Save $$ on Hardware & Power
with Impact's
“Build-Your-Own-Supercomputer” Formula
|
|
|
|
Build your own supercomputer from scratch using German engineering and quality
|
|
This “build-your-own-supercomputer” formula makes itself evident in comparison to IBM’s Watson supercomputer, the old but respected staple of supercomputers. Companies can follow this calculation to see if it’s worth it for them to stop renting an existing supercomputer and build their own from scratch using German engineering and quality. Consider the following side-by-side feature review (see detailed pricing for individual A.I. grade computer workstations
HERE -
prices as of Fall 2019
):
|
|
Watson is not a new supercomputer;
it hasn't even made the list of the 500 most powerful supercomputers. Its Linpack performance stands at 80 TeraFLOPs, which is about half as fast as the cut-off line for the "Top 500 Supercomputers" list. Watson employs a cluster of 90 “IBM Power 750” servers connected together, each of which uses:
|
|
|
|
One 3.5 GHz POWER7 eight-core processor (total speed 28GHz per unit x 90 =
2,520GHz for the whole supercomputer
|
|
|
Four threads per core,
2,880 POWER7 processor threads total for the whole supercomputer
|
|
|
About 177GB RAM each, the system consists of 90 units total and has
16 terabytes of RAM
|
|
|
IBM's master inventor and senior consultant, Tony Pearson, estimated Watson's hardware
cost at about $3,000,000
|
|
Built with Impact
IMP-3488-H1-8176-1T-2T-G4-8000. Y
ou don’t need 90 units of servers/computer workstations connected together. With each
IMP-3488-H1-8176-1T-2T-G4-8000 you get:
|
|
|
|
Two 2.1 GHz base/3.8 GHz turbo 28-core processors, 117.6GHz per unit x 22 =
2,587.2GHz for the whole Impact-based supercomputer
|
|
|
56 threads per core =
68,992 processor threads total for the whole supercomputer
|
|
|
1024GB RAM each. In total, the system consists of 15 units and has
15.3 terabytes of RAM
|
|
|
522 (=4x130.5) Tensor Deep Learning TeraFLOPS per unit = many times more per workstation unit than “IBM Power 750” server
|
|
|
1 year
for the whole system; 2 years for the motherboard used inside each workstation
|
|
|
With so many more TeraFLOPs in Impact-based supercomputer, we estimate you will require 1/3 less servers/computer workstations. So, with only 15* pcs of IMP-3488-H1-8176-1T-2T-G4-8000 (at $56K each), your required
investment cost is
only $840,000
for YOUR OWN supercomputer hardware comparable to IBM’s Watson.
|
|
Note*: This price/number of pieces is an estimate. Additional price breaks may be available for quantities of more than five workstations. If you desire the same overall main processor speed for the main comparison criteria as IBM Watson, only 22 pcs of IMP-3488-H1-8176-1T-2T-G4-8000 are necessary (at $56K each) = $1,232,000, and this supercomputer will offer much greater performance compared to IBM's Watson: 1.4 times more RAM and many more TeraFLOPs.
|
|
|
MONEY IS POWER, AND POWER IS MONEY
|
|
|
One can calculate the total power consumption of IBM Watson's 90 servers operating at the same time. Impact’s hardware-based supercomputers offer significant power consumption savings. Less servers / computer workstations mean less power consumption, resulting in even more money savings. As illustrated above, Impact was able to cut the number of connected computers by at least six. Plus, IMP-3488-H1-8176-1T-2T-G4-8000's do not require additional excessive energy for cooling, because they simply don’t produce much heat – see article below about another supercomputer that belongs to the Swiss National Supercomputing Center and review
info
about Impact’s single-rail power construction and 22 built-in temperature sensors in each workstation.
IMPACT’S A.I.-GRADE COMPUTERS DO NOT REQUIRE A WATERFALL FOR COOLING!
As the
article
states about
the Swiss National Supercomputing Center,
"Because of this extraordinarily large amount of data, the researchers needed a supercomputer to put their artificial intelligence into action. They ultimately ran their A.I. on a computer at the Swiss National Supercomputing Center in Lugano, a city in southern Switzerland which borders Italy. The supercomputers at CSCS are available to all Swiss universities and research institutions. Its machines are so powerful that, in order to stop them overheating, water from the nearby Lake Lugano is pumped in for cooling at a rate of 460 liters per second."
|
|
|
|
|
|
|