PhotoPrism is designed to use the CPU resources provided. Large libraries, face processing, thumbnails and metadata extraction can keep available cores busy, so a stronger CPU shortens maintenance windows.
For tens of thousands of photos, frequent imports and users who want fast reindexing.PhotoPrismのハードウェア要件：RAM、CPU、ストレージ、インデックス作成
PhotoPrismのRAM、CPU、SSDストレージ、インデックス作成、RAWファイル、動画トランスコード、ZimaOSのハードウェア要件について学びましょう。
PhotoPrism hardware requirements at a glance
PhotoPrism has explicit upstream hardware guidance. The general self-hosted documentation currently says at least 2 CPU cores, 3 GB physical RAM, a 64-bit OS and at least 4 GB swap. PhotoPrism Plus uses a slightly higher 4 GB physical-memory minimum.
- CPU
- Use at least 2 CPU cores. PhotoPrism scales indexing and concurrent work across available cores, so a stronger CPU materially shortens indexing, thumbnail generation, face processing and media conversion.
- RAM
- General PhotoPrism documentation currently requires at least 3 GB physical RAM; the PhotoPrism Plus requirements page specifies 4 GB. PhotoPrism recommends that RAM scale with the number of physical CPU cores, and RAW/high-resolution workloads can need more.
- Swap
- The current general setup guide says to configure at least 4 GB of swap and avoid hard memory limits, because indexing can temporarily need additional memory and otherwise cause unexpected restarts.
- Storage
- Use local SSD storage for the database, cache, thumbnails and generated storage files. PhotoPrism Plus recommends reserving roughly 50% of the originals-library size for the storage folder so the server does not unexpectedly run out of space; actual usage is often 25% or less.
- GPU / acceleration
- A GPU is not required for basic PhotoPrism. Current upstream supports Intel Quick Sync, NVIDIA and Vulkan paths for hardware video transcoding when the required devices/drivers are exposed. Advanced AI/Vision/Ollama services should be sized separately.
- Best Zima starting point
- ZimaBoard 2 832 already exceeds the 2-core and 3–4 GB memory baselines with an Intel N150 and 8 GB RAM, but add SSD storage for PhotoPrism data. ZimaCube 2 Standard is the more natural storage-first photo-library choice; Pro adds better indexing and multitasking headroom.
From official requirements to the right setup
PhotoPrism sizing starts with the official CPU/RAM baseline, then adds library size, indexing complexity, SSD IOPS, database choice, RAW/video workload and concurrent users.
-
Official requirements
Start with the current upstream baseline: at least 2 cores, 3 GB physical RAM and 64-bit OS for general PhotoPrism, plus at least 4 GB swap. If you specifically deploy PhotoPrism Plus, use its 4 GB physical-memory minimum.
-
Confirm your needs
Estimate the media workload. Large RAW files, high-resolution panoramas, video transcoding, face recognition and first-time indexing can use substantially more CPU and memory than ordinary browsing after the library is indexed.
-
Leave room to grow
Separate fast application storage from bulk originals. Put MariaDB/SQLite, thumbnails, cache and PhotoPrism storage on local SSD; keep large originals on HDD or shared storage when capacity matters more than IOPS.
-
Run it on ZimaOS
Index a representative subset and monitor CPU, physical RAM, swap use, SSD latency, storage-folder growth and video-transcoding behavior. Reduce PHOTOPRISM_WORKERS if memory pressure appears instead of assuming more CPU cores can always be used at full concurrency.
Check every playback client
- General PhotoPrism versus PhotoPrism Plus deployment
- At least 2 CPU cores and 3–4 GB physical memory
- At least 4 GB swap for the general self-hosted setup
- Total originals-library size and expected growth
- RAW, panorama and high-resolution video volume
- Local SSD for database, cache, thumbnails and storage folder
- SQLite for small/test libraries versus MariaDB for larger use
- Other ZimaOS photo, media, backup and AI workloads
Official minimum requirements
PhotoPrism currently publishes two closely related official baselines. The general self-hosted documentation uses 2 cores and 3 GB RAM, while the PhotoPrism Plus knowledge base uses 2 cores and 4 GB RAM. Both require a 64-bit environment and emphasize SSD-backed database/cache storage.
For the ZimaOS PhotoPrism App, treat 2 cores + 3 GB physical RAM + 64-bit OS + at least 4 GB swap as the general upstream baseline. Treat 4 GB physical RAM as the explicit PhotoPrism Plus minimum. In either case, 8 GB is a much more comfortable home-server starting point for indexing and co-hosted services.
| Requirement | Official minimum | What this supports |
|---|---|---|
| General PhotoPrism CPU | At least 2 cores | The current general self-hosted setup page explicitly requires at least two CPU cores. |
| General PhotoPrism RAM | At least 3 GB physical memory | The current Getting Started documentation uses 3 GB as its general baseline and says RAM should scale with the number of physical CPU cores. |
| PhotoPrism Plus RAM | At least 4 GB physical memory | The separate PhotoPrism Plus system-requirements page uses a 4 GB minimum. This is a Plus-specific higher baseline, not evidence that the general page is wrong. |
| Swap | At least 4 GB recommended by the general setup guide | PhotoPrism warns that less swap or hard memory limits can cause unexpected restarts during indexing. |
| Architecture | 64-bit OS | Current multi-arch Docker images support 64-bit AMD, Intel and ARM processors. |
| Database | SQLite 3 or MariaDB 10.5.12+ | SQLite is recommended only for small libraries/test environments where scalability and performance are not important; MariaDB is the stronger long-term path. |
When to upgrade your hardware
PhotoPrism upgrades are driven less by browsing and more by indexing, media conversion, database/cache I/O and the size of the photo/video collection.
Initial or recurring indexing keeps CPU busy for long periods
RAW, panoramas or video processing pushes memory into swap
Thumbnails, cache and database I/O make HDD-backed storage feel slow
Official documentation calls out RAW conversion, high-resolution panoramas and video transcoding as demanding workloads. If an 8 GB server repeatedly swaps during indexing, move to more RAM or reduce worker concurrency.
For photographers, RAW-heavy libraries and mixed photo/video collections.PhotoPrism strongly recommends SSD for MariaDB and its storage folder because thumbnails/cache/database operations need far more IOPS than bulk originals. Moving these active datasets to SSD can improve responsiveness more than adding RAM alone.
For large libraries currently keeping all PhotoPrism data on conventional HDD or shared storage.Plan hardware growth with confidence
PhotoPrism scales best with a fast SSD application tier, a larger bulk-originals tier and enough CPU/RAM to keep indexing workers productive without excessive swapping.
Keep database, cache and thumbnails on local SSD
PhotoPrism explicitly recommends SSD for MariaDB and the storage folder because conventional HDD IOPS are far lower. Do not place database files on unreliable flash media or shared network folders.
Use SATA/NVMe SSD for PhotoPrism storage and database; use HDD bays for the larger originals library.Reserve storage overhead beyond the originals library
PhotoPrism Plus recommends reserving about 50% of the originals-library size for the storage folder to avoid running out of space. Actual storage usage is often 25% or less, but thumbnails and transcoded media can make it much larger in some libraries.
Plan SSD/storage capacity with free-space margin instead of sizing only from the original photo files.Reduce workers when RAM is tighter than CPU
PhotoPrism recommends adjusting PHOTOPRISM_WORKERS downward if system memory does not scale with the number of CPU cores. This is often better than letting indexing concurrency force constant swap use.
Keep ZimaBoard 2 832 for moderate libraries when indexing latency is acceptable, or move to 16 GB for more worker/multitasking headroom.Enable hardware video transcoding only after validating device exposure
Current upstream can use Intel Quick Sync, NVIDIA and Vulkan paths for video transcoding, but the accelerator, driver and /dev/dri or NVIDIA device exposure must all be configured correctly.
Treat Intel iGPU/NVIDIA capability as optional acceleration rather than assuming it is active automatically in every ZimaOS container.Can it run on ZimaOS?
PhotoPrism is currently available in the ZimaOS App Store. Current Zima x86 hardware exceeds the basic PhotoPrism CPU/RAM floor, while storage layout and indexing workload determine which Zima model is the better long-term fit.
Install PhotoPrism from the ZimaOS App Store
ZimaOS currently exposes a dedicated PhotoPrism App Store page. The package can use Zima's local storage and media volumes for a private photo-library deployment.
Open PhotoPrism in the ZimaOS App StoreUse SSD-backed app data even when originals live on HDDs
PhotoPrism's official storage guidance strongly favors local SSD for database, cache, thumbnails and generated storage. This maps well to a Zima setup with SSD app data plus larger SATA/HDD photo storage.
Read PhotoPrism system requirementsValidate hardware transcoding rather than assuming it
Intel Quick Sync can accelerate compatible video transcoding only when the device and drivers are correctly exposed to PhotoPrism. The existence of an Intel iGPU in Zima hardware does not by itself prove active container acceleration.
Read PhotoPrism video-transcoding guidanceChoose Zima hardware for your PhotoPrism workload
Every current Zima configuration in this comparison has enough CPU cores and RAM to clear PhotoPrism's basic 2-core/3–4-GB requirements. The real choice is between compact SSD-attached photo hosting and a larger multi-drive photo/video library with faster indexing and more headroom.
Is this mainly a compact personal photo library, or a growing multi-drive RAW/video archive?
Compact/personal library with external SSD or a few drives
ZimaBoard 2 832 comfortably clears the upstream baseline with 4 CPU cores and 8 GB RAM. Add SSD storage because its onboard 32 GB eMMC is not an appropriate long-term PhotoPrism library/cache target. Choose 1664 for more indexing and co-hosted-app headroom.
- Personal PhotoPrism with SSD-backed app dataZimaBoard 2 832
- Larger library, more workers or more co-hosted appsZimaBoard 2 1664
Large/growing photo and video archive with many drives
Choose ZimaCube 2 when storage capacity becomes the main requirement. Standard already has enough RAM for PhotoPrism and offers a multi-drive platform; Pro adds 16 GB RAM and stronger CPU headroom for indexing, RAW/video workloads and broader services.
- Storage-first family photo/video NASZimaCube 2 Standard
- Large RAW/video library and heavier indexing/multitaskingZimaCube 2 Pro
- Only when independent GPU/AI/creator workloads also justify itZimaCube 2 Creator Pack
This is a workload guide, not a guaranteed indexing-time, photo-count or concurrent-user benchmark. Results depend on PhotoPrism edition/version, worker count, RAW/video formats, database, SSD/HDD layout, transcoding setup and other ZimaOS services.
| Zima hardware | Best for | Example workload | Core configuration | Recommended boundary | Next step |
|---|---|---|---|---|---|
| ZimaBoard 2 832 | A compact personal PhotoPrism server with added SSD/HDD storage. | Personal/family photo library, normal indexing, face recognition, web browsing and several lightweight ZimaOS apps. |
|
The 32 GB eMMC is not enough for a serious PhotoPrism library and generated storage. Large RAW/video libraries or heavy co-hosted services can also make 8 GB less comfortable during indexing. | Get Now |
| ZimaBoard 2 1664 | A compact PhotoPrism host with more RAM for indexing, MariaDB and additional home-server services. | Larger photo collections, more indexing workers, RAW/video files and multiple always-on containers. |
|
Do not claim the 1664 indexes faster solely because it has more RAM. CPU is unchanged; the extra memory mainly helps larger workers/caches and co-hosted services. | Get Now |
| ZimaCube 2 Standard | A storage-first family PhotoPrism NAS with a larger multi-drive photo/video library. | Multi-terabyte originals, family photo backups, thumbnails/cache on SSD and moderate indexing workloads. |
|
Eight GB clears the official baseline but offers less memory headroom than the 16-GB options for large RAW/video indexing or many co-hosted apps. | Get Now |
| ZimaCube 2 Pro | The most balanced current Zima configuration for a large PhotoPrism library, faster indexing and broader NAS/media use. | Large RAW/video archives, frequent imports, MariaDB, more concurrent users, backups and multiple self-hosted services. |
|
10GbE and 16 GB RAM do not guarantee a fixed indexing speed. SSD placement, workers, formats and database performance remain important. | Get Now |
| ZimaCube 2 Creator Pack | PhotoPrism on a server also used for independent local AI, advanced vision, creator and GPU workloads. | Large photo/video archive plus GPU-capable AI services, advanced Vision/Ollama workflows or other creator applications. |
|
Extreme overkill for PhotoPrism alone. Do not recommend Creator Pack just for face recognition or a large library; choose it only when independent GPU/AI/creator workloads justify the dedicated GPU and 64 GB RAM. | Get Now |
What the Press Says
Highlights from trusted reviewers worldwide.
“ZimaCube 2: Not just another NAS, tested with 25TB storage, local AI agents, 4K transcoding, and real homelab workflows.”Read full review
“The ZimaBoard 2 is a compact x86 server board that can be turned into a mini NAS, home server, media box, or self-hosting hub.”Read full review
“ZimaCube 2: A modern, high-performance NAS with plenty of room to grow—built for users who want more than basic storage.”Read full review
“Coverage focused on ZimaCube 2's open hardware foundation, no monthly fee, and self-hosting flexibility.”Read full review
Loved by the Community
Stories and reviews from people who build with Zima every day.
★★★★★
Zima Blade Little yet Powerful
Maybe I am not digital natives but I live with PCs since 12 years old in 1984 when IBM PC clone come to my home. Many years have passed and many operating system I've tried. For me Zima blade and CasaOS was a quantum leap for home PC enthusiast and server lab machine to make me stay curious and relevant for this era.
★★★★★
Very good!!
I use ZimaCube Pro as 5th Proxmox cluster node. It runs several VMs and containers, including a VM with GPU passthrough to run a self-hosted LLM. A specific LXC container runs a Samba server for NAS capabilities using four of six RAID 6 SATA HDDs with ZFS.
★★★★★
Great innovation for mini server!
It is very useful and makes a powerful mini server for many purposes, including university and college students in engineering and electronics. Thank you so much for making this server.
★★★★★
Avaliação ZimaBoard 2
Construí um servidor de uso pessoal. O desempenho está muito bom e funciona perfeitamente onde quer que eu esteja. A surpresa é não dependermos de grandes estruturas para termos nosso próprio servidor de dados. Como iniciante, estou gostando bastante do ZimaOS, pois ele é simples e eficiente.
Frequently asked questions
These answers reconcile PhotoPrism's current general and Plus requirements and separate current upstream behavior from Raspberry Pi community/testing examples.
How much RAM does PhotoPrism need?
Current general PhotoPrism documentation says at least 3 GB physical RAM, while the PhotoPrism Plus requirements page specifies 4 GB. Both also recommend more memory for larger indexing workloads. For a ZimaOS home server, 8 GB is a comfortable starting class.
Why do PhotoPrism's official pages say both 3 GB and 4 GB RAM?
They apply to different official documentation scopes. The general Getting Started page uses 3 GB physical RAM; the dedicated PhotoPrism Plus requirements page uses 4 GB. This guide keeps both numbers and does not silently collapse them.
Does PhotoPrism need swap even with enough RAM?
The current general setup guide recommends at least 4 GB swap and warns that hard memory/swap limits can cause unexpected restarts when indexing temporarily needs more memory, especially with RAW files and high-resolution panoramas.
How much extra storage does PhotoPrism need beyond my originals?
PhotoPrism Plus recommends reserving roughly 50% of the originals-library size for the storage folder so you do not unexpectedly run out of space. It says actual usage should usually be 25% or less, but thumbnails and transcoded media can push usage higher depending on formats and settings.
Should PhotoPrism use SSD or HDD storage?
Use local SSD for the database and PhotoPrism storage folder containing thumbnails, cache, configuration and generated sidecars. Conventional HDDs or shared storage are better suited to the originals folder, where capacity and sequential throughput matter more than IOPS.
Can ZimaBoard 2 832 run PhotoPrism?
Yes. Its Intel N150 has 4 CPU cores and the 832 has 8 GB RAM, so it exceeds both the general 2-core/3-GB requirement and the Plus 2-core/4-GB requirement. Add SSD/HDD storage because the 32-GB eMMC is not appropriate for a growing originals and thumbnail/cache library.
Does PhotoPrism need a GPU for face recognition or indexing?
No dedicated GPU is required for the base application. Current PhotoPrism uses CPU-based and software AI pipelines for core indexing features, while hardware acceleration is explicitly available for video transcoding through Intel, NVIDIA or Vulkan when configured. Advanced Vision/Ollama services are separate workloads.
What does the Raspberry Pi 5 guide tell us about low-end PhotoPrism hardware?
The supplied PiDIYLab article reports that an 8-GB Raspberry Pi 5 with SSD handles its tested PhotoPrism 240915 setup well, while 4 GB swaps heavily during face-enabled indexing. It also says the Pi's VideoCore GPU did not accelerate that tested inference path. This is useful device-specific evidence, but its tested PhotoPrism build predates current 2026 upstream face-detection and video-transcoding changes, so it should not override current official documentation.
What sources and further reading informed this PhotoPrism hardware guide?
The PhotoPrism Plus requirements page specifies at least 2 cores, 4 GB physical RAM, 64-bit OS, SSD-backed database/cache, and recommends reserving about 50% of originals size for the storage folder. The general PhotoPrism Getting Started page instead uses 2 cores, 3 GB physical RAM, 64-bit OS and at least 4 GB swap; this is the baseline used for the ordinary ZimaOS PhotoPrism app. The DietPi forum mainly reproduces the upstream 2-core/3-GB requirement and provides Raspberry Pi/home-server discussion rather than a separate benchmark. PiDIYLab's updated 2026 article tests PhotoPrism 240915 on a Raspberry Pi 5 8GB with SSD and reports heavier swapping on the 4GB model; because its tested PhotoPrism version predates current upstream 2026 changes, its GPU/Coral observations remain device/version-specific rather than universal current requirements.
