shaklee3

67 ms·

841.
by shaklee3 7y ago
> The motorized mount is only for pointing optimization during the installation process; it's not going to be constantly motoring around. Please cite a source on that one. If they're putting a motor on just for installation t
842.
by shaklee3 7y ago
Oneweb is not using the same technology. They're both using phased array, but SpaceX couldn't solve the low price point like oneweb, so they have it mounted on a motor. That will surely increase cost and reduce longevity. Oneweb i
843.
by shaklee3 7y ago
Elon says they're using boats now, which, if you read between the lines, it means that's not working.
844.
by shaklee3 7y ago
It's not. they're using phased array on the ground, which is far more expensive than the already-too-expensive-for-poor-countries parabolic dishes.
845.
by shaklee3 7y ago
The launch costs are nowhere near the most expensive part. By using leo they have cornered themselves into using technology that won't be available to consumers.
846.
by shaklee3 7y ago
Where, just out of curiosity?
847.
by shaklee3 7y ago
Not sure if you're aware, but it's widely publicized that SpaceX was granted licenses, waivers, and extensions that other companies were not, nor have been in the past. So yes, they have licenses, but it seems lobbying and Elon&#x
848.
by shaklee3 7y ago
Kitty doesn't seem to be available for Windows though, right?
849.
by shaklee3 7y ago
I second that, and it just seemed to be lacking features overall, and that was the direction the author wanted the project to go. It was great, but too many things missing for me.
850.
by shaklee3 7y ago
Maybe I just don't read all the articles, but the ones I've seen on there are definitely for low-level programming. Things like release/acquire semantics, writing a game engine, and AVX512 penalties.
851.
by shaklee3 7y ago
For low-level software, these are great: https://eli.thegreenplace.net/ https://travisdowns.github.io/ https://preshing.com/
852.
by shaklee3 7y ago
I have results showing otherwise. Maybe an old cuda? Here is anandtech showing almost 100TFLOPS, and they didn't every try hard to tune it: https://www.anandtech.com/show/12170/nvidia-titan-v-preview-...
853.
by shaklee3 7y ago
1) not true. I have one and hit 100TFLOPS with large enough batches in fp32 accumulate mode. Other benchmarks agree, so I'm not sure what numbers you're suggesting. 4) maybe get a cluster of Jetsons?
854.
by shaklee3 7y ago
Just out of curiosity, why is a V100 with 100TFLOPS of fp16 not good enough?
855.
by shaklee3 7y ago
Omnipath went the way infiniband is going. Ethernet has caught up, and surpassed the speeds, so using proprietary technology with fewer features isn't that attractive anymore.
856.
by shaklee3 7y ago
Ugh, speech to text. Thanks for finding that in time.
857.
by shaklee3 7y ago
It doesn't invalidate the numbers if they were in ideal conditions, but it certainly means you cannot compare them to regular human drivers. That 500k number is the average from all kinds of different driving conditions. You would have
858.
by shaklee3 7y ago
Yes, thanks for the correction. I should also add that the reason it's not fair to compare these kinds of cards is the inter-processor link is usually extremely slow. Because of that, performance with that kind of card is roughly equiv
859.
by shaklee3 7y ago
Yes, they make the K40, which has two ASICs. They do not make a current generation card with two ASICs.
860.
by shaklee3 7y ago
It's not, though. The blas and fft libraries are not only much slower (less efficient for the claimed peak of the cards), but they're missing features. For example, gemms don't work on complex half precision in any opencl lib
861.
by shaklee3 7y ago
I've seen those A100 rumors, but I think they should be taken with a grain of salt. They did not specify where the 70% speedup will be, and it's highly likely it will not be fp32 increasing by that much.
862.
by shaklee3 7y ago
That's not equivalent. That's a card with two ASICs on it, with each being 14TFLOPS. It's identical to the V100, but 2-3 years later.
863.
by shaklee3 7y ago
I would add libraries to that list (cuda equivalent).
864.
by shaklee3 7y ago
I didn't think it was too positive. He found some documentation bugs, as well as what seems like really bad locking bugs.
865.
by shaklee3 7y ago
As usual, Kyle provides an awesome write-up. As much as I miss the old, funny, prose, the level of detail is still unmatched.
866.
by shaklee3 7y ago
The article uses no factual basis to back it up, though. In order for it to be competitive against existing internet, it needs to be as cheap. Their current pizza-box phased array antenna was likely too expensive, so now they have a motoriz
867.
by shaklee3 7y ago
The latency part is true, but the cap is not: https://www.satelliteinternet.com/resources/unlimited-satell... We can argue about whether latency matters for most people, but it's not going to go anywhere.
868.
by shaklee3 7y ago
JetBlue has had ViaSat for free since 2013.
869.
by shaklee3 7y ago
It didn't mention it because that's not true. This is targeted for enterprise, and there are no plans to sell cheap internet at the price the antenna costs.
870.
by shaklee3 7y ago
That's completely false. Several providers have over 100GB/mo plans.

More ›