info
Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
I got teraflops
22 views
Skip to first unread message
J. Clarke
Nov 5, 2017, 9:46:30 PM11/5/17
to
I am sitting here laughing maniacally.
I downloaded the CUDA sdk from the nvidia site and set it up, then
found a CUDA-enabled Whetstone and ran it, and it's showing EIGHT
TERAFLOPS in single-precision Whetstone. It's down to 250 gigaFLOPS
in double-precision mode, which is expected because Nvidia crippled
double-precision on the gamer boards to preserve the market for their
Quadros.
Looking at Tom Longbottom's whetstone archive it looks like this
machine is massively more powerful in terms of vector number-crunching
than anything that existed prior to 1996.
That's running on 2500 cores so it's only really useful for things
that parallelize, however I have an application in mind that will
parallize very nicely, so it looks like I may be having a fun year.
It's going to be interesting to see how that works out in the real
world.
Just had to share.
I downloaded the CUDA sdk from the nvidia site and set it up, then
found a CUDA-enabled Whetstone and ran it, and it's showing EIGHT
TERAFLOPS in single-precision Whetstone. It's down to 250 gigaFLOPS
in double-precision mode, which is expected because Nvidia crippled
double-precision on the gamer boards to preserve the market for their
Quadros.
Looking at Tom Longbottom's whetstone archive it looks like this
machine is massively more powerful in terms of vector number-crunching
than anything that existed prior to 1996.
That's running on 2500 cores so it's only really useful for things
that parallelize, however I have an application in mind that will
parallize very nicely, so it looks like I may be having a fun year.
It's going to be interesting to see how that works out in the real
world.
Just had to share.
0 new messages