* Posts by IgorS

20 publicly visible posts • joined 7 Oct 2016

As NSA buys up Americans' browser records, Uncle Sam is asked to simply knock it off


Barking at the wrong tree

> This personal information flows from web and native apps on people's devices to app makers and their marketing partners, and then to data brokers who sell it on to others, and it can be had by Uncle Sam's agents without a warrant.

If the data is available to anyone offering money, then the users have already lost.

The fact that (US) national agencies pay for it, too, is only mildly annoying at this point.

FTC should make sure such data cannot be put for sale in the first place!

Drivers: We'll take that plain dumb car over a flashy data-spilling internet one, thanks


Re: "better insurance rates"

>I get about a $15 discount every six months

Your insurance must be really cheap for you to care about the difference!

Qualcomm in recovery position following annus horribilis


Low power is a differentiator

For mobile devices, low power is sorta obvious.

But even for desktops and datacenter, we are getting to a point where the power starts to be a real issue.

Apart from the electricity costs, all the noise created by the cooling solutions is just obnoxious.

The alternative is direct liquid cooling, with its own set of problems (and costs).

Here is where Qualcomm has a real chance.

NVIDIA has gone too much into the "high power" business lately, so they are likely vulnerable, too.

Nvidia's accelerated cadence spells trouble for AMD and Intel's AI aspirations


NVLink attacked NIC?

"The NVLink mesh is also only good for GPU-to-GPU communications"

I don't see why NVIDIA could not just add NVLink support to their NICs.

They own the whole stack, so it is within the realm of possibilities.

The larger problem I see is the partnership with other vendors, like Cray, that would have a hard time doing the same.

Intel CTO suggests using AI to port CUDA code to – surprise! – Intel chips


Niche player pushing niche solutions

If INTEL was serious about providing alternatives to NVIDIA, it would partner with the rest of not-NVIDIA GPU ecosystem.

In particular AMD, who has a very credible GPU solution on the market already!

And they have indeed invested a lot in software to support it, too.

But, no, INTEL has to go its own path and try to make its own walled garden.

(Only INTEL really supports SYCL right now!)

I get they want to make their own lock-in, but you cannot get that when you are a niche player.

They just can't get over the idea that they are not the dominant ones anymore/in this field.

Another thing you can blame AI for: Cloud grows but server shipments are down


Comparing apples to bananas

The title is like comparing apples to bananas.

"Cloud market" in $$$ vs "on premise shipments" in server units.

Only way down in the article it states "the server revenue is still on track for growth".

How about you put that in the title?

Mobileye touts bright future while Nvidia, Qualcomm win over automakers


What could possibly go wrong?

> Much like Nvidia's Drive Thor, the chip integrates multiple vehicle domains, like infotainment, ADAS, and autonomous driving into a single platform.

What could possibly go wrong?

Arm execs: We respect RISC-V but it's not a rival in the datacenter


Re: What goes around comes around

All CPUs are CISC these days.

Yes, RISC-V is technically a RISC processor, but that's just the core. Most implementations will have extension that are very "complex".

In the end, RISC-V vs ARM vs x86 performance really boils down who has the best extensions (supported in compilers and libraries), as that's where most of the compute speedup comes from.

"The base RISC-V CPU ISA is relatively light and simple, with fewer than 50 instructions, and can be extended as needed by implementers to suit their applications.... extensions include support for floating point math, atomic instructions, vector math, and so on."

Compound that 'remembers' phase transitions could have uses in computer memory


Re: up to three hours

3 hours is plenty for active working memory.

At worst, you just refresh it once an hour, or so (like we do (much more frequently) with DRAM).

Has Intel gone too far with its Ohio fab 'delay' stunt?


Nonsense article

Intel was clear that it was just postponing the ceremony, not the actual construction. If the so called journalists cannot parse the details, it is their problem, not Intel's.

As for sticking to the plan, and not actually delaying the construction start date:

That's how you distinguish a serious business from an exploitative one.

Intel would sure love the subsidies, but they seem to understand that they need the facility no matter what.

And this is good news for the government, as it is much more likely the (eventual) subsidies will actually be used for the stated purpose, and not just squandered in a minimal showcase.

Apple arms high-end MacBook Pro notebooks with M1 Pro, M1 Max processors


Great to see I am not the only one!

Not that I use it all that often, but it is more useful than the F-keys by far.

RIP ROP, COP, JOP? Intel to bring anti-exploit tech to market in this year's Tiger Lake chip family


ARM equivalent?

> Other architectures, such as Arm, have something similar

Anyone knows what the equivalent ARM mechanism is?

PC owners borg into the most powerful computer the world has ever known – all in the search for coronavirus cure


Would be amazing, if true.

Since most of the compute come from gaming hardware with very dismal fp64 FLOPS capacity, 470 fp64 PFLOPS would be more like 20 fp32 EFLOPS! (and most compute is likely not fp64)

I find that really hard to believe.

I could not find any concrete stats to back this up.

'This repository is private' – so what's it doing on the public internet, GE Aviation?


DNS problem only?

How can a DNS misconfiguration lead to this kind of problem?

Was the information on a public IP address? Were they relying on security-through-obscurity only?

Mozilla returns crypto-signed website packaging spec to sender – yes, it's Google


Re: Can we get Web caching back, please?

I am well aware of the current CDN model. And it is a security nightmare!

The content owners have to give up any control to get the needed speedup.

I was under the impression was the Google's proposal was about improving exactly this.


Re: Can we get Web caching back, please?

> > The need for integrity indeed very valuable, but that could be easily achieved in a cache-friendly manner.

> The worry isn't just that content might be read, but that it might be rewritten in transit.

Which part of the above "integrity" sentence did you not understand?

integrity == "cannot be rewritten"

And can be easily solved with e.g. "server side signing", whioch is indeed cache-friendly.

No encryption needed.


Can we get Web caching back, please?

Looks like they are trying to fix what they have broken themselves!

Before the hard push for HTTPS, most Web content could be cached, for example by Squid.

No need for complicated schemes to get great performance... all you needed was a HTTP cache somewhere close.

Then HTTPS-everywhere mania kicked in, and now every single load has to go back to the origin!

Most Web pages we consume have zero privacy needs; they come from public Web pages.

There is really no need for encryption in most of them!

The need for integrity indeed very valuable, but that could be easily achieved in a cache-friendly manner.

Why is/was that not done?

Just usual business interests of big players, or is there something else I cannot see?

2 + 2 = 4, er, 4.1, no, 4.3... Nvidia's Titan V GPUs spit out 'wrong answers' in scientific simulations


Anyone knows if this affects the server-class V100s, too?

Dirty COW explained: Get a moooo-ve on and patch Linux root hole


Re: There will always be another bug..

Why should a regular user care?

Especially for something like a networked light bulb?

As long as it does what it is advertised for, that's all the final user cares about.

The real problem is the abuse of IoT for actions that are not user visible.

Like starting DDOS attacks.

We need to make manufacturers responsible for any actions of their devices that were not explicitly advertised to the users. Then the manufacturers will start paying attention!

It is really not fair to blame the final users.

Building iRODS to take load off scientists' back


Not a single file namespace solution

I don't think IRODS share the namespace between zones.

So the picture above would not have a single, global namespace.