The bottleneck for most problems isn't CPU cycles/second, it's the bandwidth of getting data to/from those CPUs. Adding CPUs does nothing to improve performance unless you also give it a much wider I/O path to memory.
Threadripper parts have quite a lot of bandwidth. The pro parts ("Epyc") will have even more.
Threadripper is intended for the PC enthusiast market, not so much for data centers. Frankly I don't think that for even an enthusiast home user memory bandwidth will be a major differentiating point. CPU speed, number of cores, and cache size will likely matter more. (I'm not worried about any of the above. I can't find L1 or L2 cache size numbers but I found that the L3 cache is 32 MB and since AMD was very focused on instructions-per-clock I am confident they didn't skimp on the cache.)
For data centers AMD will be selling the Epyc chips, and those can support up to 2 TB of RAM per socket (i.e. a dual-socket server would max out at 4 TB of RAM). In contrast, Intel tops out at 1.5 TB, and to get that you now have to buy their special and more-expensive chips with the "M" suffix; the non-M chips top out at 768 GB of RAM.
Threadripper also has a really large number of PCI-E lanes (64) so in theory you also could set up a really wide-bandwidth SSD or RAMdisk or something.
While Zen Server parts (Epyc) look good on paper, it reamis to be seen if there will be Adoption from server makers, and demand from server purcharsers...
Well, sure. But unless the paper is a lie, those chips will do well. They will offer much-improved price/performance compared to Intel's server chips, they offer some tasty new security features (like VMs running with the in-RAM data encrypted so that there's no way for one VM to spy on another's memory), and they are doing it right when Intel is jacking their server customers on price.
corporate parts without IGP? Really? I mean, REALLY?!?!?
Does "IGP" mean integrated graphics? AMD is all over integrated graphics, they call such products "APUs" and the mobile lineup will be pretty much all APUs. So my guess is Q3 for corporate products with APUs as well. (I hope AMD supports ECC RAM on APUs, finally.)
In summary: I think violence, such as the events in Berkeley when Milo Yiannopolous was scheduled to speak, comes more often from the left than from the right. You disagree with this. I hope we can agree that violence is bad and we disapprove of it from either left or right, and leave it at that.
I think violence is not warranted to prevent words. I'm actually not sure whether you disagree with this, or not. But I hope you agree.
I see no point in further discussion; I believe it would be a waste of time for both of us. Have a nice life!
Are you trolling? I wrote that it's bad to paint all people who disagree with you with a broad brush (like "every conservative is a Nazi" or "everyone who wanted to hear Milo speak is a white supremacist") and you then posted a bunch of links about "white supremacists are bad".
You haven't shown that any of the people who wanted to see Milo speak were white nationalists, let alone all of them, and you absolutely haven't any excuse for equating an interest in watching Milo speak with the Oklahoma City bombing. Shame on you for that one.
Just in case this point is too subtle for you to understand, I'll spell it out for you: I don't approve of violence against the right, but I also don't approve of violence against the left. I said the remedy to bad speech is counter-speech, not violence, and I didn't make an exception to that for right-on-left violence.
So not walking away? Not ignoring them? You don't say they're unacceptable, but why not correct?
Sure, ignoring bad speech is fine. Making fun of the speech is also fine. What do you mean "why not correct", what do you think I meant by "counter-speech"? If someone says something dumb or bad, pointing out why it's dumb or bad is perfectly fine and a good idea. And it doesn't matter who was speaking, if one of my heroes says something dumb or bad, correct it; if someone I detest correctly points out something dumb or bad, then good for him/her.
is hauling people away violence or not?
It's violence and it's not justified for mere words. Hauling people away because they committed crimes and are going to jail, is still violence, but it's justified violence (at least if you agree that their "crimes" are actual crimes; for example, I wouldn't agree that "lese-majesty" is a real crime and people deserve to be hauled away for it). (Hauling people away is not as violent as beating them up and sending them to the hospital, but it's still violence. There is a saying, "the state holds a monopoly on violence." You might want to look that up and think about what it means.)
when is a public official obligated to listen?
Nobody is obligated to listen to any particular speech of anyone else. When President Trump talks, you don't have to listen. When you talk, President Trump doesn't have to listen.
Of course, if public officials ignore the people, they may lose their next election.
The smashing, violence, and beatings is overwhelmingly coming from the left. That's how it's been for many years.
I have to agree. My personal theory is that many of the violent people have convinced themselves that their political enemies are in fact bad people and "fair game" for anything. It goes like this: It's okay to punch a Nazi; conservatives are all Nazis... and then comes the punching.
Here is a web page linking multiple articles arguing that the violence used to prevent Milo Yiannopolous from speaking at Berkeley was justified. "Violence helped ensure safety of students" is a real headline. There was also this quote: "...some white nationalists got their ***** beat." (Just like the Nazi thing above, only this time using "white nationalist".
Someone who wanted to hear Milo speak --> white nationalist --> someone it's okay to send to the hospital.)
Also, the media coverage may tend to embolden these people. The people who smash things, light things on fire, and send people to the hospital are described as "protesters". The people who wanted to hear Milo speak are described as "alt-Right extremists". I don't want to overstate the contribution of the media but I think it's a part.
Personally I think that the correct remedy for bad speech is counter-speech. Violence isn't acceptable to prevent speech, even if you really disagree.
They still have their equivalent of Intel's IME. Until they gut that ****, I'm not buying new hardware with my own money.
According to SemiAccurate, the AMD security stuff is way better than the Intel stuff.
Why is it better than Intel?
...SemiAccurate questioned AMD about the details surrounding SME, SVE, and the PSP. On the PSP front we were similarly impressed with the answers we got. First and foremost is the simple fact that the PSP firmware must be correctly signed to run. Having the hardware that controls the root of all your platform security running only signed code seems like an obvious, basic, bare minimum requirement for any security related technology. It also seems like something even a 3rd grader would flag as mandatory, but how can we say this politely, INTEL DOES NOT DO THIS. No joke.
Still, if you want hardware that absolutely doesn't have a tricky security system, you should be trying to help the EOMA68 project succeed. I backed their Kickstarter and one of these days I will receive a little mini desktop, running Linux on an ARM core. It's the one made from stacked wood.
Whoops, sorry for the mistake. I thought Threadripper was just a code name and Ryzen Pro was the release name. But it was a dumb mistake... the table shows "Ryzen 3 Pro" with only 4 cores and no Threadripper will have that few.
AMD took forever to get Ryzen out, but they really did do a good job with the chips. No games, no tricks, and 50% more instructions per clock. And to specifically answer your question: yes, each core has its own FPU.
The most interesting thing about the new Ryzen PRO chips: much more PCI-E lanes. From an article a month ago: "...AMD committed offering all 64 PCI-E lanes and 4 DDR4 memory channels on every ThreadRipper SKU regardless of price, clockspeed, or core count. These [Intel] Core X-series chips haven't even been publicly announced for a full 24 hours and already it's clear that AMD's offering the better chip."
It's ALWAYS the fault of the person who stole the stuff. 100% of the time.
But maybe not 100% of the fault. More than one person can be at fault.
In college I took an accounting class, and the teacher's favorite subject was "Internal Controls", systems and rules set up to make sure that people can't just steal money. He gave an example:
Suppose a small company has an accounting department with poor internal controls, and the head accountant knows that if he/she just edited one spreadsheet, he could steal a whole bunch of money and the company wouldn't realize. This person shows up for work every day for 20 years and never steals anything, and then one day suddenly snaps and steals the money. Who's to blame?
Clearly the person who stole the money is to blame for stealing the money, but my accounting teacher maintained that the company is also partially to blame for putting him in that position. It's a kind of stress, to have to resist temptation all the time, and it's unfair to put people in the position of resisting it.
Similarly, I put the blame in this case on the guys who stole the vending machine food, but the vending machine should not have been so easy to cheat.
P.S. Presumably they were paid well enough that they could afford to pay for vending machine food, so I'm not very sympathetic. And people who could entrap themselves by serially stealing petty things from a vending machine would seem to be high risks for being suborned by outside parties, so it's probably for the best if they aren't working in the CIA anymore.
I predict that Python, already popular, will become even more popular in the coming years. "Dominant"? Eh, you decide.
First of all, you guys moaning about white space? Just get over it, or go sulk in a corner or something. Python has always been the way it is, it's not going to change, and most people just don't care, or even like it. There are various "lint" tools for Python and any of them can complain about tabs (and most or all of them do by default). And I have been programming in Python for as long as Git has been around, and I have never had a "nightmare" merge caused by tabs and spaces. You're selling a problem I've never seen in the real world.
Those of you complaining about static typing? It's now available as an option. Python supports "type annotations" that have no meaning at runtime, but can be used by "lint" sort of tools to do static type checks. Large companies like Google and Microsoft were starting to require comments with type annotations and rolling their own static checking tools, so the Benevolent Dictator for Life ruled that Python should gain a built-in standardized way to do it.
"But both of the above suggest running a lint-like tool... it should be in the compiler." Well, sorry, but I am definitely willing to live with having the lint be optional. C used to be that way, you know, and your IDE or CI system can enforce use of the lint tools if you care.
"But the GIL! It's so awful!" Actually, spinning up multiple processes, or running code on a GPU, are both good solutions for many or most problems. Also, there is a serious project to remove the GIL, called "the GILectomy", being run by a Python core developer named Larry Hastings. Check YouTube for his lectures on it and how it's doing. He has made progress recently so there is hope. I'll say it again, multi-process is not bad at all in Python, and I have a bunch of production code that spins up about 30 processes at a time to do embarrassingly parallel tasks, so I personally don't care that much about the GIL, but you can't say nobody is working on it. It just turns out that the GIL is a really efficient way to solve a bunch of problems, and every attempt so far to remove it has made Python really slow.
With the major complaints mentioned, let me now mention some reasons for increased growth of the language.
The major growth area is in sciences. If you are studying astronomy, and probably most other sciences, Python is the single best language you can learn. SciPy is great: you can write pleasant Python code and it runs at FORTRAN speed, because the guts of SciPy are old (well-understood and well-debugged) FORTRAN libraries. I was able to go to the SciPy conference a few years ago and I learned that scientists who are abandoning wacky or proprietary languages are all converging on Python to gain the benefits of SciPy. For example, all the people working on the data from the Hubble space telescope all use Python now (instead of IDL, the original language when the Hubble first launched).
I would go so far as to say that for the sciences, Python will "dominate". It's achieved first-mover advantage, and it's still improving so it will be difficult to displace. (For example, a library called Pandas offers data frames, an idea copied from R. If Python successfully copies all the best parts of it's competition it will continue to dominate in science.)
Also, as a Python developer and as someone who has gone to the Python conference for years, I am seeing signs that the split between Python 2.x and Python 3.x is finally about to end, with most people finally moving to Python 3.x. The latest release of Python 3.x has much more excitement around it than any previous release. It now uses less memory, yet runs faster, than Python 2.x; and it has some nifty new features for async code, as well as a new language hook that PyCharm uses for debugging 50x faster than previous Python versions. My own company has started the switch to 3.x and we are far from alone.
I've posted this before, on Slashdot, but I can't find it now. I very much want to see a standard digital album format. Something simple and open, like a zip file with a standard folder structure and files easy to find within it.
I want high-resolution images of the album art, the original liner notes in both EPUB and PDF format, lyrics (ideally with optional time marks so that the lyrics can be displayed properly as the song plays) and room for extras. Like, I once saw a GIF of the cover of Wish You Were Here with an animation so the flames looked like they were burning... what the heck, let's have the album art optionally moving.
If I have a home entertainment system hooked up to a large screen in my living room, the album art can show while the music plays.
The old beautiful album art? I want to see it bigger than ever and in digital quality. I'd love to look at the details in the cover of The Lamb Lies Down on Broadway.
I figured Apple would have done this by now, and it would be patented and locked down and only work right on an AppleTV. But nobody has done it.
Heck, if this was done properly, I might be willing to buy again some albums I bought on CD and ripped, just because I can afford it and the extras would be nice.
And it would be interesting to see new albums designed from the start to be in this format, doing things I'm not clever enough to dream about yet.
USB C is, finally, USB done right. The connector is small, which is good for small devices; there is only one connector (no A and B variants); since there's only one connector you don't need a huge variety of cables (just USB A to USB C plus USB C to USB C and you are covered for 99.9% of scenarios); the plug doesn't have a "top" or "bottom", it just plugs in; and it was even designed to deliver useful amounts of power (enough to charge a small laptop).
My phone and my wife's phone are USB C and I just bought a Samsung Chromebook Plus, which charges by USB C (and it has two USB C ports, making it better than the Apple netbook). I'm planning to ask my employer to give me a laptop with USB C ports. I'm just waiting for a compact camera that uses USB C for charging and data and I'll buy that too. If I can manage it I won't buy another gadget with Micro USB or Mini USB ever again.
So congratulations, Microsoft! You managed to give me yet another reason to not buy your mobile devices!
P.S. I'm waiting for someone to make a kit that includes two or three USB C cables (USB C on both ends) plus a bunch of adapters: USB C to USB A, USB C to Mini USB, USB C to Micro USB, USB C to USB B, USB C to Ethernet jack, etc. Plus a USB to serial and USB to parallel and USB to IDE and SATA. It would be one kit that would let you connect almost anything to your laptop.
Sure public transportation is fine. You said Uber and Lyft.
Dude, I'm not the Emperor of the World. I'm making my predictions of the future. Whether you like it or not, whether I like it or not, I predict that Uber and Lyft and such services are going to become more common in the future (especially if there are self-driving cars). It's too bad you don't approve, I guess, but there it is.
(In principle a city or county could operate a service similar to Uber or Lyft. Perhaps that would make you happier?)
They have at least thought of these things?? Wow, drink the kool-aid some more.
If I'm understanding you correctly, you believe that Tesla is spending thousands of dollars per car to provide the hardware for self-driving, without having thought of common problems like driving in snow in the winter. They've been testing this stuff for years, but presumably you believe their testing was flawed and/or inadequate. I'd be interested to find out from you what Tesla did wrong, and why they are wrong to think that their hardware is adequate.
Would you say the Tesla hardware is completely useless, or would you say that there are some circumstances under which the Tesla hardware can do a useful job of driving the car? Also, in your opinion, are the various videos of self-driving Teslas all faked? And thevideos where a Tesla equipped with "Autopilot" takes action on its own to avoid an accident, are those faked?
It will be a sad world if people ever have to rely on Uber and Lyft to get from place to place.
I suspect that you are older than a Millenial and do not live in a dense urban area. (Personally, I'm older than a Millenial, I don't live in a dense area, and I very much treasure owning my own car.)
When I was a teen I was just counting the days until I got a permit to be able to drive a car; now Millenials are increasingly not bothering to get driver's license and insurance, and taking bus/Uber/Lyft when they want to go somewhere.
And there are people who live in dense urban areas who would find it a hassle to park a car, and prefer not to own a car there. More, there are cities that are actively trying to reduce the number of cars on their roads.
Tesla has not demonstrated that the sensors they are shipping will be able to handle all cases.
Okay, we get it, you're skeptical of the full self-driving features.
Will they be aimed low enough to stop to allow a rabbit to cross the road safely or are we just running over animals now? Will they scan the contour of the road so they can drive properly through ice ruts or around deep potholes? I didn't think they had that kind of tech yet.
Frankly I don't know the answers to these questions, but if Tesla thinks their current sensor tech is sufficient for full self-driving, my guess is they have at least thought of each of these things.
My guess, and it is just a guess, is that the ultrasonic sensors would be used to watch for ice ruts and potholes; that the testing program has already included people driving the test cars on roads with ice ruts and potholes; and the forward radar would likely do a better job of spotting a rabbit than a tired human at night. I don't think anyone is claiming that the self-driving features would completely eliminate all road kills, but equally I doubt self-driving cars will be worse than humans.
There are a lot of promises that the pro-automated crowd are making over and over. Two of the main ones are, 1) they will make driving safer, and 2) insurance will be cheap because they will be low-incident.
This will only happen if 95% of all people can afford these things.
If true automated cars truly work, then services like Uber and Lyft have the potential to become even cheaper. The cars get more expensive but no human drivers need be paid. It may be that if ride services are ubiquitous and cheap, people will use them more and not want to tie up their own money in a car.
Also, insurance for you may become cheaper if you use self-driving even if nobody else does. If you let the car drive you safely home you aren't getting into an accident due to being sleepy or drunk. From what I have read, insurance on Tesla cars is surprisingly inexpensive since the cars are so safe. (Although I think if you buy a "P" model with the "Ludicrous Speed" crazy acceleration, insurance rates go up.)
I consider it somewhat of a cop-out to just assume that any sensor on a fighter jet can apply when some of those sensors may be worth thousands of dollars.
And I consider it somewhat strange that you are arguing about whether Tesla can ship sensors that Tesla is already shipping as standard equipment on every car. The Tesla web site promises that the full self-driving sensors will be on every Model 3, that's the less-expensive model.
Tesla is making a bet that including all those sensors will pay off eventually. That bet may or may not be a good idea, but we are past the point of arguing over whether Tesla can do it. They are already doing it.
I just put some frozen (straw)berries in bowl, put it in the microwave to thaw, add some yogurt, and eat it.
That doesn't sound like it makes strawberry frozen yogurt. I like strawberry frozen yogurt, and the way I make it, it includes a whole lot of strawberries.
Having lots of fibre and vitamins in your diet is good. But a juicer is basically a machine for separating the fibre from the juice, and it also separates the skin of the fruit which often contains a lot of vitamins.
Juice by itself is sweet and tasty, but it basically gets all of its calories from carbohydrates, and without any fibre the carbs will hit your system quickly. The glycemic index of carrot juice is very high, while eating carrots will not usually have much effect on your blood sugar. (See the difference between glycemic index and glycemic load: https://en.wikipedia.org/wiki/Glycemic_load)
If you want to enjoy a tasty beverage that is better for you than juice, I recommend you get a VitaMix. The VitaMix company has been around for decades, and their product is an extremely expensive blender that is IMHO worth the expense. A VitaMix is so much more powerful than a cheap blender that it can do things the cheap blender cannot do.
So a typical fruit smoothie will start with some juice or even water and then throw fruit in, where the 2 horsepower motor at full speed breaks the fibre, skin, and even seeds down to the point where you don't really even know they are there. If you want to add a pleasing orange color to a smoothie, throw in some carrots; the texture will be a bit thicker and the color will be orange but you won't find any carrot chunks.
VitaMix has competitors, and some of the competitors may be as good. BlendTec and Ninja seem to have similar horsepower. I'm only recommending VitaMix since I have had one in my house for like two decades now. We use it so much that, despite the high purchase cost, it has a very low cost per operating minute... our food processor cost less but we hardly use the thing.
My favorite recipe: put a cup of plain yogurt into the blender, and add a spoonful of sugar and a squeeze of lime juice. Then dump a 10-ounce bag of frozen organic strawberries in (still frozen!), and run the blender on "high" while using the "tamper" to push the berries down into the blades. When the texture is smooth, serve. Don't overblend because you don't want to heat up the mixture. It's a tasty sweet dessert, and much healthier than any strawberry frozen yogurt you can buy.
Here's an example of throwing various hunks of vegetables into the thing to make a vegetable smoothie. After that demo the next demo is a sweet fruit smoothie. https://youtu.be/1qemLSu63d0?t=1m36s
P.S. LOL, YouTube appears to have a channel called "Blender Babes" where young females demonstrate blenders. I wonder what would happen if they tried to get a booth at a conference that bans "booth babes"... would they not be allowed at their own booth? https://www.youtube.com/watch?v=nXG65NgmrIM
How do you know what equipment is adequate for self-driving, when no self-driving cars have actually been made?
Are you serious?
Tesla has been testing experimental self-driving cars for years now. The Tesla "Autopilot" is a limited self-driving system that actual customers are using on actual roads, right now; and Tesla has been working on full self-driving features even though a fully autonomous car isn't legal yet.
Watch the demo video on this page: https://www.tesla.com/autopilot What's neat is that they also show feeds from three of the car's cameras, augmented with annotations from the computer vision system showing what the car is tracking in its environment. Note how the human keeps his hands off the wheel. Note how the car slows when unexpected pedestrians pop up at one point. Also note how, after the human leaves, the car parks itself.
Unless you are going to argue that the above video was faked, then clearly self-driving cars have actually been made.
There are limits on how long the drivers can be actively driving
True for human drivers. Put a sleeper cab on the thing, let the human sleep while the truck drives itself, and the truck could drive round the clock, just pausing for battery swaps.
That's assuming they want to keep a human on the truck to handle unexpected breakdowns or whatever. If they are willing to leave the truck completely unattended and just send someone out to check on it when it breaks down, then the truck can still drive around the clock and you don't even need the sleeper cab.
How do they get the batteries to the hubs? Manually driven vehicles?
Probably. But who cares?
Once the batteries are there, they will be used for swaps. Every time a battery is taken away, another battery is left behind. So these hubs will be charging hubs, hooked up to the electrical grid.
Now, let's assume that Mr. Anonymous Coward is correct and it really does take 6 hours to charge the battery. If truck A does a swap, and 6 hours later truck B needs a swap, the battery from truck A will be charged. A remote hub that never gets more than one truck every 6 hours could squeak by with a single on-site battery, but of course spares would always be a good idea. Anyway it won't be difficult to work out how many battery packs to stock at the various hubs.
Nothing says long haul trucking like a vehicle with a 200 mile range and a 6 hour recharge time.
I guarantee you that this thing is going to have a fast-swap battery pack.
The Model S already has a battery pack that can be swapped in about 90 seconds by a computer-controlled machine. It turned out that very few Model S owners wanted to pay for the fast battery swap service; the Supercharger service is adequate to most people's needs. (By the way, the Supercharger is much faster than your suggested 6 hours of charge time, for existing cars at least.)
So if range and charging time is an issue, companies will have the option of buying extra batteries and setting up battery-swap hubs at key locations on long haul routes. Or Tesla will do it like they tried for the Model S.
And hey what do you know, Tesla is investing heavily in a battery "gigafactory" and is going to bring the cost of batteries down as much as possible, as soon as possible.
So your joke was amusing but you have not actually identified a real problem. It's almost like Tesla knows what it's doing.
It has to have the capacity for a driverless upgrade out of the gate
Please note that Tesla is now building every new car (Model S, Model 3, and Model X) with full self-driving hardware. This includes 8 cameras, 12 ultrasonic sensors, a forward-facing radar, and computers adequate for self-driving (they claim 40x more processing power than the previous "Autopilot" computers). In the future, every Tesla car sold this year could be software-upgraded to full self-driving.
So, call me crazy, but I think Tesla might have thought of your point and is probably on top of it.
The aircraft will be a battery-first series hybrid, or an electric-powered aircraft with a range extender -- sort of like General Motors' Chevy Volt. All of the propulsion will come from the electric motor, said Kumar, and if there's enough battery power to run the entire flight, the jet fuel won't need to kick in. The company will also offer all-electric options.
Developers will never, ever build enough units to drop rental prices. That would be stupid.
This idea only makes sense if these "developers" you name can keep some kind of monopoly. If just anyone can build more units and enter the market, then your idea falls apart. Your idea is that by underbuilding, the developers can collect a premium... but the higher the premium, the more attractive the market becomes, and the more likely some new players are to try to build more units and enter the market.
If the developers have the ability to keep new units from being built, then sure. It's why I'm a big fan of the free market, and not at all a fan of crony capitalism.
There is no incentive whatsoever to build cheap apartments.
Oh, yeah? If there is an underserved market segment, someone could make money by serving it. Your statement makes exactly as much sense as saying there is no incentive whatsoever to build a fast-food hamburger shack instead of a fancy steakhouse. The steakhouse can charge more per meal, but would leave money on the table from the lower end of the market. Someone who wants all the money is incented to build both.
It is true that there are forces working against people who want to build affordable housing. One obvious way to make affordable apartments is to design them to be super space-efficient and just pack more of them into the same space; Seattle has made that practice illegal.
But that's not the free market failing. That's government putting its boot on the neck of the people who want to build small apartments.
It could very well be that the current property owners used crony capitalism to get the city of Seattle to block the tiny apartment projects. In a truly free market they would be unable to do that... someone would build apartments for the low end of the market as well as the high end.
Slashdot Mirror
The bottleneck for most problems isn't CPU cycles/second, it's the bandwidth of getting data to/from those CPUs. Adding CPUs does nothing to improve performance unless you also give it a much wider I/O path to memory.
Threadripper parts have quite a lot of bandwidth. The pro parts ("Epyc") will have even more.
Threadripper is intended for the PC enthusiast market, not so much for data centers. Frankly I don't think that for even an enthusiast home user memory bandwidth will be a major differentiating point. CPU speed, number of cores, and cache size will likely matter more. (I'm not worried about any of the above. I can't find L1 or L2 cache size numbers but I found that the L3 cache is 32 MB and since AMD was very focused on instructions-per-clock I am confident they didn't skimp on the cache.)
For data centers AMD will be selling the Epyc chips, and those can support up to 2 TB of RAM per socket (i.e. a dual-socket server would max out at 4 TB of RAM). In contrast, Intel tops out at 1.5 TB, and to get that you now have to buy their special and more-expensive chips with the "M" suffix; the non-M chips top out at 768 GB of RAM.
https://semiaccurate.com/2017/07/11/intel-launches-purley-aka-metal-xeons/
Threadripper also has a really large number of PCI-E lanes (64) so in theory you also could set up a really wide-bandwidth SSD or RAMdisk or something.
AMD has nothing for the Laptop Market in the Zen Class Architecture.
Coming in Q3. In other words, 2-4 months from now.
Laptops refresh twice a year, and the Ryzen launch wasn't in time for the last laptop refresh. No big deal; they're coming.
https://semiaccurate.com/2017/05/22/amd-talks-threadripper-ryzen-mobile-ryzen-pro/
While Zen Server parts (Epyc) look good on paper, it reamis to be seen if there will be Adoption from server makers, and demand from server purcharsers...
Well, sure. But unless the paper is a lie, those chips will do well. They will offer much-improved price/performance compared to Intel's server chips, they offer some tasty new security features (like VMs running with the in-RAM data encrypted so that there's no way for one VM to spy on another's memory), and they are doing it right when Intel is jacking their server customers on price.
corporate parts without IGP? Really? I mean, REALLY?!?!?
Does "IGP" mean integrated graphics? AMD is all over integrated graphics, they call such products "APUs" and the mobile lineup will be pretty much all APUs. So my guess is Q3 for corporate products with APUs as well. (I hope AMD supports ECC RAM on APUs, finally.)
In summary: I think violence, such as the events in Berkeley when Milo Yiannopolous was scheduled to speak, comes more often from the left than from the right. You disagree with this. I hope we can agree that violence is bad and we disapprove of it from either left or right, and leave it at that.
I think violence is not warranted to prevent words. I'm actually not sure whether you disagree with this, or not. But I hope you agree.
I see no point in further discussion; I believe it would be a waste of time for both of us. Have a nice life!
Hi, anonymous.
Are you trolling? I wrote that it's bad to paint all people who disagree with you with a broad brush (like "every conservative is a Nazi" or "everyone who wanted to hear Milo speak is a white supremacist") and you then posted a bunch of links about "white supremacists are bad".
You haven't shown that any of the people who wanted to see Milo speak were white nationalists, let alone all of them, and you absolutely haven't any excuse for equating an interest in watching Milo speak with the Oklahoma City bombing. Shame on you for that one.
Just in case this point is too subtle for you to understand, I'll spell it out for you: I don't approve of violence against the right, but I also don't approve of violence against the left. I said the remedy to bad speech is counter-speech, not violence, and I didn't make an exception to that for right-on-left violence.
So not walking away? Not ignoring them? You don't say they're unacceptable, but why not correct?
Sure, ignoring bad speech is fine. Making fun of the speech is also fine. What do you mean "why not correct", what do you think I meant by "counter-speech"? If someone says something dumb or bad, pointing out why it's dumb or bad is perfectly fine and a good idea. And it doesn't matter who was speaking, if one of my heroes says something dumb or bad, correct it; if someone I detest correctly points out something dumb or bad, then good for him/her.
is hauling people away violence or not?
It's violence and it's not justified for mere words. Hauling people away because they committed crimes and are going to jail, is still violence, but it's justified violence (at least if you agree that their "crimes" are actual crimes; for example, I wouldn't agree that "lese-majesty" is a real crime and people deserve to be hauled away for it). (Hauling people away is not as violent as beating them up and sending them to the hospital, but it's still violence. There is a saying, "the state holds a monopoly on violence." You might want to look that up and think about what it means.)
when is a public official obligated to listen?
Nobody is obligated to listen to any particular speech of anyone else. When President Trump talks, you don't have to listen. When you talk, President Trump doesn't have to listen.
Of course, if public officials ignore the people, they may lose their next election.
The smashing, violence, and beatings is overwhelmingly coming from the left. That's how it's been for many years.
I have to agree. My personal theory is that many of the violent people have convinced themselves that their political enemies are in fact bad people and "fair game" for anything. It goes like this: It's okay to punch a Nazi; conservatives are all Nazis... and then comes the punching.
Here is a web page linking multiple articles arguing that the violence used to prevent Milo Yiannopolous from speaking at Berkeley was justified. "Violence helped ensure safety of students" is a real headline. There was also this quote: "...some white nationalists got their ***** beat." (Just like the Nazi thing above, only this time using "white nationalist".
Someone who wanted to hear Milo speak --> white nationalist --> someone it's okay to send to the hospital.)
http://www.dailycal.org/2017/02/07/violence-self-defense/
Also, the media coverage may tend to embolden these people. The people who smash things, light things on fire, and send people to the hospital are described as "protesters". The people who wanted to hear Milo speak are described as "alt-Right extremists". I don't want to overstate the contribution of the media but I think it's a part.
Personally I think that the correct remedy for bad speech is counter-speech. Violence isn't acceptable to prevent speech, even if you really disagree.
They still have their equivalent of Intel's IME. Until they gut that ****, I'm not buying new hardware with my own money.
According to SemiAccurate, the AMD security stuff is way better than the Intel stuff.
Why is it better than Intel?
http://semiaccurate.com/2017/06/22/amds-epyc-major-advance-security/
Still, if you want hardware that absolutely doesn't have a tricky security system, you should be trying to help the EOMA68 project succeed. I backed their Kickstarter and one of these days I will receive a little mini desktop, running Linux on an ARM core. It's the one made from stacked wood.
https://www.crowdsupply.com/eoma68/micro-desktop
ryzen pro is different than threadripper.
Whoops, sorry for the mistake. I thought Threadripper was just a code name and Ryzen Pro was the release name. But it was a dumb mistake... the table shows "Ryzen 3 Pro" with only 4 cores and no Threadripper will have that few.
Mea culpa.
AMD took forever to get Ryzen out, but they really did do a good job with the chips. No games, no tricks, and 50% more instructions per clock. And to specifically answer your question: yes, each core has its own FPU.
http://www.pcworld.com/article/3176907/components-processors/ryzen-cpus-explained-everything-you-need-to-know-about-amds-disruptive-multicore-chips.html
The most interesting thing about the new Ryzen PRO chips: much more PCI-E lanes. From an article a month ago: "...AMD committed offering all 64 PCI-E lanes and 4 DDR4 memory channels on every ThreadRipper SKU regardless of price, clockspeed, or core count. These [Intel] Core X-series chips haven't even been publicly announced for a full 24 hours and already it's clear that AMD's offering the better chip."
http://semiaccurate.com/2017/05/31/amds-ryzen-threadripper-brings-socket-tr4-x399-chipset/
It's ALWAYS the fault of the person who stole the stuff. 100% of the time.
But maybe not 100% of the fault. More than one person can be at fault.
In college I took an accounting class, and the teacher's favorite subject was "Internal Controls", systems and rules set up to make sure that people can't just steal money. He gave an example:
Suppose a small company has an accounting department with poor internal controls, and the head accountant knows that if he/she just edited one spreadsheet, he could steal a whole bunch of money and the company wouldn't realize. This person shows up for work every day for 20 years and never steals anything, and then one day suddenly snaps and steals the money. Who's to blame?
Clearly the person who stole the money is to blame for stealing the money, but my accounting teacher maintained that the company is also partially to blame for putting him in that position. It's a kind of stress, to have to resist temptation all the time, and it's unfair to put people in the position of resisting it.
Similarly, I put the blame in this case on the guys who stole the vending machine food, but the vending machine should not have been so easy to cheat.
P.S. Presumably they were paid well enough that they could afford to pay for vending machine food, so I'm not very sympathetic. And people who could entrap themselves by serially stealing petty things from a vending machine would seem to be high risks for being suborned by outside parties, so it's probably for the best if they aren't working in the CIA anymore.
I predict that Python, already popular, will become even more popular in the coming years. "Dominant"? Eh, you decide.
First of all, you guys moaning about white space? Just get over it, or go sulk in a corner or something. Python has always been the way it is, it's not going to change, and most people just don't care, or even like it. There are various "lint" tools for Python and any of them can complain about tabs (and most or all of them do by default). And I have been programming in Python for as long as Git has been around, and I have never had a "nightmare" merge caused by tabs and spaces. You're selling a problem I've never seen in the real world.
Those of you complaining about static typing? It's now available as an option. Python supports "type annotations" that have no meaning at runtime, but can be used by "lint" sort of tools to do static type checks. Large companies like Google and Microsoft were starting to require comments with type annotations and rolling their own static checking tools, so the Benevolent Dictator for Life ruled that Python should gain a built-in standardized way to do it.
"But both of the above suggest running a lint-like tool... it should be in the compiler." Well, sorry, but I am definitely willing to live with having the lint be optional. C used to be that way, you know, and your IDE or CI system can enforce use of the lint tools if you care.
"But the GIL! It's so awful!" Actually, spinning up multiple processes, or running code on a GPU, are both good solutions for many or most problems. Also, there is a serious project to remove the GIL, called "the GILectomy", being run by a Python core developer named Larry Hastings. Check YouTube for his lectures on it and how it's doing. He has made progress recently so there is hope. I'll say it again, multi-process is not bad at all in Python, and I have a bunch of production code that spins up about 30 processes at a time to do embarrassingly parallel tasks, so I personally don't care that much about the GIL, but you can't say nobody is working on it. It just turns out that the GIL is a really efficient way to solve a bunch of problems, and every attempt so far to remove it has made Python really slow.
With the major complaints mentioned, let me now mention some reasons for increased growth of the language.
The major growth area is in sciences. If you are studying astronomy, and probably most other sciences, Python is the single best language you can learn. SciPy is great: you can write pleasant Python code and it runs at FORTRAN speed, because the guts of SciPy are old (well-understood and well-debugged) FORTRAN libraries. I was able to go to the SciPy conference a few years ago and I learned that scientists who are abandoning wacky or proprietary languages are all converging on Python to gain the benefits of SciPy. For example, all the people working on the data from the Hubble space telescope all use Python now (instead of IDL, the original language when the Hubble first launched).
I would go so far as to say that for the sciences, Python will "dominate". It's achieved first-mover advantage, and it's still improving so it will be difficult to displace. (For example, a library called Pandas offers data frames, an idea copied from R. If Python successfully copies all the best parts of it's competition it will continue to dominate in science.)
Also, as a Python developer and as someone who has gone to the Python conference for years, I am seeing signs that the split between Python 2.x and Python 3.x is finally about to end, with most people finally moving to Python 3.x. The latest release of Python 3.x has much more excitement around it than any previous release. It now uses less memory, yet runs faster, than Python 2.x; and it has some nifty new features for async code, as well as a new language hook that PyCharm uses for debugging 50x faster than previous Python versions. My own company has started the switch to 3.x and we are far from alone.
So Python is a
I've posted this before, on Slashdot, but I can't find it now. I very much want to see a standard digital album format. Something simple and open, like a zip file with a standard folder structure and files easy to find within it.
I want high-resolution images of the album art, the original liner notes in both EPUB and PDF format, lyrics (ideally with optional time marks so that the lyrics can be displayed properly as the song plays) and room for extras. Like, I once saw a GIF of the cover of Wish You Were Here with an animation so the flames looked like they were burning... what the heck, let's have the album art optionally moving.
If I have a home entertainment system hooked up to a large screen in my living room, the album art can show while the music plays.
The old beautiful album art? I want to see it bigger than ever and in digital quality. I'd love to look at the details in the cover of The Lamb Lies Down on Broadway.
I figured Apple would have done this by now, and it would be patented and locked down and only work right on an AppleTV. But nobody has done it.
Heck, if this was done properly, I might be willing to buy again some albums I bought on CD and ripped, just because I can afford it and the extras would be nice.
And it would be interesting to see new albums designed from the start to be in this format, doing things I'm not clever enough to dream about yet.
USB is serial, so a $200-300 converter box is required
I don't know where you got that $200-$300 price estimate, but you are way high. Here's one for $10:
Sabrent USB to parallel adapter, $9.99 quantity 1
I almost added "...and a USB to SCSI adapter" but SCSI isn't common at all anymore, and the few such adapters I have found really did cost hundreds.
USB C is, finally, USB done right. The connector is small, which is good for small devices; there is only one connector (no A and B variants); since there's only one connector you don't need a huge variety of cables (just USB A to USB C plus USB C to USB C and you are covered for 99.9% of scenarios); the plug doesn't have a "top" or "bottom", it just plugs in; and it was even designed to deliver useful amounts of power (enough to charge a small laptop).
My phone and my wife's phone are USB C and I just bought a Samsung Chromebook Plus, which charges by USB C (and it has two USB C ports, making it better than the Apple netbook). I'm planning to ask my employer to give me a laptop with USB C ports. I'm just waiting for a compact camera that uses USB C for charging and data and I'll buy that too. If I can manage it I won't buy another gadget with Micro USB or Mini USB ever again.
So congratulations, Microsoft! You managed to give me yet another reason to not buy your mobile devices!
P.S. I'm waiting for someone to make a kit that includes two or three USB C cables (USB C on both ends) plus a bunch of adapters: USB C to USB A, USB C to Mini USB, USB C to Micro USB, USB C to USB B, USB C to Ethernet jack, etc. Plus a USB to serial and USB to parallel and USB to IDE and SATA. It would be one kit that would let you connect almost anything to your laptop.
Sure public transportation is fine. You said Uber and Lyft.
Dude, I'm not the Emperor of the World. I'm making my predictions of the future. Whether you like it or not, whether I like it or not, I predict that Uber and Lyft and such services are going to become more common in the future (especially if there are self-driving cars). It's too bad you don't approve, I guess, but there it is.
(In principle a city or county could operate a service similar to Uber or Lyft. Perhaps that would make you happier?)
They have at least thought of these things?? Wow, drink the kool-aid some more.
If I'm understanding you correctly, you believe that Tesla is spending thousands of dollars per car to provide the hardware for self-driving, without having thought of common problems like driving in snow in the winter. They've been testing this stuff for years, but presumably you believe their testing was flawed and/or inadequate. I'd be interested to find out from you what Tesla did wrong, and why they are wrong to think that their hardware is adequate.
Would you say the Tesla hardware is completely useless, or would you say that there are some circumstances under which the Tesla hardware can do a useful job of driving the car? Also, in your opinion, are the various videos of self-driving Teslas all faked? And the videos where a Tesla equipped with "Autopilot" takes action on its own to avoid an accident, are those faked?
It will be a sad world if people ever have to rely on Uber and Lyft to get from place to place.
I suspect that you are older than a Millenial and do not live in a dense urban area. (Personally, I'm older than a Millenial, I don't live in a dense area, and I very much treasure owning my own car.)
When I was a teen I was just counting the days until I got a permit to be able to drive a car; now Millenials are increasingly not bothering to get driver's license and insurance, and taking bus/Uber/Lyft when they want to go somewhere.
And there are people who live in dense urban areas who would find it a hassle to park a car, and prefer not to own a car there. More, there are cities that are actively trying to reduce the number of cars on their roads.
Tesla has not demonstrated that the sensors they are shipping will be able to handle all cases.
Okay, we get it, you're skeptical of the full self-driving features.
Will they be aimed low enough to stop to allow a rabbit to cross the road safely or are we just running over animals now? Will they scan the contour of the road so they can drive properly through ice ruts or around deep potholes? I didn't think they had that kind of tech yet.
Frankly I don't know the answers to these questions, but if Tesla thinks their current sensor tech is sufficient for full self-driving, my guess is they have at least thought of each of these things.
My guess, and it is just a guess, is that the ultrasonic sensors would be used to watch for ice ruts and potholes; that the testing program has already included people driving the test cars on roads with ice ruts and potholes; and the forward radar would likely do a better job of spotting a rabbit than a tired human at night. I don't think anyone is claiming that the self-driving features would completely eliminate all road kills, but equally I doubt self-driving cars will be worse than humans.
https://www.tesla.com/blog/upgrading-autopilot-seeing-world-radar
https://www.tesla.com/blog/all-tesla-cars-being-produced-now-have-full-self-driving-hardware
There are a lot of promises that the pro-automated crowd are making over and over. Two of the main ones are, 1) they will make driving safer, and 2) insurance will be cheap because they will be low-incident.
This will only happen if 95% of all people can afford these things.
If true automated cars truly work, then services like Uber and Lyft have the potential to become even cheaper. The cars get more expensive but no human drivers need be paid. It may be that if ride services are ubiquitous and cheap, people will use them more and not want to tie up their own money in a car.
Also, insurance for you may become cheaper if you use self-driving even if nobody else does. If you let the car drive you safely home you aren't getting into an accident due to being sleepy or drunk. From what I have read, insurance on Tesla cars is surprisingly inexpensive since the cars are so safe. (Although I think if you buy a "P" model with the "Ludicrous Speed" crazy acceleration, insurance rates go up.)
I consider it somewhat of a cop-out to just assume that any sensor on a fighter jet can apply when some of those sensors may be worth thousands of dollars.
And I consider it somewhat strange that you are arguing about whether Tesla can ship sensors that Tesla is already shipping as standard equipment on every car. The Tesla web site promises that the full self-driving sensors will be on every Model 3, that's the less-expensive model.
Tesla is making a bet that including all those sensors will pay off eventually. That bet may or may not be a good idea, but we are past the point of arguing over whether Tesla can do it. They are already doing it.
Sounds like a lot of work.
It really isn't.
I just put some frozen (straw)berries in bowl, put it in the microwave to thaw, add some yogurt, and eat it.
That doesn't sound like it makes strawberry frozen yogurt. I like strawberry frozen yogurt, and the way I make it, it includes a whole lot of strawberries.
Having lots of fibre and vitamins in your diet is good. But a juicer is basically a machine for separating the fibre from the juice, and it also separates the skin of the fruit which often contains a lot of vitamins.
Juice by itself is sweet and tasty, but it basically gets all of its calories from carbohydrates, and without any fibre the carbs will hit your system quickly. The glycemic index of carrot juice is very high, while eating carrots will not usually have much effect on your blood sugar. (See the difference between glycemic index and glycemic load: https://en.wikipedia.org/wiki/Glycemic_load)
If you want to enjoy a tasty beverage that is better for you than juice, I recommend you get a VitaMix. The VitaMix company has been around for decades, and their product is an extremely expensive blender that is IMHO worth the expense. A VitaMix is so much more powerful than a cheap blender that it can do things the cheap blender cannot do.
So a typical fruit smoothie will start with some juice or even water and then throw fruit in, where the 2 horsepower motor at full speed breaks the fibre, skin, and even seeds down to the point where you don't really even know they are there. If you want to add a pleasing orange color to a smoothie, throw in some carrots; the texture will be a bit thicker and the color will be orange but you won't find any carrot chunks.
VitaMix has competitors, and some of the competitors may be as good. BlendTec and Ninja seem to have similar horsepower. I'm only recommending VitaMix since I have had one in my house for like two decades now. We use it so much that, despite the high purchase cost, it has a very low cost per operating minute... our food processor cost less but we hardly use the thing.
My favorite recipe: put a cup of plain yogurt into the blender, and add a spoonful of sugar and a squeeze of lime juice. Then dump a 10-ounce bag of frozen organic strawberries in (still frozen!), and run the blender on "high" while using the "tamper" to push the berries down into the blades. When the texture is smooth, serve. Don't overblend because you don't want to heat up the mixture. It's a tasty sweet dessert, and much healthier than any strawberry frozen yogurt you can buy.
Here's an example of throwing various hunks of vegetables into the thing to make a vegetable smoothie. After that demo the next demo is a sweet fruit smoothie. https://youtu.be/1qemLSu63d0?t=1m36s
P.S. LOL, YouTube appears to have a channel called "Blender Babes" where young females demonstrate blenders. I wonder what would happen if they tried to get a booth at a conference that bans "booth babes"... would they not be allowed at their own booth? https://www.youtube.com/watch?v=nXG65NgmrIM
How do you know what equipment is adequate for self-driving, when no self-driving cars have actually been made?
Are you serious?
Tesla has been testing experimental self-driving cars for years now. The Tesla "Autopilot" is a limited self-driving system that actual customers are using on actual roads, right now; and Tesla has been working on full self-driving features even though a fully autonomous car isn't legal yet.
Watch the demo video on this page: https://www.tesla.com/autopilot What's neat is that they also show feeds from three of the car's cameras, augmented with annotations from the computer vision system showing what the car is tracking in its environment. Note how the human keeps his hands off the wheel. Note how the car slows when unexpected pedestrians pop up at one point. Also note how, after the human leaves, the car parks itself.
Unless you are going to argue that the above video was faked, then clearly self-driving cars have actually been made.
There are limits on how long the drivers can be actively driving
True for human drivers. Put a sleeper cab on the thing, let the human sleep while the truck drives itself, and the truck could drive round the clock, just pausing for battery swaps.
That's assuming they want to keep a human on the truck to handle unexpected breakdowns or whatever. If they are willing to leave the truck completely unattended and just send someone out to check on it when it breaks down, then the truck can still drive around the clock and you don't even need the sleeper cab.
How do they get the batteries to the hubs? Manually driven vehicles?
Probably. But who cares?
Once the batteries are there, they will be used for swaps. Every time a battery is taken away, another battery is left behind. So these hubs will be charging hubs, hooked up to the electrical grid.
Now, let's assume that Mr. Anonymous Coward is correct and it really does take 6 hours to charge the battery. If truck A does a swap, and 6 hours later truck B needs a swap, the battery from truck A will be charged. A remote hub that never gets more than one truck every 6 hours could squeak by with a single on-site battery, but of course spares would always be a good idea. Anyway it won't be difficult to work out how many battery packs to stock at the various hubs.
Nothing says long haul trucking like a vehicle with a 200 mile range and a 6 hour recharge time.
I guarantee you that this thing is going to have a fast-swap battery pack.
The Model S already has a battery pack that can be swapped in about 90 seconds by a computer-controlled machine. It turned out that very few Model S owners wanted to pay for the fast battery swap service; the Supercharger service is adequate to most people's needs. (By the way, the Supercharger is much faster than your suggested 6 hours of charge time, for existing cars at least.)
So if range and charging time is an issue, companies will have the option of buying extra batteries and setting up battery-swap hubs at key locations on long haul routes. Or Tesla will do it like they tried for the Model S.
And hey what do you know, Tesla is investing heavily in a battery "gigafactory" and is going to bring the cost of batteries down as much as possible, as soon as possible.
So your joke was amusing but you have not actually identified a real problem. It's almost like Tesla knows what it's doing.
It has to have the capacity for a driverless upgrade out of the gate
Please note that Tesla is now building every new car (Model S, Model 3, and Model X) with full self-driving hardware. This includes 8 cameras, 12 ultrasonic sensors, a forward-facing radar, and computers adequate for self-driving (they claim 40x more processing power than the previous "Autopilot" computers). In the future, every Tesla car sold this year could be software-upgraded to full self-driving.
So, call me crazy, but I think Tesla might have thought of your point and is probably on top of it.
I found an article with a better description of the proposed technology.
https://www.greentechmedia.com/articles/read/this-startup-is-building-an-electric-airplane
Developers will never, ever build enough units to drop rental prices. That would be stupid.
This idea only makes sense if these "developers" you name can keep some kind of monopoly. If just anyone can build more units and enter the market, then your idea falls apart. Your idea is that by underbuilding, the developers can collect a premium... but the higher the premium, the more attractive the market becomes, and the more likely some new players are to try to build more units and enter the market.
If the developers have the ability to keep new units from being built, then sure. It's why I'm a big fan of the free market, and not at all a fan of crony capitalism.
There is no incentive whatsoever to build cheap apartments.
Oh, yeah? If there is an underserved market segment, someone could make money by serving it. Your statement makes exactly as much sense as saying there is no incentive whatsoever to build a fast-food hamburger shack instead of a fancy steakhouse. The steakhouse can charge more per meal, but would leave money on the table from the lower end of the market. Someone who wants all the money is incented to build both.
It is true that there are forces working against people who want to build affordable housing. One obvious way to make affordable apartments is to design them to be super space-efficient and just pack more of them into the same space; Seattle has made that practice illegal.
http://www.sightline.org/2016/09/06/how-seattle-killed-micro-housing/
But that's not the free market failing. That's government putting its boot on the neck of the people who want to build small apartments.
It could very well be that the current property owners used crony capitalism to get the city of Seattle to block the tiny apartment projects. In a truly free market they would be unable to do that... someone would build apartments for the low end of the market as well as the high end.