twtw

51 ms·

481.
by twtw 8y ago
Fun fact: the FFT (the algorithm that is actually used in applications today) was originally developed around 1805 by Gauss, but he didn't publish the work. Sorta crazy since it precedes Fourier's work.
482.
by twtw 8y ago
Never. Not going to happen.
483.
by twtw 8y ago
Analog electronics is basically only used nowadays when digital literally cannot accomplish the task (RF, SERDES, ADC/DAC). I haven't seen any asynchronous designs that can do things sync designs can't do, just that claim to
484.
by twtw 8y ago
He said the word in a meeting regarding the use of "sensitive words" in comedy.
485.
by twtw 8y ago
A net neutrality bill already passed the senate. With a new house (and speaker) it could potentially go through.
486.
by twtw 8y ago
> Only use threads where true CPU concurrency is needed. This is the case much more often now than it was in 1995.
487.
Genomics Code Exceeds ExaOps on Summit Supercomputer(olcf.ornl.gov)
7 points by twtw 8y ago|0 comments
488.
by twtw 8y ago
could you specify what generation you are talking about? There have been a lot of GPUs.
489.
by twtw 8y ago
All the current rumors and info I've seen indicate that Volta will never be in a consumer graphics card, so it has already launched (tesla v100 was available ~July 2017).
490.
by twtw 8y ago
It's definitely possible, but I just want to point out for other readers that it's not like the collaborators haven't thought about this stuff. Quoting from the article at https://arxiv.org/abs/1805.12028
491.
by twtw 8y ago
It depends highly on the region and culture. In some areas of the U.S. these expectations are still around (and quite achievable).
492.
CUDA: New Features and Beyond (GTC 2018) [pdf](on-demand.gputechconf.com)
3 points by twtw 8y ago|0 comments
493.
by twtw 8y ago
They were expecting to see about 1600 events, and instead they saw about 2000 events. The significance of the excess in miniboone is therefore around 4.5 sigma.
494.
by twtw 8y ago
I'm an electrical engineer oriented towards signal processing and controls, and I'm also incapable of understanding Matlab's niche and effectiveness. I wish Matlab could just vanish. It's a glorified calculator that has
495.
by twtw 8y ago
That's what I find particularly cool about stuff like nvidia DGX systems: they are great at machine learning stuff, but they also have ridiculously high performance for more traditional scientific computing (GROMACS, QUDA, etc). A DGX-
496.
by twtw 8y ago
As a point of interest, ARM is using NVIDIA's open source Deep Learning Accelerator (NVDLA) IP in Project Trillium. https://nvidianews.nvidia.com/news/nvidia-and-arm-partner-to... http://nvdla.org
497.
by twtw 8y ago
> absolutely no idea what you are doing. Isn't that sort of the idea of a textbook? The authors emphasize how poor of a model the normal distribution is in that case less than a full page after introducing it as a "simple hypot
498.
by twtw 8y ago
The submitted title doesn't correspond to the true title.
499.
Nvidia launches colossal HGX-2 cloud server to power HPC and AI(techcrunch.com)
60 points by twtw 8y ago|21 comments
500.
Deep RL in a Handful of Trials Using Probabilistic Dynamics Models(sites.google.com)
1 points by twtw 8y ago|0 comments
501.
by twtw 8y ago
The info page states that they are all periodic solutions.
502.
by twtw 8y ago
Regular cruise control is a feedback system, comparing the current speed to the desired speed constantly and adjusting the throttle to make the two match. A fixed throttle would result in wildly varying speed based on road grade and other f
503.
by twtw 8y ago
I had the same response. AlphaZero was published like 5 months ago. Saying they've reached the end of the line because they haven't matched AlphaZero in six months is lame.
504.
by twtw 8y ago
> Nvidia car could not drive literally ten miles without a disengagement. From the same source as the author cites, that's because their test runs are typically 5 miles and resuming manual control at the end of a test counts as a di
505.
On to the Next Thing – Matt Pharr Joins Nvidia(pharr.org)
3 points by twtw 8y ago|0 comments
506.
by twtw 8y ago
Why?
507.
by twtw 8y ago
Google started playing this game with the tpuv2, wherein they define a "TPU" as whatever they want to make it sound suitably impressive (I.e. Cloud TPU is four chips). This in turn led to nvidia calling the DGX-2 "the world&#
508.
by twtw 8y ago
Instead of being able to buy your own nvidia gpu and run any cuda or opencl on it, now you can run only tensor flow on a tpu only in Google cloud. How fantastic.
509.
by twtw 8y ago
No it doesn't. xor swap works fine when the two values are equal. The problem is when &a == &b.
510.
by twtw 8y ago
Great work, RiseML. This benchmark is sincerely appreciated. I wonder whether NVLink would make any difference for Resnet-50. Does anyone know whether these implementations require any inter-GPU communication?

More ›