AMD currently has nothing to offer for the ultralow power niche (equivalent of Intel Y series processors). It is also behind the performance curve in the mainstream laptop niche (Intel U series); Mobile Ryzen isn't terrible but it's not yet a serious challenger to a mobile i7. Finally, they don't have anything to go up against the high power H-series; the best they could do is a down-clocked variant of a desktop Ryzen but then you would also need a separate GPU. I don't expect AMD to launch a serious challenge to those markets until Zen 2 and its accompanying 7nm process tech are ready next year; they just aren't yet where they need to be on power consumption.
AMD has solid offerings for the desktop and server markets. It's nice to see real competition in the market for PC processors again. But they've still got work to do. The good news is that they have a road map in place for doing it, and the company has been executing its plans well in recent years.
That's version 1. Version 1 of any Microsoft product has problems. If it follows past history, Windows on ARM won't really hit its stride until version 3, except that it won't be called that because we now have Windows 10 everywhere and forever.
It's a little more complicated than the feature size. Intel's 10nm process is about as dense as the 7nm processes at Samsung or TSMC. But that doesn't change the fact that Intel isn't shipping its 10nm in quantity yet, while those others are shipping 7nm. GlobalFoundries will probably also be shipping 7nm before Intel has its 10nm going.
Intel has fallen behind in process technology. That's an unaccustomed position for them; their ability to manufacture (as opposed to design) chips better than anybody else has been a key part of their success for many years. I'm not writing off Intel, but they've got some work to do.
Can you say camping trip? I knew you could. Lots more stuff you need to carry. You can get the carry weight down by buying the right gear, but that involves serious expense. Just as with bicycles and laptop computers, less weight equals more money.
Or how about a trip where you are going to DO something and need the equipment for the activity? Or a visit to family when you are carrying presents? A trip to a formal event where you need fancy clothes for the occasion? Or just a trip to a place where the weather is highly variable, so you need cold weather and rain gear that you may or may not use? (On that camping trip you probably WILL need them; if you're up in the mountains it's not unheard of to have 90 degree temperatures at midday and 40 at night.
And we haven't even addressed the question of special needs. If you need to transport any adaptive or medical equipment, the weight can add up quickly.
My personal goal is to use at least 80% of what I pack. I'll never reach 100% because I pack a spare change of clothes for anything longer than an overnight, and because I may have things for weather contingencies. And I always overpack socks and underwear because things can happen.
The tire pressure warning is indeed required now in the US. So is the one for seat belts. Not only are the speedometer and odometer required, but there are accuracy regulations for them. I don't believe there is a mandate for any of the other gauges, but the market isn't likely to accept a car without a fuel gauge, and the temperature gauge is a good idea for any car with an engine that uses an aluminum block. (Engine overheating is a more severe problem for aluminum engines than for the old steel block ones, because a single incidence of severe overheating can deform the engine block and instantly turn the engine into a heavy chunk of scrap metal.) Backup cameras are the most recently added instrumentation requirement.
In a modern car, especially an EV which by necessity is already going to have a lot of electronics, it's probably best to integrate all the information into a single display. That's what the Tesla Model 3 has done, though I'm not convinced they made the right choice of location and size.
A more serious problem with the user interface of the Model 3 is the elimination of most of the switches and knobs and replacing them with a touchscreen interface. The touchscreen is difficult for the driver to operate because it requires taking eyes off the road. Tesla's plan appears to be to use voice control instead, but most of that isn't ready yet.
Rev counters, aka tachometers, are useful in a gas-powered car, especially one with a small engine, so you know when to shift. They are useless in an electric car. EVs only have one gear; they don't need more because electric motors don't have a torque curve, they have a torque straight line.
It's smaller than the Fortwo. It has a smaller motor and its speed is more limited. Unlike the Fortwo it's not likely to meet current safety standards for a car in the US. Finally, it's cheaper. The price translates to about $15,000 without any tax credits; if it's ever allowed to be sold here the price after incentives will be quite low.
A problem with partial voucher systems is that not all students cost the same amount to educate. The vouchers are usually worth the city-wide average cost of educating a student, but special needs students are far more expensive and mainstream students are cheaper. The voucher schools weasel out of taking the special needs students, one way or another, and are effectively overpaid for the students they do take. Meanwhile, the remaining public system effectively has even less money per student to work with because they're stuck with all the expensive students.
A full voucher system eliminates that issue, sort of. But how do you get those high cost students educated? I suppose you could have a system where the reimbursement for high cost students is larger, but I don't know of any school voucher system that has done that.
A related problem: what do you do with students that nobody wants, perhaps because of disciplinary issues? They're required to go to school by law, but how do they avoid breaking the law if nobody will take them?
The homeless are more visible in San Francisco because the friendly climate makes it easier to live outdoors. They don't have to deal with extreme heat or cold and the amount of rain is modest. The actual problem is not worse than in other major cities.
Rain increases the path loss so you receive less signal. You could get back signal strength by using an antenna with more gain. That higher gain antenna will be larger and will require more careful aiming by the installer, and would thus be more costly.
How serious the problem is depends on where you live. TV satellites are in geostationary orbit above the equator; the farther away from the equator you live, the higher the path loss. The satellites serving the US are mostly above the Mountain time zone; Dish Network also has a couple that are over the Atlantic. If you're someplace else, again the path to the satellite is longer and the signal is weaker. Some people may have some amount of vegetation obstructing the path to the satellite; the amount of signal loss from that increases MUCH more quickly with rain. (Wet foliage is worse than wet air.)
It comes down to the question of what level of reliability customers are willing to pay for. If the reliability of the usual free installation is inadequate, you can do better by buying a larger replacement antenna. And you can clear away any trees that are near your dish.
They're giving in to the upsell and buying now because of the situation with the tax credit. By the time Tesla actually ships the $35K model, Teslas will no longer be eligible because the company will have sold too many electric cars. So you're really comparing buying a $36.5K car now vs a $35K car later, and given that choice buying the LR version is a no brainer.
I have to disagree on #12. It's important to have at least one browser that is NOT based on the dominant code base. Otherwise the dominant browser effectively becomes the standard. There is also Edge but it's Windows-only. Safari/WebKit has diverged substantially from Chrome by now but they have common roots. Firefox and its forks are the only completely independent cross-platform alternatives.
Just about all the major releases show up at all the chains. But the more limited releases don't. Buying into a one-chain pass means there are movies you won't get to see without spending extra. And in some cases you might not get to see the major film you want in your desired format. For example, back when the Hobbit movies were coming out, the first one was only in HFR 3D at AMC and the other two were only in HFR 3D at Regal. (Yes, I saw all three that way. I know a lot of people hated HFR but I liked it. Also, HFR and 3D are synergistic; HFR makes 3D work better.)
This is a major shift in the power balance of the CPU business.
For many years, one of the key things that kept Intel on top is that they were the best in the business at making chips. Not at designing an architecture, and not always at executing that architecture (like the era of Pentium 4 vs Athlon 64), but their chip fabrication technology was on top.
But now they find themselves in the unaccustomed position of playing catch-up. Samsung and TSMC are already shipping 7nm, with GlobalFoundries close behind, while Intel won't hit volume production of its 10nm process for another year. (Intel's 10nm is about the same density as the 7nm from those other companies despite the larger feature size so a direct comparison is fair.) This is the biggest business challenge that Intel has faced in a long time; I think they will find a way through it, but they may never return to the same level of market dominance.
Here in Boston we have two of the big mainstream theater chains (AMC and Regal) plus a new small one (Showplace Icon). Across the river in Cambridge there is an art house chain (Landmark), an independent art house (Brattle), and two small local chains (Frame One Theatres and Apple Cinemas). Neighboring Brookline has another independent art house (Coolidge Corner). The suburbs of Boston have yet another medium-sized chain (National Entertainment, dba Showcase and Cinema De Lux), two IMAX theaters run by the Jordan's Furniture chain (!), and additional AMC locations and independent art houses.
Lots of choices here. That substantially decreases the appeal of single chain offerings like AMC A-List.
There are a few theater chains that MoviePass is working with directly, mostly the ones with e-ticketing options. Landmark Theatres is one. But they don't have any special deals with the big players like AMC, Regal, and Cinemark.
It wasn't just to prevent people selling the tickets. I heard reports that some people were using MoviePass to buy tickets for movies they had no intention of watching just to get the loyalty points from the theater. If it was an urban theater near your place of work that was rather painless to do.
The i9 in the MacBook Pro is the 8950HK, a six core mobile CPU, not an eight core desktop part.
The fastest Mobile Ryzen currently available is the 2700U, a four core CPU. Between the lower core count and fewer instructions per MHz than Intel, it is not competitive in performance with the 8950HK. A slightly higher performance 2800H has been leaked but not officially announced; it's still a four core CPU but with higher clocks and a GPU with more stream units enabled. Those tweaks still won't make it competitive with Intel's H series parts, so it's hard to imagine much market interest and it may never get released.
The only current alternative would be to design around a downclocked desktop part like the 2700 and also include a separate GPU. (That model of MBP already has one, but using a CPU rather than an APU would mean that the GPU would have to be on full time.) That would give you eight cores and a good GPU, but the requirement to use the GPU full time would result in far worse battery life, not a tradeoff that Apple would care to make.
AMD may have something that's competitive at the high end of laptops next year when the Zen 2 chips are released. For now, AMD isn't an option for performance laptops. The 2500U and 2700U are decent mainstream laptop processors, offering more graphics performance but less processor performance than their counterparts from Intel; systems featuring those chips have started to appear but sales seem slow so far.
In the first implementation of C++ it WAS compiled by a compile that could only compile C. That first implementation was a preprocessor that turned your C++ program into somewhat hard to read C code, which was then run through the C compiler.
Nowadays you are far more likely to do the opposite: compile your C code with a compiler that is written in C++. That includes current versions of GCC and LLVM/Clang.
I wish I hadn't already commented on this subject so I could mod this comment up.
Michael Abrash is somebody who LITERALLY wrote the book on writing high performance code. More than one, actually. I also remember his posts on the subject on the late lamented BIX. He is the antithesis of an incompetent programmer.
Say programmers who know how to balance development time against run time.
Some programs are run ONCE and discarded. Or more accurately run a bunch of times during development and testing, and then there is one final run on the real data. That's typical for a program that is converting data from some old format to a newer one. Obviously, spending any time optimizing such a program would be utterly pointless, as the extra time spent on development would far outweigh any savings you would achieve during the program's single use. (Not to mention that 90% of the run time of such a program is probably spent on input and output, so improving the algorithms won't buy you much.)
At the other end of the scale, there is code that does crucial inner loop calculations. That code gets run a lot, and shaving even a little bit of time off its run time adds up to a lot. Code like that is a prime target for optimization.
All the other use cases lie somewhere in between. And each calls for a different amount of effort put into improvement. The mark of a good programmer and designer is knowing where to spend the effort and where not to.
No, Apple gave us only PART of what we wanted. A proper thermal solution is the missing part of the equation. But to do that, Apple would have had to actually put some engineering time and effort into the update of the MacBook Pro, rather than just shoving a newer processor and DDR4 RAM into the box. And they might have had to compromise the thinness and lightness of the system or make the fans louder.
Slashdot Mirror
AMD currently has nothing to offer for the ultralow power niche (equivalent of Intel Y series processors). It is also behind the performance curve in the mainstream laptop niche (Intel U series); Mobile Ryzen isn't terrible but it's not yet a serious challenger to a mobile i7. Finally, they don't have anything to go up against the high power H-series; the best they could do is a down-clocked variant of a desktop Ryzen but then you would also need a separate GPU. I don't expect AMD to launch a serious challenge to those markets until Zen 2 and its accompanying 7nm process tech are ready next year; they just aren't yet where they need to be on power consumption.
AMD has solid offerings for the desktop and server markets. It's nice to see real competition in the market for PC processors again. But they've still got work to do. The good news is that they have a road map in place for doing it, and the company has been executing its plans well in recent years.
That's version 1. Version 1 of any Microsoft product has problems. If it follows past history, Windows on ARM won't really hit its stride until version 3, except that it won't be called that because we now have Windows 10 everywhere and forever.
Besides Apple, it's also inside FreeNAS and pfSense. It's not as big a success as Linux but that is not equal to failure.
It's a little more complicated than the feature size. Intel's 10nm process is about as dense as the 7nm processes at Samsung or TSMC. But that doesn't change the fact that Intel isn't shipping its 10nm in quantity yet, while those others are shipping 7nm. GlobalFoundries will probably also be shipping 7nm before Intel has its 10nm going.
Intel has fallen behind in process technology. That's an unaccustomed position for them; their ability to manufacture (as opposed to design) chips better than anybody else has been a key part of their success for many years. I'm not writing off Intel, but they've got some work to do.
Can you say camping trip? I knew you could. Lots more stuff you need to carry. You can get the carry weight down by buying the right gear, but that involves serious expense. Just as with bicycles and laptop computers, less weight equals more money.
Or how about a trip where you are going to DO something and need the equipment for the activity? Or a visit to family when you are carrying presents? A trip to a formal event where you need fancy clothes for the occasion? Or just a trip to a place where the weather is highly variable, so you need cold weather and rain gear that you may or may not use? (On that camping trip you probably WILL need them; if you're up in the mountains it's not unheard of to have 90 degree temperatures at midday and 40 at night.
And we haven't even addressed the question of special needs. If you need to transport any adaptive or medical equipment, the weight can add up quickly.
My personal goal is to use at least 80% of what I pack. I'll never reach 100% because I pack a spare change of clothes for anything longer than an overnight, and because I may have things for weather contingencies. And I always overpack socks and underwear because things can happen.
The tire pressure warning is indeed required now in the US. So is the one for seat belts. Not only are the speedometer and odometer required, but there are accuracy regulations for them. I don't believe there is a mandate for any of the other gauges, but the market isn't likely to accept a car without a fuel gauge, and the temperature gauge is a good idea for any car with an engine that uses an aluminum block. (Engine overheating is a more severe problem for aluminum engines than for the old steel block ones, because a single incidence of severe overheating can deform the engine block and instantly turn the engine into a heavy chunk of scrap metal.) Backup cameras are the most recently added instrumentation requirement.
In a modern car, especially an EV which by necessity is already going to have a lot of electronics, it's probably best to integrate all the information into a single display. That's what the Tesla Model 3 has done, though I'm not convinced they made the right choice of location and size.
A more serious problem with the user interface of the Model 3 is the elimination of most of the switches and knobs and replacing them with a touchscreen interface. The touchscreen is difficult for the driver to operate because it requires taking eyes off the road. Tesla's plan appears to be to use voice control instead, but most of that isn't ready yet.
Rev counters, aka tachometers, are useful in a gas-powered car, especially one with a small engine, so you know when to shift. They are useless in an electric car. EVs only have one gear; they don't need more because electric motors don't have a torque curve, they have a torque straight line.
No cameras is no longer an option in the US. Backup cameras are mandatory as of May 2018.
It's smaller than the Fortwo. It has a smaller motor and its speed is more limited. Unlike the Fortwo it's not likely to meet current safety standards for a car in the US. Finally, it's cheaper. The price translates to about $15,000 without any tax credits; if it's ever allowed to be sold here the price after incentives will be quite low.
Baby, meet bath water.
A problem with partial voucher systems is that not all students cost the same amount to educate. The vouchers are usually worth the city-wide average cost of educating a student, but special needs students are far more expensive and mainstream students are cheaper. The voucher schools weasel out of taking the special needs students, one way or another, and are effectively overpaid for the students they do take. Meanwhile, the remaining public system effectively has even less money per student to work with because they're stuck with all the expensive students.
A full voucher system eliminates that issue, sort of. But how do you get those high cost students educated? I suppose you could have a system where the reimbursement for high cost students is larger, but I don't know of any school voucher system that has done that.
A related problem: what do you do with students that nobody wants, perhaps because of disciplinary issues? They're required to go to school by law, but how do they avoid breaking the law if nobody will take them?
The homeless are more visible in San Francisco because the friendly climate makes it easier to live outdoors. They don't have to deal with extreme heat or cold and the amount of rain is modest. The actual problem is not worse than in other major cities.
That is fixable with proper system design.
Rain increases the path loss so you receive less signal. You could get back signal strength by using an antenna with more gain. That higher gain antenna will be larger and will require more careful aiming by the installer, and would thus be more costly.
How serious the problem is depends on where you live. TV satellites are in geostationary orbit above the equator; the farther away from the equator you live, the higher the path loss. The satellites serving the US are mostly above the Mountain time zone; Dish Network also has a couple that are over the Atlantic. If you're someplace else, again the path to the satellite is longer and the signal is weaker. Some people may have some amount of vegetation obstructing the path to the satellite; the amount of signal loss from that increases MUCH more quickly with rain. (Wet foliage is worse than wet air.)
It comes down to the question of what level of reliability customers are willing to pay for. If the reliability of the usual free installation is inadequate, you can do better by buying a larger replacement antenna. And you can clear away any trees that are near your dish.
They're giving in to the upsell and buying now because of the situation with the tax credit. By the time Tesla actually ships the $35K model, Teslas will no longer be eligible because the company will have sold too many electric cars. So you're really comparing buying a $36.5K car now vs a $35K car later, and given that choice buying the LR version is a no brainer.
I have to disagree on #12. It's important to have at least one browser that is NOT based on the dominant code base. Otherwise the dominant browser effectively becomes the standard. There is also Edge but it's Windows-only. Safari/WebKit has diverged substantially from Chrome by now but they have common roots. Firefox and its forks are the only completely independent cross-platform alternatives.
Just about all the major releases show up at all the chains. But the more limited releases don't. Buying into a one-chain pass means there are movies you won't get to see without spending extra. And in some cases you might not get to see the major film you want in your desired format. For example, back when the Hobbit movies were coming out, the first one was only in HFR 3D at AMC and the other two were only in HFR 3D at Regal. (Yes, I saw all three that way. I know a lot of people hated HFR but I liked it. Also, HFR and 3D are synergistic; HFR makes 3D work better.)
This is a major shift in the power balance of the CPU business.
For many years, one of the key things that kept Intel on top is that they were the best in the business at making chips. Not at designing an architecture, and not always at executing that architecture (like the era of Pentium 4 vs Athlon 64), but their chip fabrication technology was on top.
But now they find themselves in the unaccustomed position of playing catch-up. Samsung and TSMC are already shipping 7nm, with GlobalFoundries close behind, while Intel won't hit volume production of its 10nm process for another year. (Intel's 10nm is about the same density as the 7nm from those other companies despite the larger feature size so a direct comparison is fair.) This is the biggest business challenge that Intel has faced in a long time; I think they will find a way through it, but they may never return to the same level of market dominance.
Here in Boston we have two of the big mainstream theater chains (AMC and Regal) plus a new small one (Showplace Icon). Across the river in Cambridge there is an art house chain (Landmark), an independent art house (Brattle), and two small local chains (Frame One Theatres and Apple Cinemas). Neighboring Brookline has another independent art house (Coolidge Corner). The suburbs of Boston have yet another medium-sized chain (National Entertainment, dba Showcase and Cinema De Lux), two IMAX theaters run by the Jordan's Furniture chain (!), and additional AMC locations and independent art houses.
Lots of choices here. That substantially decreases the appeal of single chain offerings like AMC A-List.
There are a few theater chains that MoviePass is working with directly, mostly the ones with e-ticketing options. Landmark Theatres is one. But they don't have any special deals with the big players like AMC, Regal, and Cinemark.
It wasn't just to prevent people selling the tickets. I heard reports that some people were using MoviePass to buy tickets for movies they had no intention of watching just to get the loyalty points from the theater. If it was an urban theater near your place of work that was rather painless to do.
The i9 in the MacBook Pro is the 8950HK, a six core mobile CPU, not an eight core desktop part.
The fastest Mobile Ryzen currently available is the 2700U, a four core CPU. Between the lower core count and fewer instructions per MHz than Intel, it is not competitive in performance with the 8950HK. A slightly higher performance 2800H has been leaked but not officially announced; it's still a four core CPU but with higher clocks and a GPU with more stream units enabled. Those tweaks still won't make it competitive with Intel's H series parts, so it's hard to imagine much market interest and it may never get released.
The only current alternative would be to design around a downclocked desktop part like the 2700 and also include a separate GPU. (That model of MBP already has one, but using a CPU rather than an APU would mean that the GPU would have to be on full time.) That would give you eight cores and a good GPU, but the requirement to use the GPU full time would result in far worse battery life, not a tradeoff that Apple would care to make.
AMD may have something that's competitive at the high end of laptops next year when the Zen 2 chips are released. For now, AMD isn't an option for performance laptops. The 2500U and 2700U are decent mainstream laptop processors, offering more graphics performance but less processor performance than their counterparts from Intel; systems featuring those chips have started to appear but sales seem slow so far.
In the first implementation of C++ it WAS compiled by a compile that could only compile C. That first implementation was a preprocessor that turned your C++ program into somewhat hard to read C code, which was then run through the C compiler.
Nowadays you are far more likely to do the opposite: compile your C code with a compiler that is written in C++. That includes current versions of GCC and LLVM/Clang.
I wish I hadn't already commented on this subject so I could mod this comment up.
Michael Abrash is somebody who LITERALLY wrote the book on writing high performance code. More than one, actually. I also remember his posts on the subject on the late lamented BIX. He is the antithesis of an incompetent programmer.
Say programmers who know how to balance development time against run time.
Some programs are run ONCE and discarded. Or more accurately run a bunch of times during development and testing, and then there is one final run on the real data. That's typical for a program that is converting data from some old format to a newer one. Obviously, spending any time optimizing such a program would be utterly pointless, as the extra time spent on development would far outweigh any savings you would achieve during the program's single use. (Not to mention that 90% of the run time of such a program is probably spent on input and output, so improving the algorithms won't buy you much.)
At the other end of the scale, there is code that does crucial inner loop calculations. That code gets run a lot, and shaving even a little bit of time off its run time adds up to a lot. Code like that is a prime target for optimization.
All the other use cases lie somewhere in between. And each calls for a different amount of effort put into improvement. The mark of a good programmer and designer is knowing where to spend the effort and where not to.
No, Apple gave us only PART of what we wanted. A proper thermal solution is the missing part of the equation. But to do that, Apple would have had to actually put some engineering time and effort into the update of the MacBook Pro, rather than just shoving a newer processor and DDR4 RAM into the box. And they might have had to compromise the thinness and lightness of the system or make the fans louder.