From 3fd8fa1c5cfa498aa696f3818a601272dc336d0b Mon Sep 17 00:00:00 2001 From: Koushik Dutta Date: Thu, 26 Dec 2024 21:54:29 -0800 Subject: [PATCH] guide --- buyers-guide/servers.md | 16 ++++++++++------ 1 file changed, 10 insertions(+), 6 deletions(-) diff --git a/buyers-guide/servers.md b/buyers-guide/servers.md index f99fa68..3c62114 100644 --- a/buyers-guide/servers.md +++ b/buyers-guide/servers.md @@ -47,11 +47,11 @@ The [Object Detection Benchmark](https://scripts.scrypted.app/object-detection-b |Name|Rating|Notes| |-|-|-| -|TPU: Apple Mac Mini M1/M2|🥇|Low power consumption, fast, and expensive. A single package off the shelf solution for a large number of cameras.| -|GPU: Intel 13500H|🥈|The GPU on the H series Intel chips have 2-3x the typical number of cores. They are blazing fast at low power. Highly recommended.| -|GPU: AMD 6900XT or NVIDIA 3080|🥈|The Scrypted Desktop App running on systems using these GPU for Object Detection perform great, though at a much higher performance per watt.| +|NPU: Apple Mac Mini M1/M2|🥇|Low power consumption, fast, and expensive. Can easily handle a large number of cameras.| +|NPU: Intel Core Ultra 125H|🥇|The GPU on the H series Intel chips have 2-3x the typical number of cores. The Core Ultra series includes Intel's new NPU. Low power consumption, fast, and moderately priced. Can easily handle a large number of cameras.| +|GPU: Intel 12600H|🥈|The GPU on the H series Intel chips have 2-3x the typical number of cores. They are blazing fast at low power. Highly recommended.| |GPU: Intel N100|🥉|Fantastic Object Detection Performance in a tiny package. Highly recommended.| -|GPU: NVIDIA 4XXX|🥇|The NVIDIA 4XXX series has incredible decode and detect performance at the but at a high price and power usage. Generally not recommended for purchase, great if on hand.| +|GPU: NVIDIA|🥇|The NVIDIA 3000 and 4000 series has incredible decode and detect performance but at a high price and power consumption. Generally not recommended for purchase, great if on hand.| |TPU: M.2/PCI Coral|🥉|The M.2 dual TPU model is is not supported by many systems. **Google Abandonware. Do not buy if purchasing new hardware**.| |TPU: USB Coral|🥉|High latency and crash prone. Google Abandonware. **Google Abandonware. Do not buy if purchasing new hardware**.| @@ -63,9 +63,13 @@ All servers are recommended to have at least 8GB of RAM, 16GB preferred. ## Apple Mac Mini M1/M2 -This system contains a CPU, GPU, and a TPU (neural cores). An Apple Silicon Mac Mini is the most capable Scrypted NVR system available. +This system contains a CPU, GPU, and a TPU (neural cores). -## Intel 13500H +## Intel Core Ultra 125H + +This Intel 12th generation NUC contains a CPU and laptop class GPU and NPU in a tiny package. The H series chip has 80 GPU cores compared to the 24 on the Intel N100. The Intel Core Ultra 125H is the top recommendation for large scale deployments. + +## Intel 12600H This Intel 12th generation NUC contains a CPU and laptop class GPU in a tiny package. The H series chip has 80 GPU cores compared to the 24 on the Intel N100.