My point is, are those 3 8core rack mount machines plus racks plus external cooling plus rack mount management infrastructure plus the gpu limitation really a cost savings over buying one 24core workstation that has much more gpu expandability and the flexibility to be a backup workstation? Not to mention simplifying the network bottleneck and render management problems...
Not to mention the problem of housing rack mount machines that due to their form factor tend to have small inefficient fans that run at very high speed and produce a lot of noise. You tend to want to put them in an environment that isolates their high noise level but then that usually means a small enclosed environment that itself needs cooling management to keep the nodes from overheating...
Basically for a small facility that doesn't have a machine room, rack mount render nodes might not be a great match.
Also be careful with older pc servers when trying to get the best cpu performance, what are the exact specs of the servers you may buy? If they are older architecture or low clock speed you may not be getting the performance you expect.
Not to mention the problem of housing rack mount machines that due to their form factor tend to have small inefficient fans that run at very high speed and produce a lot of noise. You tend to want to put them in an environment that isolates their high noise level but then that usually means a small enclosed environment that itself needs cooling management to keep the nodes from overheating...
Basically for a small facility that doesn't have a machine room, rack mount render nodes might not be a great match.
Also be careful with older pc servers when trying to get the best cpu performance, what are the exact specs of the servers you may buy? If they are older architecture or low clock speed you may not be getting the performance you expect.
Comment