What's new
  • Please do not post any links until you have 3 posts as they will automatically be rejected to prevent SPAM. Many words are also blocked due to being used in SPAM Messages. Thanks!

Method identified to double computer processing speeds

LaughingCrow

Well-known member
Joined
Feb 8, 2010
Messages
1,458
Location
Southern Ontario
Can imagine that there might be some companies that wouldn't want this to affect their hardware sales...

In their paper, Tseng and UCR computer science graduate student Kuan-Chieh Hsu introduce what they call “simultaneous and heterogeneous multithreading” or SHMT. They describe their development of a proposed SHMT framework on an embedded system platform that simultaneously uses a multi-core ARM processor, an NVIDIA GPU, and a Tensor Processing Unit hardware accelerator.

The system achieved a 1.96 times speedup and a 51% reduction in energy consumption.
 

Izerous

Well-known member
Folding Team
Joined
Feb 7, 2019
Messages
3,678
Location
Edmonton
I won't claim to have read the whole thing but I skimmed over it. One consideration i didn't see mentioned (and even searched for it afterwards) is security of any kind. Already have vulnerabilities like meltdown, spectre, and many many more and further enhancing co-processing between the different components you open yourself up to further security vulnerabilities.

Taken at face value though companies like Amazon AWS I could imagine would potentially push for this. Not because of the performance boost but even more so because of the power efficiency gains he was claiming to see.

"SHMT with QAWS-TS reduces energy consumption and EDP by 51.0% and 78.0% on average, respectively."

You take a data center and even make a 5% power efficiency gain that is a massive difference in power consumption from the computers themselves... that also reduces the heat output... in turn reduces the work load on the cooling systems further reducing the consumption of the center as a whole. So to be claiming 50% ish efficiency gains would be a massive net effect when spread across a whole data center.
 

Marzipan

Well-known member
Joined
Nov 21, 2007
Messages
12,152
Location
Prince Rupert, British Columbia, Canuckistan
too technical for me...but does SHMT have a better comparative similarity to AMD's SMT over Intel's HT? by this i am thinking about whether it would be easier for AMD to implement than Intel. or is it just semantics with acronym's and doesn't matter either way, LoL!
 

Izerous

Well-known member
Folding Team
Joined
Feb 7, 2019
Messages
3,678
Location
Edmonton
too technical for me...but does SHMT have a better comparative similarity to AMD's SMT over Intel's HT? by this i am thinking about whether it would be easier for AMD to implement than Intel. or is it just semantics with acronym's and doesn't matter either way, LoL!
Simple version spread the CPU load to the GPU to increase processing capacity even if it is not necessarily a GPU specific load as we would think of it today instead of letting the GPU sit there idle.
 

Latest posts

Top