Device performance and responsiveness

Start with Auto#

Choose Studio menu → Preferences → Performance and inspect the device/performance controls. Auto is the factory preference. Performance choices balance update frequency, interaction responsiveness and resource use. A preferred high-quality setting does not guarantee that every device can sustain it.

What acceleration can do#

Render Preview uses browser GPU rendering where supported. GPU lighting can accelerate supported lighting work; unsupported content falls back to the CPU worker. The application cannot grant itself unrestricted system resources or force a particular physical GPU. Browser acceleration, drivers, power settings and available memory still matter.

Diagnose a slow scene#

1. In Preferences → Performance, return to Auto; it applies immediately. Reduce Quality in HDRi Canvas or DPR in the Render Preview header as needed.
2. Disable costly screen-space effects such as SSR, AO or depth of field while placing lights.
3. Use fewer simultaneously visible lights and a smaller model when possible.
4. Disable live updates while dragging if your device cannot keep up; the final position still updates after the gesture.
5. Run the device check available in Preferences → Performance and note the reported renderer, effective quality and warnings.

Exports use original assets and their selected output resolution; lower interactive quality does not itself lower your requested export resolution.

Device-local preferences#

Usage Mode, Interface Zoom, grid visibility, interaction preferences and performance choices are saved for this browser. They are separate from the portable project. Auto is the default performance mode; live dragging and GPU lighting are enabled by default where supported.

If GPU rendering fails#

Use a current desktop browser with graphics acceleration available and restart it after changing its acceleration setting. If the browser or graphics driver cannot create the required context, the application cannot repair that driver. Save your project and use troubleshooting to gather a reproducible report.

Preferences edits apply and save immediately. Leave the movable window open while comparing the preview, allow rendering to settle, then run the device check. Direct toolbar controls remain synchronized.

Instances and NURBS#

Instances reuse object and inheritance indexes, skip unchanged assembly updates and detach unused source copies. A source-level CPU benchmark covers 500 nested placements; actual frame rate still depends on GPU, draw calls, materials and model complexity. The optional NURBS panel generates guides only on request. Start with Draft for dense CAD files, and disable control nets or isocurves when they are not useful.

Large Rhino files and hierarchy panels#

Geometry Tree asks Load geometry hierarchy above 250 scene objects. Search also opens matching results. Opening the list never changes model visibility. Linked block contents start collapsed; use their disclosure arrows or Expand all to inspect them. The tree displays 80 rows per page, and Instances displays 50 placements per page. Search and selection operate on the full logical list; selecting an object reveals its page in an already loaded tree.

Unopened inactive tabs do not mount their contents. Opened inactive or collapsed panels retain their local state without rebuilding their rows on camera updates. Geometry Tree and Instances also reuse their rendered content during unrelated camera updates.

Ground shadows reuse their map while only the camera moves. Model transforms, visibility/material changes, animation, environment changes and ground settings refresh it. This helps when ground shadows are enabled; it does not reduce the source model's triangle count or simplify geometry.

Chrome GPU setup#

In Chrome, open Preferences → System → Use graphics acceleration when available, enable it, and relaunch. Enter chrome://gpu in the address bar and check WebGL's hardware-acceleration status. Auto, High Performance and Maximum Quality already request high-performance graphics from the browser; this is a preference, not a guarantee of a particular GPU.

There is no per-site setting in HDRi Studio to grant Chrome extra video memory. Use Auto, reduce DPR or expensive effects when necessary, and keep the graphics driver/browser current. GPU lighting acceleration is a separate setting from the 3D viewport. See Chrome's graphics troubleshooting guidance and WebGL memory budgeting.

The device check now reports CPU draw time, draw calls and triangles summed across all passes, alongside GPU time when available. Multipass effects, shadows and transparent surfaces can submit the same geometry more than once, so these counters can exceed the model's unique triangle count. CPU draw time excludes React panel rendering; it is not a complete main-thread profile. The check measures your current scene and resolution; source-level benchmarks are not a guarantee of viewport FPS.

Coalesced project writes#

Document edits share a 1.6-second idle autosave delay, reducing repeated writes during a continuous edit. Opening or selecting a panel does not itself change the project. This is a write-frequency improvement; no new browser-storage or GPU frame-rate benchmark is claimed for this update.

Explicit viewport density#

Auto uses the selected Preview Frame Limit (60 FPS by default) and keeps preview density stable during camera movement. Auto and Maximum Quality have an 8-megapixel drawing budget; graphics allocation limits can also cap the selected DPR. High Performance and Battery Saver may temporarily reduce interactive density based on measured rendering cost, then restore full permitted detail after interaction settles. Idle gaps between requested frames do not count as slow rendering. Mode also controls update frequency, lighting scheduling, cache budget and GPU preference. These are budgets, not a guarantee of the selected FPS on every scene or device. Choose a lower DPR when a large panel or expensive effects make interaction slow. HDRi Canvas Quality selects environment resolution separately.

Floating preview diagnostics#

Enable Render Performance in Panels or Render Performance panel in Preferences → Performance. Both controls stay synchronized; closing a floating monitor turns both off. It starts as a floating panel and can dock, stack, join portrait tab groups or open in a separate window. Drag its title bar to move freely; use its six-dot grip to dock or stack.

Use the Compact performance view icon to keep FPS, status and the history chart while hiding detailed counters and explanatory text. Show performance details restores frame/CPU/GPU timing, draw calls, geometry/resource counters, DPR and drawing-buffer resolution. This choice is remembered on this device. The chevron collapses the whole panel to its heading while docked. Floating performance has no collapse button; use Compact performance view to hide its details. Its compact, reset, options and close actions share square hit areas. The reset icon clears history.

Unsupported GPU measurements remain unavailable; WebGL does not expose exact GPU-memory bytes. Hiding or collapsing the monitor stops its UI subscription, while the shared telemetry history remains available when reopened.

The preview bypasses a neutral grading pass and avoids empty helper-scene draws and duplicate synchronization. These reduce unnecessary work; actual FPS still depends on the model, active effects, DPR and graphics device. Screen DPR does not mean one physical pixel per CSS pixel on a Retina display.

Anti-aliasing and optional geometry passes#

Balanced anti-aliasing keeps the existing idle rendering policy. It uses a single FXAA pass at all pixel densities, avoiding duplicate native multisampling. SMAA and explicit MSAA can be selected per project. Edge Rounding is Off by default and adds a shared geometry-buffer pass only when used. Background masking also needs this buffer when background tone mapping is disabled or clipping is enabled. Hardware-dependent FPS, driver memory and visual quality should be checked on the target device; no universal speedup is claimed.

Grid rendering cost#

The finite grid uses one two-triangle surface with shader-drawn lines. Increasing cell count does not allocate more line geometry. Line-density fading reduces distant clutter. This improves the grid's appearance and bounds its geometry cost; no universal preview frame-rate increase is claimed. See grid settings.

Restore Performance defaults#

Choose Reset Performance inside the Performance tab. It restores Auto, supported GPU lighting enabled, the performance monitor off, live dragging enabled, HDR preview quality 2K · Balanced and Screen DPR. Other Preferences tabs and export settings remain unchanged. Preview quality and DPR belong to the current document; the other choices are device preferences. The device-check report clears when its performance mode, acceleration, quality or DPR changes; run Check My Device again after the preview settles. Use the notification's Undo to restore the previous Performance values. See scoped reset and Undo.

Scene Settings reuses unchanged slider controls during camera updates. Its disclosure menus ignore value-only changes, and Preferences device readings update independently of the editor. This reduces CPU interface work while panels are open; it does not change shader quality or render resolution.

Preview frame limit and measured FPS#

Choose Scene Settings → Render Canvas → Preview Frame Limit: 30, 60, 90, 120, 150, 180 or 240 FPS. This is a portable document setting and is independent of video export FPS. The default remains 60. Battery Saver limits the preview to 30 FPS. Your browser, monitor refresh rate, power policy and rendering workload can impose a lower limit.

Render Performance reports actual rendered-frame intervals, including rates above 60, and shows Idle when a settled scene no longer needs frames. Selecting 120 on a 60 Hz display does not create 120 visible frames. Temporal AA adds work while refining a still image and stops after its configured sample count; it is a quality option rather than a universal speed boost.