There's a fbdev based windowing system in the pipes called DiniX (Dinix Is Not X). It's basicly a framebuffer multiplexor, so it sould be just about as efficient as raw fbdev access.
When I worked at the Nuclear Physics Group at UNH we put together a 60 or so node dual celery-450 cluster. It was for crunching LOTS of data so we ended up putting serveral ide cards in, and striping a bunch of maxtors. At 4 disks per card, you can fit quite a few. And it's a LOT cheaper than scsi:)
There was even discussion of gigabit for the server to provide enough bandwidth. I think they decided that size mattered more than speed and went with 100bT hubs.
btw, the ide raid stuff i've heard about (e.g. promise) is just a ide card and software raid in the driver. linux has it's own software raid, no need for a special controller.
Indeed, such winmodems as you describe really are crummy, but i've heard that others are just DSPs connected through the PCI bus. Bit of an improvement over uart, wouldn't you say?
I think the MWave (combo sound card/modem) is such a beast. Problem is (last i checked), it's not at all linux-friendly.
Ah, but notice it's sparc solaris, not ia32 solaris...
They're willing to port to solaris because a browser doesn't do much good to a server and it lets them claim that it's now a cross platform browser. ia32 solaris runs on the same hardware as nt, so perhaps is more of a competitor. An intel port also runs the risk of being run under emulation on linux.
Besides, I hear they did a pretty crummy job anyway.
High quality card, very fast, has a glx driver now, and is developed by an traditional open source project. They haven't even gotten to multitexturing yet and it's already faster than the windows driver in 1024x768. And this is the driver that John Carmack himself is working on.
Matrox has a reputation for quality cards and cooperation with open source types. Full specs == better drivers.
Slashdot Mirror
There's a fbdev based windowing system in the pipes called DiniX (Dinix Is Not X). It's basicly a framebuffer multiplexor, so it sould be just about as efficient as raw fbdev access.
Nice to see someone pointing out the important distinction.
When I worked at the Nuclear Physics Group at UNH we put together a 60 or so node dual celery-450 cluster. It was for crunching LOTS of data so we ended up putting serveral ide cards in, and striping a bunch of maxtors. At 4 disks per card, you can fit quite a few. And it's a LOT cheaper than scsi :)
There was even discussion of gigabit for the server to provide enough bandwidth. I think they decided that size mattered more than speed and went with 100bT hubs.
how does 4*40 == 1600?
btw, the ide raid stuff i've heard about (e.g. promise) is just a ide card and software raid in the driver. linux has it's own software raid, no need for a special controller.
Indeed, such winmodems as you describe really are crummy, but i've heard that others are just DSPs connected through the PCI bus. Bit of an improvement over uart, wouldn't you say?
;)
I think the MWave (combo sound card/modem) is such a beast. Problem is (last i checked), it's not at all linux-friendly.
Besides, i can't wait to ditch my last ISA card
Ah, but notice it's sparc solaris, not ia32 solaris...
They're willing to port to solaris because a browser doesn't do much good to a server and it lets them claim that it's now a cross platform browser. ia32 solaris runs on the same hardware as nt, so perhaps is more of a competitor. An intel port also runs the risk of being run under emulation on linux.
Besides, I hear they did a pretty crummy job anyway.
G400, hands down.
High quality card, very fast, has a glx driver now, and is developed by an traditional open source project. They haven't even gotten to multitexturing yet and it's already faster than the windows driver in 1024x768. And this is the driver that John Carmack himself is working on.
Matrox has a reputation for quality cards and cooperation with open source types. Full specs == better drivers.
See glx.on.openprojects.net for more info.