Hi,
I played with the idea last year to build a dedicated single or twin GPU rig for my ML experiments. I didn't build that one in the end because the GPU upgrade helped my trusty old machine a lot. But now its finally time to retire this thing, and without overbuilding it, I need a new 'daily driver'. Doing mostly software development related stuff, sometimes a VM to test things, NumPy, a little ML.
8 - 12 Cores
64 -128 GB Ram
1TB NVME
Nvidia GPU
AVX512 would be nice, but not 'must'.
Option 1:
Xeon Gold QS + C612
(2 AVX512 Pipelines)
Option 2:
9800X + X299
Option 2.1:
Xeon W2xxx + C422
(Xeon W is expensive, Ram is 'reasonable')
Option 3:
Ryzen 3900X + Z570
While doing the math, the three options are not too far apart from a money-wise.
The 9800X is 'cheap' in terms of Intel, the Gold QS can be assembled with 'used ECC memory'.
Any other options or ideas? At the moment, I think about Option 2...
Cheers!
Styp
I played with the idea last year to build a dedicated single or twin GPU rig for my ML experiments. I didn't build that one in the end because the GPU upgrade helped my trusty old machine a lot. But now its finally time to retire this thing, and without overbuilding it, I need a new 'daily driver'. Doing mostly software development related stuff, sometimes a VM to test things, NumPy, a little ML.
8 - 12 Cores
64 -128 GB Ram
1TB NVME
Nvidia GPU
AVX512 would be nice, but not 'must'.
Option 1:
Xeon Gold QS + C612
(2 AVX512 Pipelines)
Option 2:
9800X + X299
Option 2.1:
Xeon W2xxx + C422
(Xeon W is expensive, Ram is 'reasonable')
Option 3:
Ryzen 3900X + Z570
While doing the math, the three options are not too far apart from a money-wise.
The 9800X is 'cheap' in terms of Intel, the Gold QS can be assembled with 'used ECC memory'.
Any other options or ideas? At the moment, I think about Option 2...
Cheers!
Styp