Bandwidth Requirements for Remote Mac Sessions

Before committing a team of five to remote Mac rentals, you need to know whether your office uplink can carry five simultaneous VNC sessions. The answer depends on what those sessions are doing, what quality settings the VNC client uses, and whether latency, not bandwidth, is the actual bottleneck. This guide gives you a sizing table for common workloads, a method to measure your real per-session consumption, and the client-side knobs to turn when you are over budget.

What Drives VNC Bandwidth

VNC implements the RFB protocol (RFC 6143), whose single display primitive is "put a rectangle of pixel data at a given x,y position." Unlike protocols that transmit drawing commands, VNC transmits pixel data, which is why its consumption swings widely with what the screen is actually doing.

Three factors dominate:

Screen change rate. A developer staring at a text editor changes a few hundred pixels per keystroke. An animator scrubbing a timeline can change every pixel on a 1440p display dozens of times per second. The difference is several orders of magnitude.

Colour depth. A 24-bit true-colour session sends three bytes per changed pixel before compression. Dropping to 16-bit or 8-bit cuts that proportionally, at the cost of photo and gradient fidelity.

Encoding. RFC 6143 defines Raw, CopyRect, Hextile, TRLE and ZRLE. ZRLE applies zlib compression over tiled run-length encoded rectangles and is the usual default for modern servers. Tight (a common extension) is similar in practice. Raw should never be used over the open internet.

Measuring Real Session Bandwidth

A synthetic speed test tells you the maximum throughput between your office and the data centre. It does not tell you what a live VNC session actually uses. Measure both at once.

Step 1 — Install iperf3 on both sides. On the rental Mac open a terminal; on the $29 virtual Intel macOS plan that terminal is reached through the browser console, which is the verified access path for that plan. On physical i3 or M-series rentals, use the shell access confirmed in your quote.

brew install iperf3

On your local machine, install via Homebrew on macOS or sudo apt install iperf3 on Debian or Ubuntu.

Step 2 — Decide which direction you can test. iperf3 needs an inbound TCP port on the server side. The $29 virtual Intel macOS does not expose arbitrary inbound ports, so run iperf3 as a client from inside the VM against a public reflector, or reverse the test with -R. On a physical rental where your quote confirms port forwarding, run the server there and the client on your laptop:

iperf3 -s -p 5201

Step 3 — While a real VNC session is open and running your actual workload, run a capped client test for thirty seconds:

iperf3 -c your-rental-host -p 5201 -t 30 -b 20M -R

Watch retransmits and jitter in the iperf3 output. They tell you about congestion, not raw throughput.

Step 4 — Capture the VNC traffic separately. On your local machine open Wireshark and filter by tcp.port == 5900, or whichever port your VNC server listens on. In Statistics → I/O Graphs, read actual bytes per second during coding, design, and video scrubbing. That is your real per-session consumption for each workload.

Bandwidth Sizing Table by Workload

Treat these ranges as starting estimates calibrated against the method above, at 1080p. Rerun the capture on your own workload before sizing a line.

These numbers are per session. Total bandwidth is the sum across concurrent sessions.

Reducing Bandwidth With VNC Client Settings

Before upgrading an internet line, try reducing session quality. Most workloads tolerate lower settings with minimal visual impact.

Colour depth. TigerVNC's vncviewer exposes LowColorLevel: 0 is 8 colours, 1 is 64 colours, 2 is 256 colours. Confirm the exact flag against your installed vncviewer -help:

vncviewer -FullColor=0 -LowColorLevel=2 your-rental-host:0

Resolution. Ask the rental Mac to run at 1280×800 instead of 1920×1200. In macOS System Settings → Displays, pick a lower scaled resolution. 1280×800 is about 44% of 1920×1200 pixels, so expect roughly that reduction in change-region bytes.

Encoding. Prefer ZRLE or Tight. Avoid Raw. CopyRect helps for window moves but is not a substitute for a general encoder.

Frame rate. Apple's built-in Screen Sharing adapts its cadence to the client's requested update rate; see Apple's Remote Desktop guide for the macOS server behaviour. Third-party VNC servers often expose an explicit maximum update rate you can cap.

Multiple Simultaneous Sessions

For N users, the data-centre uplink must carry the sum of their active sessions, and your office router must handle the same aggregate inbound.

A 50 Mbit/s symmetric fibre line is comfortable for five coding sessions and starts to strain under five simultaneous high-quality video scrubbing sessions. Add roughly 20% headroom for spikes, background OS updates, and simultaneous browser activity.

When Bandwidth Is Not the Bottleneck

If a VNC session feels sluggish despite adequate throughput, latency is usually the cause. Each frame requires the server to detect a change, encode it, and transmit it before the client renders. At 150 ms round-trip, even a text cursor feels delayed.

Run ping your-rental-host and read the round-trip time. Under 30 ms feels natural. 50 to 80 ms is tolerable for coding. Above 100 ms, drag-to-scroll and live typing start to drag regardless of bandwidth.

A Figment rental in a region close to your users reduces that round-trip independently of line speed. A London team will feel the difference between a European data centre and a US West Coast one even when both sides have gigabit fibre.

Picking a Figment Plan for Interactive Use

For interactive VNC work the right starting point is a Figment Mac that fits the workload, not the cheapest line item. The $29 virtual Intel macOS (2 vCPU, 8 GiB RAM, 80 GiB disk) is accessed through the browser console and suits terminal and lightweight IDE work on supported Intel toolchains, including Xcode 26 or earlier; it has no GPU acceleration or audio. For designers in Figma, native Xcode device simulation on Apple silicon, or video scrubbing, request an M-series rental from $95 per month; exact chip, RAM and disk are confirmed in the quote. The i3 at $49 and physical M-series plans currently collect configuration requests until supply is confirmed, so ask for availability before you plan a start date.

Start a request at /mac-hosting/start, or read more at /mac-hosting.

Related guides:

• Remote Mac for distributed teams
• Remote Mac support session guide
• GitHub Actions self-hosted runner on a Mac