Expand all | Collapse all

Hardware Configuration

  • 1.  Hardware Configuration

    Posted 01-28-2019 23:54

    Hi all!

    I’m up to test drive the omnisci bd and was wondering about hardware config requirements!
    I saw the https://www.omnisci.com/docs/latest/4_hardware_configuration_guide.html but there is no mention of regular GPU type card like the 1080ti or the new rtx2080ti. On the https://lambdalabs.com/blog/best-gpu-tensorflow-2080-ti-vs-v100-vs-titan-v-vs-1080-ti-benchmark/
    The rtx2080ti conclusion is the following:

    • 80% as fast as the Tesla V100 with FP32, 82% as fast with FP16, and ~1/5 of the cost.
      Will the omnisci bd will work on regular rtx2080ti and 1080ti?
      Will the omnisci db will work on other than nvidia card like the ATI vega 64 card?

    All comments and experience/test results with hardware other than what is on the hardware_configuration_guide will be appreciated.


  • 2.  RE: Hardware Configuration

    Posted 01-29-2019 03:44

    Hi @spazio,

    I’m running the omnisci database on both cards you mentioned and it’s working flawlessly.
    In the end, the database works with any Nvidia card using Kepler, Pascal, Volta or Turing and being developed with Nvidia Cuda it doesn’t work on at cards.

    The performance on both cards it’s amazing; I did some benchmarking with analytical queries on a typical star schema and a single gtx1080ti is 75 times faster than an Oracle 12c with the in-memory option running on 16c/32t blade running the queries serially and 100 times faster when I tried to run queries concurrently.
    I won’t report any number about Exasol, because the difference is so embarrassing, that leads me to believe I did something wrong.

  • 3.  RE: Hardware Configuration

    Posted 01-29-2019 09:00

    Hi @ Aznable

    Many thanks for your inputs! I will sure now go ahead and take this DB for a spin!