OTOY Forums

Hi Guys Since Tegra X1 has over a teraflop of Power is it possible to run octane on the chip?

it should theoretically work, I think

Poodle wrote:Hi Guys Since Tegra X1 has over a teraflop of Power is it possible to run octane on the chip?

We don't support ARM CPUs. And the 1TFlop NVIDIA mentions is for half-precision floating point operations. Octane uses single-precision floating point operations and in one place even double-precision.

abstrax wrote: We don't support ARM CPUs. And the 1TFlop NVIDIA mentions is for half-precision floating point operations. Octane uses single-precision floating point operations and in one place even double-precision.

Hi, Abstrax, could You be more specific what processes run on DP?

glimpse wrote:

abstrax wrote: We don't support ARM CPUs. And the 1TFlop NVIDIA mentions is for half-precision floating point operations. Octane uses single-precision floating point operations and in one place even double-precision.

Hi, Abstrax, could You be more specific what processes run on DP?

Hair.

abstrax wrote:

Poodle wrote:Hi Guys Since Tegra X1 has over a teraflop of Power is it possible to run octane on the chip?

We don't support ARM CPUs. And the 1TFlop NVIDIA mentions is for half-precision floating point operations. Octane uses single-precision floating point operations and in one place even double-precision.

Then You would pretty much need a bunch of X1's just to "try" to tie the performance on a Titan, I say this because X1 is quiet and doesnt make any sound, Really NVIDIA can do an ARM chip that performs like a bunch of TITANs, engineers can make this work but they just increase the power a little bit every year to make a lot of profits, otherwise they lose money.

Poodle wrote:Then You would pretty much need a bunch of X1's just to "try" to tie the performance on a Titan, I say this because X1 is quiet and doesnt make any sound, Really NVIDIA can do an ARM chip that performs like a bunch of TITANs, engineers can make this work but they just increase the power a little bit every year to make a lot of profits, otherwise they lose money.

No, that's not correct. The tech that ends up in modern desktop GPUs is derived from mobile GPU technology. At NVIDIA this development started with Kepler and has continued with Maxwell. But if you want the power of a Titan you have to pay the (energy) price. There is no way around it. There are also other things to consider like memory bandwidth, cache sizes, PCIe speeds etc.

So in a nutshell: Current GPU technology is pretty good for what it achieves per Watt and if NVIDIA could release the power of a Titan without the effort they do to cool it, they would have done it already.

Like They have the knowledge on how to do it. Because they haven't release something doesn't mean that they can't do it or know how to.

Poodle wrote:.. Because they haven't release something doesn't mean that they can't do it or know how to.

if They would know how to..they would release that & charge Goooood money for that solution!
..if that doesn't exist..that's for a reason..whether it's power consumption, heat or other issues..