The problem is AMD is using an outdated architecture. More cores != more speed for general use.
Ahh, but the problem with Intel is price. Everything else is great - just not price.
Very, very few. A single fast core can outperform a few slow cores in general usage and AMD seems only concerned with getting more and more cores on a single CPU die which really doesn't translate to great performance in the real world for general use.
True, but more cores for less money lets you spend money on more I/O - and your CPU strength really doesn't matter when disk I/O holds you up.
So regardless of whether 4 cores or 6 cores is better, cheaper IS better because it lets you spend more on HDDs or SSDs, which will have a far bigger impact on workstation use.
But if I could get 12 Phenom II cores in a system... that would be a dream come true for x264 encoding.
Only the ones with updateable firmware and/or operating systems.
The n900 has a DSP doesn't it? Basically a second CPU, extremely optimized for decoding video. Supporting a new codec is as easy as understanding the DSP's instruction set, and writing a decoder.
We are talking about the internet. The web. And marketshare. Sorenson Spark and On2 VP6 are the winners, and h264 is tiny, almost vanishing by comparison.
Could you back that up with some examples? Every site I know of that does video streaming uses H.264
Interesting thought, just for comparison I did some research: it seems the PS2 has sold around twice as many as the Wii. I am not sure when the PS2 sales started slowing down, but there is definitely room for more Wii sales.
I disagree with this Michael Pachter guy. Announcing a Wii 2 would kill sales - everyone stops buying when a better successor is just around the corner.
Seems like it'll be a while yet before they completely saturate the market. I can't wait for the price to drop into the $150 range.
In other words, it is true file sharers are leeches on society who take without giving back, but they aren't the ones who caused the problems in the recording industry. The industry brought it on themselves.
In other words, they don't matter. Don't factor into the equation. Ignore them.
P.S. I've seen articles suggesting that a pirated copy (of a game) is worth about 1/20th of a legit copy, because of word of mouth.;)
But if the game sucks, then the word of mouth will hurt rather than help, which leaves you with... Spore.
Universal Iron rule of the Internet: Everyone would be happy to pay for X, but they're only willing to pay half of what's being asked. Songs are a buck? 50c please. Netflix is $10 a month? I'll only pay $5 a month, and only if there's a bigger selection. An iPad will be $999? Well I'd happily pay $500, and only if it isn't crippled with Apple's retard-o-platform!
Paying $2/epsiode is not cheap. I would pay $1 for an hour long show (42 minutes in reality) as long as it is commercial free. IF you try to sell me commercials, forget it! 30 minute shows I would pay $.50-$.75, but again, only for a commercial free version.
I'd pay $0.05/ep, and if I liked it, I'd donate anywhere from $0.10 to $0.50 (this is to-stream; not to-own)
But there's no way I'm paying for those lame recap episodes that shows are constantly pumping out.
I think a pay-to-stream service could be quite popular. If it were like Hulu, minus the ads, but with credit linked to your account... you could transfer money for free to content producers. Some episodes could be free (first few of a season), while others could have a mandated donation of $0.05 or $0.10. New episodes could be $0.50 for two weeks, then drop to $0.10 - or whatever the content producer decides. The trick is making it easy to pay - you do that by letting the payment rules be configured on your account page, and then just displaying a small notice when payment is automatically sent.
To take the fear out of paying for lemon episodes, the funds (or points) could be transfered part-way through viewing... or perhaps there could be an easy way to undo a donation, even if you fully watched a show. (but only within 24h) This could be used to indicate to producers which episodes didn't go over so well - and to maintain relevance, it could factor in the viewer's behaviour, to filter out viewers that already donate very little (or nothing) on average. A trend engine like that would be valuable for figuring out what viewers want to see, what they want to pay for, and how much they're willing to pay.
The content producers could have access to all sorts of statistics, including average donations per episode/show.
And the viewers could get other perks - perhaps if they donated enough (highest 10%), the system could offer them bonus episodes as thanks - episodes normally reserved for DVD extras.
And finally, it all comes down to what the producer is comfortable with. If a show or movie was expensive to create, they could choose not to stream it to viewers with an average donation of $0.00 - or even one less than $0.20. But rather than cut all those viewers out, the intuitive Hulu-like UI could simply display an information popup, requesting payment. Click okay, and $0.20 is deducted, then it starts streaming. For expensive just-released films (like Avatar?), the producer could scale this to whatever they feel is fair ($2.00?), and track everything with the statistics/trend-engine.
There's lots of ways for a pay-to-stream service to evolve. The trick is making everything very, very convenient. Convenience when it comes to viewing, payment, and statistics, increases the perceived value for both viewers and producers. In the end, many viewers will want to support the shows and episodes that they like, especially since it will lead to better TV.
Paying $2/epsiode is not cheap. I would pay $1 for an hour long show (42 minutes in reality) as long as it is commercial free.
One of the major benefits of a system like this, is it lets you pay $1 for a show you like, $2 for a show you really like, and at the same time, I pay $0.10. (unless I choose to donate more) The current system has us paying $0 + $0
I disagree. Then again, I also tend to multitask - watch TV, fiddle with my phone, work on the other computer, etc...
I was referring to Crysis only. The link proves that SSDs have massive benefits for some tasks, even if you can't notice it easily by counting in your head.;)
I have heard that alt+tabbing back into games can drop from 10 seconds to 2. That's a 500% performance increase, with even a modest SSD. Okay, it's only 8 seconds, so it hardly matters - but do try to understand what the numbers reflect. SSDs are very powerful underneath.
Most of my games utilize 2-4 cores, so I'd probably get a responsiveness improvement from having six. I understand what you're getting at, though. One of my favourite games is TF2, and I doubt I'll ever see it using more than 2.
- Hoshi's universal translator from Enterprise. It can hear a few phrases in an alien language and then be able to translate back a response that includes words and semantic structures it never heard yet in that language. Note that it didn't even need to be told a translation for that original sample. It could just hear "bbzzt klick klickety-klick hrr bzzt" in some insectoid language and just figure out what it means and, for that matter, what the whole rest of the language is like.
I believe we'll eventually have translators like that - but they won't function like that.
The translator will passively scan the nearby creature's brain, examine and interpret the structure, then crunch numbers until it can guess the language, perhaps cross referencing with the creature's memories, etc. until it has a profile built.
Of course, to get there we'll need a revolutionary speed increase, rather than evolutionary. Something that kicks us from ghz to thz or beyond. Building a profile of a creature's mind would probably take hundreds of terabytes of RAM, too.
But if we look at the speed and capacity increases over the past 40 years... maybe, just maybe!
If you didn't notice an improvement from a quad, then you aren't using your computer fully. That's the same reason lots of people can get away with using netbooks. I personally noticed a big responsiveness improvement when I went from dual-core to quad, and encoding times dropped quite a bit. I suspect your computer usage is somewhere in the middle.
Don't forget to prioritize I/O. A cheap AMD Quad + SSD will give incredible responsiveness compared to a more expensive Quad + regular HDD. Or if you need tons of space, like me, go for multiple HDDs. Those spinning things are still the slowest component that we have to deal with... so speeding them up is more important than a 10% faster CPU or better RAM.
What I was getting at was the PS3 has quite a beefy processor for the price. For an underfunded department, it's a very cheap alternative. Great if your problem fits in the available RAM.
And if it doesn't, and RAM is your only factor, there's x86 Tyan boards with 48-96GB of RAM. They barely cost $1000.
I think like most mainframes, I/O is probably a big part of supercomputers - though not often mentioned. It's easier to throw the amount of RAM or raw core count around.
Accurate speech recognition is easy. We know how to do it - we just don't have the speed required, yet.
It'd be a lot faster if we had a database of all the variations in how we speak most syllables - but I suspect such a database would easily surpass 4GB, for a single person. Add in the differences between people, and you'd have hundreds of terabytes of slightly different ways to pronounce things - for a single language. Actually, dialects can vary a lot, so it's probably far far higher.
Once we have such a database, perhaps some enterprising company will distill all that data into something easier to process? It happened for video - what was once stored analog, or close to, is now stored digital in formats like H.264, often getting 1:50 or 1:100 compression, with no perceivable quality loss. Once our desktops have 256GB of RAM, and 256-core CPUs, someone or some company will take that data, analyze it, and output something that can do accurate speech recognition with a meagre amount of RAM and CPU time. Perhaps 4-8 cores, and 2-4GB of RAM.
And then we'll all wonder why we didn't have perfect speech recognition in 2010, forgetting all the processing power required to distill that audio data into something meaningful.
I'm considering migrating from Ubuntu to Debian stable for my NAS.
9.10 was pretty funny. It thought my 3114 PCI card was an nVidia RAID Array. It only took them a few weeks to get that fixed, but I still reverted to 9.04, and it's got me thinking about Debian instead.
The thing about Ubuntu and its sister distros is... don't be an early adopter.
Whenever a new release comes out, it's packed with hundreds of strange bugs. The forums get flooded by people with all sorts of issues. Then within a few months, 95% of them are fixed.
The best time to hop on 10.04 will be 2 months from now. Ubuntu is firm about their 6 month releases, but it'd really work better if they went with 8 months. Do a feature freeze after 4, then spend 4 getting it rock solid stable, and doing proper regression testing.
We're now quoting android development costs at 4x's that of the iPhone because of QA.
I know someone that's in the exact same line of work. (iPhone and Android apps) His company quotes 2.5x the cost, so while the issue isn't as bad as you make it sound, it is an issue.
On the other hand, how much bigger of an install base do you get? There's more Android phones out there - they're just fragmented a lot more than iPhones are.
His company is migrating to some framework that apparently targets iPhone, Android (All), and I think either WinCE 6 or WebOS - can't remember which. It reduces dev time a lot, at the cost of some battery life - but most clients will take that tradeoff to save money.
...until I can get a decent (120GB+) sized SSD that doesn't cost as much as a new video card?
2012.
In late 2010/2011 we'll probably see the old Indilinx stock being sold off for cheap. People will jump on it.
Then by 2012 once all the new processes for NAND and controller fabrication have matured, and yields are very high, we'll see controllers like the Sandforce becoming affordable - truly decent performance, if I ever saw any.;)
I suspect that sometime in 2012 Sandforce SSDs will hit 100GB for $100, or 200GB for $200. I might be wrong, but that's my guess.
I do not think many SSDs will drop into the sub-$80 range. In 2008 we had $70 32GB Patriot WARPs. Now we have $120 30/32GB SSDs - far faster, but more expensive per GB. NAND will drop in price shortly, but not enough to put SSDs in the sub-$50 range. For that price range, we'll have to stick with SDHC/SDXC, and HDDs.
Slashdot Mirror
Seems like you can easily use all those cores.
Modern games can use 4 cores fairly well. Sometimes well enough to impact UI responsiveness.
Therefore: Six!
The problem is AMD is using an outdated architecture. More cores != more speed for general use.
Ahh, but the problem with Intel is price. Everything else is great - just not price.
Very, very few. A single fast core can outperform a few slow cores in general usage and AMD seems only concerned with getting more and more cores on a single CPU die which really doesn't translate to great performance in the real world for general use.
True, but more cores for less money lets you spend money on more I/O - and your CPU strength really doesn't matter when disk I/O holds you up.
So regardless of whether 4 cores or 6 cores is better, cheaper IS better because it lets you spend more on HDDs or SSDs, which will have a far bigger impact on workstation use.
But if I could get 12 Phenom II cores in a system... that would be a dream come true for x264 encoding.
Only the ones with updateable firmware and/or operating systems.
The n900 has a DSP doesn't it? Basically a second CPU, extremely optimized for decoding video. Supporting a new codec is as easy as understanding the DSP's instruction set, and writing a decoder.
We are talking about the internet. The web. And marketshare. Sorenson Spark and On2 VP6 are the winners, and h264 is tiny, almost vanishing by comparison.
Could you back that up with some examples? Every site I know of that does video streaming uses H.264
Interesting thought, just for comparison I did some research: it seems the PS2 has sold around twice as many as the Wii. I am not sure when the PS2 sales started slowing down, but there is definitely room for more Wii sales.
I disagree with this Michael Pachter guy. Announcing a Wii 2 would kill sales - everyone stops buying when a better successor is just around the corner.
Seems like it'll be a while yet before they completely saturate the market. I can't wait for the price to drop into the $150 range.
Your solution won't work if the router is compromised. :/
In other words, it is true file sharers are leeches on society who take without giving back, but they aren't the ones who caused the problems in the recording industry. The industry brought it on themselves.
In other words, they don't matter. Don't factor into the equation. Ignore them.
P.S. I've seen articles suggesting that a pirated copy (of a game) is worth about 1/20th of a legit copy, because of word of mouth. ;)
But if the game sucks, then the word of mouth will hurt rather than help, which leaves you with... Spore.
Universal Iron rule of the Internet: Everyone would be happy to pay for X, but they're only willing to pay half of what's being asked. Songs are a buck? 50c please. Netflix is $10 a month? I'll only pay $5 a month, and only if there's a bigger selection. An iPad will be $999? Well I'd happily pay $500, and only if it isn't crippled with Apple's retard-o-platform!
Well that's easy to solve.
GOLD membership! $5/mo, 50% off every episode!
Or you can opt for the $0/mo regular membership.
Paying $2/epsiode is not cheap. I would pay $1 for an hour long show (42 minutes in reality) as long as it is commercial free. IF you try to sell me commercials, forget it! 30 minute shows I would pay $.50-$.75, but again, only for a commercial free version.
I'd pay $0.05/ep, and if I liked it, I'd donate anywhere from $0.10 to $0.50 (this is to-stream; not to-own)
But there's no way I'm paying for those lame recap episodes that shows are constantly pumping out.
I think a pay-to-stream service could be quite popular. If it were like Hulu, minus the ads, but with credit linked to your account... you could transfer money for free to content producers. Some episodes could be free (first few of a season), while others could have a mandated donation of $0.05 or $0.10. New episodes could be $0.50 for two weeks, then drop to $0.10 - or whatever the content producer decides. The trick is making it easy to pay - you do that by letting the payment rules be configured on your account page, and then just displaying a small notice when payment is automatically sent.
To take the fear out of paying for lemon episodes, the funds (or points) could be transfered part-way through viewing... or perhaps there could be an easy way to undo a donation, even if you fully watched a show. (but only within 24h) This could be used to indicate to producers which episodes didn't go over so well - and to maintain relevance, it could factor in the viewer's behaviour, to filter out viewers that already donate very little (or nothing) on average. A trend engine like that would be valuable for figuring out what viewers want to see, what they want to pay for, and how much they're willing to pay.
The content producers could have access to all sorts of statistics, including average donations per episode/show.
And the viewers could get other perks - perhaps if they donated enough (highest 10%), the system could offer them bonus episodes as thanks - episodes normally reserved for DVD extras.
And finally, it all comes down to what the producer is comfortable with. If a show or movie was expensive to create, they could choose not to stream it to viewers with an average donation of $0.00 - or even one less than $0.20. But rather than cut all those viewers out, the intuitive Hulu-like UI could simply display an information popup, requesting payment. Click okay, and $0.20 is deducted, then it starts streaming. For expensive just-released films (like Avatar?), the producer could scale this to whatever they feel is fair ($2.00?), and track everything with the statistics/trend-engine.
There's lots of ways for a pay-to-stream service to evolve. The trick is making everything very, very convenient. Convenience when it comes to viewing, payment, and statistics, increases the perceived value for both viewers and producers. In the end, many viewers will want to support the shows and episodes that they like, especially since it will lead to better TV.
Paying $2/epsiode is not cheap. I would pay $1 for an hour long show (42 minutes in reality) as long as it is commercial free.
One of the major benefits of a system like this, is it lets you pay $1 for a show you like, $2 for a show you really like, and at the same time, I pay $0.10. (unless I choose to donate more) The current system has us paying $0 + $0
See what I'm getting at?
Firefox has better compatibility than IE6, now. :/
I disagree. Then again, I also tend to multitask - watch TV, fiddle with my phone, work on the other computer, etc...
I was referring to Crysis only. The link proves that SSDs have massive benefits for some tasks, even if you can't notice it easily by counting in your head. ;)
I have heard that alt+tabbing back into games can drop from 10 seconds to 2. That's a 500% performance increase, with even a modest SSD. Okay, it's only 8 seconds, so it hardly matters - but do try to understand what the numbers reflect. SSDs are very powerful underneath.
Most of my games utilize 2-4 cores, so I'd probably get a responsiveness improvement from having six. I understand what you're getting at, though. One of my favourite games is TF2, and I doubt I'll ever see it using more than 2.
- Hoshi's universal translator from Enterprise. It can hear a few phrases in an alien language and then be able to translate back a response that includes words and semantic structures it never heard yet in that language. Note that it didn't even need to be told a translation for that original sample. It could just hear "bbzzt klick klickety-klick hrr bzzt" in some insectoid language and just figure out what it means and, for that matter, what the whole rest of the language is like.
I believe we'll eventually have translators like that - but they won't function like that.
The translator will passively scan the nearby creature's brain, examine and interpret the structure, then crunch numbers until it can guess the language, perhaps cross referencing with the creature's memories, etc. until it has a profile built.
Of course, to get there we'll need a revolutionary speed increase, rather than evolutionary. Something that kicks us from ghz to thz or beyond. Building a profile of a creature's mind would probably take hundreds of terabytes of RAM, too.
But if we look at the speed and capacity increases over the past 40 years... maybe, just maybe!
There's plenty of unfixed bugs listed for the betas and RC, which make it into the final release.
Correcting them before release wouldn't hurt, even if more are found afterwards.
PCI-X video card
It's PCIe - not PCI-X
If you didn't notice an improvement from a quad, then you aren't using your computer fully. That's the same reason lots of people can get away with using netbooks. I personally noticed a big responsiveness improvement when I went from dual-core to quad, and encoding times dropped quite a bit. I suspect your computer usage is somewhere in the middle.
Don't forget to prioritize I/O. A cheap AMD Quad + SSD will give incredible responsiveness compared to a more expensive Quad + regular HDD. Or if you need tons of space, like me, go for multiple HDDs. Those spinning things are still the slowest component that we have to deal with... so speeding them up is more important than a 10% faster CPU or better RAM.
Some games also benefit immensely from faster I/O. Check out the Min and Average FPS for Crysis: http://www.anandtech.com/show/2614/14
SSD is very playable, but HDD isn't really.
True, true.
What I was getting at was the PS3 has quite a beefy processor for the price. For an underfunded department, it's a very cheap alternative. Great if your problem fits in the available RAM.
And if it doesn't, and RAM is your only factor, there's x86 Tyan boards with 48-96GB of RAM. They barely cost $1000.
I think like most mainframes, I/O is probably a big part of supercomputers - though not often mentioned. It's easier to throw the amount of RAM or raw core count around.
Accurate speech recognition is easy. We know how to do it - we just don't have the speed required, yet.
It'd be a lot faster if we had a database of all the variations in how we speak most syllables - but I suspect such a database would easily surpass 4GB, for a single person. Add in the differences between people, and you'd have hundreds of terabytes of slightly different ways to pronounce things - for a single language. Actually, dialects can vary a lot, so it's probably far far higher.
Once we have such a database, perhaps some enterprising company will distill all that data into something easier to process? It happened for video - what was once stored analog, or close to, is now stored digital in formats like H.264, often getting 1:50 or 1:100 compression, with no perceivable quality loss. Once our desktops have 256GB of RAM, and 256-core CPUs, someone or some company will take that data, analyze it, and output something that can do accurate speech recognition with a meagre amount of RAM and CPU time. Perhaps 4-8 cores, and 2-4GB of RAM.
And then we'll all wonder why we didn't have perfect speech recognition in 2010, forgetting all the processing power required to distill that audio data into something meaningful.
Sounds like it worked out well for you.
It doesn't for everyone, unfortunately.
The Slashdot Trolls all agree; Ubuntu is the worst OS ever made, and only caters to retards!
Any time I discuss any OS, I get modded informative/insightful, then modded down as troll.
I think it must be my writing style. I post facts and harsh criticisms about all operating systems and their feature regressions. I must be a troll.
Because of the ease of use/management.
Ubuntu was - very briefly - a drop-in replacement for XP, which almost anyone could figure out.
Now Canonical is making some silly design choices that are pushing away from that. People are getting confused, rather like the XP->Vista switch.
I'm considering migrating from Ubuntu to Debian stable for my NAS.
9.10 was pretty funny. It thought my 3114 PCI card was an nVidia RAID Array. It only took them a few weeks to get that fixed, but I still reverted to 9.04, and it's got me thinking about Debian instead.
The thing about Ubuntu and its sister distros is... don't be an early adopter.
Whenever a new release comes out, it's packed with hundreds of strange bugs. The forums get flooded by people with all sorts of issues. Then within a few months, 95% of them are fixed.
The best time to hop on 10.04 will be 2 months from now. Ubuntu is firm about their 6 month releases, but it'd really work better if they went with 8 months. Do a feature freeze after 4, then spend 4 getting it rock solid stable, and doing proper regression testing.
THAT would make me happy. ;)
We're now quoting android development costs at 4x's that of the iPhone because of QA.
I know someone that's in the exact same line of work. (iPhone and Android apps) His company quotes 2.5x the cost, so while the issue isn't as bad as you make it sound, it is an issue.
On the other hand, how much bigger of an install base do you get? There's more Android phones out there - they're just fragmented a lot more than iPhones are.
His company is migrating to some framework that apparently targets iPhone, Android (All), and I think either WinCE 6 or WebOS - can't remember which. It reduces dev time a lot, at the cost of some battery life - but most clients will take that tradeoff to save money.
...until I can get a decent (120GB+) sized SSD that doesn't cost as much as a new video card?
2012.
In late 2010/2011 we'll probably see the old Indilinx stock being sold off for cheap. People will jump on it.
Then by 2012 once all the new processes for NAND and controller fabrication have matured, and yields are very high, we'll see controllers like the Sandforce becoming affordable - truly decent performance, if I ever saw any. ;)
I suspect that sometime in 2012 Sandforce SSDs will hit 100GB for $100, or 200GB for $200. I might be wrong, but that's my guess.
I do not think many SSDs will drop into the sub-$80 range. In 2008 we had $70 32GB Patriot WARPs. Now we have $120 30/32GB SSDs - far faster, but more expensive per GB. NAND will drop in price shortly, but not enough to put SSDs in the sub-$50 range. For that price range, we'll have to stick with SDHC/SDXC, and HDDs.
New controllers, please. Buying new ones for every patient is probably cheaper than sterilizing the old ones.
And some enterprising individual could make money selling the used ones on eBay.
$2000? Try $10k.
But it will require a barrage of tests to get certified as medical equipment.
Interestingly, we see the same thing in other fields - like supercomputers.
72 core system - Price $23,695.00
I wouldn't hazard a guess on what their 5832 core system costs - but a few PS3's would be far far cheaper. (though perhaps not an option anymore)