← Back to context Comment by bdhcuidbebe 4 months ago What about the Coral? I been running frigate with mine for 2 years. 7 comments bdhcuidbebe Reply zamalek 4 months ago What's the deal with memory? Does it use the system RAM? I've been considering a cluster for local inference. ReadEvalPost 4 months ago The data has to go across PCI / USB to get to the internal (tiny) SRAM. It's not a good choice for clustering, you should just use a GPU for that. utopcell 4 months ago It has a tiny amount, I believe it was 8MiB. That's Mega, not Giga. bdhcuidbebe 4 months ago Dont know, sorry. But the performance is not that awesome, but works well enough for my setup doing realtime object detection with a few cameras.My main reason for getting one was power efficency compared to a traditional GPU for this task. Fnoord 4 months ago I have a Coral M.2 and an Intel Xe iGPU on the same machine. Both passthrough to VM, Proxmox, iGPU with SR-IOV. On Proxmox I can see power usage of one camera for SR-IOV device utilizing intel_gpu_top: less than 0.1W. 2 replies →
zamalek 4 months ago What's the deal with memory? Does it use the system RAM? I've been considering a cluster for local inference. ReadEvalPost 4 months ago The data has to go across PCI / USB to get to the internal (tiny) SRAM. It's not a good choice for clustering, you should just use a GPU for that. utopcell 4 months ago It has a tiny amount, I believe it was 8MiB. That's Mega, not Giga. bdhcuidbebe 4 months ago Dont know, sorry. But the performance is not that awesome, but works well enough for my setup doing realtime object detection with a few cameras.My main reason for getting one was power efficency compared to a traditional GPU for this task. Fnoord 4 months ago I have a Coral M.2 and an Intel Xe iGPU on the same machine. Both passthrough to VM, Proxmox, iGPU with SR-IOV. On Proxmox I can see power usage of one camera for SR-IOV device utilizing intel_gpu_top: less than 0.1W. 2 replies →
ReadEvalPost 4 months ago The data has to go across PCI / USB to get to the internal (tiny) SRAM. It's not a good choice for clustering, you should just use a GPU for that.
bdhcuidbebe 4 months ago Dont know, sorry. But the performance is not that awesome, but works well enough for my setup doing realtime object detection with a few cameras.My main reason for getting one was power efficency compared to a traditional GPU for this task. Fnoord 4 months ago I have a Coral M.2 and an Intel Xe iGPU on the same machine. Both passthrough to VM, Proxmox, iGPU with SR-IOV. On Proxmox I can see power usage of one camera for SR-IOV device utilizing intel_gpu_top: less than 0.1W. 2 replies →
Fnoord 4 months ago I have a Coral M.2 and an Intel Xe iGPU on the same machine. Both passthrough to VM, Proxmox, iGPU with SR-IOV. On Proxmox I can see power usage of one camera for SR-IOV device utilizing intel_gpu_top: less than 0.1W. 2 replies →
What's the deal with memory? Does it use the system RAM? I've been considering a cluster for local inference.
The data has to go across PCI / USB to get to the internal (tiny) SRAM. It's not a good choice for clustering, you should just use a GPU for that.
It has a tiny amount, I believe it was 8MiB. That's Mega, not Giga.
Dont know, sorry. But the performance is not that awesome, but works well enough for my setup doing realtime object detection with a few cameras.
My main reason for getting one was power efficency compared to a traditional GPU for this task.
I have a Coral M.2 and an Intel Xe iGPU on the same machine. Both passthrough to VM, Proxmox, iGPU with SR-IOV. On Proxmox I can see power usage of one camera for SR-IOV device utilizing intel_gpu_top: less than 0.1W.
2 replies →