HYDRADOCS
ManualGPUELECTRIC DREAMS NIGHT BENCHMARK
GPU · Benchmark

ELECTRIC DREAMS NIGHT BENCHMARK

What it is. A GPU benchmark built into HYDRA: a night forest scene in Unreal Engine 5.8.4 with Lumen, hardware ray tracing, Nanite and MegaLights, always rendered at 2560 × 1440. The score is the number of frames rendered during a fixed flight through the scene, and a run can be published to an online scoreboard.

Why use it. In a few minutes it shows what a GPU setting really gives in a modern, heavy game engine. Compare runs before and after a change, see the clocks, voltages and power behind each score, and see where your chip and your score stand among cards of the same model.

Works with GeForce RTX · Radeon RX with 12 GB of VRAM or more Render 2560 × 1440 · DirectX 12 Opens from TESTS → Electric Dreams Benchmark Updated 31 Oct 2026

Run the benchmark

  1. Open the TESTS tab and click Electric Dreams Benchmark in the GPU section.
  2. Choose the graphics card in the list next to Run; the GPU active in HYDRA is selected by default. Close other demanding programs for a fair result, then click Run.
  3. The shaders compile first. Then the measured flight runs with an overlay that shows FPS, frame time, the 1% low and the elapsed and total time. Esc exits.
  4. The result screen shows the score, the average FPS, the 1% low and FPS over the run. Back in HYDRA, the run appears in the table.
Electric Dreams Night loading screen compiling shaders: pass 1, 22 %, 0:47 left.
Shaders compile before the measured run.
Night forest stream scene with the benchmark overlay: 41.6 FPS, 24.0 ms, 1% low 31.9, 2:25 of 3:37.
During the run: FPS, frame time, 1% low and elapsed | total time.
Result screen: score 8 440 frames rendered, 38.7 average FPS, 31.5 1% low and an FPS chart over 3:37.
The result: the score is the number of frames rendered.
  • The benchmark needs a GPU with at least 12 GB of VRAM; on Radeon, an RX 6800, RX 7700 XT, RX 9060 XT or faster.
  • It always renders at 2560 × 1440: in full screen on a 2560 × 1440 monitor, in a borderless 2560 × 1440 window on any other. A smaller monitor shows only part of the window, but the GPU load is the same, so scores stay comparable. If Windows makes the window smaller, HYDRA stops the run.
  • The flight has a fixed length (3:37 in the example), so a faster GPU renders more frames and gets a higher score.

The scoreboard

Electric Dreams Night scoreboard in HYDRA: online top runs for the RTX 5090, the score of the selected run, the silicon map, the chip verdict and the score distribution.
The online scoreboard for the RTX 5090. Hover a number to find it in the list; click the screenshot to see it full size.
  1. 1ONLINE. On: the world top 100 for the selected GPU, one best run per account, with your unpublished run from this session at its expected place (~N). Off: all runs saved on this PC, without contacting the server.
  2. 2GPU filter. All GPUs, only NVIDIA, only Radeon or one model.
  3. 3Counters. How many runs and GPU models the online board, or this PC, holds.
  4. 4Compare. Tick two runs here, then click Compare.
  5. 5Silicon. The chip's factory frequency from NVIDIA's model, not a test result. See Silicon map.
  6. 6Compare and Show more. Compare the two ticked runs, or open the charts and details of the selected run.
  7. 7GPU and Run. The card that runs the benchmark, and the start button.
  8. 8Publish. Sends the run you measured in this session to the online scoreboard under your nickname.
  9. 9Score. The score of the selected run and the benchmark version.
  10. 10Silicon map. Speed against leakage of the selected chip among chips of the same model.
  11. 11Chip verdict. What the chip's place on the map means for air cooling, undervolting and LN2.
  12. 12Score distribution. How the scores of this GPU model spread; the bar of the selected run is red.
  • Only a run measured in the current session can be published. The board keeps one best run per account.
  • The nickname belongs to your HYDRA account and is set in SETTINGS: up to 24 characters, letters, digits, spaces and . _ -. If you have none yet, HYDRA asks for it when you publish.
  • The online scoreboard can also be viewed in a browser, without HYDRA, at hydraboard.1usmus.dev.

Silicon map

The Silicon column and the map use the chip passport that NVIDIA stores in the VBIOS of every card. They describe the chip itself: overclocking, cooling and settings do not change them.

  • Silicon is the chip's Stock frequency at 60 °C from NVIDIA's factory model: the frequency at the voltage ceiling NVIDIA sets for this chip. The more the chip leaks, the lower that ceiling.
  • The map puts the chip on two scales within NVIDIA's range for the model: speed upwards, leakage to the right. Compare only cards of the same GPU model.
  • When the online board has at least 5 different chips of the model, the line under the verdict says how many of them yours beats. With fewer, it gives the chip's place in NVIDIA's range.
  • Some models have no leakage rating, such as the RTX 4080 below; then only speed counts. Radeon cards have no chip passport, and the Silicon column shows a dash.
VerdictSpeedLeakageMeaning
Golden for airHighLowHigh voltage ceiling and the most MHz per watt. Best for daily overclocking and undervolting.
Fast chipHighMediumStrong on air with a decent voltage ceiling. Worth a try on LN2.
LN2 candidateHighHighCold removes the leakage and the speed stays. On air: heat and a low voltage ceiling.
Balanced chipMediumLowLow leakage: good on air and for undervolting. Average on LN2.
Average chipMediumMediumA typical sample of this GPU model.
Warm chipMediumHighLower voltage ceiling and more heat. Good cooling pays off.
Cool and frugalLowLowUndervolts well, but the frequency ceiling is low.
Modest chipLowMediumBelow-average speed. Undervolting suits it best.
Modest sampleLowHighLow voltage ceiling and runs warm. Undervolting suits it best.
Local Electric Dreams runs on this PC with ONLINE off; an RTX 4080 run is selected and its silicon map shows speed only.
ONLINE off: the runs saved on this PC. The RTX 4080 model has no leakage rating, so its map shows speed only.

Compare runs

Tick two runs in the Compare column and click Compare. Both runs must use GPUs from the same vendor, NVIDIA or Radeon.

Comparison of two RTX 5090 runs: core frequency over the run for both, a Min, Avg, Max and Median table, and averaged results side by side.
Two RTX 5090 runs compared by core frequency. Hover a number to find it in the list; click the screenshot to see it full size.
  1. 1Value at the cursor. The time and the values of both runs where the mouse pointer is on the chart.
  2. 2Chart value. What the chart shows over the run: FPS, FPS per 100 W, clocks, voltages, temperatures or power.
  3. 3Chart. The chosen value over the run, one line per run.
  4. 4Statistics. Minimum, average, maximum and median of the chosen value for each run.
  5. 5Runs and values. GPU, silicon, CPU, RAM, driver, Windows build, benchmark version and score of both runs, then their results or settings.
  6. 6Swap sides and Settings. Swap runs #1 and #2, or switch the table between the results and the settings of each run.
  7. 7Go back. Returns to the scoreboard.
  • Green marks the better value in each row: higher for the score, FPS and clocks, lower for voltages, temperatures, power and limits.
  • Results shows the averages of the run; on NVIDIA: FPS and 1% low, core and XBAR clocks, NVVDD and MSVDD, core and hotspot temperature, video memory clock and temperature, and board power. Settings shows what was set in HYDRA; on NVIDIA: core, XBAR, VRAM and SYS offsets, XBAR Ratio, NVVDD and MSVDD Lim, and temperature and power limits.
  • Show more on the scoreboard opens the same view for a single run.
The same comparison with the table switched to Settings: offsets, XBAR Ratio, NVVDD and MSVDD limits and the power limit of both runs.
The table switched to Settings: what was set in HYDRA for each run.
Run details of a single RTX 5090 run: core frequency over the run and its averaged results.
Show more: the charts and details of one run.