I put my Steam library on my NAS, and it was both brilliant and stupid
At a glance:
- Using iSCSI over a 10G network, a TrueNAS-powered TerraMaster F4-424 Max NAS can run Steam games with load times only 2–3 seconds slower than local SSDs in demanding titles like The Witcher 3 and Cyberpunk 2077.
- Multiplayer titles show mixed results: Helldivers 2 and Lethal Company stay responsive, but Elden Ring: Nightreign and Far Far West introduce noticeable latency in early rounds.
- The setup requires a block-level protocol (iSCSI), not SMB or NFS, and works best with SSD-backed iSCSI datasets on at least 5GbE (or 2.5GbE) networks.
Why iSCSI and not SMB or NFS
Ayush Pande typically runs his NAS shares over SMB and NFS, which handle file-level sharing fine for backups, media archiving, and Linux-centric projects. But when he tried routing Steam game files through those same protocols, the results were disastrous. Load times ballooned, some titles crashed on launch, and others simply refused to boot when their datasets were accessed over SMB or NFS. The problem is latency: file-sharing protocols add overhead that gaming workloads simply cannot tolerate.
That's where iSCSI changed the equation. Unlike SMB or NFS, iSCSI is a block-sharing protocol — it tricks the client machine into believing a network drive is just another local disk. That illusion eliminates the protocol-level lag that file-sharing stacks introduce. Pande connected his primary gaming PC to a TrueNAS-powered TerraMaster F4-424 Max over a 10G Zyxel switch, using a reused PCIe Gen 4 SSD as the iSCSI dataset. The SSD itself can push 6,000 MB/s, but the 10G network stack caps throughput at roughly 1,250 MB/s. TrueNAS's built-in ZFS cache smooths out the bottleneck, keeping gameplay and load times from taking a severe hit.
Benchmark results: single-player titles
Pande ran timed benchmarks comparing internal SSD performance against the iSCSI share (NAS SSD). The numbers tell a surprisingly encouraging story for a network-attached setup.
| Game | Internal SSD | iSCSI share (NAS SSD) |
|---|---|---|
| Clair Obscur: Expedition 33 (Loading) | 5.14s | 7.31s |
| Final Fantasy XVI (Loading) | 6.77s | 9.28s |
| Final Fantasy XVI (Fast-travel) | 3.92s | 4.17s |
| The Witcher 3: Wild Hunt (Loading) | 34.82s | 36.47s |
| The Witcher 3: Wild Hunt (Fast-travel) | 5.24s | 6.58s |
| Cyberpunk 2077 (Loading) | 8.91s | 11.42s |
| Cyberpunk 2077 (Fast-travel) | 7.7s | 9.59s |
| Sunless Skies (Loading) | 6.98s | 7.29s |
| Hades II (Loading) | 4.2s | 4.74s |
For indie and offline titles, the iSCSI setup was virtually indistinguishable from local storage. Even graphically intensive games like The Witcher 3: Wild Hunt, Cyberpunk 2077, and Final Fantasy XVI showed only marginal differences. Clair Obscur: Expedition 33 — described as the most demanding title in the author's library — added only a few extra seconds to boot time. Lightweight games like Sunless Skies and Hades II showed almost no gap at all. Textures in open-world games might load a millisecond or two slower, but overall fluidity was excellent.
Multiplayer performance: a mixed bag
Co-op and multiplayer titles introduced more variability. Elden Ring: Nightreign showed the biggest gap: 11.28s on internal SSD versus 14.71s over iSCSI, with latency noticeable in the first round with random players. Helldivers 2 and Lethal Company (running at least fifty mods) were more forgiving, generating maps quickly and feeling responsive. Far Far West had manageable latency in its first two rounds. Slay the Spire 2 loaded decks slightly slower at the end of rounds in a handful of fights.
| Game | Internal SSD | iSCSI share (NAS SSD) |
|---|---|---|
| Elden Ring: Nightreign | 11.28s | 14.71s |
| Helldivers 2 | 11.5s | 15.46s |
| Far Far West | 6.03s | 8.42s |
| Lethal Company | 17.93s | 21.18s |
Pande chalks the worst latency spikes up to random network hiccups rather than a fundamental iSCSI limitation, but acknowledges that ranked multiplayer should be approached with caution — or at least with a heads-up to your squad.
What you need to make this work
The experiment hinges on a few non-negotiable pieces. First, you need iSCSI, not SMB or NFS — the block-level protocol is what makes the latency acceptable for gaming. Second, an SSD-backed iSCSI dataset matters: Pande reused a PCIe Gen 4 SSD from an old gaming laptop, and he explicitly recommends sticking to SSDs over HDDs for the dataset. Third, network speed matters — 5GbE (or even 2.5GbE) support on the NAS, the gaming PC, and the switch is the minimum for a usable experience. His setup uses a 10G Zyxel switch connecting a TrueNAS SCALE system to his primary PC.
On the software side, TrueNAS has a straightforward iSCSI creation wizard, and pairing the NAS share with a Windows gaming PC is as simple as connecting via the iSCSI Initiator and formatting it through Disk Management. Steam's Storage tab lets you migrate games between drives, including the freshly configured iSCSI disk, without any extra gymnastics.
Who this is for (and who should skip it)
If you're running out of storage slots on your primary gaming machine and want to offload massive titles that eat hundreds of gigabytes each, this setup is genuinely viable — provided you stick to offline or non-competitive multiplayer games. The convenience of freeing up internal drive space without re-downloading terabytes of game data is real. But if you need rock-solid latency for competitive ranked play, or if your network infrastructure tops out at 1GbE, the iSCSI-over-NAS approach will feel more like a novelty than a daily driver.
Pande sums it up bluntly: the project is "somewhat viable," and he doesn't regret it. The TerraMaster F4-424 Max running TrueNAS SCALE proved more than capable of hiding behind a 10G switch and a good SSD cache, even when asked to serve up Cyberpunk 2077 and Clair Obscur: Expedition 33.
FAQ
What protocol should you use to run Steam games from a NAS?
How much slower were game load times on the iSCSI NAS setup compared to a local SSD?
What network speed and hardware do you need for NAS gaming over iSCSI?
More in the feed
Prepared by the editorial stack from public data and external sources.
Original article