Professional
A 40-node HP ProLiant render farm with 2.4 PFLOPS of GPU power – on-premises, in our own server room. Fixed and scalable capacity, with no cloud dependency and no hidden fees.
Have a question? Contact our expert.
40 nodes · HP ProLiant DL380 Gen10 · 80× NVIDIA Quadro RTX · 2.4 PFLOPS · 40 GbE · 1.5 PB Hitachi HNAS ·
One render farm, three strengths
Server Fleet
40 nodes · HP ProLiant DL380 Gen10. Per node: 2× Intel Xeon Platinum 8180 – 56 cores / 112 threads, 256–512 GB RAM.
GPU Capacity
2× NVIDIA Quadro RTX per node – 80 professional GPUs in total and 2.4 PFLOPS of FP32 peak performance.
On-Premises
The hardware is physically on site: owned and operated by us. No cloud dependency, no hidden traffic fees.
Numbers that
40 nodes, 80 GPUs, petabyte-scale storage – in a single, locally operated render farm.
= 2.4 PFLOPS
40 nodes in total
professional class
40 × 256 GB in total
2× Xeon Platinum 8180
Hitachi HNAS platform
fast parallel I/O
+ 16G Fibre Channel SAN
Network, storage and render buffer
Network
Storage
Render buffer
40 GbE backbone and dedicated SAN
High-bandwidth inter-node communication, complemented by a dedicated 16G Fibre Channel SAN storage network for the NAS.
- 40 GbE backbone
- 16G Fibre Channel SAN
- Bootless nodes
Hitachi NAS Platform (HNAS)
0.5–1.5 PB of file-level workspace: scenes, projects and asset files in one place. The nodes have no local working storage – all data comes from central storage.
- 0.5–1.5 PB capacity
- File-level workspace
- Central data source
0.5 PB dedicated render buffer
Intermediate files and render output with fast parallel reads and writes. The separate buffer protects the main storage system during large jobs.
- Fast parallel I/O
- Isolated buffer
- Protected main NAS
On-Premises – the difference
Five aspects in which an on-premises render farm differs from cloud solutions (AWS, Azure, GCP…).
Hardware – physical or virtualized?
MassFrame: physically present, real servers in our server room.
Cloud: virtualized resources – a shared pool with variable performance.
Ownership – dedicated or shared?
MassFrame: our own hardware – non-virtualized, non-shared resources.
Cloud: shared resources – capacity is decided by the provider.
Cost – predictable or usage-based?
MassFrame: fixed and scalable capacity, allocated according to your deadlines.
Cloud: traffic and storage fees – unpredictable costs for long render jobs.
Data transfer – local or uploaded?
MassFrame: all data comes from central storage over a 40 GbE backbone.
Cloud: network latency – uploading large scenes can be slow.
Control – in-house team or vendor?
MassFrame: operated by our own IT team – full control, no vendor lock-in.
Cloud: vendor dependency – pricing and capacity are decided by the provider.