The last AMD system I built is quieter and uses about an order of magnitude less power than my PS3 (~30W idle; about the same as a friend's Atom board manages).
Re:Zsh has had these features for years
on
BASH 4.0 Released
·
· Score: 1
Ah, bit more sensible if doxygen is configured WITHOUT_DOXYWIZARD, ta.
Re:Zsh has had these features for years
on
BASH 4.0 Released
·
· Score: 2, Interesting
PAE has driver issues, still doesn't allow individual processes to use more than 4GB (in a way anyone is actually going to bother with), and is a hideous hack reminiscent of the windowing extensions used in the days when the 640k barrier was a concern.
Just 4 Gb of RAM, a 32-bit address, and make it as fast as you can. Forget about that 64-bit bullshit, I'm not running the Social Security database
Um. You can't actually have 4GB of memory with 32bit address space; anywhere between 0.5-2G is going to be eaten up by PCI address space and video memory. On top of that, you can only actually use a maximum of 3GB in a single process; more often 2GB. Painfully limited now; after the years it'll take to develop a brand new x86 it'll be utterly pathetic.
In fact, given the current sweet spot for decent desktop machines is 4-6GB, it's pathetic now. Why would anyone develop such a chip for the 2011 market? It's not like limiting yourself to 32bit is going to magically make it faster or make the cores much smaller.
To go with that, let's have some thousands of cores for number crunching. Mega cores, giga cores, you can never have enough cores for number crunching. But these cores need not have 64-bit capability
Uhm, right. Yes, we have a use for loads of cores too, but if you tell us we can't plug more than 4GB into it, we're going to laugh at you and go back to our 64GB T5240. And yes, we need fast 64bit integer operations too, thanks.
I play my PS3 on the same display I play my PC games on, so it's not really that silly. For you maybe, if you're playing sat 7ft away from a 40" TV, not for me playing 3ft away from a 30" monitor.
Yup; for example, GTA4 on the PS3 looks extra-blurry compared with the 360 because it's actually upscaled from 1120x630. And while the 360 will manage 720p, it'll just upscale for 1080p.
Key word being "slight". No gamer is going to notice the few percent performance reduction ECC introduces. I play a lot of games and I sure prefer having more piece of mind over the safety of my data than getting 122FPS instead of 118FPS in Left 4 Dead. If I *do* turn retarded, I can always turn it off.
Yes, they're mysteriously changing the solder they use with G92's and G94's which happens to be the same as that they use for mobile chips. I wonder why.
nVidia and their partners are understandably not particularly interested in sharing what failure rates they're seeing, but things like this certainly match my experience; I've never had a graphics card go bad, and when I hear of others with failures it's typically the result of overclocking, or something like a fan failing.
Now my 8800GTS 512 goes bad after 7 months, a friend with a different make 8800GTS 512 fails after 6 months, and forums are kicking up a fuss with similar failures. Not exactly rock solid evidence, but enough for me to avoid nVidia for the time being.
SVK has issues with revision numbers because they're the native way SVN indexes revisions by; doesn't SVK provide something more global?
bzr uses chained hashing just like every other DVCS; your numbers can get out of sync with upstream if you've committed different things locally, but in that case your repository's diverged anyway and you'll either have to merge or rebase it to get back in sync; the merge changes your revisions to dotted form, 534.1.1 or so, while rebase moves the revisions to the end.
I bought an ATI when my 8800GTS 512 died; I didn't want to play the lottery as to whether the replacement would have the same manufacturing defects or not.
nVidia are going to have to do something pretty special to attract me back after that; putting two of their power hungry barely-fabricatable huge monolithic GPUs on a single card just isn't it.
ATI have a free GPU-accelerated video encoder which supports a range of formats. nVidia have one too, but it only supports H.264 and with a smaller range of options even for that, and it's not even free to use.
Not to mention a good selection of AMD boards support ECC memory (which adds about £4 to the £37 of 4GB DDR2-800) all across the price range; want a Core i7 with ECC support? That'll be £250 for the motherboard and £120 for 3GB of DDR3. And good luck finding a Core 2 board with ECC support.
You might say "meh, why do you need ECC?"; I think it's quite important when you can fit 12-16GB in a machine, and I'd rather the first sign of a memory problem be a MCE noted in my system log rather than repeated crashes or silent data loss.
Given that I can have it with AMD for an extra £4 with more than adequate performance, while I can barely even *find* it with the already more expensive Intel, well...
If he and his team are experienced with Ruby, why would they switch not just frameworks, but language? Python's not especially faster, it doesn't have a particularly better threading model (in fact, once you factor in JRuby, it's much worse), and the language itself isn't going to do anything to reduce "constriction", especially if they're unfamiliar with it.
If it's Rails you find "... constricting" (whatever that means), there are at least a dozen other Ruby frameworks out there, which would be rather easier to transplant existing code to.
But no, sure, your little 2 line "Rails suxxx! use a completely different language and my favourite framework for no reason I'll actually explain" sure does deserve to be +5 Informative.
Not had any problem with ATI's drivers myself; they feel pretty much like nVidia's, only with a very slightly less awful control panel app. I've been all-nVidia since the GeForce 256, and switching to a HD4870 has been about the least painful hardware change ever; everything works, it's a fair bit faster, and new drivers keep coming out that supposedly make things even faster.
Crysis is the exception rather than the rule, though; I have a system you could build for about $600 and Crysis is pretty much the only game I can't manage at 2560*1600. No great loss.
Regardless, it runs fine at a more modest resolution, especially Warhead which is somewhat better optimized. Dead Space, Far Cry 2, PoP, Left 4 Dead, The Witcher, ArmA, BioShock, Stalker; all quite happy at 2560*1600 and typically with maximum settings (though I don't usually bother with AA).
Which is, of course, the entire point of the article; why spend $2000+ when $600 can get you something with 90% of the utility?
Slashdot Mirror
The last AMD system I built is quieter and uses about an order of magnitude less power than my PS3 (~30W idle; about the same as a friend's Atom board manages).
Ah, bit more sensible if doxygen is configured WITHOUT_DOXYWIZARD, ta.
A shell wants to suck in half of qt? :/
4GB of DDR2-800 costs £35. Most people going for higher end kit with Core i7 get 6GB.
PAE has driver issues, still doesn't allow individual processes to use more than 4GB (in a way anyone is actually going to bother with), and is a hideous hack reminiscent of the windowing extensions used in the days when the 640k barrier was a concern.
Just 4 Gb of RAM, a 32-bit address, and make it as fast as you can. Forget about that 64-bit bullshit, I'm not running the Social Security database
Um. You can't actually have 4GB of memory with 32bit address space; anywhere between 0.5-2G is going to be eaten up by PCI address space and video memory. On top of that, you can only actually use a maximum of 3GB in a single process; more often 2GB. Painfully limited now; after the years it'll take to develop a brand new x86 it'll be utterly pathetic.
In fact, given the current sweet spot for decent desktop machines is 4-6GB, it's pathetic now. Why would anyone develop such a chip for the 2011 market? It's not like limiting yourself to 32bit is going to magically make it faster or make the cores much smaller.
To go with that, let's have some thousands of cores for number crunching. Mega cores, giga cores, you can never have enough cores for number crunching. But these cores need not have 64-bit capability
Uhm, right. Yes, we have a use for loads of cores too, but if you tell us we can't plug more than 4GB into it, we're going to laugh at you and go back to our 64GB T5240. And yes, we need fast 64bit integer operations too, thanks.
Wow, what does line noise look like in your reality?
Nope, I'm pretty sure neither of those look like line noise.
You can use C with JRuby too, using the exact same interface as MRI and Rubinius.
Um, it's pretty competitive against Python and PHP. I guess they're pretty much doomed against Smalltalk and JS too? Oh, wait..
I love Talisker, nice and peaty. Balvenie is awesome if you like something sweeter and more drinkable.
I play my PS3 on the same display I play my PC games on, so it's not really that silly. For you maybe, if you're playing sat 7ft away from a 40" TV, not for me playing 3ft away from a 30" monitor.
Yup; for example, GTA4 on the PS3 looks extra-blurry compared with the 360 because it's actually upscaled from 1120x630. And while the 360 will manage 720p, it'll just upscale for 1080p.
Intel's new SSD's have MFT-style stuff on the controllers; they have very fast random writes.
In contrast, the Intel SSD does about 8,500 4kB random writes per second.
Key word being "slight". No gamer is going to notice the few percent performance reduction ECC introduces. I play a lot of games and I sure prefer having more piece of mind over the safety of my data than getting 122FPS instead of 118FPS in Left 4 Dead. If I *do* turn retarded, I can always turn it off.
Yes, they're mysteriously changing the solder they use with G92's and G94's which happens to be the same as that they use for mobile chips. I wonder why.
nVidia and their partners are understandably not particularly interested in sharing what failure rates they're seeing, but things like this certainly match my experience; I've never had a graphics card go bad, and when I hear of others with failures it's typically the result of overclocking, or something like a fan failing.
Now my 8800GTS 512 goes bad after 7 months, a friend with a different make 8800GTS 512 fails after 6 months, and forums are kicking up a fuss with similar failures. Not exactly rock solid evidence, but enough for me to avoid nVidia for the time being.
SVK has issues with revision numbers because they're the native way SVN indexes revisions by; doesn't SVK provide something more global?
bzr uses chained hashing just like every other DVCS; your numbers can get out of sync with upstream if you've committed different things locally, but in that case your repository's diverged anyway and you'll either have to merge or rebase it to get back in sync; the merge changes your revisions to dotted form, 534.1.1 or so, while rebase moves the revisions to the end.
Nope, they all had the defective solder.
From what I've read, the current Phenom II's are limited to DDR2 and AM2+; the AM2+/AM3 DDR2/DDR3 supporting ones are due next month.
I bought an ATI when my 8800GTS 512 died; I didn't want to play the lottery as to whether the replacement would have the same manufacturing defects or not.
nVidia are going to have to do something pretty special to attract me back after that; putting two of their power hungry barely-fabricatable huge monolithic GPUs on a single card just isn't it.
ATI have a free GPU-accelerated video encoder which supports a range of formats. nVidia have one too, but it only supports H.264 and with a smaller range of options even for that, and it's not even free to use.
I wonder if the backwards compatability will work both ways - could an AM2+ processor be used in an AM3 motherboard?
Nope, since they lack the DDR3 support in their memory controller.
Not to mention a good selection of AMD boards support ECC memory (which adds about £4 to the £37 of 4GB DDR2-800) all across the price range; want a Core i7 with ECC support? That'll be £250 for the motherboard and £120 for 3GB of DDR3. And good luck finding a Core 2 board with ECC support.
You might say "meh, why do you need ECC?"; I think it's quite important when you can fit 12-16GB in a machine, and I'd rather the first sign of a memory problem be a MCE noted in my system log rather than repeated crashes or silent data loss.
Given that I can have it with AMD for an extra £4 with more than adequate performance, while I can barely even *find* it with the already more expensive Intel, well...
If he and his team are experienced with Ruby, why would they switch not just frameworks, but language? Python's not especially faster, it doesn't have a particularly better threading model (in fact, once you factor in JRuby, it's much worse), and the language itself isn't going to do anything to reduce "constriction", especially if they're unfamiliar with it.
If it's Rails you find "... constricting" (whatever that means), there are at least a dozen other Ruby frameworks out there, which would be rather easier to transplant existing code to.
But no, sure, your little 2 line "Rails suxxx! use a completely different language and my favourite framework for no reason I'll actually explain" sure does deserve to be +5 Informative.
Not had any problem with ATI's drivers myself; they feel pretty much like nVidia's, only with a very slightly less awful control panel app. I've been all-nVidia since the GeForce 256, and switching to a HD4870 has been about the least painful hardware change ever; everything works, it's a fair bit faster, and new drivers keep coming out that supposedly make things even faster.
Crysis is the exception rather than the rule, though; I have a system you could build for about $600 and Crysis is pretty much the only game I can't manage at 2560*1600. No great loss.
Regardless, it runs fine at a more modest resolution, especially Warhead which is somewhat better optimized. Dead Space, Far Cry 2, PoP, Left 4 Dead, The Witcher, ArmA, BioShock, Stalker; all quite happy at 2560*1600 and typically with maximum settings (though I don't usually bother with AA).
Which is, of course, the entire point of the article; why spend $2000+ when $600 can get you something with 90% of the utility?