Flash used to have a limit of about 500,000 read/writes. That limit has since been surpassed. I gather it can exceed 1 million now, though Wikipedia still says the former.
This is a common mistake. Modern flash rather unintuitively has a worse limit as the density goes up. Current high density SLC is at about 100k writes and MLC is at about 10k writes.
Can I just say, it's about time they brought out a version that could compare with existing low-end laptop drives in terms of capacity. If you ask me, that's what was really holding back the big-spenders from buying into this tech.
Absolutely! The power savings of flash-based bulk storage vs spinning media are amazing portable devices. The main problem is a big stack of flash is only just becoming cheap enough to be viable.
I don't get it. Modern flash has 1M+ write cycles, and we might presume that there is some rudimentary write balancing in these drives.
Strangely enough, modern flash is about 100k write cycles for high density SLC NAND and 10k writes for MLC NAND. Newer flash actually gets worse as the densities get better.
Even so, with proper wear leveling and sufficient redundancy you can achieve failure rates better than a spinning media. In fact, you can pick the numbers to achieve any arbitrary failure rate.
As for speed - you're correct, no single flash chip is 50MB/sec, but you can stack many of them in parallel and get that. That's a common way of doing it.
I think you're being overly harsh and pessimistic with your figures. There are some workloads you obviously shouldn't pair with a NAND flash, but quite frankly gaming isn't going to stress these things.
What nobody is pointing out is that a standard windows install will thrash the hell out of a Solid state drive. There is a reason you need to balance your writes and not treat a SS disk like a hard drive. I destroyed a Solid state IDE drive back 6 years ago (you have been able to buy them for over 15 years now) by installing windows on it. the swap space died within weeks.
These days (well, since YEARS ago now) we have this thing called Wear Leveling which means you can't wear out NAND flash by simply writing over the same portion over and over again. The writes get spread around other areas instead.
It hasn't been possible to kill a (decent) solid state drive like this in a very long time now. Please don't misinform people.
By all means, point me to some benchmarks. Your claims fly in the face of every bit of embedded work I've ever done.
So do yours. Provide ME with benchmarks. You only have to look at the architecture of both parts to work out that an ARM11 is clearly better clock-for-clock at general purpose applications. Perhaps you have only ever seen an ARM paired with a low performance external bus, or running at a low clock rate. That's not unusual, because that's perfectly normal if you actually wanted something low power. The Geode spends 3 cycles of its pipeline just decoding damn x86, for heaven's sake. Even after it does that - unlike other x86 chips - it's still only issuing 1 instruction per clock, which means that the 2 register instruction format of x86 makes it even worse. ARM packs a lot of punch per instruction compared to x86.
In fact a couple years ago I figured that drastically underclocking a PIII to the point that it used ~200mA like many ARM chips would result in a chip that could easily outperform anything available out there. It's just not practical for embedded systems because of the price of a PIII, and the rather complex supporting hardware for such desktop CPUs, as compared with ($5) ARMs. Yes, even in the most power-frugal situations, price is an even more major concern.
You are off by so many orders of magnitude on those figures.
A Pentium 3 underclocked is basically a very large, very expensive 8 ohm resistor. Even when you don't clock the thing, it's going to be sucking a few 100mW in leakage. They aren't made on low power silicon processes, and to be fair - they aren't designed to be. Nearly every ARM part I've worked with in the last 5 years has idled at about 0.1-10mW (depending on peripherals). They generally go about 0.5-1.0mW per MHz after that. That is an ORDER OF MAGNITUDE better than what a Pentium 3 manages. Never, ever say that an x86 is competitive with an ARM when underclocked in front of someone in the industry without expecting to get laughed at!
As for price - a typical $5 ARM chip will have an ARM1176 plus EVERY SINGLE PERIPHERAL YOU NEED integrated into it. Some have analogs integrated, some don't. A typical x86 chip will have maybe an IDE and VGA integrated and bugger all else (even the Geode has very little compared to most ARMs). It'll have an external PCI bus, which is a power hog and hugely expensive to buy companion chips for.
Again, you are an order of magnitude off on the BOM cost of an ARM vs x86.
Emulating floating point with integers is painfully slow. Of course Excel would be okay with an emulated FPU, but the kinds of multimedia being done on the OLPC are very float-intensive. Try playing a few videos on your ARM test systems and see just how fast of a CPU you need... Hint, it'll require a number of clock cycles, far in excess of a Geode.
We're not going to be playing 1080p H.264 videos on this thing. An ARM1176 is perfectly capable of playing back VGA resolution H.264 at about 400-500MHz depending on what you stick on it for SDRAM and how much cache it has. Been there, done that. This is for the OLPC. Who the hell wants to play silly resolution and/or bitrate videos on this thing? Who the hell cares about it doing lots of nifty multimedia anyway? This thing is enabling people to have computers in third world countries. Fitness for purpose.
And if you're really fussed about playing back multimedia, there are plenty of ARM-based CPUs around that have hardware video and/or 3D accelerators in them. What do you think is in an iPhone? That thing can play back 720p HD H.264 videos without strain. ARM1176 with accel. Pisses all over any x86 part for power consumption and integration.
I haven't had the opportunity to work with ARMs having an FPU, but everything I've heard (second hand) says they are extremely basic, and very very slow.
Then quite frankly you haven't experience with ARMs in the last 8 years. The FPUs are basic because they are
Seconded. When the kids get exposed to assembly, why not have it a clean modern design rather than the ugly crap that is x86? I'm guessing that there are many more choices for x86 supplier (especially for slower chips), and the licensing cost is less if you want to fab your own. There are hundreds (1000s?) of ARM vendors around, but only a tiny handful of x86 vendors. They're usually incredibly cheap too. You would never want to fab your own in any case - that's something only an elite, small group of big manufacturers are capable of these days.
I'm also puzzled as to why they didn't use one of: XScale (Intel or Marvell), Freescale, Samsung, TI, Toshiba and countless other big names that sell ARM based, high performance, low power CPUs, usually with a TON of integrated peripherals. I know at least 4 CPUs that are just plain better in every single way (even performance) than that Geode, but they all lack one feature: they don't run x86 natively.
It makes me mad - this isn't the 1990s any more. Portable code isn't hard. These days you can cross-compile the majority of Linux and GNU tools for an ARM platform without an error or problem at all. You'd probably be surprised just how much of your own C/C++/whatever runs unmodified on ARM (hint: all of it, if you didn't do anything illegal).
It's way past time that people got over instruction set differences and went for the best chip for the job.
ARM and SH are both very low power, but that is entirely at the expense of performance. A good trade-off for embedded systems that don't need much processing power, but certainly not for multimedia applications. As soon as you start trying to do floating point calculations, watch your ARM/SH CPU grind to a halt. They certainly do less per clock than even older x86 CPUs, and are a long way behind fairly modern x86 CPUs like the Geode. That goes double for Intel's clock-inflated XScale CPUs, pushing 1GHz.
Wrong, wrong and wrong.
SH isn't much of a performer, but ARM certainly is - especially anything from ARM11 and beyond. The AMD Geode chip in the OLPC is 433MHz and single issue. Pretty much all modern low power ARM chips are at least that clock speed.
So, compare what they do with that clock speed? If you were to compare general/integer computation between even an old ARM9 and the Geode, an ARM9 will beat it clock-for-clock. Modern desktop x86 chips are only fast because they do a ton of translation from x86 code and shove it down multiple pipelines. Geode only has one execution pipeline.
We're also talking about a VERY large difference in power consumption per "MIP" here too. From the Geode datasheet, it will be chewing nearly 1W of core power at 433MHz (650mA, 1.45V core). A comparable performance ARM would be more like 250mW-500mW.
Floating point? We're not in the dark ages any more in the embedded community. For starters: WE DON'T NEED IT. Emulation suffices pretty damn well for every day usage. It's not like Excel requires 8GFLOPS to recalculate 1,000 cells. These machines aren't for games either (and hell, you can do games without an FPU anyway - we managed for decades perfectly well). If you still really, really must waste a big bunch of silicon area to hardware floating point, then there's plenty of solutions out there for ARM which are 1-2 MACs per clock. That'll match the Geode easily. Oh, and also vector stuff should you want that.
Sorry, but you really aren't giving ARM (and XScale) based stuff enough credit. For the OLPC platform, it would have been ideal, and would have performed equally well if not better. I suspect the real reason they chose AMD was fear of a non-x86 platform and not "fitness for purpose" at all.
Every politician seems to be using this logical fallacy these days, to the extent that they're pretty much one-trick ponies. I can hardly stomach any interviews with a politician because they just launch straight into a false dichotomy at the earliest opportunity, immediately muddying the discussion, turning it into an ideological debate rather than arguing the problem at hand, and diverting from any fault that may be attributable to them.
It sickens me. What annoys me more is that journalists (including the interviewers) love it too. They just carry on with their interview and debate the extremes. It makes for heated debates, and bigger headlines. It all avoids doing any actual fucking journalism.
The world isn't full of extremes but unfortunately it's currently being run by people with 1-bit vision and 1-bit responses. It's all going to end in a 1-bit result if it carries on - and I don't think it'll be the good one.
I am very much aware that comparing power:mhz is often an incredibly bad thing to do, but I am also very much aware of the internal architecture of these two cores.
Via is single issue like ARM11. (At least, last I checked, but maybe they've changed that)
ARM1176 (a common variant) has hit-under-miss caches, some SIMD extensions (slow compared to Via, though), and a decent FPU (optional). No write order dependency issues on ARM. Very comparable to Via.
ARM11 has much lower memory latency than Via. It only takes 3 clocks to drag something out of memory while execution continues, you get write-to-read forwarding and a lot more registers which means less memory hits overall anyway.
The bus speeds of both are comparable. External memory is implementation dependent, but can also be matched.
So they're actually pretty similar in a lot of ways. I guess I should restrict my usage of "better" to the usual target market: ARM11 is a better core at being a web browser or general multimedia device. It'll do the same job at lower power. In addition, you could run an ARM11 comfortably past 600MHz without a fan or even a heatsink. If you felt like splashing out, you could get a 4 core variant that would cream pretty much anything Via offers. Obviously 4 times the peak power, but still the amazing idle power consumption most ARMs give you.
Again, x86 is still the obvious choice in the market due to a SOFTWARE problem. I say "problem", but what I really mean is the unwillingness of vendors to write portable code and realise that recompiling for multiple platforms stopped being a major headache about 10 years ago.
It's pretty sad that an ISA is still a barrier to this kind of thing. There is a major example in the volume market where this did work out, though: Apple switching to x86. Hell, they even had to switch endian and everything worked. If anyone ever tells you that switching ISA is prohibitive, go point them at Apple, because it worked out damn well for them, and you can always use Apple as some kind of Godwin's Law because nobody will ever argue they're wrong.
But, with all that, seems like hybrid drives would be the way to go right now.. after all, there's no limit on READING from solid state drives, just writing.
Actually, there is a limit on reading. It's called "read disturbance". It's a tiny effect on the more expensive SLC flash variants, but an enormous problem on MLC flash. MLC uses a hilariously scary technique of storing multiple bits per cell - multiple voltage levels, roughly speaking. It only takes a very small amount of current flowing through the same block to cause a disturbance that shifts a couple of electrons here and there and eventually changes the level of a cell.
On MLC, that happens at a frightening rate - something like 10-100k reads within the same block will cause it to grow an error. It's even common on the higher density SLC around at the moment, though I don't have a figure readily available (it's up in the millions, but that's not negligable for reads). That's easily detected and recovered using ECC, but the side-effect is you have to rewrite it. So in order to read a block... sometimes you have to write! This means that continuous reads of flash DO in fact wear it out. Again, the huge figures on read-disturbance aren't as good as they seem, beceause reads aren't wear leveled like writes are. You can re-read the same few blocks 10k times in a very short space of time and there's nothing the controller can do about it other than caching.
In fact, for SSD, I would assume they'd use the cheapest = highest density flash available to them. I would hate to think they use MLC, because that technology is a bunch of trash for general purpose usage as far as I'm concerned. But even SLC these days is starting to behave very much like video tape.
Everyone seems to be assuming that flash densities will go up and they'll overtake disk drives. What will actually happen is the endurance will drop through the floor as the densities increase. I wonder if there's a proof out there which shows that ECC can more than compensate for the loss of endurance vs density. If it doesn't, then flash has a brick wall it'll never go past.
How does this chip compare in performance per watt against ARM, PowerPC and the like?
Pathetically badly. Most modern low power ARM variants are in the range 0.3-0.5mW/MHz. At 500MHz you'd see them chewing up about 150-250mW. Last I checked the Via x86 chips were single issue, so it's not too unfair to compare an ARM11 (or similar) against them. Quite frankly an ARM11 will outperform the Via chip and run lower power.
The idle power figure is a joke. I can't recall the last time I used an ARM chip that idled at 100mW. More like 1-10mW. Still, it's nice to see an x86 chip get into sub-watt territory.
Of course, ARM doesn't run native x86... and that's pretty much the only reason there's such a large market for these Via x86 chips. It's also the reason you never see them in deeply embedded systems where people don't really care so much about what ISA you're running.
and at only 0.3W that's much lower power than most ARM chips
Most ARM chips I'm working with are more like 10-20 milliwatts when throttled and still doing useful work, and also in the ballpark of 0.3mW/MHz (which is what this new chip is if you work their numbers out). Pretty much any ARM chip you find aimed at the low power embedded market has a ton of features and overall design that make the entire SYSTEM much more efficient than something designed for extreme performance. Although, to be honest, I'm quite impressed it still maintains 0.3mW/MHz when it has such a peak potential. If it holds to be true.
No, this doesn't beat ARM for power consumption at the low end, but it'll beat the crap out of them at peak performance when required. Even so, I doubt you'll find it beating an ARM in mW/MIP for anything in the "sweet spot" of "normal" consumer computing, for example video and audio playback, or office document and web browsing stuff.
I'm not totally sure what this thing is targeted at... although it does open up a new point on the big dotted graph of power vs MIPs.
This sounds very familiar to me (I work with a lot of deeply embedded systems). What they've probably got is a clock in the WiFi which is referenced from the CPU clock. It could be the core clock for it or maybe used as a reference to generate the 2.4GHz signal.
So, when you change the CPU clock to 333MHz, the hard-wired multipliers for the WiFi clock don't work.
It could also be that the clock jitter at 333MHz is greater than at 222MHz, so the WiFi doesn't work even if the clock dividers/multipliers can be adjusted.
It could also be that they need to increase the core voltage to manage 333MHz, and that breaks some other aspect of the WiFi. Typically RF parts are very sensitive to these changes and the signal will end up garbled.
It could also be that the power regulator can't actually supply the total system power load required for 33MHz CPU and WiFi at the same time (WiFi is quite power hungry).
There's plenty of non-conspiracy reasons why this could be the way it is, and all of them are quite acceptable seeing as the part was only intended to be 222MHz in the first place. The fact that something doesn't work at 333MHz kind of validates the original rating.
The difference is it is called "CHEATING". You may have heard of that term somewhere and promptly forgot about it because it gets in the way of your ego.
You site the same tired old pathetic and incorrect crap about how "Blizzard could end gold farming if they wanted". No, they can't because there's always players like you out there who are determined to cheat and have no sense of ethics.
The game defines the act of linking real world wealth for in-game wealth as cheating. They make the rules. You are breaking them. You are cheating.
Why do people like you not understand these simple logical links?
Oh YAWN YAWN YAWN that's a perfect cut-and-paste apologist response about mudflation.
It DOES happen and it's not because MUD economies are broken. Give them some credit. Every MUD has money sinks as well as unlimited resources. You know what the factor is which moderates it all? TIME.
Gaining money takes time and effort. That determines the price of goods. That's time and effort you spent that someone else didn't have to. Simple (real) example: if you can kill mobs for 100g/hour, or mine 20 "khorium" bars per hour, then khorium bars are going to be worth 5g each, plus a fee for having to specialise to mine it in the first place.
What also needs bearing in mind is that often these activities involve doing repetitive/boring/unrewarding tasks. That usually increases the price some more still - and in fact this is the most profitable thing to do - because people are most willing to spend their gold on goods which they don't want to farm for themselves. Or goods which requires severe specialisation and time sinks to craft or obtain.
It is the gold selling industry that is the problem. They don't value their time as much as you do. They don't value it at all, in fact. The people they hire are lowest wage workers who spent their entire day doing the same thing over, and over again. They don't have some other fun questing to do, or raid to do, or some item they're working towards. No, they just have to do that to get their real life wages.
A recent problem is that often these farmers are bots. That is a clear violation of ToS - and people that buy from them are scum. They have virtually no cost to run and have no concept of how valuable their time is.
And so the economy starts getting fucked. Gold is sold straight to a player. Players start running around with ridiculous amounts of gold. It's usually very, very obvious when people do this because they end up with massive amounts of high end gear that would have taken masses of time and experience to obtain, but SOMEHOW don't have the playing skills or knowledge that you would have figured out having done that... The effect is that the cost of items goes up, because suddenly everyone has a lot of gold.
And then all the people with weak personalities cry that everyone else has more gear than them, and they can't afford to keep up. They go buy gold, and make the problem worse. More demand for goods, more people with money... higher cost of goods again. The cycle continues. The weak personalities say it's the only way you can keep up... etc... etc...
So yes, there is inflation, but ironically it's caused by the gold selling industry. So the next time you buy gold from them because the item you want is expensive, just think about how you did it to yourself when you first bought gold, how you've been caught like a fool in a nice little trap, and how you're screwing up the game for the rest of us. Die in a fire.
Q&A process / final signoff
Gold master creation (at least 6-8 weeks before launch day for our projects if I recall correctly)
TRC certification
ESRB board ratings
Disk duplication
Oops! Disk duplication took 1 day extra. DOH! It's September 26th shipping.
Or alternatively, your beta testing and bug fixing just had 1 less day to complete, because you told everyone you're going to ship to customer hands on exactly September 25th morning and your boxing and packaging plant will miss your run if you don't. I bet one major bug just got reclassified minor.
I'm not being pessimistic - this is what the real state of affairs is out there. You have been lucky in your job picks so far.
My own experience is not actually the video game industry, it's the MP3 player industry:) Same consumer oriented product design, same numb skull management.
I silently laugh every time I see someone propose an exact release date - sometimes even as ridiculous as morning or afternoon - which is 4 months in the future. Yes, software is so exact a job that you judge how long it'll take to write to within 0.4%!
To be fair, what's supposed to happen is you add on about 25-50% to your predicted hand-waving estimate, and when the release date gets near and you're running short of time, in order:
Cut features
Prioritize bugs - serious or very obvious cosmetic ones first (reviews always pick up on those), minor or cosmetic and hard to spot ones last. QA review board decides whether the remaining cosmetic ones are OK to ship with. Yes, people do ship with known bugs, but if they're rare or just minor glitches then it's better than shipping months later, and they can always be patched in the field. Note - I'm not talking about security products here:)
Absolutely last thing - ask people to work extra hours for perks such as free food, extra time off or even overtime pay.
What really happens (from long experience) is actually, in order:
Ask people to work extra hours for no return, which gets you very grumpy employees getting fat on pizza and losing their social life. Usually goes hand-in-hand with a deadline that was unrealistic when it was set 6 months before.
Add extra features. Yes, when a product launch is looming, for some reason people always think up extra features to stick in rather than cutting them. Something is always a new must-have feature, and there is a disconnect in the brains of the management who ask for them.
Start reclassifying serious bugs as minor. Reclassify minor bugs as cosmetic. Ignore cosmetic bugs. Cosmetic bugs are now features. There, no more bugs!
It pisses me off to see this happening elsewhere, and even more so when I realise this is just the way things are in the software industry. It's mostly run by people who haven't got a clue how software or QA works at all.
Re:Shame it's a rubbish report
on
The BBC On RMT
·
· Score: 2, Insightful
Not to mention that for 3 minutes of the 3:17 clip, they go on about the whole thing as if it's actually something allowed by the game. It isn't. Blizzard invented the game, they invented the rules, and they say it's cheating, so by definition it's cheating.
Again, they only give 15 seconds of the damn report to that. They give no time to mention the hideous pay and conditions the gold farmers have to work under either...
Shame it's a rubbish report
on
The BBC On RMT
·
· Score: 2, Insightful
Sadly the report didn't really understand the issues either. This is a typical shitty "fair two sided report" you get from the BBC these days. Someone must have told them a while ago that all reports must be fair and impartial, but these days all the second rate journalists take that to stupid extremes.
This gives far too much credibility to the arguments gold sellers and gold buyers make. They are so, so easy to counter and refute, but did you see how much air time they gave the Blizzard rep? 15 seconds.
Did you see the shitty argument the "Gamer" called "Ryan" made? He has a full time job, and therefore he can't complete. Oh my god, you aren't a gamer, Ryan, you're just a delusional loser with an inadequacy complex. Then the BBC report bought the whole argument and went with it...
The whole "job vs gold" argument is a false dichotomy invented by the people that buy and sell gold. It doesn't exist. You can play the game perfectly well along side a full time job, still have a social life, and still "compete" (as the "gamer" "ryan" puts it). Some people suffer from an inadequacy complex more than others, I guess.
We don't need people that that kind of mentality in the game at all. They are the problem.
There are people who understand how gold sellers are BAD for the game, and there are those who either use them or don't play the game. It's always like that - supporters are either somebody writing a template essay at school on the matter (and do the usual inconclusive style essay that schools love), or they're active users.
The problem is that people who use gold sellers aren't actually playing the game. Most of the time they aren't any good at the game, ironically because they bought their way to their status rather than earned it through their own achievements. I shudder to think what these people must be like in real life.
I'm guessing it's that asshole author of the WoWGlider bot, actually. Quite an evident bias and quite blatant half-truths in that summary. It's the same for every anti-Blizzard story that comes out about WoW - usually it's the virtual gold industry posting anonymous stories about how great they are and how Blizzard is being heavy handed on their poor, innocent selves. Rubbish.
He's ruining the game so he can make some profit out of it. I have absolutely no sympathy whatsoever, even if what Blizzard is doing is dubious. He certainly doesn't have clean hands himself.
There are ways around this. A common trick used to exchange gold cross-faction (not involving gold sellers at all) is to stick a "plain letter" (a very cheap item) on the cross-faction auction house for the amount of gold you want. The player on the other faction then buys it.
Same applies to same-faction trading. In order to really stop gold changing hands, you would need to remove the auction houses. It would also render a lot of the profession system useless because you couldn't do enchants/crafting and get a fee.
No, then the game would suck even more than it currently does. Sony put a nail in the coffin of EQ1/2 when they did that and pretty much every authoritative commentator out there said it was yet another in a long string of extremely dumb moves by them.
There simply isn't any benefit to this. The solution, which I would never have suggested a year ago, is for them to stop bothering with the gold sellers and to start banning the gold users. Unfortunately the problem with WoW at the moment is the user base. There has been an explosion of spoiled emo brats with too much money, too little ethics and far too much mouth. They need to be killed off.
Slashdot Mirror
This is a common mistake. Modern flash rather unintuitively has a worse limit as the density goes up. Current high density SLC is at about 100k writes and MLC is at about 10k writes.
Can I just say, it's about time they brought out a version that could compare with existing low-end laptop drives in terms of capacity. If you ask me, that's what was really holding back the big-spenders from buying into this tech.Absolutely! The power savings of flash-based bulk storage vs spinning media are amazing portable devices. The main problem is a big stack of flash is only just becoming cheap enough to be viable.
Strangely enough, modern flash is about 100k write cycles for high density SLC NAND and 10k writes for MLC NAND. Newer flash actually gets worse as the densities get better.
Even so, with proper wear leveling and sufficient redundancy you can achieve failure rates better than a spinning media. In fact, you can pick the numbers to achieve any arbitrary failure rate.
As for speed - you're correct, no single flash chip is 50MB/sec, but you can stack many of them in parallel and get that. That's a common way of doing it.
I think you're being overly harsh and pessimistic with your figures. There are some workloads you obviously shouldn't pair with a NAND flash, but quite frankly gaming isn't going to stress these things.
These days (well, since YEARS ago now) we have this thing called Wear Leveling which means you can't wear out NAND flash by simply writing over the same portion over and over again. The writes get spread around other areas instead.
It hasn't been possible to kill a (decent) solid state drive like this in a very long time now. Please don't misinform people.
By all means, point me to some benchmarks. Your claims fly in the face of every bit of embedded work I've ever done.
So do yours. Provide ME with benchmarks. You only have to look at the architecture of both parts to work out that an ARM11 is clearly better clock-for-clock at general purpose applications. Perhaps you have only ever seen an ARM paired with a low performance external bus, or running at a low clock rate. That's not unusual, because that's perfectly normal if you actually wanted something low power. The Geode spends 3 cycles of its pipeline just decoding damn x86, for heaven's sake. Even after it does that - unlike other x86 chips - it's still only issuing 1 instruction per clock, which means that the 2 register instruction format of x86 makes it even worse. ARM packs a lot of punch per instruction compared to x86.
In fact a couple years ago I figured that drastically underclocking a PIII to the point that it used ~200mA like many ARM chips would result in a chip that could easily outperform anything available out there. It's just not practical for embedded systems because of the price of a PIII, and the rather complex supporting hardware for such desktop CPUs, as compared with ($5) ARMs. Yes, even in the most power-frugal situations, price is an even more major concern.
You are off by so many orders of magnitude on those figures.
A Pentium 3 underclocked is basically a very large, very expensive 8 ohm resistor. Even when you don't clock the thing, it's going to be sucking a few 100mW in leakage. They aren't made on low power silicon processes, and to be fair - they aren't designed to be. Nearly every ARM part I've worked with in the last 5 years has idled at about 0.1-10mW (depending on peripherals). They generally go about 0.5-1.0mW per MHz after that. That is an ORDER OF MAGNITUDE better than what a Pentium 3 manages. Never, ever say that an x86 is competitive with an ARM when underclocked in front of someone in the industry without expecting to get laughed at!
As for price - a typical $5 ARM chip will have an ARM1176 plus EVERY SINGLE PERIPHERAL YOU NEED integrated into it. Some have analogs integrated, some don't. A typical x86 chip will have maybe an IDE and VGA integrated and bugger all else (even the Geode has very little compared to most ARMs). It'll have an external PCI bus, which is a power hog and hugely expensive to buy companion chips for.
Again, you are an order of magnitude off on the BOM cost of an ARM vs x86.
Emulating floating point with integers is painfully slow. Of course Excel would be okay with an emulated FPU, but the kinds of multimedia being done on the OLPC are very float-intensive. Try playing a few videos on your ARM test systems and see just how fast of a CPU you need... Hint, it'll require a number of clock cycles, far in excess of a Geode.
We're not going to be playing 1080p H.264 videos on this thing. An ARM1176 is perfectly capable of playing back VGA resolution H.264 at about 400-500MHz depending on what you stick on it for SDRAM and how much cache it has. Been there, done that. This is for the OLPC. Who the hell wants to play silly resolution and/or bitrate videos on this thing? Who the hell cares about it doing lots of nifty multimedia anyway? This thing is enabling people to have computers in third world countries. Fitness for purpose. And if you're really fussed about playing back multimedia, there are plenty of ARM-based CPUs around that have hardware video and/or 3D accelerators in them. What do you think is in an iPhone? That thing can play back 720p HD H.264 videos without strain. ARM1176 with accel. Pisses all over any x86 part for power consumption and integration.
I haven't had the opportunity to work with ARMs having an FPU, but everything I've heard (second hand) says they are extremely basic, and very very slow.
Then quite frankly you haven't experience with ARMs in the last 8 years. The FPUs are basic because they are
I'm also puzzled as to why they didn't use one of: XScale (Intel or Marvell), Freescale, Samsung, TI, Toshiba and countless other big names that sell ARM based, high performance, low power CPUs, usually with a TON of integrated peripherals. I know at least 4 CPUs that are just plain better in every single way (even performance) than that Geode, but they all lack one feature: they don't run x86 natively.
It makes me mad - this isn't the 1990s any more. Portable code isn't hard. These days you can cross-compile the majority of Linux and GNU tools for an ARM platform without an error or problem at all. You'd probably be surprised just how much of your own C/C++/whatever runs unmodified on ARM (hint: all of it, if you didn't do anything illegal).
It's way past time that people got over instruction set differences and went for the best chip for the job.
Wrong, wrong and wrong.
SH isn't much of a performer, but ARM certainly is - especially anything from ARM11 and beyond. The AMD Geode chip in the OLPC is 433MHz and single issue. Pretty much all modern low power ARM chips are at least that clock speed.
So, compare what they do with that clock speed? If you were to compare general/integer computation between even an old ARM9 and the Geode, an ARM9 will beat it clock-for-clock. Modern desktop x86 chips are only fast because they do a ton of translation from x86 code and shove it down multiple pipelines. Geode only has one execution pipeline.
We're also talking about a VERY large difference in power consumption per "MIP" here too. From the Geode datasheet, it will be chewing nearly 1W of core power at 433MHz (650mA, 1.45V core). A comparable performance ARM would be more like 250mW-500mW.
Floating point? We're not in the dark ages any more in the embedded community. For starters: WE DON'T NEED IT. Emulation suffices pretty damn well for every day usage. It's not like Excel requires 8GFLOPS to recalculate 1,000 cells. These machines aren't for games either (and hell, you can do games without an FPU anyway - we managed for decades perfectly well). If you still really, really must waste a big bunch of silicon area to hardware floating point, then there's plenty of solutions out there for ARM which are 1-2 MACs per clock. That'll match the Geode easily. Oh, and also vector stuff should you want that.
Sorry, but you really aren't giving ARM (and XScale) based stuff enough credit. For the OLPC platform, it would have been ideal, and would have performed equally well if not better. I suspect the real reason they chose AMD was fear of a non-x86 platform and not "fitness for purpose" at all.
Every politician seems to be using this logical fallacy these days, to the extent that they're pretty much one-trick ponies. I can hardly stomach any interviews with a politician because they just launch straight into a false dichotomy at the earliest opportunity, immediately muddying the discussion, turning it into an ideological debate rather than arguing the problem at hand, and diverting from any fault that may be attributable to them.
It sickens me. What annoys me more is that journalists (including the interviewers) love it too. They just carry on with their interview and debate the extremes. It makes for heated debates, and bigger headlines. It all avoids doing any actual fucking journalism.
The world isn't full of extremes but unfortunately it's currently being run by people with 1-bit vision and 1-bit responses. It's all going to end in a 1-bit result if it carries on - and I don't think it'll be the good one.
So they're actually pretty similar in a lot of ways. I guess I should restrict my usage of "better" to the usual target market: ARM11 is a better core at being a web browser or general multimedia device. It'll do the same job at lower power. In addition, you could run an ARM11 comfortably past 600MHz without a fan or even a heatsink. If you felt like splashing out, you could get a 4 core variant that would cream pretty much anything Via offers. Obviously 4 times the peak power, but still the amazing idle power consumption most ARMs give you.
Again, x86 is still the obvious choice in the market due to a SOFTWARE problem. I say "problem", but what I really mean is the unwillingness of vendors to write portable code and realise that recompiling for multiple platforms stopped being a major headache about 10 years ago.
It's pretty sad that an ISA is still a barrier to this kind of thing. There is a major example in the volume market where this did work out, though: Apple switching to x86. Hell, they even had to switch endian and everything worked. If anyone ever tells you that switching ISA is prohibitive, go point them at Apple, because it worked out damn well for them, and you can always use Apple as some kind of Godwin's Law because nobody will ever argue they're wrong.
But, with all that, seems like hybrid drives would be the way to go right now.. after all, there's no limit on READING from solid state drives, just writing.
Actually, there is a limit on reading. It's called "read disturbance". It's a tiny effect on the more expensive SLC flash variants, but an enormous problem on MLC flash. MLC uses a hilariously scary technique of storing multiple bits per cell - multiple voltage levels, roughly speaking. It only takes a very small amount of current flowing through the same block to cause a disturbance that shifts a couple of electrons here and there and eventually changes the level of a cell.
On MLC, that happens at a frightening rate - something like 10-100k reads within the same block will cause it to grow an error. It's even common on the higher density SLC around at the moment, though I don't have a figure readily available (it's up in the millions, but that's not negligable for reads). That's easily detected and recovered using ECC, but the side-effect is you have to rewrite it. So in order to read a block... sometimes you have to write! This means that continuous reads of flash DO in fact wear it out. Again, the huge figures on read-disturbance aren't as good as they seem, beceause reads aren't wear leveled like writes are. You can re-read the same few blocks 10k times in a very short space of time and there's nothing the controller can do about it other than caching.
In fact, for SSD, I would assume they'd use the cheapest = highest density flash available to them. I would hate to think they use MLC, because that technology is a bunch of trash for general purpose usage as far as I'm concerned. But even SLC these days is starting to behave very much like video tape.
Everyone seems to be assuming that flash densities will go up and they'll overtake disk drives. What will actually happen is the endurance will drop through the floor as the densities increase. I wonder if there's a proof out there which shows that ECC can more than compensate for the loss of endurance vs density. If it doesn't, then flash has a brick wall it'll never go past.
How does this chip compare in performance per watt against ARM, PowerPC and the like?
Pathetically badly. Most modern low power ARM variants are in the range 0.3-0.5mW/MHz. At 500MHz you'd see them chewing up about 150-250mW. Last I checked the Via x86 chips were single issue, so it's not too unfair to compare an ARM11 (or similar) against them. Quite frankly an ARM11 will outperform the Via chip and run lower power.
The idle power figure is a joke. I can't recall the last time I used an ARM chip that idled at 100mW. More like 1-10mW. Still, it's nice to see an x86 chip get into sub-watt territory.
Of course, ARM doesn't run native x86... and that's pretty much the only reason there's such a large market for these Via x86 chips. It's also the reason you never see them in deeply embedded systems where people don't really care so much about what ISA you're running.
and at only 0.3W that's much lower power than most ARM chips
Most ARM chips I'm working with are more like 10-20 milliwatts when throttled and still doing useful work, and also in the ballpark of 0.3mW/MHz (which is what this new chip is if you work their numbers out). Pretty much any ARM chip you find aimed at the low power embedded market has a ton of features and overall design that make the entire SYSTEM much more efficient than something designed for extreme performance. Although, to be honest, I'm quite impressed it still maintains 0.3mW/MHz when it has such a peak potential. If it holds to be true.
No, this doesn't beat ARM for power consumption at the low end, but it'll beat the crap out of them at peak performance when required. Even so, I doubt you'll find it beating an ARM in mW/MIP for anything in the "sweet spot" of "normal" consumer computing, for example video and audio playback, or office document and web browsing stuff.
I'm not totally sure what this thing is targeted at... although it does open up a new point on the big dotted graph of power vs MIPs.
So, when you change the CPU clock to 333MHz, the hard-wired multipliers for the WiFi clock don't work.
It could also be that the clock jitter at 333MHz is greater than at 222MHz, so the WiFi doesn't work even if the clock dividers/multipliers can be adjusted.
It could also be that they need to increase the core voltage to manage 333MHz, and that breaks some other aspect of the WiFi. Typically RF parts are very sensitive to these changes and the signal will end up garbled.
It could also be that the power regulator can't actually supply the total system power load required for 33MHz CPU and WiFi at the same time (WiFi is quite power hungry).
There's plenty of non-conspiracy reasons why this could be the way it is, and all of them are quite acceptable seeing as the part was only intended to be 222MHz in the first place. The fact that something doesn't work at 333MHz kind of validates the original rating.
You site the same tired old pathetic and incorrect crap about how "Blizzard could end gold farming if they wanted". No, they can't because there's always players like you out there who are determined to cheat and have no sense of ethics.
The game defines the act of linking real world wealth for in-game wealth as cheating. They make the rules. You are breaking them. You are cheating.
Why do people like you not understand these simple logical links?
It DOES happen and it's not because MUD economies are broken. Give them some credit. Every MUD has money sinks as well as unlimited resources. You know what the factor is which moderates it all? TIME.
Gaining money takes time and effort. That determines the price of goods. That's time and effort you spent that someone else didn't have to. Simple (real) example: if you can kill mobs for 100g/hour, or mine 20 "khorium" bars per hour, then khorium bars are going to be worth 5g each, plus a fee for having to specialise to mine it in the first place.
What also needs bearing in mind is that often these activities involve doing repetitive/boring/unrewarding tasks. That usually increases the price some more still - and in fact this is the most profitable thing to do - because people are most willing to spend their gold on goods which they don't want to farm for themselves. Or goods which requires severe specialisation and time sinks to craft or obtain.
It is the gold selling industry that is the problem. They don't value their time as much as you do. They don't value it at all, in fact. The people they hire are lowest wage workers who spent their entire day doing the same thing over, and over again. They don't have some other fun questing to do, or raid to do, or some item they're working towards. No, they just have to do that to get their real life wages.
A recent problem is that often these farmers are bots. That is a clear violation of ToS - and people that buy from them are scum. They have virtually no cost to run and have no concept of how valuable their time is.
And so the economy starts getting fucked. Gold is sold straight to a player. Players start running around with ridiculous amounts of gold. It's usually very, very obvious when people do this because they end up with massive amounts of high end gear that would have taken masses of time and experience to obtain, but SOMEHOW don't have the playing skills or knowledge that you would have figured out having done that... The effect is that the cost of items goes up, because suddenly everyone has a lot of gold.
And then all the people with weak personalities cry that everyone else has more gear than them, and they can't afford to keep up. They go buy gold, and make the problem worse. More demand for goods, more people with money... higher cost of goods again. The cycle continues. The weak personalities say it's the only way you can keep up... etc... etc...
So yes, there is inflation, but ironically it's caused by the gold selling industry. So the next time you buy gold from them because the item you want is expensive, just think about how you did it to yourself when you first bought gold, how you've been caught like a fool in a nice little trap, and how you're screwing up the game for the rest of us. Die in a fire.
You talk about them a lot, yet give none.
Gold master creation (at least 6-8 weeks before launch day for our projects if I recall correctly)
TRC certification
ESRB board ratings
Disk duplication
Oops! Disk duplication took 1 day extra. DOH! It's September 26th shipping.
Or alternatively, your beta testing and bug fixing just had 1 less day to complete, because you told everyone you're going to ship to customer hands on exactly September 25th morning and your boxing and packaging plant will miss your run if you don't. I bet one major bug just got reclassified minor.
I'm not being pessimistic - this is what the real state of affairs is out there. You have been lucky in your job picks so far.
My own experience is not actually the video game industry, it's the MP3 player industry :) Same consumer oriented product design, same numb skull management.
To be fair, what's supposed to happen is you add on about 25-50% to your predicted hand-waving estimate, and when the release date gets near and you're running short of time, in order:
What really happens (from long experience) is actually, in order:
It pisses me off to see this happening elsewhere, and even more so when I realise this is just the way things are in the software industry. It's mostly run by people who haven't got a clue how software or QA works at all.
Again, they only give 15 seconds of the damn report to that. They give no time to mention the hideous pay and conditions the gold farmers have to work under either...
This gives far too much credibility to the arguments gold sellers and gold buyers make. They are so, so easy to counter and refute, but did you see how much air time they gave the Blizzard rep? 15 seconds.
Did you see the shitty argument the "Gamer" called "Ryan" made? He has a full time job, and therefore he can't complete. Oh my god, you aren't a gamer, Ryan, you're just a delusional loser with an inadequacy complex. Then the BBC report bought the whole argument and went with it...
The whole "job vs gold" argument is a false dichotomy invented by the people that buy and sell gold. It doesn't exist. You can play the game perfectly well along side a full time job, still have a social life, and still "compete" (as the "gamer" "ryan" puts it). Some people suffer from an inadequacy complex more than others, I guess.
There are people who understand how gold sellers are BAD for the game, and there are those who either use them or don't play the game. It's always like that - supporters are either somebody writing a template essay at school on the matter (and do the usual inconclusive style essay that schools love), or they're active users.
The problem is that people who use gold sellers aren't actually playing the game. Most of the time they aren't any good at the game, ironically because they bought their way to their status rather than earned it through their own achievements. I shudder to think what these people must be like in real life.
He's ruining the game so he can make some profit out of it. I have absolutely no sympathy whatsoever, even if what Blizzard is doing is dubious. He certainly doesn't have clean hands himself.
I hate defeatists.
Same applies to same-faction trading. In order to really stop gold changing hands, you would need to remove the auction houses. It would also render a lot of the profession system useless because you couldn't do enchants/crafting and get a fee.
Have you played WoW?
There simply isn't any benefit to this. The solution, which I would never have suggested a year ago, is for them to stop bothering with the gold sellers and to start banning the gold users. Unfortunately the problem with WoW at the moment is the user base. There has been an explosion of spoiled emo brats with too much money, too little ethics and far too much mouth. They need to be killed off.