I'm looking at buying a used or "refurbished" workstation. The specs are:
2.70GHz Intel Xeon 8-Core E5-2680
32GB RAM
256 GB SSD
The above, all assembled and cleaned up is $600. (It's an HP workstation with "industrial" build quality.)
I will be adding a new GTX 10-series Graphic card.
I'm wondering if the Xeon is still reasonable compared to newer chips from Intel. My application runs dozens of compute intensive threads. I need high compute throughput. I'm also on a budget. Can I get more "compute power" for the same cost in different system? Are there higher performance systems available for $600 or less?
If you want to know exactly what I'm doing, it's two things, not at the same time....
1) training a Deep Learning network, for that the graphic card matters the most. A GTX1080ti is best but I'll likely have to "make do" with something less expensive. No monitor will be attached. The machine runs headless, using the GTX for computing not graphics.
2) Used to control a Robot using ROS. This is a great use of multithreading, I'll be running dozens of compute tasks that all communicate. So I have use for many cores and lots of RAM.
2.70GHz Intel Xeon 8-Core E5-2680
32GB RAM
256 GB SSD
The above, all assembled and cleaned up is $600. (It's an HP workstation with "industrial" build quality.)
I will be adding a new GTX 10-series Graphic card.
I'm wondering if the Xeon is still reasonable compared to newer chips from Intel. My application runs dozens of compute intensive threads. I need high compute throughput. I'm also on a budget. Can I get more "compute power" for the same cost in different system? Are there higher performance systems available for $600 or less?
If you want to know exactly what I'm doing, it's two things, not at the same time....
1) training a Deep Learning network, for that the graphic card matters the most. A GTX1080ti is best but I'll likely have to "make do" with something less expensive. No monitor will be attached. The machine runs headless, using the GTX for computing not graphics.
2) Used to control a Robot using ROS. This is a great use of multithreading, I'll be running dozens of compute tasks that all communicate. So I have use for many cores and lots of RAM.