deafcalculus

How does failover across zones work?

The blog post says there's one static ip per zone. I suppose www.mydomain should have multiple A records each pointing to an elastic ip in a zone. What happens when one zone entirely fails? Does it need a DNS change at this point? Or does the NLB have a different IP with which it can do BGP failover?

In this context, I intended it to mean a combination of addition, multiplication, and a small set of relatively simple non-linear functions.

Consciousness is likely just a whole bunch of computation.

I suspect "What is consciousness?" will go the way of "What is life?". We more or less understand things that make up a bacteria. Those components aren't alive although the bacteria is. So, it's just a matter of definition.

The high bandwidth price is mostly to drive away seedboxes and porn sites. Huge discounts are likely available to big customers.

Eliminate the driver and Uber loses the already weak marketplace network effect. Why would a taxi fleet service come to Uber rather than having their own app?

Why doesn't PS4 use LEDBAT for background downloads? Wouldn't this address the latency problem without sacrificing download speeds? AFAIK, Macs do this at least for OS updates.

Same here. I can access http://web.archive.org/ on ACT and on my university campus internet which peers directly with the National Internet Backbone (NIB).

Do you think there's another housing bubble now?

What's the killer app for mouse level AI?

I suspect the high bandwidth price is a targeting tool and is primarily for repelling those wanting to host seed boxes, porn sites and the like. You can probably get a much better deal if you're paying > 10k$/mo.

Even then, cloud bandwidth is insanely expensive. For example, Hetzner offers 1.3$/TB (if you happen to exceed their generous 30 TB quota). In comparison, Amazon is 70x more expensive at 90$/TB.

I love the intel graphics linux driver. It's the only GPU with which I've been able to get HDMI working on a laptop in Linux.

What I don't like is Intel spending so much area for the iGPU[1]. Why should an iGPU be consuming nearly the same area as 4 cores? An area optimised iGPU that can only do a few GFLOPS is good enough, and where it isn't (gaming, deep learning), a discrete GPU with wide memory will be needed anyway.

[1] http://www.anandtech.com/show/10968/the-intel-core-i7-7700k-...

It's high time Intel started adding more cores to consumer CPUs rather than spending half the silicon area on a crappy integrated GPU. It's only thanks to Ryzen that this is happening.

That's about what I did for a while. Then realized Twitter is a better fit for public posts. Now, I visit Facebook in an incognito window only when my mom tells me to check out something.

Smoothies aren't as good as eating the whole fruit! Fruit is better than juice because it has fiber. The insoluble fiber becomes something like a mesh inside the intestine and prevents food from getting digested too soon. When you put the fruit through the blender, the insoluble fiber is ripped apart and you lose a lot of its benefit. Dr. Lustig, an endocrinologist at UCSF, talks about this in one of his lectures. It's on YouTube, but I can't remember exactly which lecture.

I voted for Alexandra Elbakyan.

Her blog - https://engineuring.wordpress.com/ Her email - mindwrapper@gmail.com

The new APIs (Vulkan, DX12, Metal) are mostly about performance. I can't think of any major thing that you can't implement in WebGL. But, a portable 3d API would be awesome for native apps. And, why not have that be the 3d API for the web too? Like WebGL / OpenGL ES 2.

This is awesome. IMHO, reducing the complexity of resource management in the app is the right way to go for a web API, and Metal is closer to what a web API should be than Vulkan.

Porting an app from OpenGL to Metal is much easier than porting to Vulkan. A naive port to Metal often performs better than the GL version, whereas getting performance out of a Vulkan port is a lot harder [1]. So, I really hope the proposed API will be a C-based API that can also serve as a middle layer on top of Vulkan/D3D12 for lesser mortals like me in writing cross-platform native apps.

[1] aras-p is a Unity dev: https://twitter.com/aras_p/status/628569113053528065 and https://twitter.com/aras_p/status/628569238794543104

Turn to Twitter.

Vote to serve only magnet links.

What can each one of those 2.5 million "logic elements" do? Last time I used an FPGA, they were mostly made up of 4-bit LUTs.

How many NOT operations can this do per cycle (and per second)? I realise FPGAs aren't the most suited for this, but the raw number is useful when thinking about how much better the FPGA is compared to a GPU for simple ops.

Any chance Google will match this price for their coldline storage? I was planning to archive a few TBs in Google coldline, but Glacier is now cheaper and has a sane retrieval pricing model.

developer tools, operating system and software ecosystem, probably other IPs (like GPUs) integrate well with ARM but that might not be the case for a brand new processor, and a history of success (selling processor IP is enterprise sales).

The difference really is more in the architecture than in the programming model or the instruction set. SIMT programs, such as OpenCL kernels, can indeed be accelerated with SSE/AVX, but it is not that efficient (in silicon utilization). Intel's cores are fat cores optimized for single threaded programs. But if you don't care about single threaded performance and optimize for throughput in SIMT programs, the fat core is not a great idea. It is more efficient to hide latency to DRAM with computation rather than through caches, super-scalar and out-of-order execution, register renaming and so on. Expand Intel's SMT (hyperthreading) from 2-threads to 1024-threads, get rid of all the stuff that's there to hide latency and use the saved silicon area to add more cores, and you're looking at something that resembles a GPU.

Do you use compute shaders / CUDA? I'm really surprised at the error rate. I've used fragment shaders in OpenGL ES 2.0 for compute on mobile platforms, and the "errors" turned out to be dithering.

With the right extensions, you can render to a texture which can be sampled from in subsequent draw calls.

The framebuffer object extensions that allow you to write to 32-bit RGBA textures are widely supported on desktops and mobiles (OpenGL ES 2). But floating point textures are not. So, shaders resort to encoding 32-bit floats in RGBA textures. This unfortunately isn't a simple cast. More here: http://aras-p.info/blog/2009/07/30/encoding-floats-to-rgba-t...

Could you please hide comments at depth > 2 by default? Comments on popular posts tend to be deeply nested.

The one thing I've found baffling about both AWS and GCE is the bandwidth pricing. At about 0.1$/GB, it's 10x more expensive than competitors like linode. Why is that?

Bandwidth cost is the primary reason I'm afraid of hosting side projects on AWS/GCE.

I wonder if even that's a good idea for Sony seeing how Microsoft faced a severe fragmentation issue with Kinect twice. If Sony offers VR as an optional add-on, most devs will throw in half assed VR features, which is not good for any one. IMO, a separate PSVR that's completely unconnected to the PS4 might be better for developers, gamers, and the future of VR.