPerf Storage throughput result.
Rank #4
- Value NAS with RAID for centralized storage and backup for all your devices. Check out the LS 700 for enhanced features, cloud capabilities, macOS 26, and up to 7x faster performance than the LS 200.
- Connect the LinkStation to your router and enjoy shared network storage for your devices. The NAS is compatible with Windows and macOS*, and Buffalo's US-based support is on-hand 24/7 for installation walkthroughs. *Only for macOS 15 (Sequoia) and earlier. For macOS 26, check out our LS 700 series.
- Subscription-Free Personal Cloud – Store, back up, and manage all your videos, music, and photos and access them anytime without paying any monthly fees.
- Storage Purpose-Built for Data Security – A NAS designed to keep your data safe, the LS200 features a closed system to reduce vulnerabilities from 3rd party apps, SSL encryption for secure file transfers, and RAID for redundancy.
- Back Up Multiple Computers & Devices – NAS Navigator management utility and PC backup software included. NAS Navigator 2 for macOS 15 and earlier. You can set up automated backups of data on your computers.
The Storage Performance Council’s active results table uses SPC-1 version 3 for active published results and distinguishes accepted results from submissions still under review. The status matters:
| System | SPC-1 result | Status and date |
|---|---|---|
| ExponTech WDS V3 | 27,201,325 IOPS | Accepted; submitted in 2023 and accepted November 19, 2023. |
| FlashNexus FN8200 | 30,002,765 IOPS | Submission dated February 25, 2025; marked “Submitted for Review,” not accepted. |
The larger pending figure is not the accepted SPC-1 record. Older SPC-1 results remain historical, so check both the benchmark version and current status when consulting the results table.
Best Value
- [Quad-Core Processor for Daily Tasks] Powered by Realtek RTD1619B quad-core 1.7GHz CPU, ensuring stable performance for file storage, backup, and personal cloud usage.
- [Fast 2.5GbE Network Upgrade] Equipped with 2.5-Gigabit Ethernet, delivering faster transfer speeds than traditional Gigabit NAS for home data management.
- [1GB DDR4 Efficient Memory] Optimized 1GB DDR4 memory ensures stable system performance for basic multitasking and background services.
- [Personal Cloud Backup Solution] Create your own private cloud to store, access, and manage files securely from anywhere with full data ownership.
- [Tool-Free Installation for Fast & Easy Setup] Designed for user convenience with tool-free drive installation, allowing quick HDD/SSD setup without screws—perfect for beginners and IT professionals.
What current submissions can—and cannot—say about other vendors
Several prominent AI-storage vendors did not submit to MLPerf Storage v3.0. Current coverage reports that DDN has v2.0 results but did not submit to v3.0; Hammerspace also sat out v3.0; WEKA’s most recent audited submission is v1.0; and VAST has not submitted to MLPerf Storage. A separate comparison of the v3.0 round counts 143 results and notes that several prominent v2.0 submitters did not participate.
Missing current-round data is not evidence of inferior performance. It means there is no v3.0 result in this comparison for that product. Vendor platforms, product status, and deployment descriptions in current coverage should be treated as attributed reporting unless supported by a specific audited benchmark entry.
A practical way to choose a storage array
Start with the bottleneck in the target workload, then compare systems on the same evidence basis. A high score in one category does not settle performance in another.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsQuick Recap
- Name the workload. Decide whether the priority is training data reads, checkpoint write and recovery read, vector database work, KV-cache access, or business-application random I/O.
- Set the test conditions. Match workload, accelerator type, protocol, and tested system configuration before comparing throughput.
- Check the operating target. Consider the required accelerator scale, sustained bandwidth, checkpoint recovery needs, relevant small-file or metadata behavior, and whether the application uses POSIX or S3.
- Assess efficiency and footprint. Use capacity-normalized bandwidth, rack density, and power efficiency only where the result reports them in a comparable way.
- Verify provenance. Separate benchmark submissions from internal vendor claims and deployment examples; inspect the configuration and acceptance status behind a headline number.
- Confirm product fit outside the benchmark. The figures establish workload-specific performance under test, not a complete verdict on deployment requirements, operations, or suitability for every enterprise array buyer.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
