AstralStorm: or maybe per-cpu clockdown...
DanaG: Frankly, I don't mind the lack of 3D so much -- but this eating 30 watts on battery (for whole system) compared to the 16-19 watts in Windows, sucks.
AstralStorm: per shader pipeline that is
AstralStorm: DanaG: well, AMD should provide the docs for that... right now there are none
DanaG: Oh yeah, another oddity about radeonhd: doesn't export DPI properly.
DanaG: It gets all correct info from the EDID...
DanaG: ... but then seems to say... "oh, screw this, I'm calling it 96."
DanaG: hmm, so there's some sort of info there.
AstralStorm: yes, but info is not enough
AstralStorm: that is, being able to fetch it is not
AstralStorm: esp. when its format is undefined
AstralStorm: radeonhd driver seems to choke on some "e" entry of EDID
AstralStorm: thus throw it away
AstralStorm: (still radeon driver fails to parse my EDID completely, not just partially)
DanaG: "radeon" gets dpi perfectly, oddly enough.
AstralStorm: but not modes (for my card)
AstralStorm: actually, I suspect both mostly fail reading EDID
DanaG: Hmm, it works perfectly fine for me with radeon.
AstralStorm: (except top resolution mode)
AstralStorm: here, radeon grabs only full resolution of the panel
AstralStorm: while radeonhd grabs full resolutions and all others
AstralStorm: (though it does have a few "above-native" resolutions in the list too)
AstralStorm: funny part is, these almost work
AstralStorm: i suspect this is the panel scaling ;>
kcodyjr: EDID is mostly fail
AstralStorm: 2048x1536Scaled 60.5 - while my panel is 1920x1200
kcodyjr: the reality is that manufacturers are about as reliable about EDID correctness as cops are about donut diets
AstralStorm: funny part is that I can get... antialiasing
AstralStorm: with such modes
kcodyjr: AstralStorm, that sounds like a modeline problem, if the 1920x1200 is the real resolution
kcodyjr: you'll never beat driving the display at its native resolution
AstralStorm: the native is detected
kcodyjr: and generally monitor circuitry sucks at scaling
AstralStorm: and added as 0 position to the list
kcodyjr: good it should be
AstralStorm: the "Scaled" modes are scaled by GPU I guess
AstralStorm: not by panel
kcodyjr: hmm. i've never tried. and i never would... ;)
AstralStorm: (yes, panel scaling sucks here - does nearest neighbor)
AstralStorm: difference is easily visible, e.g. 800x600 mode vs 800x600 Scaled mode
kcodyjr: i can see using the gpu to upscale to deal with naive software
kcodyjr: but i can't see using a higher than reality resolution
kcodyjr: DanaG, power saving just isn't done for r600+ yet
kcodyjr: let them figure out how to use the chip fully before worrying about how to deactivate parts of it
AstralStorm: kcodyjr: higher res does bring you some measure of antialiasing
AstralStorm: and actually larger surface to work with
AstralStorm: think of it as zoom-out
DanaG: Any ETA for power management? Scale of months is fine. And I mean for either radeon or radeonhd -- doesn't matter which.
kcodyjr: DanaG, i have no idea but i doubt it's on the radar yet, they have a lot of work ahead still
DanaG: Another advantage of really high DPI: really tiny pixels == really really smooth fonts.
AstralStorm: DanaG: but this is scaled
AstralStorm: so it's not higher dpi
AstralStorm: it's "fake" high DPI
AstralStorm: you get fake subpixels
kcodyjr: it tricks it into giving you FSAA without just turning on FSAA
kcodyjr: and actually that does break real subpixel rendering
DanaG: I have real high-dpi. 147.
AstralStorm: I also have, 133 dpi
kcodyjr: so, the best formula, use the real resolution and give it the right subpixel order
AstralStorm: the trick is that no subpixel order looks right to me ;P
AstralStorm: the color fringes look bad either way (and the screen is RGB order)
DanaG: And picture 1366x768 on 10.1" -- that'd be like 180 DPI.
AstralStorm: DanaG: meh, I have Nokia n810
AstralStorm: that tablet has 225 dpi
kcodyjr: no idea what mine is because i've got bigger fish to fry. but i have noticed that fglrx and radeonhd and radeon all pick it up differently.
AstralStorm: on a 4,3" screen though
AstralStorm: btw, Scaled modes do linear interpolation apparently
AstralStorm: at least looks like it
AstralStorm: I'd expect AMD/ATI to use something better ;p
kcodyjr: it only makes sense to change your framebuffer resolution when the driven device really can change its physical resolutions, that means crt only
kcodyjr: and even then, there is at most exactly one resolution which lines up with the phosphor dot pitch, so even flexible systems have an optimal resolution that's independent of its dpi measurement
AstralStorm: I mean, that's for resolutions lower than native
AstralStorm: which are useful
kcodyjr: ahh. yes i'd expect lower than native to do better too
AstralStorm: but using bilinear? come on, they can do better than this
kcodyjr: perhaps it's configurable and the driver guys just haven't found the switch yet
kcodyjr: betcha windows is doing bicubic ;)
kcodyjr: and really it isn't hard to do in a shader anyway, so i can see an ordinary accelerated driver accomplishing it
AstralStorm: no, it is not doing bicubic
AstralStorm: it also does bilinear :>
AstralStorm: because it's done in hardware on output stage
kcodyjr: oh? good, that gives oss a chance to outshine.
AstralStorm: not some shader
AstralStorm: still better than nearest neighbor my panel does
kcodyjr: flexible resolution can be just as easily done by having a native frontbuffer and an adjustable backbuffer, and use a bicubic texture copy during the vblank
kcodyjr: i repeat: nirvana is achieved when the physical device is being pushed at its native/optimal resolution, and the driver knows those data, regardless of what the application (X server) is seeing
AstralStorm: but that is non-free
kcodyjr: but it's cheap
AstralStorm: yes, still, not free
kcodyjr: and dedicated hardware is disappearing, so it's "cheap" or "not at all"
AstralStorm: while this bilinear is effectively free
AstralStorm: (in hardware)
DanaG: annoying thing in Windows drivers: no way to dither to fake color depth.
kcodyjr: don't count on that staying that way with the r6xx and newer
AstralStorm: kcodyjr: I am on r6xx
kcodyjr: AstralStorm, i said don't count on it to stay that way
kcodyjr: every new 100 or so they drop more dedicated circuitry
AstralStorm: well, I think they'll keep it though
kcodyjr: there is likely to come a point where it will do what the monitor says and that's it
AstralStorm: because otherwise they have to add rendering to a texture for the whole pipeline
AstralStorm: which is likely to be slower than render to screen
kcodyjr: the last generation has seen a 2x increase in memory bandwidth and a 10x increase in shader power
AstralStorm: and eat some 8-10MB more
kcodyjr: that's the trend
DanaG: My HD3650 is kickass compared to my old Go 7600.
kcodyjr: the way it's going is to do everything with shaders
kcodyjr: i don't like it either, but that's how it's going
AstralStorm: DanaG: heh, obviously
AstralStorm: that's three or so generations of gfx chips
kcodyjr: yeah that's what, r100 to r600?
DanaG: Oh heck, I even like ATI WIndows better than nvidia windows.
AstralStorm: your HD3650 is a nice collection of stream processors
DanaG: When I got the 7600, I actually actively disliked the drivers compared to the 9800 Pro I was using before.
kcodyjr: nvidia is for people that would sacrifice quality to avoid f**king with it at any level
DanaG: Oh, and the shimmering!
AstralStorm: well, nvidia sucks at high resolutions
AstralStorm: and their binary driver has fairly slow 2D
kcodyjr: AstralStorm, let me fix that for you:
kcodyjr: well, nvidia sucks
AstralStorm: I mean, after a month of work, nouveau beat them
kcodyjr: sorry, that was just fruit hanging so low i hit my head on it.
DanaG: custom resolution in windows: nvidia: go to controller, add custom resolution, and it makes you test it -- and invariably fails.
DanaG: ATI: look for DALNonstandardModesBCD, and edit it. Disable and re-enable video card in device manager. Done.
AstralStorm: hmm, I should do that
AstralStorm: btw, scaling doesn't seem to work > 1600x1200 in windoze
AstralStorm: uhm, I mean, there's no such mode as
DanaG: syntax is hh hh vv vv 00 00 00 dd
DanaG: horiz, vert, color depth
DanaG: er, last should be rr for refresh.
AstralStorm: which looks like the sweet spot in many games
AstralStorm: because 1440xfoo is fast
AstralStorm: and native starts getting slow
DanaG: runs crysis 1280x800 with antialiasing (transparent AA!), with vsync on, smoothly.
AstralStorm: DanaG: well, why AA when you can do 1440x900 w/o and get better results?
DanaG: Old gave me like 1 frame every 1.5 seconds, even at 1024x640.
DanaG: Jaggies on leaves.
AstralStorm: so what? better have a few jaggies than lower quality all around
DanaG: I don't even plan to buy the game, anyway.
AstralStorm: I do my testing on NWN2
AstralStorm: which has a horribly slow engine
AstralStorm: so slow, that overclocking CPU buys me 5-8 FPS
AstralStorm: (but the cpu is slow for the gfx chip, all right ;p cheap laptop)
DanaG: also likes not having to worry about crappy substrate.
AstralStorm: you know, HD3850 paired with Turion X2 Ultra ZM-82
AstralStorm: which is not exactly best combo
AstralStorm: the cpu is the slowest part ;P
AstralStorm: MSI weird cheapness
AstralStorm: but I couldn't get anything better really for the price
AstralStorm: ATI-containing laptops are rare here
AstralStorm: they all get 3450 or such
AstralStorm: which is well, a class down
AstralStorm: and it was hard in its own right to get high-dpi screen
DanaG: I got myself an EliteBook 8530w.
DanaG: Had to custom-order to get 1920x1200 with the FireGL.
DanaG: HP only offered either lower-res with ATI, or higher-res with NVIDIA -- and I was dead set on both together.
DanaG: It's a damned good laptop, though.
DanaG: For some reason, fglrx newer than 8.543 doesn't work even in Intrepid.
AstralStorm: how much did it cost"
AstralStorm: (because this MSI GT735 is normal order - but there are variants with different resolution screen, two cpu variants and bluray)
AstralStorm: it's 2.5x the price
AstralStorm: definitely out of my range
DanaG: Oh, my laptop was 2200 with customize-to-order and ADP and computrace warranty.
DanaG: Yeah. Business laptop.
DanaG: There's a 14" and a 12" version, too.
DanaG: Check the "smart buy" models, also.
DanaG: the "p" is cheaper than the "w".
nanonyme: Hrm, starting to wonder if the guys who write articles on Phoronix about fglrx are in fact people used to Windows computers. Can't really understand why someone says a new driver release has nothing interesting when it has only bug fixes...
Larrxi: why does I get the following with my X1900 XT? (WW) RADEONHD: No matching Device section for instance (BusID PCI:1:0:0) found
Larrxi: (EE) No devices detected.
Larrxi: I removed a busid option and now it works
Larrxi: is dri needed for direct rendering?
MrSunshine: what gpus are open spec now ?
MrSunshine: rv350 ? :)
Larrxi: dont try to get a answer in this channel :P
Larrxi: I swapped my 8800 gts 512 for an x1900 xt at blocket yesterday :)
MostAwesomeDude: Larrxi: Direct rendering can be done with a software rasterizer.
MostAwesomeDude: MrSunshine: r3xx-r7xx.
MrSunshine: r3xx includes rv350 or what? ... im thinking the v isnt there? :)
MostAwesomeDude: Yeah, we have rv350.
MrSunshine: is there documentation of the 2d parts of the ati cards somewhere also ? :)
MrSunshine: all i can find is the 3d parts
MostAwesomeDude: There's modesetting docs for r3xx.
Larrxi: MostAwesomeDude, but dri is the fastest one?
MostAwesomeDude: Larrxi: The DRI driver is accelerated in hardware; the software rasterizer is not.
Larrxi: thank you
Veza: Is it normal for radeonhd to probe monitor data 20+ times on startup? (grep Monitor /var/log/Xorg.0.log | wc)
adamk: Veza: Any chance you are using KDE4 ?
Veza: adamk: no, gnome.
adamk: Well, there went that idea.
kensai: hi guys, I have been suing radeonhd in my laptop, anyone know how one can fix the problem with this drivers making my laptop run hot?
yangman: kensai: power management isn't supported yet
kensai: just came out of my fingers :D
kensai: need to get stuck with the very poor catalyst drivers then
kensai: or try radeonhd from git
kensai: maybe there is something :D
yangman: no, there's no support at all
kensai: I'm sad, I can't leave the bad proprietary drivers
yangman: mobility X1xxx or HD?
kensai: I was confident that support on linux would have improved when I bought the laptop
kensai: now, I have been proved wrong
yangman: it's certainly improved
yangman: it doesn't happen overnight
kensai: yeah, but it is improving so so so slowly
yangman: patches welcome ;)
yangman: power management docs are very high on the list of things to come from AMD
kensai: haha, you know, I understand, I'm an archlinux developer
kensai: and I tell that to the users all the time
kensai: I mean, the radeonhd drivers are awesome, they do a lot and are way more stable than catalyst
kensai: my problem is with catalyst drivers, they progress so slowly
agd5f: the hard part is, I've yet to find any good documentation for power management :/
kensai: and 9.1 and 9.2 ahs been worse than 8.12 on my hd3200
yangman: agd5f: :(
kcodyjr: balancing an elephant on the head of a pin is more stable than catalyst.
kensai: indeed, I am in talks with another arch developer to remove catalyst from the main repositories
kensai: catalyst 9.1 and .2 doesn't run on our 64 bit systems
kensai: they need an ugly workaround of symlinking
kcodyjr: i suppose catalyst is useful for professional cad/cam applications, but it just sux for an end user
kcodyjr: much as i'd love to bash though, i have to try and keep thinking - gotta implement cubic spline interpolation, bleh
kcodyjr: damn the math libraries for all being GPL ;)
kcodyjr: the number of times, this project, i've had to step away from the keyboard and pick up pencil and paper... grr
kensai: what do you think of a laptop running at 73 celsius?
kensai: is that acceptable?
kensai: I mean having it run around 70 to 80 celsius?
brot: depending on how much it can take.
kensai: well is an amd turion x2 64
kcodyjr: the CPU, that's high but not outrageous. anywhere else in the system, you're probably clogged with dust.
kensai: nah, the laptop is pretty much new
kensai: in an idle system it runs at 70 celsius
kensai: compiling it can get to 80
kensai: that is with radeonhd
kensai: on the other hand with catalyst it runs between 60 to 70 celsius
kensai: so the difference is 10 grades
kcodyjr: suppose that isn't a surprise, radeonhd isn't doing any pm at all yet
kensai: yeah, I have been told that
kcodyjr: and ain't gonna until amd releases the docs for that, plus cut the guys some slack and give them time...
kensai: but is up to 80 acceptable for an amd turion x2 64 bit laptop
kcodyjr: i wouldn't know, but i'd also expect the bios to bitch if it isn't. check the thresholds, if any, in the bios setup.
kensai: I can't access the bios in this hp laptop :D
kensai: you know, vendors lock in haha
kensai: well, I am googling, to see how much is acceptable on this cpu
kcodyjr: anyone good with math? i can't figure out how to resolve a circular dependency: http://en.wikipedia.org/wiki/Spline_interpolation#Interpolation_using_natural_cubic_spline
kensai: max operating temperature is 95 degrees !!!
kensai: so I'm ok using radeonhd since they just get up to 80 degrees
kcodyjr: impressive. it can almost boil water. it can definitely cook an egg.
kcodyjr: funny, they used to refer to underpowered automobiles as 'stoves', implying their best use was to generate heat rather than going anywhere...
kensai: so off I goa nd reinstall my 64bit arch on my laptop using radeonhd
kcodyjr: have fun
kensai: I am just so going to love this :D
kcodyjr: i think i may have found an algorithm...
kensai: then it is a happy day my frined :D
kcodyjr: ;) i read fluent typo, don't worry about it
kensai: damn typos
kcodyjr: if i assume n == 3 (four data points) that yields two z_i= equations, one for i==1 and one for i==2, and it will give the correct result for i=0.
kensai: at last, my system is complete again!
kensai: who guesses the movie :D
kcodyjr: no idea
kensai: sweeney todd :D
kensai: at last, my arm is complete again :D
kcodyjr: only saw part of it
kensai: don't worry I'm this crazy
kensai: well, I have seen the 60s theater, the 2001 concert and the 2007 movie
kensai: the best, the 2007 movie
kensai: since I'm young :D
kensai: well, at least ati has better open source drivers than nvidia
kensai: but well, nvidia has way better proprietary driver than ati
kcodyjr: pretty easy to write a decent driver for a turd of a chip.
kensai: define turd
kensai: oh, this is not ubiquity :D
kcodyjr: do you seriously need me to define 'turd' ? ;)
kensai: let me check wordreference
kensai: turd (tūrd)
kensai: n. Vulgar
kensai: 1. A piece of excrement.
kensai: 2. Slang A contemptible person.
kensai: 1 or 2?
kcodyjr: i'm trying to think through that algorithm to be sure it's valid
kensai: don't make your brain run at 95 degrees
kcodyjr: specifically, if i set z_3 == 0 temporarily, calculate z_1 and z_2, keep z_1, throw away z_2 to be recalc'ed at the next cycle... is that going to yield the same result for z_1 as if i hadn't cheated on z_3
kensai: now, that is a strange language
kcodyjr: what i said, or math in general?
kensai: math in general
kcodyjr: precise but obscure
kensai: that is why I'm not a programmer, I jsut do the basic programming
kensai: nothing fancy, or professional
kensai: well, got to go, nice talking to all of you !!!
kcodyjr: i'm an i.t. guy during the day
kensai: hey, I am an it guy as well
kensai: but just hate programming, so if you hate programming, just be a network administrator like me :D
kcodyjr: i'm a network engineer and unix admin
kcodyjr: but i like programming. it's just that math class was 10+ years ago.
kensai: nice, that is so where I want to get
kcodyjr: look for small companies then. big ones are afraid of open source networking.
kensai: yeah, so I have noticed, I work at a big companie where I live
kensai: they are so closed to suggestions
kensai: well, now I'm off