Yowza, thanks NetAvenger and Foolhardy for two very informative posts! At long last I know why pulling stuff back in from swap takes so damn long.
I didn't really get interested in the Linux VM and block layers until fairly recently, but I believe that Linux also keeps pages in swap and RAM as of 2.6 (IIRC there used to be a swappiness parameter to govern how agressively pages were written to swap until someone figured out how to do it automatically). Swap prefetch definitely does this, although less noticeable for me after I splurged on an extra 2GB for my workstation.
Finally I'd like to say I imagine there are few technially minded people here who aren't looking at some of Vista's cooler features with great interest. There's obviously some funky stuff going on under the hood (I've been wondering for years why delayed autostart of daemons hadn't been implemented, at least finally someone is using it mainstream) - just a shame, IMHO, that it has an incredibly frustrating and dare I say it bloated UI bolted to the top of it. Address space randomisation and defaulting NX to "on" should hopefully make for fewer or less feasible exploits.... I know there's some work gone into disk and RAM based prefetching for Linux (heck, even seen some people have logon scripts that cat soon-to-be-needed files so they're more likely to be at least partly cached in RAM) but have yet to see anything mainstream.
P.S. I find CK's patchset utterly superb for desktop use and swap prefetch gave me noticeably better responsiveness (yes, highly subjective and non-benchmarkable I know) when pulling the system back from a large video render or whatever. Just a shame to see Con a little disheartened at failing to get swap prefetch merged into mainline.
Not strictly related, I just like to plug my fave kernel variant;
Con Kolivas maintains a great kernel patchset containing one of his own patches, swap prefetch.
It basically goes that, on a Linux system, most of you RAM is going to be used up caching stuff or being used for running applications. If you're not doing anything very demanding or you're away from the computer, background processes are going to eat up RAM doing stuff so your running apps get swapped out. As soon as the system drops back to idle again, the previously swapped apps are then copied back to memory so when you get back to your machine it's as if nothing happened.
Not as advanced as Vista's pre-emptive (I think) superfetch stuff, but a start anyway.
P.S. can anyone confirm or deny whether previous versions of NT held stuff in RAM? I've always been a bot confused by the ubiquitous amounts of "free memory" that could be put to better use; is it just that task manager doesn't report that memory as being used?
As others have pointed out, the AMD64 on-die mem controller means that memory access latency is significantly lower, making fetches from memory much faster and obviating the need for huge caches to make up for a cache miss. Again as has been pointed out, benches comparing the 512kB AMD64's to the 1MB AMD64's of the same clock and family typically show no difference, so cache is clearly not a bottleneck (except in the very few maths-eavy benches where the app fits entirely into a 1MB cache but not the 512k one).
Also, Intel's 4-core chips do not share cache; each dual core die has a cache that is shared between two cores so it would be correct to define these 8MB chips as having 2x4MB caches. Potentially big difference there, as an app that was running across all four cores would have to use a slower interconect in order for dies 0 and 1 to talk to dies 2 and 3. It also means that a single threaded process could not use more than 4MB of cache.
...since I imagine most people posting here are UK/Europe residents and already familiar with the UK's ID card debacle; the government has basically told us all we're getting biometric ID cards hooked into a national DB, at our own expense, whilst also costing the taxpayer billions in the contruction of the backend system. Despite general uproar from anyone who knows anything about privacy and/or implementing such things on a nationwide scale (not to mention the people pointing out the flaws in the proposed implementations), the government has basically steamrolled the whole thing through. This petition is yet another thing in a long line of criticism that has been judiciously ignored by the bods in power. There have been plenty of god-awful public investigations down by the government that usually go something along the lines of "would you support technology that would prevent terrorist attacks and stop paedophiles killing your children?", during which everyne answers "yes" so the gov can claim overwhelming public support for national ID cards. UK sites like The Register have been running stories on the ID card disaster-in-waiting for several years now (as I'm sure many of you are aware, UK government has a history of successive failures as far as implementing nationwide IT projects).
Democracy in action, I'm sure. Unfortunately a great mass of the uninformed public have been sufficiently brainwashed to believe that such an ID system will be as near to infallible as you can get, so it's going to take fairly herculean efforts to get other political parties to step away from implementing the same ID scheme. Ignorance is strength. Freedom is slavery.
In slightly related news, More4 (UK digital channel) showed a surprisingly watchable documentary on modern privacy and surveillance issues in a prog called Suspect Nation. I can't find any official sources so I'll reluctantly post a link to a torrent here. Well worth downloading to show to like-minded friends who aren't aware how pervasive some of this technology is.
Disclaimer: I am very much anti ID and even go to the bother of registering my Oyster card (RFID card used for travel on London transport which tracks every journey you make) under a false address, and always pay for them in cash. I've had some people brand me a bit of a conspiraloon, but you don't need to study much history before you find out it's a bad idea to give governments too much information and power, no matter how benevolent the intentions.
...and if someone can point out why XP is utterly incapable of running DX10 when Vista is, or why Company of Heroes refuses to install on Windows 2000 but runs just fine on XP.
Hint: the answer to both questions is firstly a) Payola and b) greedy fucking monopolist bastards.
I've also recently switched to an Intel GFX card for my Myth backend - integrated GMA X3000 in a GigaByte 965G-DS3. Nice board and, in general, nice graphics - after a bit of tinkering getting xorg working with DRI was pretty easy (although fiddling with 915resolution to get my 1680x1050 TFT working at native res was a bit of a pain, but then I guess that's the "attraction" of using Gentoo;)).
However, like Andy Dodd point out there are several glaring omissions in the driver; the biggest one for me is that XvMC doesn't work (meaning I can't play back HD video). If Intel were really serious about the Linux market, XvMC would be included in the drivers for the 9xx chipsets, but they're not. I haven't experimented with things like S3TC since I'm not much of a games buff (and I have a dual-boot workstation with an nVidia card in it anyway), but I wonder how much stuff Intel will leave out of their open source drivers in the future - if not throug lazyness, then by being strong-armed by the mafiaa's of this world.
What with all the fandango about HDCP and protected audio/video paths I'm fairly certain that HD will never be officially supported under Linux, although I'm not too versed in how the protection scheme works (please correct my assumptions if I'm wrong) - I'm guessing that that video driver has to auth against the chipset to verify that it's "protected", whcih then auths against the monitor to enable HD output for stuff like HDDVD. Since I can't see any manufacturer in their right mind open-sourcing the routine for key exchange (since it's security through obscurity if my understanding of priv/public key crypto is up to snuff), then to me it seems the options are:
1) Require a closed source driver, and mandate that video/audio playback mechanisms also support the same key exchange system (which would, again, have to be closed source for fear of teh crackerz) 2) Just not bother supporting HD stuff at all, requiring AACS and the rest to be cracked wide open before they're playable on Linux 3) Just not bother supporting Linux
Can anyone more adept with the HDCP situation (or debacle as I like to think of it) explain how things like open source video drivers are able to exist in an HDCP environment?
I relaise I've gone slightly OT, but it's taken me the best part of an hour to write this post as I'm at work and everyone wants something fixed...
As an aside, I wonder if it's possible to create a Linux bootable CD/DVD that's capable of starting up a la Knoppix, but then running the pre-existing windows install from the HDD in a VT environment...? That could make for an incredible recovery/debugging disc...
Unfortunately, Alan Cox is not a billionaire. If he were I have no doubt that he could retroactively sue aforementioned coroprations and turn a tidy profit settling out of court, but as it is he's just a regular Joe meddling in the big boys' court system.
Unless he's granted a huge wad of cash by someone like RedHat (who I can't see wanting to be involved in a more-political-than-anything-else patent dispute that much), I don't really see this going anywhere, and in the grand scheme of things I don't think that's neccesarily a bad thing. One the one hand it's showing what a stupid mess software patents can be, on the other hand it's using them as a weapon, and this is one of the things I thought the majority of the FOSS community was against...?
Thanks for that - unfirtunately the frequent "minimal" use flags aren't really documented that well and you have to go out of your way to find out what they're for... wish they'd introduce somethign a bit more self-explanatory, e.g. clientonly or suchlike...
That'll teach me not to read the parent post properly... damn thing was marked "informative" when I read it. Good job I made a fool of myself instead of moderating;)
Ditto for my experience. On getting a shiny new desktop system with an almost-top-of-the-line A64 3500, I was excited as hell to try out 64bit Linux.
Indeed, once it was all up and running, I did notice some great speed benefits - OpenSSL simply flies, and many CPU-heavy multimedia apps such as LAME and mplayer showed slight improvments.
But the userland was another story. Since so much near-essential stuff* is still 32bit (thanks, mainly to sloppy coding [hello OOo...!] and closed source stuff [Flash]) you pretty much have to keep an almost entire 32b userland running at the same time as your 64bit DE. I guess I was kinda lucky in that Opera is my fave browser, and it being 32bit only I had no problems with 32bit flash, and it was still fairly quick to load. Starting a 32bit FF when you're running a 64bit KDE installation required me to load all of the 32b GTK libs before firefox could begin to think about executing. Combine that with the (slightly) inflated binary sizes of 64b executables and the higher memory requirements of running applications in a 64b envionment and you end up with a machine that is slower overall.
So, to cut a long story short, marginal improvements in some apps were countered by horrendous usability problems on the desktop.
If you're running a server (or any other machine that doesn't require 32bt niceties), 64bit Linux just rocks.
DISCLAIMER: I relaise that issues with 64b Linux on the desktop aren't Linux's fault, just like XP64's problems with drivers aren't MS's fault either. They're just issues that most would like to avoid
* Yes, I'm aware that Flash can be ignored by many people, but it's a deal breaker for the WAF. Similarly OpenOffice, whilst techically replaceable with things like Koffice that don't hog so many resources, is still the de-facto standard that all other office apps have to measure up to.
Maybe your notepad.exe is a different one from mine, but I can imagine very few bona fide developers (apart from the very newest) using notepad, since IMHO it's unusably bad for all but the simplest of stuff. No syntax highlighting and annoying cursor placement issues being two of the most obvious problems. There are a million and one text editors superior to notepad available for win32 that are just as esy to use fr base functionality.
OTOT, no-one would think of comparing vi/emacs/whatever to notepad. A more valid desription would be, say, kedit and notepad.
Surely a much simpler way of doing things would involve tweaking the package management software to automaticaly install associated -devel packages, with an option somewhere to not install things. This way, people unfamilar with the need for -devel packages don't notice, advanced users can tweak their system to use minimal resources, and the package maintainers don't have to change diddly squat in their (no doubt already complicated) job of packaging up the latest and greatest. IMHO a packging format shoud handle nothing but packages, and all of the clever stuff like dependancies, repositories etc should be left to higher level tools.
I'm a gentoo user (at home), and as much as I love the fiddling I do with it, it's a major pain to have to install the entirity of MySQL just to get MySQL client support, or some other thing for a particular set of headers - it's vastly inefficient with disc space and CPU cycles. I love the way Debian does away with the need for all of these stupid build-time dependencies by splitting the package's various compile-time options into seperate packages - automagic installation of -devel packages would be a boon for any n00b blundering into his first botched compile on Linux.
I love my 6310(i). Many a provider has tried to tempt me away with colour screens, the ability to watch music videos amd to play garbled-sounding MP3's to the annoyance of everyone else on the train. Many service providers have even gotten quite rude when I tell them that I do not want a phone that doesn't last as long as my 6310.
The battery still lasts a week (that's 7 days) standby time after three years of constant use. This includes about 30 mins worth of voice calls and 5-10 texts a day. I bought a spare battery for it two years ago and haven't used it yet. I've dropped the phone onto a concrete floor from a first floor (that's second floor to non-Brits) window. The thing flew to pieces. Put it back together again and it works fine. The screen wasn't even scratched. I've lost count of the number of times it's gotten soaked when I've been caught in the rain or had beer poured all over it. The thing is god damn indestructible. They keys still work brilliantly.
Compare with the Blackberry 8700 my company bought me (a modern equivalent of the 6310, since the 6310 was Nokia's almost-top-of-the-line business-class phone in it's day); on a good week the battery will last 3-4 days between charges, but if you have a game or two of Brickbreaker and surf slashdot a bit on your evening commute (or even just type lots of emails), you can cut that down to 1 day. The keys are too small, but that can be expected. The rocker wheel is already feeling spongy after 6 months. The case is already badly scratched in places after a few not-very-nasty knocks. It doesn't work with Linux either...
My friend bought a 6310 soon after me, but gave it up last year for a RAZR. Which broke. Switched back to the Nokia until he got a new Sony Ericsson something-or-other... which had a defective charging circuit. Switched back to the Nokia until he could find a phone that he liked. Currently has an LG something-or-other which looks very nice, but I'll bet anyone a tenner that next time I see him he'll be using his Nokia:D
You'll pry my 6310 from my cold, dead hands. Sorry Nokia, but I've seen nowt to touch it since.
Here's a thought - would it be possible to create a FUSE module that would be able to present BDB file internals as filesystem objects? True, it'd be slow but at leas you'd have the advantage of being able to directly read and manipulate BDB info. You could probably do the same thing for SQL.
Heck, FUSE might even make it possible to embedded a ReiserFS partition as a file. You'd still have to go through the host filesystem, mind.
Heh, I already have a 3800... and a 4200... I can't believe I was contemplating replacing at least one of them for one of the new AM2 energy efficent varieties (new mobo, new memory, new coolers... ARGH!).
I'm just really tempted by the "fake" Kentsfield quad cores since H264 encoding takes an age, even with both cores flat out, but I'm still not sold on Core2 wholsesale since it's only energy efficient when it's running full tilt. Maybe I'll get a Kentsfield for my workstation and keep my servers (idle 95% of the time) on AMD64 for now... or maybe I'll just die of silicon-induced indecision before then.
Strikes me that Intel is running out of buzzwords! Was the marketing dept. severely depleted in the last round of purges?
THe next 12 months or so are going to be a very interesting time for the CPU world. All Intel needs to do it get their chips' idling power down into the same ballpark as AMD, and AMD need that 65nm process in volume *now*! I've actually been finding myself forcing myself not to look at computer stores and upgrade my workstation because I know that six months down the line there'll be something orders of magnitudes better on the scene...
[i]Practically any website in the UK takes your credit card before giving you access to a promotional offer. Amazon, Ebay, HMV, you name it, they do it.
So either you are an Alien whose veins are filled with acid and that just landed on planet Earth, or you are being horrifcally pedantic.[/i]
I'm not in the habit of paying attention to any promotions because I generally couldn't care less about them. Despite that fact, I'm pretty sure that Amazon, eBay and HMV let you look at stuff online before you're required to sign up and fork over your CC numbers. Why is eMusic being so different? What if I don't want to sign up for the promotion (which I don't) and just want to see if the promotion is worth signing up for?
[i]This site, and the internet in general are full of reviews about emusic, so your nonsense about it being an scam site reflects more about your laziness...[/i]
Wow, who's being an alien now? Why the fuck should I have to do reserach just to browse a fucking catalogue? I've known what eMusic was since about 1 month after it was launched in the US. And no, I wasn't calling it a scam site. I'm saying that to the casual browser (who may not have the slightest idea what it is) is likely to be put off by it asking for their credit card. If I directed you to some random site asking for CC details, you'd say yes I suppose?
[i]...that about a bussiness trying to provide a legitimate service. Because I suppose you did contact them in order for them to change this? Did you? No? Oh well...[/i]
So now I should be emailing them to ask this shit? Sorry, I made my descisions WRT to online music years ago when I signed up with Bleep. No fuss and definitely no arsing around with email to tell them to make their catalogue available online - if a burgeoning online music store omits such an obvious step as making their catalogue available, I doubt a solitary email is going to help. This isn't a bugtraq for OpenOffice, it's a damned [i]shop[/i]. It's not meant to be complicated.
At the end of the day - do I care enough about what I've read about eMusic to jump through all these silly artificial hoops? No, I don't.
I don't know what the eMusic in the states is like, but I have a major issue with the one in the UK in that I have to hand over my credit card details before I even have half a clue what they're selling. All they offer without an account/login is the ability to... er... create an account and log in. Woudl you hand over your credit card on walking into a shop "just in case" you wanted to buy something? Keeping it secret suggests that either their catalogue is crap, they're secretive lock-in merchants (hello AOL) or (dare I say it) a scam site.
Until then, I'll fulfill my music downloading tastes with unencrypted MP3, AAC and FLAC from Bleep, Tunetribe and 4AD.
Please note: I'm not knocking eMusic or legal downloads in general (indeed, I spend about £20 a month buying tunes online compared to £0 on music two years ago). I just don't trust the way their UK store seems to be working. And yes, I did want a chance to try it out.
Completely agree, it's a much overlooked area of CPU reviews. Try finding real-world power consumption figures for the Intel E6300 - not easy. Everyone only seems to give a crap about overclocking the new XTXXLLLLGTSFX2006 these days;)
Many of the reviews I've seen show that the AMD systems consume significantly less power at idle than the equivalent C2D system. Whilst the C2D is pretty much undoubtedly the faster of the two arches, I'm still pretty staggered by the energy efficiency of the AMD64. As anothe poster pointed out, AMD's cherry picked ADD chips (well, the 3800 X2 ADD anyway) consume utterly tiny amounts of power, even on an appallingly stone aged 90nm lith process;) I can't wait for AMD's 65nm to start shipping once their process is all sorted out, since 90nm SOI has worked so well for them. That said, since the ADD series are cherry picked they're expensive and hard to come by - last I heard they were of limited availability in Germany.
Since most computers I own spend 90% of their time idling away at 1-5% load, I'm sticking with my AMD's for the time being - I'm not a gamer (although I do alot of video endocoding and have been contemplating a C2D system for my main workstation) so balls to the wall performance is not my highest priority, and keeping the costs of running my PVR's down is quite important to me.
As an aside, has anyone seen any benches for a Merom chip outside of a laptop? I've been thinking abut using one of these as a new Myth frontend, but the chips are like rocking-horse poo at the moment and I'm not aware of any UK stockists.
Anyway, like I said last time - yay for competition! For the first time in years both CPU companies are releasing some pretty interesting kit. Prices for both chips are incredibly low, and given that you can grab an X2 (i.e. more than enough CPU grunt to run anything quickly, including vista, except the latest games at max settings) for less than £100 makes this a great time for customers.
Anyway, enough rambling, time to drool over more CPU specs...:D
Slashdot Mirror
Yowza, thanks NetAvenger and Foolhardy for two very informative posts! At long last I know why pulling stuff back in from swap takes so damn long.
I didn't really get interested in the Linux VM and block layers until fairly recently, but I believe that Linux also keeps pages in swap and RAM as of 2.6 (IIRC there used to be a swappiness parameter to govern how agressively pages were written to swap until someone figured out how to do it automatically). Swap prefetch definitely does this, although less noticeable for me after I splurged on an extra 2GB for my workstation.
Finally I'd like to say I imagine there are few technially minded people here who aren't looking at some of Vista's cooler features with great interest. There's obviously some funky stuff going on under the hood (I've been wondering for years why delayed autostart of daemons hadn't been implemented, at least finally someone is using it mainstream) - just a shame, IMHO, that it has an incredibly frustrating and dare I say it bloated UI bolted to the top of it. Address space randomisation and defaulting NX to "on" should hopefully make for fewer or less feasible exploits.... I know there's some work gone into disk and RAM based prefetching for Linux (heck, even seen some people have logon scripts that cat soon-to-be-needed files so they're more likely to be at least partly cached in RAM) but have yet to see anything mainstream.
P.S. I find CK's patchset utterly superb for desktop use and swap prefetch gave me noticeably better responsiveness (yes, highly subjective and non-benchmarkable I know) when pulling the system back from a large video render or whatever. Just a shame to see Con a little disheartened at failing to get swap prefetch merged into mainline.
Not strictly related, I just like to plug my fave kernel variant;
Con Kolivas maintains a great kernel patchset containing one of his own patches, swap prefetch.
It basically goes that, on a Linux system, most of you RAM is going to be used up caching stuff or being used for running applications. If you're not doing anything very demanding or you're away from the computer, background processes are going to eat up RAM doing stuff so your running apps get swapped out. As soon as the system drops back to idle again, the previously swapped apps are then copied back to memory so when you get back to your machine it's as if nothing happened.
Not as advanced as Vista's pre-emptive (I think) superfetch stuff, but a start anyway.
P.S. can anyone confirm or deny whether previous versions of NT held stuff in RAM? I've always been a bot confused by the ubiquitous amounts of "free memory" that could be put to better use; is it just that task manager doesn't report that memory as being used?
As others have pointed out, the AMD64 on-die mem controller means that memory access latency is significantly lower, making fetches from memory much faster and obviating the need for huge caches to make up for a cache miss. Again as has been pointed out, benches comparing the 512kB AMD64's to the 1MB AMD64's of the same clock and family typically show no difference, so cache is clearly not a bottleneck (except in the very few maths-eavy benches where the app fits entirely into a 1MB cache but not the 512k one).
Also, Intel's 4-core chips do not share cache; each dual core die has a cache that is shared between two cores so it would be correct to define these 8MB chips as having 2x4MB caches. Potentially big difference there, as an app that was running across all four cores would have to use a slower interconect in order for dies 0 and 1 to talk to dies 2 and 3. It also means that a single threaded process could not use more than 4MB of cache.
...since I imagine most people posting here are UK/Europe residents and already familiar with the UK's ID card debacle; the government has basically told us all we're getting biometric ID cards hooked into a national DB, at our own expense, whilst also costing the taxpayer billions in the contruction of the backend system. Despite general uproar from anyone who knows anything about privacy and/or implementing such things on a nationwide scale (not to mention the people pointing out the flaws in the proposed implementations), the government has basically steamrolled the whole thing through. This petition is yet another thing in a long line of criticism that has been judiciously ignored by the bods in power. There have been plenty of god-awful public investigations down by the government that usually go something along the lines of "would you support technology that would prevent terrorist attacks and stop paedophiles killing your children?", during which everyne answers "yes" so the gov can claim overwhelming public support for national ID cards. UK sites like The Register have been running stories on the ID card disaster-in-waiting for several years now (as I'm sure many of you are aware, UK government has a history of successive failures as far as implementing nationwide IT projects).
Democracy in action, I'm sure. Unfortunately a great mass of the uninformed public have been sufficiently brainwashed to believe that such an ID system will be as near to infallible as you can get, so it's going to take fairly herculean efforts to get other political parties to step away from implementing the same ID scheme. Ignorance is strength. Freedom is slavery.
In slightly related news, More4 (UK digital channel) showed a surprisingly watchable documentary on modern privacy and surveillance issues in a prog called Suspect Nation. I can't find any official sources so I'll reluctantly post a link to a torrent here. Well worth downloading to show to like-minded friends who aren't aware how pervasive some of this technology is.
Disclaimer: I am very much anti ID and even go to the bother of registering my Oyster card (RFID card used for travel on London transport which tracks every journey you make) under a false address, and always pay for them in cash. I've had some people brand me a bit of a conspiraloon, but you don't need to study much history before you find out it's a bad idea to give governments too much information and power, no matter how benevolent the intentions.
...and if someone can point out why XP is utterly incapable of running DX10 when Vista is, or why Company of Heroes refuses to install on Windows 2000 but runs just fine on XP.
Hint: the answer to both questions is firstly a) Payola and b) greedy fucking monopolist bastards.
I've also recently switched to an Intel GFX card for my Myth backend - integrated GMA X3000 in a GigaByte 965G-DS3. Nice board and, in general, nice graphics - after a bit of tinkering getting xorg working with DRI was pretty easy (although fiddling with 915resolution to get my 1680x1050 TFT working at native res was a bit of a pain, but then I guess that's the "attraction" of using Gentoo ;)).
However, like Andy Dodd point out there are several glaring omissions in the driver; the biggest one for me is that XvMC doesn't work (meaning I can't play back HD video). If Intel were really serious about the Linux market, XvMC would be included in the drivers for the 9xx chipsets, but they're not. I haven't experimented with things like S3TC since I'm not much of a games buff (and I have a dual-boot workstation with an nVidia card in it anyway), but I wonder how much stuff Intel will leave out of their open source drivers in the future - if not throug lazyness, then by being strong-armed by the mafiaa's of this world.
What with all the fandango about HDCP and protected audio/video paths I'm fairly certain that HD will never be officially supported under Linux, although I'm not too versed in how the protection scheme works (please correct my assumptions if I'm wrong) - I'm guessing that that video driver has to auth against the chipset to verify that it's "protected", whcih then auths against the monitor to enable HD output for stuff like HDDVD. Since I can't see any manufacturer in their right mind open-sourcing the routine for key exchange (since it's security through obscurity if my understanding of priv/public key crypto is up to snuff), then to me it seems the options are:
1) Require a closed source driver, and mandate that video/audio playback mechanisms also support the same key exchange system (which would, again, have to be closed source for fear of teh crackerz)
2) Just not bother supporting HD stuff at all, requiring AACS and the rest to be cracked wide open before they're playable on Linux
3) Just not bother supporting Linux
Can anyone more adept with the HDCP situation (or debacle as I like to think of it) explain how things like open source video drivers are able to exist in an HDCP environment?
I relaise I've gone slightly OT, but it's taken me the best part of an hour to write this post as I'm at work and everyone wants something fixed...
The problem is that once you've done a ping sweep of the IPv6 network, the first lot of snowflakes have melted (along with the DHCP server).
As an aside, I wonder if it's possible to create a Linux bootable CD/DVD that's capable of starting up a la Knoppix, but then running the pre-existing windows install from the HDD in a VT environment...? That could make for an incredible recovery/debugging disc...
FFS! I've told cybrthng a billion times not to exaggerate!
Unfortunately, Alan Cox is not a billionaire. If he were I have no doubt that he could retroactively sue aforementioned coroprations and turn a tidy profit settling out of court, but as it is he's just a regular Joe meddling in the big boys' court system.
Unless he's granted a huge wad of cash by someone like RedHat (who I can't see wanting to be involved in a more-political-than-anything-else patent dispute that much), I don't really see this going anywhere, and in the grand scheme of things I don't think that's neccesarily a bad thing. One the one hand it's showing what a stupid mess software patents can be, on the other hand it's using them as a weapon, and this is one of the things I thought the majority of the FOSS community was against...?
Thanks for that - unfirtunately the frequent "minimal" use flags aren't really documented that well and you have to go out of your way to find out what they're for... wish they'd introduce somethign a bit more self-explanatory, e.g. clientonly or suchlike...
That'll teach me not to read the parent post properly... damn thing was marked "informative" when I read it. Good job I made a fool of myself instead of moderating ;)
Ditto for my experience. On getting a shiny new desktop system with an almost-top-of-the-line A64 3500, I was excited as hell to try out 64bit Linux.
Indeed, once it was all up and running, I did notice some great speed benefits - OpenSSL simply flies, and many CPU-heavy multimedia apps such as LAME and mplayer showed slight improvments.
But the userland was another story. Since so much near-essential stuff* is still 32bit (thanks, mainly to sloppy coding [hello OOo...!] and closed source stuff [Flash]) you pretty much have to keep an almost entire 32b userland running at the same time as your 64bit DE. I guess I was kinda lucky in that Opera is my fave browser, and it being 32bit only I had no problems with 32bit flash, and it was still fairly quick to load. Starting a 32bit FF when you're running a 64bit KDE installation required me to load all of the 32b GTK libs before firefox could begin to think about executing. Combine that with the (slightly) inflated binary sizes of 64b executables and the higher memory requirements of running applications in a 64b envionment and you end up with a machine that is slower overall.
So, to cut a long story short, marginal improvements in some apps were countered by horrendous usability problems on the desktop.
If you're running a server (or any other machine that doesn't require 32bt niceties), 64bit Linux just rocks.
DISCLAIMER: I relaise that issues with 64b Linux on the desktop aren't Linux's fault, just like XP64's problems with drivers aren't MS's fault either. They're just issues that most would like to avoid
* Yes, I'm aware that Flash can be ignored by many people, but it's a deal breaker for the WAF. Similarly OpenOffice, whilst techically replaceable with things like Koffice that don't hog so many resources, is still the de-facto standard that all other office apps have to measure up to.
Maybe your notepad.exe is a different one from mine, but I can imagine very few bona fide developers (apart from the very newest) using notepad, since IMHO it's unusably bad for all but the simplest of stuff. No syntax highlighting and annoying cursor placement issues being two of the most obvious problems. There are a million and one text editors superior to notepad available for win32 that are just as esy to use fr base functionality.
OTOT, no-one would think of comparing vi/emacs/whatever to notepad. A more valid desription would be, say, kedit and notepad.
Surely a much simpler way of doing things would involve tweaking the package management software to automaticaly install associated -devel packages, with an option somewhere to not install things. This way, people unfamilar with the need for -devel packages don't notice, advanced users can tweak their system to use minimal resources, and the package maintainers don't have to change diddly squat in their (no doubt already complicated) job of packaging up the latest and greatest. IMHO a packging format shoud handle nothing but packages, and all of the clever stuff like dependancies, repositories etc should be left to higher level tools.
I'm a gentoo user (at home), and as much as I love the fiddling I do with it, it's a major pain to have to install the entirity of MySQL just to get MySQL client support, or some other thing for a particular set of headers - it's vastly inefficient with disc space and CPU cycles. I love the way Debian does away with the need for all of these stupid build-time dependencies by splitting the package's various compile-time options into seperate packages - automagic installation of -devel packages would be a boon for any n00b blundering into his first botched compile on Linux.
Wow, I was about to make the exact same post.
:D
I love my 6310(i). Many a provider has tried to tempt me away with colour screens, the ability to watch music videos amd to play garbled-sounding MP3's to the annoyance of everyone else on the train. Many service providers have even gotten quite rude when I tell them that I do not want a phone that doesn't last as long as my 6310.
The battery still lasts a week (that's 7 days) standby time after three years of constant use. This includes about 30 mins worth of voice calls and 5-10 texts a day. I bought a spare battery for it two years ago and haven't used it yet.
I've dropped the phone onto a concrete floor from a first floor (that's second floor to non-Brits) window. The thing flew to pieces. Put it back together again and it works fine. The screen wasn't even scratched.
I've lost count of the number of times it's gotten soaked when I've been caught in the rain or had beer poured all over it. The thing is god damn indestructible. They keys still work brilliantly.
Compare with the Blackberry 8700 my company bought me (a modern equivalent of the 6310, since the 6310 was Nokia's almost-top-of-the-line business-class phone in it's day); on a good week the battery will last 3-4 days between charges, but if you have a game or two of Brickbreaker and surf slashdot a bit on your evening commute (or even just type lots of emails), you can cut that down to 1 day. The keys are too small, but that can be expected. The rocker wheel is already feeling spongy after 6 months. The case is already badly scratched in places after a few not-very-nasty knocks. It doesn't work with Linux either...
My friend bought a 6310 soon after me, but gave it up last year for a RAZR. Which broke. Switched back to the Nokia until he got a new Sony Ericsson something-or-other... which had a defective charging circuit. Switched back to the Nokia until he could find a phone that he liked. Currently has an LG something-or-other which looks very nice, but I'll bet anyone a tenner that next time I see him he'll be using his Nokia
You'll pry my 6310 from my cold, dead hands. Sorry Nokia, but I've seen nowt to touch it since.
Soon the number of cores in my desktop machine will surpass the number of blades in my shaving razor.
Mateorabi's law, anyone?
Here's a thought - would it be possible to create a FUSE module that would be able to present BDB file internals as filesystem objects? True, it'd be slow but at leas you'd have the advantage of being able to directly read and manipulate BDB info. You could probably do the same thing for SQL.
Heck, FUSE might even make it possible to embedded a ReiserFS partition as a file. You'd still have to go through the host filesystem, mind.
You don't have to imagine, but you'll be in breach of the EULA if you don't.
Heh, I already have a 3800... and a 4200... I can't believe I was contemplating replacing at least one of them for one of the new AM2 energy efficent varieties (new mobo, new memory, new coolers... ARGH!).
I'm just really tempted by the "fake" Kentsfield quad cores since H264 encoding takes an age, even with both cores flat out, but I'm still not sold on Core2 wholsesale since it's only energy efficient when it's running full tilt. Maybe I'll get a Kentsfield for my workstation and keep my servers (idle 95% of the time) on AMD64 for now... or maybe I'll just die of silicon-induced indecision before then.
Penryn new instructions = PNI = SSE4
Prescott new instructions = PNI = SSE3
Therefore SSE3 = SSE4...?
Strikes me that Intel is running out of buzzwords! Was the marketing dept. severely depleted in the last round of purges?
THe next 12 months or so are going to be a very interesting time for the CPU world. All Intel needs to do it get their chips' idling power down into the same ballpark as AMD, and AMD need that 65nm process in volume *now*! I've actually been finding myself forcing myself not to look at computer stores and upgrade my workstation because I know that six months down the line there'll be something orders of magnitudes better on the scene...
[i]Practically any website in the UK takes your credit card before giving you access to a promotional offer. Amazon, Ebay, HMV, you name it, they do it.
So either you are an Alien whose veins are filled with acid and that just landed on planet Earth, or you are being horrifcally pedantic.[/i]
I'm not in the habit of paying attention to any promotions because I generally couldn't care less about them. Despite that fact, I'm pretty sure that Amazon, eBay and HMV let you look at stuff online before you're required to sign up and fork over your CC numbers. Why is eMusic being so different? What if I don't want to sign up for the promotion (which I don't) and just want to see if the promotion is worth signing up for?
[i]This site, and the internet in general are full of reviews about emusic, so your nonsense about it being an scam site reflects more about your laziness...[/i]
Wow, who's being an alien now? Why the fuck should I have to do reserach just to browse a fucking catalogue? I've known what eMusic was since about 1 month after it was launched in the US. And no, I wasn't calling it a scam site. I'm saying that to the casual browser (who may not have the slightest idea what it is) is likely to be put off by it asking for their credit card. If I directed you to some random site asking for CC details, you'd say yes I suppose?
[i]...that about a bussiness trying to provide a legitimate service. Because I suppose you did contact them in order for them to change this? Did you? No? Oh well...[/i]
So now I should be emailing them to ask this shit? Sorry, I made my descisions WRT to online music years ago when I signed up with Bleep. No fuss and definitely no arsing around with email to tell them to make their catalogue available online - if a burgeoning online music store omits such an obvious step as making their catalogue available, I doubt a solitary email is going to help. This isn't a bugtraq for OpenOffice, it's a damned [i]shop[/i]. It's not meant to be complicated.
At the end of the day - do I care enough about what I've read about eMusic to jump through all these silly artificial hoops? No, I don't.
I don't know what the eMusic in the states is like, but I have a major issue with the one in the UK in that I have to hand over my credit card details before I even have half a clue what they're selling. All they offer without an account/login is the ability to... er... create an account and log in. Woudl you hand over your credit card on walking into a shop "just in case" you wanted to buy something? Keeping it secret suggests that either their catalogue is crap, they're secretive lock-in merchants (hello AOL) or (dare I say it) a scam site.
Until then, I'll fulfill my music downloading tastes with unencrypted MP3, AAC and FLAC from Bleep, Tunetribe and 4AD.
Please note: I'm not knocking eMusic or legal downloads in general (indeed, I spend about £20 a month buying tunes online compared to £0 on music two years ago). I just don't trust the way their UK store seems to be working. And yes, I did want a chance to try it out.
Completely agree, it's a much overlooked area of CPU reviews. Try finding real-world power consumption figures for the Intel E6300 - not easy. Everyone only seems to give a crap about overclocking the new XTXXLLLLGTSFX2006 these days ;)
For a great little run down on CPU power across a wide range of chips, try this one for size: http://www.lostcircuits.com/cpu/low_e/
Many of the reviews I've seen show that the AMD systems consume significantly less power at idle than the equivalent C2D system. Whilst the C2D is pretty much undoubtedly the faster of the two arches, I'm still pretty staggered by the energy efficiency of the AMD64. As anothe poster pointed out, AMD's cherry picked ADD chips (well, the 3800 X2 ADD anyway) consume utterly tiny amounts of power, even on an appallingly stone aged 90nm lith process ;) I can't wait for AMD's 65nm to start shipping once their process is all sorted out, since 90nm SOI has worked so well for them. That said, since the ADD series are cherry picked they're expensive and hard to come by - last I heard they were of limited availability in Germany.
:D
Since most computers I own spend 90% of their time idling away at 1-5% load, I'm sticking with my AMD's for the time being - I'm not a gamer (although I do alot of video endocoding and have been contemplating a C2D system for my main workstation) so balls to the wall performance is not my highest priority, and keeping the costs of running my PVR's down is quite important to me.
As an aside, has anyone seen any benches for a Merom chip outside of a laptop? I've been thinking abut using one of these as a new Myth frontend, but the chips are like rocking-horse poo at the moment and I'm not aware of any UK stockists.
Anyway, like I said last time - yay for competition! For the first time in years both CPU companies are releasing some pretty interesting kit. Prices for both chips are incredibly low, and given that you can grab an X2 (i.e. more than enough CPU grunt to run anything quickly, including vista, except the latest games at max settings) for less than £100 makes this a great time for customers.
Anyway, enough rambling, time to drool over more CPU specs...