I’ve been looking at high end OpenCL 1.2 compatible GPU’s and it seems the difference in prices between professional targeted and private targeted GPU’s are quite high compared to potential benefit.
For example the Radon HD7990 has about 6GB of memory, rated to perform 8.2 teraflops and costs about 1300usd. The Firepro W9000 costs 3400 usd, has 6 GB of memory and is capable of 4 teraflops.
There are other examples for nVidia (Quadro vs GeForce) and also for CPU’s (Xeon vs Core i7).
So why would one get a W9000, when you could instead buy three HD7990’s for about the same price? Any thoughts?
There are a few differences between workstation grade GPUs and consumer ones. The main ones are that normally FP64 (double precision) performance is generally faster, the chips are rated for higher workloads (e.g. continually running at close to 100% utilisation), and have ECC (error correcting) memory (so less likely to make computational errors).
Whether these are worth the extra cost is a matter of debate. It also used to be the case that Linux/Unix drivers for the consumer cards were not up to scratch, but vendors have improved on that a lot in the last couple of years.
Also bear in mind that in your example, the W9000 is a couple of years old and has just been replaced with the W9100.