Re: Off-Topic: The Machine
On Wed, Jun 25, 2014 at 3:52 PM, Steven D'Aprano st...@pearwood.info wrote: Taking it with a generous grain of salt is one thing, but outright rejecting it is a bit harsh. I understand that HP has actually demonstrated the Machine, so unless they faked the demo, the basic facts are probably more-or-less correct. Like all benchmarks used in advertising, it's going to focus on what the machine does well, regardless of how closely that parallels real-world usage. Legit ones attempt to ensure that there's at least some correlation, but even then, it's impossible to be totally fair. So more-or-less correct may be true, but I still take *all* such benchmarks with the aforementioned salt. Also, mobile phones don't waste most of their power doing calculating and handling terabytes of data, but the RF and display consumes the most of power. Therefore, even if you could scale the CPU down your phone would still not go 2-3 months on a single charge. Fair point. But given how much smart phones get used for playing games these days, I think the savings would still be considerable. Plus, most of computing is just doing the same thing over and over again. The improvements done to the CPU might well be able to be applied, in a different form, to other parts of the device. Sure, the screen has to emit light, which costs power; but if computation is cheap enough, it might be possible to calculate exactly how much light is falling on the screen, and back down the brightness automatically when you move into shadow. Engineering is generally about trading one resource for another, so gains in one area can result in gains in others too. Of course, it's always possible for beautiful engineering to be destroyed by stupid politicking. But here's hoping. ChrisA -- https://mail.python.org/mailman/listinfo/python-list
Re: Off-Topic: The Machine
Steven D'Aprano wrote: Heh, yes, it's a puff-piece, based on HP's publicity, not an in-depth review. Considering that The Machine isn't publicly available yet, that's hardly surprising. There's a talk here that goes into a bit more detail, although still not much: https://www.youtube.com/watch?v=JzbMSR9vA-c The basic ideas seem to be: 1) An extremely large number of CPU cores, many of them specialised for particular tasks. 2) A single form of high speed, non-volatile memory, of very large capacity, replacing cache, RAM, disk, flash, etc. 3) A high-speed optical connection between the CPUs and the memory. They claim to be able to retrieve any desired byte out of a petabyte of storage in 250ns. That's nice, but the question that comes to my mind is: What happens when a zillion cores are all competing for high-speed access to that memory? -- Greg -- https://mail.python.org/mailman/listinfo/python-list
Re: Off-Topic: The Machine
On Wed, Jun 25, 2014 at 5:31 PM, Gregory Ewing greg.ew...@canterbury.ac.nz wrote: That's nice, but the question that comes to my mind is: What happens when a zillion cores are all competing for high-speed access to that memory? And what happens if some of those cores are corrupt? Can you initiate a core transfer? Will there be a stalemate? Are there any trained Stalemate Resolution Associates around? What if there's a fire in the Stalemate Resolution Annex? Seriously, though, how would you go about debugging a system so massively parallel? If most of the CPU cores are special-purpose, it's going to require a whole new method of testing. I can just imagine configure scripts having to be taught to cope with a new form of CPU bug (in the same way that some still test for the Pentium division bug). ChrisA -- https://mail.python.org/mailman/listinfo/python-list
Re: Off-Topic: The Machine
On 24.06.2014 03:23, Steven D'Aprano wrote: http://www.iflscience.com/technology/new-type-computer-capable- calculating-640tbs-data-one-billionth-second-could Relevance: The Machine uses *eighty times less power* for the same amount of computing power as conventional architectures. If they could shrink it down to a mobile phone, your phone might last 2-3 months on a single charge. The article is highly unscientific and unspecific. It does not elaborate on what it means to calculate a terabyte of data nor does it specify what it means to handle a petabyte of data. They're terms used by ignorants, written for ignorants. So all in all I think it's safe to discard the article. Also, mobile phones don't waste most of their power doing calculating and handling terabytes of data, but the RF and display consumes the most of power. Therefore, even if you could scale the CPU down your phone would still not go 2-3 months on a single charge. Cheers, Joe -- Wo hattest Du das Beben nochmal GENAU vorhergesagt? Zumindest nicht öffentlich! Ah, der neueste und bis heute genialste Streich unsere großen Kosmologen: Die Geheim-Vorhersage. - Karl Kaos über Rüdiger Thomas in dsa hidbv3$om2$1...@speranza.aioe.org -- https://mail.python.org/mailman/listinfo/python-list
Re: Off-Topic: The Machine
On Tue, 24 Jun 2014 13:06:56 +0200, Johannes Bauer wrote: On 24.06.2014 03:23, Steven D'Aprano wrote: http://www.iflscience.com/technology/new-type-computer-capable- calculating-640tbs-data-one-billionth-second-could Relevance: The Machine uses *eighty times less power* for the same amount of computing power as conventional architectures. If they could shrink it down to a mobile phone, your phone might last 2-3 months on a single charge. The article is highly unscientific and unspecific. It does not elaborate on what it means to calculate a terabyte of data nor does it specify what it means to handle a petabyte of data. They're terms used by ignorants, written for ignorants. Heh, yes, it's a puff-piece, based on HP's publicity, not an in-depth review. Considering that The Machine isn't publicly available yet, that's hardly surprising. So all in all I think it's safe to discard the article. Taking it with a generous grain of salt is one thing, but outright rejecting it is a bit harsh. I understand that HP has actually demonstrated the Machine, so unless they faked the demo, the basic facts are probably more-or-less correct. Also, mobile phones don't waste most of their power doing calculating and handling terabytes of data, but the RF and display consumes the most of power. Therefore, even if you could scale the CPU down your phone would still not go 2-3 months on a single charge. Fair point. But given how much smart phones get used for playing games these days, I think the savings would still be considerable. -- Steven -- https://mail.python.org/mailman/listinfo/python-list
Off-Topic: The Machine
Following up on an earlier thread which started as a discussion on Apple's new language Swift and (d)evolved into a discussion about energy efficiency of computers, I came across this announcement of a new type of computer architecture invented by HP: The Machine. http://www.iflscience.com/technology/new-type-computer-capable- calculating-640tbs-data-one-billionth-second-could Relevance: The Machine uses *eighty times less power* for the same amount of computing power as conventional architectures. If they could shrink it down to a mobile phone, your phone might last 2-3 months on a single charge. -- Steven -- https://mail.python.org/mailman/listinfo/python-list