My old Pentium 90 with no L2 cache croaked out ~18 fps, so you likely managed somewhere in the mid-20s. But over a 28.8k modem, that'd have been sketchy even if you weren't running Windows too. Sympathies - I remember trying to connect to a server in my hometown on a 14.4 modem back in those days. It was basically unplayable until QuakeWorld came along, but the improvement wasn't large enough to keep me from buying a 33.6 kbps modem a little while later. The night I broke down, bought a NIC, and went to a friend's LAN party was a transformative experience.
I think most nice 386 'boards maxed out at 16 MB RAM, which would have been garishly expensive in the late '80s to early '90s. You'd have to pare a modern kernel down until you sweated blood just to make a command-line install squeeze into that space without paging out to swap... As has been suggested by others, the 2.4 kernel was probably the last realistic choice, and some would argue that going further back would be wise.
Re:What was the last version which actually did?
on
Linux Nukes 386 Support
·
· Score: 4, Informative
Ack, I was wrong: according to this handy timeline, the last release was 2.4.37 at the very beginning of 2011. Were one to roll his own distribution and cross-compile, you could still make a surprisingly modern Linux run on a 386...
Re:What was the last version which actually did?
on
Linux Nukes 386 Support
·
· Score: 3, Insightful
I think 2.4 was the last safe bet for 386's, but the mainline support for that terminated about seven years ago. All things considered, the 386 had a pretty fucking good run.
With a two party system co-opted by special interests, we were left with the option of a hammer and one nail, or a nailgun. With those as our only realistic options, you go with the guy with the hammer, because at least his aim might be off.
I think your reading's the only one that makes sense. Lower sales volume = higher prices. $150,000 per movie's just a bit excessive as a judgment, though - outside of some peculiar fetishes, most porn's both abundant and highly similar. Ding this guy an order of magnitude less money and let everyone get on with their lives - this just makes the porn industry look even skeezier than usual, and appears set to ruin some guy's life forever.
Don't underestimate these things for video encoding or playing around with scientific computing. They're great for embarrassingly parallel computing problems, and the price is very good.
"Shinier graphics" = hardware that you paid for, which works with its advertised feature set? Jesus, way to trivialize an entire range of useful, revenue-generating products in the professional world.
I remember running into the RDTSC problem in Windows XP x64 with Left 4 Dead back in 2009. It's an irritation for anything as latency-sensitive as a first-person shooter, but still not a deal-breaker on the order of PRMan's accusation.
Well, for servers I wouldn't count them out. I'm planning to build an AM3+ Piledriver box because I process a lot of data that owes itself to massive parallelization (hello, seismic depth migration and DVD/Blu-ray ripping), and the price point's just too good to pass up. Given the current price delta between an FX 6100 and an entry-level i5, I can afford to use ECC memory... But in the consumer sector I'm a bit afraid for AMD - Llano's still appearing at the low end of their FM2 offerings, which highlights just how many problems the new architecture's presenting.
Because the risk you speak of doesn't exist any more. The manufacturers of those excruciating old chipsets aren't in the running: as best I recall VIA only manufactures chipsets for its own products, and SiS got out of the motherboard racket altogether. AMD's chipsets are good, reliable, and feature-competitive.
That wasn't Lawrence Fishburne, that was Delroy Lindo. And yes, The Core is one of the most hideously inaccurate, ostensibly scientific films ever made. What galls is that the film itself isn't awful in terms of character development or plotting - it's just oriented around a series of terribly wrong fundamental assumptions, and then ties itself into progressively more ridiculous knots to support them.
On the other hand, it is riotously funny to watch with a room full of tipsy geologists.
Among other things, Slackware's the only distro I ran into that compiled Seismic Unix out of the box, without the need to install any dependencies. I'll grant that the lack of dependency resolution's kind of irritating, but it's knock-on-wood reliable, stable as hell, usually more current than Debian Stable, and is a terrific base for a workstation or server without a lot of fluff. Slackbuilds is a godsend too, and the community's second-to-none.
Yes, but most of the existing solutions for doing so result in lousy-looking, inefficiently large video files. Some have blamed GPU precision issues (especially prior to 64-bit float support in OpenGL 4.x GPUs), and some have blamed lousy H.264 implementations within the encoders. The truth's likely somewhere in the middle. Intel's QuickSync seems pretty good for a black box solution, but most of the screengrabs from those videos still look like the kind of thing I'd want to see on a cell phone instead of my HDTV. My money's currently on the x264 project's addition of OpenCL support, which looks to add a 33-100+% speedup on what's already one of the best video encoders around.
Actually, in Intel's funhouse world of advertising, the i3 ("three stars") is midrange - the crippled Celeron ("one star", no hyperthreading, single or dual core, low clocks, hobbled cache) and unremarkable Pentium ("two stars", dual core, no hyperthreading) parts are still lower.
Intel Quick Sync is also proprietary and is mostly famous because it offers better video quality than other, earlier hardware video encoders with terrible H.264 implementations. Talk OpenCL, and talk about Handbrake's upcoming release with partial encoding offload via OpenCL, and we'll see how Intel holds up.
I don't even really think MP3s qualify as demanding these days; a single core Atom can manage MP3 encoding at high quality settings at better than real-time, and even though most available encoders aren't multithreaded an awful lot of people will be limited by the speed of their optical drive before their CPUs really come into the picture. DVD and Blu-ray transcoding have stepped into the role of multimedia-centric CPU flogger in its stead. That - and scientific computing - are what's leading me to consider building a 100% new system for the first time in five years...
Slashdot Mirror
My old Pentium 90 with no L2 cache croaked out ~18 fps, so you likely managed somewhere in the mid-20s. But over a 28.8k modem, that'd have been sketchy even if you weren't running Windows too. Sympathies - I remember trying to connect to a server in my hometown on a 14.4 modem back in those days. It was basically unplayable until QuakeWorld came along, but the improvement wasn't large enough to keep me from buying a 33.6 kbps modem a little while later. The night I broke down, bought a NIC, and went to a friend's LAN party was a transformative experience.
I think most nice 386 'boards maxed out at 16 MB RAM, which would have been garishly expensive in the late '80s to early '90s. You'd have to pare a modern kernel down until you sweated blood just to make a command-line install squeeze into that space without paging out to swap... As has been suggested by others, the 2.4 kernel was probably the last realistic choice, and some would argue that going further back would be wise.
Ack, I was wrong: according to this handy timeline, the last release was 2.4.37 at the very beginning of 2011. Were one to roll his own distribution and cross-compile, you could still make a surprisingly modern Linux run on a 386...
I think 2.4 was the last safe bet for 386's, but the mainline support for that terminated about seven years ago. All things considered, the 386 had a pretty fucking good run.
"Gibs" like "giblets." Simple.
With a two party system co-opted by special interests, we were left with the option of a hammer and one nail, or a nailgun. With those as our only realistic options, you go with the guy with the hammer, because at least his aim might be off.
I think your reading's the only one that makes sense. Lower sales volume = higher prices. $150,000 per movie's just a bit excessive as a judgment, though - outside of some peculiar fetishes, most porn's both abundant and highly similar. Ding this guy an order of magnitude less money and let everyone get on with their lives - this just makes the porn industry look even skeezier than usual, and appears set to ruin some guy's life forever.
Predictable behavior and high-quality manufacturing, too.
Forgive my cynicism, but I'm pretty sure There's a Black Man in My Wife's Ass didn't cost more to make than Forrest Gump, or even something like They Live.
Don't underestimate these things for video encoding or playing around with scientific computing. They're great for embarrassingly parallel computing problems, and the price is very good.
What about one of the snap-around Bluetooth keyboards for the iPhone?
"Shinier graphics" = hardware that you paid for, which works with its advertised feature set? Jesus, way to trivialize an entire range of useful, revenue-generating products in the professional world.
I was always partial to "Fartbeak," myself, mostly because I wonder what the mascot would look like.
I remember running into the RDTSC problem in Windows XP x64 with Left 4 Dead back in 2009. It's an irritation for anything as latency-sensitive as a first-person shooter, but still not a deal-breaker on the order of PRMan's accusation.
Well, for servers I wouldn't count them out. I'm planning to build an AM3+ Piledriver box because I process a lot of data that owes itself to massive parallelization (hello, seismic depth migration and DVD/Blu-ray ripping), and the price point's just too good to pass up. Given the current price delta between an FX 6100 and an entry-level i5, I can afford to use ECC memory... But in the consumer sector I'm a bit afraid for AMD - Llano's still appearing at the low end of their FM2 offerings, which highlights just how many problems the new architecture's presenting.
Citations beyond a single allegation would be helpful, here.
Because the risk you speak of doesn't exist any more. The manufacturers of those excruciating old chipsets aren't in the running: as best I recall VIA only manufactures chipsets for its own products, and SiS got out of the motherboard racket altogether. AMD's chipsets are good, reliable, and feature-competitive.
See, my personal tipping point was the "microwaves" coming through the Earth's magnetic field and burning through the Golden Gate Bridge.
That wasn't Lawrence Fishburne, that was Delroy Lindo. And yes, The Core is one of the most hideously inaccurate, ostensibly scientific films ever made. What galls is that the film itself isn't awful in terms of character development or plotting - it's just oriented around a series of terribly wrong fundamental assumptions, and then ties itself into progressively more ridiculous knots to support them.
On the other hand, it is riotously funny to watch with a room full of tipsy geologists.
If it works for him without requiring him to cross-compile, then what's the issue? It makes him happy. Why henpeck?
Among other things, Slackware's the only distro I ran into that compiled Seismic Unix out of the box, without the need to install any dependencies. I'll grant that the lack of dependency resolution's kind of irritating, but it's knock-on-wood reliable, stable as hell, usually more current than Debian Stable, and is a terrific base for a workstation or server without a lot of fluff. Slackbuilds is a godsend too, and the community's second-to-none.
Yes, but most of the existing solutions for doing so result in lousy-looking, inefficiently large video files. Some have blamed GPU precision issues (especially prior to 64-bit float support in OpenGL 4.x GPUs), and some have blamed lousy H.264 implementations within the encoders. The truth's likely somewhere in the middle. Intel's QuickSync seems pretty good for a black box solution, but most of the screengrabs from those videos still look like the kind of thing I'd want to see on a cell phone instead of my HDTV. My money's currently on the x264 project's addition of OpenCL support, which looks to add a 33-100+% speedup on what's already one of the best video encoders around.
Actually, in Intel's funhouse world of advertising, the i3 ("three stars") is midrange - the crippled Celeron ("one star", no hyperthreading, single or dual core, low clocks, hobbled cache) and unremarkable Pentium ("two stars", dual core, no hyperthreading) parts are still lower.
Intel Quick Sync is also proprietary and is mostly famous because it offers better video quality than other, earlier hardware video encoders with terrible H.264 implementations. Talk OpenCL, and talk about Handbrake's upcoming release with partial encoding offload via OpenCL, and we'll see how Intel holds up.
I don't even really think MP3s qualify as demanding these days; a single core Atom can manage MP3 encoding at high quality settings at better than real-time, and even though most available encoders aren't multithreaded an awful lot of people will be limited by the speed of their optical drive before their CPUs really come into the picture. DVD and Blu-ray transcoding have stepped into the role of multimedia-centric CPU flogger in its stead. That - and scientific computing - are what's leading me to consider building a 100% new system for the first time in five years...