US Dethrones China With IBM Summit Supercomputer

Status
Not open for further replies.

DerekA_C

Prominent
Mar 1, 2017
177
0
690
1
So when they push it to the max is is more than double the performance of China's, sweet deal to bad Intel built China's super computer shame on Intel such damn sell outs of this great nation. Let China make their own CPU's GPU's Memory and other components and leave USA products in USA stop selling out it is ridiculous. MAGA!!!!
 

maniac62

Distinguished
Mar 8, 2010
11
0
18,510
0

From Wikipedia:
The Sunway TaihuLight uses a total of 40,960 Chinese-designed SW26010 manycore 64-bit RISC processors based on the Sunway architecture...
 

bit_user

Splendid
Ambassador
27,648 Tesla V100 accelerators
Which should be good for 3.46 deep learning exaflops. I'm a little surprised they're not touting that, given how much Nvidia likes to talk about the deep learning performance of the V100.

And... exaflops? Okay, it's not exactly general-purpose, but still... has anything else yet reached exaflop territory?

Okay, speaking of general-purpose, the general-purpose fp16 performance should be around 824 petaflops.
 

stdragon

Commendable
Apr 5, 2018
1,551
4
1,660
196
When the next one is named "Skynet", I highly recommend we pull the plug on it ASAP! But, well, we've all seen how that movie played out.
 

milkod2001

Distinguished
Apr 20, 2010
231
0
18,710
14
''was conceived to research the fields of human diseases such as Alzheimer and cancer, astrophysics, fusion energy, and climate change''So many supercomputers and nothing groundbreaking came from it regarding that area of research for last 20 years. I bet they use it for something 'more interesting' like processing our personal data, global surveillance and stupid rush to get the best Ai advantage.
 

Brian_R170

Honorable
Jun 24, 2014
288
2
10,785
0


You are thinking of the Tianhe-2 which used Intel chips and topped the supercomputer list in 2014. As a result, in 2015, the US Government banned Intel, NVidia, and AMD from selling chips to China for use in supercomputers. Less than two years later, China had created the Sunway TaihuLight that is much faster and uses only Chinese chips (this is the supercomputer referenced in this article). Exactly what you suggested came to pass: China made their own chips and it was a win-win... for China. The ban did absolutely nothing to stop China's supercomputer ambitions and US companies lost out on the sale of hundreds of millions of dollars in high-tech equipment.
 

bit_user

Splendid
Ambassador

No, the ban wasn't simply because China built the Tianhe-2. Rather, it was for non-public reasons probably relating to hacking or maybe IP theft.

Such bans are sanctions of last-resort, since it just gets China to accelerate development of its own technology. Then, not only do you lose the option of doing it again, but now there's another Chinese competitor to a US company and... oh, look - your trade deficit just got bigger!


Fortunately, it's also much less general than if it had used off-the-shelf CPUs and GPUs. So, I'd say it's only faster on paper. But that's enough to make their point.
 

stdragon

Commendable
Apr 5, 2018
1,551
4
1,660
196
If you know anything about the Chinese (the nation, not at the individual level per se) is that they're notorious for engaging in espionage and IP theft; both at the civil and governmental level. The military technology secrets that they've stolen from the US armed forces is enough to make anyone cry. Part because the Chinese are darned good at it, and partly because of our own incompetence at keeping the stuff secure at all levels.

Our (US) government fears back-doors embedded in Chinese IC hardware. And given our own attempt at the Clipper Chip in the 90s, I get the irony and hypocrisy of it all at the same time. Nonetheless, it is a valid concern at the national security level.

Regardless of banning Chinese products, they will build and export them around the world regardless. For one, because it's cheaper. And secondly, probably because the CCP would mandate spying on their own citizens anyways. Remember, their government is totalitarian. To embed a "snitch" in every box isn't all that far fetched of an idea for an Orwellian society.
 

bit_user

Splendid
Ambassador

OMG, if you just look at all the devices that are already using SoCs from Rockchip, Allwinner, and others, the Chinese chipmakers are already making major inroads to the point that a ban of Chinese SoCs would cripple many US tech companies.


It's only a matter of time, if not already a reality on their phones.
 

bit_user

Splendid
Ambassador

Is that what the entire thing cost? Because just spending that much on the GPUs would mean $7234 each. Even for the 16 GB Tesla V100, that's quite a decent price.

...and if they build the entire thing for that cost, then it's a steal!
 

kyotokid

Distinguished
Jan 26, 2010
244
0
18,680
0

...guess where Mr. Trump gets many of his furnishings and most supplies for his resorts from.

If you shop at a Dollar Store or Dollar Tree, guess where a lot of the products sold there come from as well.

Oh, and as others mentioned, China does make their own chips and CPUs.
 

kyotokid

Distinguished
Jan 26, 2010
244
0
18,680
0
...anyway, they finally got it up and running. I've been following the development of Summit for a while and this is a powerhouse. Those multiple NvidiaTesla 100s linked together via NVLink in each node can pool their memory resources (this is where the HBM memory total comes from). Each of them has 64 tensor cores which support FP64 computation and there are over 400 such nodes total. Deep learning? How about deep thinking?

Oh and I believe the Tesla V100 has 16 GB of HBM2. The 10,000$ Quadro GV100 has 32 GB. But at what, 6 cards per node fully crosslinked on NVLink boards, that means 96 GB of VRAM per node.

...which leads me to wonder, with all those Tesla GPUs NVLinked together, would it be able to render one of my epic Iray scenes in under an hour? ;-)
 

bit_user

Splendid
Ambassador

My bad. I guess I just assumed they'd use the 32 GB version (yes, it exists).

https://www.nvidia.com/en-us/data-center/tesla-v100/
 
Status
Not open for further replies.

ASK THE COMMUNITY

TRENDING THREADS