Answer to common questions about joining the Render Compute Network
Last updated
The Render Compute Network compute nodes waitlist is now open. In order to join the waitlist, an application form is required. It includes a benchmark assessment and speed test, and other required fields.
Below are some frequently asked questions about the waitlist application form.
Refer to for full list of requirements:
GPU with a compute score at or above the RTX 3050, from recent GPU generations up to RTX 5090 (see the approved list)
64GB+ RAM and 2TB+ SSD
Linux OS (Ubuntu 22.04 or 24.04 preferred)
100 Mbps+ download / 75 Mbps+ upload
Docker + NVIDIA Container Toolkit installed
IMPORTANT NOTE: This is a call for existing hardware. We strongly advise against purchasing new hardware. Joining the waitlist does not guarantee a spot in the first (or subsequent) cohorts. There are no guarantees about the amount of work that will be assigned to nodes via the network.
FREQUENTLY ASKED QUESTIONS
Should I run Windows or Linux?
Linux or WSL are acceptable. For optimal compatibility we recommend native linux, ubuntu 22.04 or 24.04.
The benchmark doesn’t see my x hardware (e.g. 4090 GPU)? Is that normal?
Yes, this is normal. We have data about expected GPU scores via OctaneBench. The benchmark assessment is able to record the type of GPU being assessed.
My computer was using AMD CPU/GPU and not my RTX5090 for the benchmark test. Json script wasn’t generated.
We recommend running the benchmark assessment from the shell. It should provide the necessary information about where the json file was saved. Alternatively, check the working folder.
Can nodes be multi-GPU? X4 RTX4090 for example?
We don’t recommend multi-GPU nodes at this time.
It says my 9800XD, 96GB DDR5 and 2x 4090s was like a 7 year old Mac - what's happening?
The built-in comparison data generates this output due to the limited sample size of initial data. As additional examples are collected, comparison results will update. This should not impact the ability to run the tests and upload the data.
Can I use 2 ssd 1TB or must it be 2TB ssd to submit?
I can’t find the benchmark result file. Where did it go?
It should either save in the Desktop bench_output folder or home directory for Mac/Linux. Alternatively, check whether it saved to the current folder. It is meant to be run as a command line tool not as a GUI. It should display all the information while it’s running. We recommend running the benchmark assessment from a terminal/shell or cmd prompt, which should tell you what it’s doing and where the JSON file is being saved.
Example
See example run below for the latest version using testing_ variant.
charlie@M4 ~ % /Applications/Bench.app/Contents/MacOS/bench
=============================================
bench - cross-platform system benchmark
v0.2.1 (build: 2025-05-08)
=============================================
Running all benchmarks...
System Information:
==================
OS: macOS 24.5.0
Hostname: M4.local
CPU Information:
Model: Apple M4 Max
Architecture: arm64
Physical Cores: 16
Logical Cores: 16
Base Clock: 3200 MHz
Memory Information:
Total Physical: 128.00 GB
Available Physical: 88.09 GB
Total Virtual: 128.00 GB
Available Virtual: 88.09 GB
Disk Information:
/: 3.63 TB
/: 3.63 TB
Temp Directory: /tmp
GPU Information:
Name: Apple M4 Max
Vendor: Apple
Driver Version: N/A
Running benchmark: CPU Matrix Multiplication
Description: Measures CPU performance for matrix multiplication operations (FLOPS)
[==================================================] 100%
[==================================================] 100%
Result: 4.36459 GFLOPS
Running benchmark: CPU Threading Performance
Description: Measures CPU's multi-threading performance
Result: 34428.4 MegaItems/sec
Running benchmark: CPU Vector Operations
Description: Measures performance of CPU for vector operations (SIMD-friendly)
Result: 112.857 GB/s
Running benchmark: RAM Bandwidth
Description: Measures memory read/write bandwidth
[==================================================] 100%
[==================================================] 100%
Result: 125.637 GB/s
Running benchmark: RAM Latency
Description: Measures memory access latency using pointer chasing
[==================================================] 100%
[==================================================] 100%
Result: 11.9266 ns
Running benchmark: Disk Read Speed
Description: Measures sequential disk read bandwidth
[==================================================] 100%
[==================================================] 100%
Result: 18556.2 MB/s
Test file removed: /tmp/bench_disk_read_test.dat
Running benchmark: Disk Write Speed
Description: Measures sequential disk write bandwidth
[==================================================] 100%
[==================================================] 100%
Result: 7392.24 MB/s
Running benchmark: Disk Random I/O
Description: Measures random disk access performance (IOPS)
[==================================================] 100%
[==================================================] 100%
Result: 552133 IOPS
Running benchmark: GPU Matrix Multiplication
Description: Measures GPU performance for matrix multiplication operations (GFLOPS)
[==================================================] 100%
[==================================================] 100%
Result: 4.33622 GFLOPS
Running benchmark: GPU Memory Bandwidth
Description: Measures GPU memory read/write bandwidth
Result: 0 GB/s
Benchmark Results Summary:
=========================
Benchmark Score Units Rating
----------------------------------------------------------------------
CPU Matrix Multiplication 4.36 GFLOPS **---
CPU Threading Performance 34428.41 MegaItems/sec ***+-
CPU Vector Operations 112.86 GB/s ****-
RAM Bandwidth 125.64 GB/s **+--
RAM Latency 11.93 ns **---
Disk Read Speed 18556.19 MB/s **+--
Disk Write Speed 7392.24 MB/s **---
Disk Random I/O 552132.99 IOPS **+--
GPU Matrix Multiplication 4.34 GFLOPS -----
GPU Memory Bandwidth 0.00 GB/s -----
Overall Score: 118.77 points
Performance Category: Very Good - Better than reference system
(Reference System: Apple M3 Max (2023))
Comparisons with other reference systems:
- Faster than Intel Core i9-13900K (2023) (134%)
- Faster than AMD Ryzen 9 7950X (2023) (139%)
- Significantly faster than Intel Core i5-13600K (2023) (506%)
- Significantly faster than Apple M1 (2020) (243%)
- Significantly faster than Intel Core i3-12100 (2022) (592%)
- Significantly faster than Raspberry Pi 4 (2019) (3779%)
Benchmark Results Graph:
------------------------
GB/s:
CPU Vector Operations |########################################## | 112.86 GB/s
RAM Bandwidth |############################################### | 125.64 GB/s
GPU Memory Bandwidth | | 0.00 GB/s
GFLOPS:
CPU Matrix Multiplication |############################################### | 4.36 GFLOPS
GPU Matrix Multiplication |############################################### | 4.34 GFLOPS
IOPS:
Disk Random I/O |############################################### | 552132.99 IOPS
MB/s:
Disk Read Speed |############################################### | 18556.19 MB/s
Disk Write Speed |################## | 7392.24 MB/s
MegaItems/sec:
CPU Threading Performance |############################################### | 34428.41 MegaItems/sec
ns:
RAM Latency |############################################### | 11.93 ns
macOS build: Saving output to /Users/charlie/Documents/bench_output/bench_results_2025-05-22_13-14-36.json
Results saved to JSON file 'bench_results_2025-05-22_13-14-36.json'
As stated in the requirements, 1TB NVMe SSD per node. See under the Storage (HDD/SSD) minimum requirements section in Appendix 1.
