Global Illumination with Cosine-Weighted Important Sampling // 1 Sample per Hemisphere// 16 depth // Low-discrepancy 256-1024 SPP // Shadows 4-32 SPP // 95% Confidence // MAX_DELTA 0.01 // 12 Threads // Core i7-10750H // Render-time 00:16:22 // Needs a Denoiser and MIS

MULTI-THREADING IN BUCKETS DONE LETS GOOOOOOOOOO // Global Illumination with Cosine-Weighted Important Sampling // 64 Samples per Hemisphere // Low-discrepancy 16-64 SPP // Shadows 4-32 SPP // 95% Confidence // MAX_DELTA 0.01 // 12 Threads // Core i7-10750H // Render-time 00:16:15

LETS GOOO // Global Illumination with Cosine-Weighted Important Sampling // 32 samples per hemi // Low-discrepancy 16-32 SPP // Shadows 8-32 SPP // 95% Confidence // MAX_DELTA 0.01 // No Multi-threading // Core i7-10750H // Render-time 00:29:30

After dividing the result by the number of bounces looks like I was right but is still a bit lighter than before overall.

Finally did GI. One sample, but I think it actually bounced and took the sum of 5 bounces.

Monte-Carlo Global Illumination // Low-discrepancy Adaptive AA 100,000 SPP // 99.99% Confidence Interval // MAX_DELTA 0.01 // No Multi-threading // Core i7-10750H // Render-time 00:00:01