Run 4K Remux at Scale on a TrueNAS Tower
A 6-8 bay TrueNAS tower for a 4K remux library and many concurrent streams
As an Amazon Associate we earn from qualifying purchases. Prices and availability come from Amazon.com as of the last update and may change.
The high-end Plex server is a 4K-remux-at-scale machine: a 14th-generation Core i5 (or a Xeon E for ECC) whose UHD 770 iGPU still does the Plex transcoding while its cores serve many concurrent streams and run TrueNAS or Unraid, in a six-to-eight-bay tower with ECC memory and 10GbE, for roughly $1,500 to $2,500 before drives. Build it when you are running a large, high-bitrate 4K remux library with many viewers, and you want the data integrity and network throughput of a proper NAS underneath Plex. It is the most machine on this page, and the only tier where a discrete HBA, ECC and 10GbE are worth the cost.
Who this build is for
This is for the enthusiast running Plex on top of a real NAS: a large 4K remux collection measured in dozens of terabytes, many people streaming at once, and a wish for the reliability that ECC memory and ZFS bring. The transcoding needs are real but, crucially, are still handled by the same iGPU as the 4K tier; what changes here is everything around it, because serving many high-bitrate streams from a large protected pool is a NAS problem as much as a Plex one.
It is overkill for anyone whose library and audience are modest. If you do not need six or more bays, ECC, or 10GbE, the 4K build does the transcoding for far less money and power. The high-end tier earns its keep only when the storage scale, the concurrent-stream count and the integrity requirements all point the same way. Be honest that this is a NAS-plus-Plex build, not just a faster Plex box.
The components, and why the iGPU still leads
Even at this tier, start with the transcoding engine, because the surprise is that it does not change much: a 14th-generation Core i5 keeps the UHD 770 iGPU doing all the Plex transcoding through Quick Sync, exactly as the 4K tier does. You do not add a graphics card. What the bigger CPU buys is cores, to serve many concurrent streams, run the ZFS filesystem, and host containers, while the iGPU quietly tone-maps and transcodes in hardware. Choosing a Xeon E instead trades a little iGPU pedigree for straightforward ECC support; either way the integrated graphics is what serves Plex.
The rest of the build is where the money goes, and each part has a job. A W680 motherboard enables ECC with a Core i5 and provides the PCIe lanes for an HBA and a network card. ECC UDIMMs, 32 to 64GB, protect the ZFS pool from silent bit-flips and feed the ARC read cache that keeps a large library responsive. An LSI or Broadcom HBA flashed to IT mode adds the eight-plus drive ports a six-to-eight-bay build needs beyond the board's SATA. A 10GbE or 2.5GbE NIC moves 4K remux bitrates and library transfers without the 1GbE ceiling. A Fractal Define 7-class tower, a 650W 80+ Gold supply and six or more enterprise-class CMR drives finish a machine built to run for years.
What the extra money buys over the 4K build
| High-End adds | Why it matters | |
|---|---|---|
| ECC memory + W680 | Error-correcting RAM | Catches silent bit-flips before they reach a large ZFS pool |
| HBA in IT mode | 8+ drive ports | Feeds six-to-eight bays that the board's SATA cannot |
| 10GbE / 2.5GbE NIC | Faster networking | Moves 4K remux bitrates and big transfers past the 1GbE limit |
| More cores | i5-14500 / Xeon E | Serves many concurrent streams plus TrueNAS/Unraid and apps |
Power draw and running cost
This is the tier where the bill becomes noticeable, because you are running a full desktop CPU, six-plus spinning drives, an HBA and a 10GbE card around the clock. Idle with the drives spun up the build draws roughly 60 to 90 watts; under load with several streams and disk activity it can reach 110 to 140 watts. Taking about 110 watts as a 24/7 average at $0.17 per kilowatt-hour, that is around $165 a year, and more if your pool is large and rarely idle.
Enterprise drives and an HBA are the biggest additions to that figure, so if power matters, favour higher-capacity CMR drives to hit your target terabytes with fewer spindles. The running cost is the price of scale and integrity; it is worth it for a large, protected, many-stream library, and hard to justify for anything smaller.
Upgrade path
This platform is built to grow. The tower and HBA leave room for more drives, you can add vdevs to a ZFS pool or drives to an Unraid array over time, and the socketed CPU can step up to a Core i7 or a higher Xeon if your stream count climbs. Memory can grow to feed a bigger ARC as the pool expands, and a second NVMe can mirror the boot drive for redundancy on TrueNAS.
Size the storage with the ZFS or drives calculator and prefer dual parity (RAID-Z2) on wide vdevs of large drives, because resilvering a big pool is where single parity gets dangerous. Recertified Exos is usually the cheapest route to this much CMR; check it against the shucking tracker before each drive purchase.
When a different tier makes sense
Drop to 4K
- You need 4K transcoding but not six-plus bays
- ECC and 10GbE are not requirements for you
- You want lower power, cost and complexity
Go further (DIY)
- You need double-digit bays or multiple vdevs
- You want a separate transcode GPU for very high stream counts
- You are building a rack server, not a tower
Verdict
For a large 4K remux library with many viewers and a demand for integrity, this is the build, and the reason to spend the money is the NAS around Plex, not the transcoding, which the same iGPU handles as the tier below. ECC, an HBA, 10GbE and six-to-eight bays add up to a machine that runs for years and protects a big pool, at about $165 a year to run.
If you do not need that scale, the 4K build transcodes just as well for far less. Buy the high-end tier when the storage size, the stream count and the integrity requirements all justify it, and use the parts list above, treating the HBA and NIC as the optional-but-recommended parts they are marked as.
All parts in detail
The total sums the in-stock components verified live on Amazon.com. Drive cost is separate: size the array with the drives calculator and cut the $/TB via the shucking tracker.
Frequently asked questions
Do I need a GPU for a high-end Plex server?
No, not for transcoding. Even at this tier the integrated UHD 770 on the Core i5 does the Plex transcoding through Quick Sync, and the extra spend goes on cores, ECC, an HBA and networking. A discrete transcode GPU only makes sense for unusually high concurrent 4K stream counts, well beyond a typical home server.
Does a Plex server need ECC RAM?
Plex does not require it, but this build pairs Plex with a large ZFS pool, and ECC catches silent memory bit-flips before they can corrupt data on that pool. For a big, long-lived library you care about, ECC on a W680 or Xeon platform is the safer choice, which is why it is core to the high-end tier and optional below it.
Why an HBA instead of the motherboard's SATA ports?
A six-to-eight-bay build needs more drive ports than a typical motherboard provides. An LSI or Broadcom HBA flashed to IT mode adds eight or more SATA/SAS ports and presents the drives directly to ZFS or Unraid, which is what you want for a software RAID pool. It is marked optional because smaller arrays fit on the board's SATA.
Do I need 10GbE for Plex?
Not for streaming to individual clients, which rarely saturate 1GbE, but a large server benefits from 10GbE or 2.5GbE for fast library transfers, backups and moving high-bitrate 4K remux around your network. It is one of the optional-but-recommended parts on this tier, worthwhile once your pool and usage are large.
TrueNAS or Unraid for a high-end Plex build?
Both work. TrueNAS with ZFS gives the strongest data integrity and performance and pairs naturally with ECC, while Unraid offers easy mixed-drive expansion and single-drive fault isolation. Choose TrueNAS for a performance-and-integrity pool, Unraid for flexible, gradual growth; the hardware here suits either.
How many drives should a high-end Plex server start with?
A common sweet spot is six drives in one RAID-Z2 vdev, which balances usable capacity, dual-failure protection and rebuild time for a large 4K library. Size it for your target capacity with the ZFS calculator, and prefer higher-capacity drives to hit that target with fewer spindles and less power.