Phoronix IRC Log: 2009-07-27
Kano: hi michaellarabel
michaellarabel: hi kano
Kano: any news about b4 or final of pts?
Kano: or do i have to package git ;)
michaellarabel: package git for now
michaellarabel: The final plans have been for 4 Aug and that is not changing.
vadi2: Does anyone have any resources / guides to overclocking on linux for newbies?
maccam94: vadi2: overclocking what?
vadi2: The video card to begin with
maccam94: nvidia or ati?
vadi2: Want to find out where my bottleneck is :-/ nvidia on intel
vadi2: 8600m gt
maccam94: you can't overclock mobile graphics cards
maccam94: i wouldn't advise OC'ing a laptop in general
vadi2: Why? It's fan is generally not on
maccam94: they already pack about as much heat as they can in that small of a place
maccam94: your graphics card changes its clock rate dynamically, depending on the load
maccam94: so if you start playing a full screen 3D game for a bit, the fan will come on full
maccam94: the cpu also changes clockspeed depending on load
vadi2: I want more :( and I've had the laptop for about 2 years now, so I'm ok with shortening the lifetime
maligor: and there's the business with clock gating too when idle
maligor: better to play games on a desktop :P
vadi2: No, I like my mobile gaming
vadi2: I just want moar. Probably getting a new laptop soon
vadi2: going to go play a game and monitor the card temp while at it
vadi2: the fan is indeed on full, buts its sitting at 68C okay
maligor: efficency probably hasn't gone much up from when you bought that laptop
maligor: so either you get a hotter laptop or one that performs relative equal
maligor: shader perf might be better on some newer hw tho
maligor: and they implement the compute shader now on nvidia gtx I think
maccam94: performance has gone up. battery life has not ;-)
maligor: guess what that means
maligor: unless battery life has stuck on level, the heat generation will go up :P
vadi2: well, there is a shader clock in in the nvclock tihng
vadi2: but I'm not sure what its about since afaik there is only the gpu and memory clocks
maccam94: batteries and heat aren't really related...
maccam94: batteries haven't advanced a whole lot in the last few years
maligor: sure they are
maccam94: processors and graphics cards keep using smaller, more efficient components, so the speed increase stays within the same thermal/power profile
maligor: most energy consumed is converted into heat
maccam94: they high tech batteries you hear about haven't made it into laptops yet, except for the new apple macbooks
maccam94: i'm not arguing that
maccam94: i'm saying they've just maintained the same energy consumption
maligor: Wouldn't touch apple stuff even with a long stick
maligor: vadi2, I stand corrected
maccam94: laptop batteries have stayed in the 6 to 9 cell range for years
maligor: he clearly knows his laptops
maligor: the real question is how to fit a GTX 295 into a laptop
maccam94: they can't
maccam94: the current mobile GTX cards aren't really even GTX chips
maccam94: they're rebranded 9xxx chips
maligor: you could make a atom system board in a >17" laptop
maligor: and put over half the space below the keyboard for the graphics card
maccam94: which are essentially rebranded 8xxx
maligor: maccam94, I know
maligor: that's why I said what I said
maligor: I suppose they have smaller fab now tho
maccam94: you need a large amount of space in even a desktop to run a GTX series card
maligor: do you know if the same is true for amd mobile 4k?
maccam94: no way you can run a dual gpu card in a laptop
maligor: maccam94, actually...
maccam94: are you talking about the radeon 4xxx?
maligor: there are sli laptops
maccam94: i've got a 17" laptop
maccam94: yes i know
maccam94: but they haven't gotten to a small enough manufacturing process for the gtx chips yet
maccam94: i believe the chips themselves are much larger than previous generations
maligor: just wondering if mobile 4xxx is rebranded or not
maligor: I know amd does it too to some extent
maccam94: i think i read recently that their names are slightly adjusted
maccam94: but not at all to the extent of the nvidia mobile chips
maligor: even amd has Mobility Radeon HD 4850 X2
maccam94: " ...a Mobility Radeon HD 4850 that differs from its desktop counterpart in clock speed alone, not architecture."
maligor: never understood people who buy ridiculous graphics chips in laptops
maccam94: my laptop has a 9800M GT
maligor: heh, then I don't understand you
maccam94: 1920x1200 screen
maligor: it must require some pretty impressive and heavy cooling solution
maligor: the whole laptop must weigh like a brick cart
maccam94: it's got a few fans on the underside, and a fairly large heatsink to the back
maccam94: i've got a 5796, it's pretty much the same model
maccam94: builds muscle :-P
maccam94: the way i see it, i carry it around in a backpack when i move it most of the time anyway, so it doesn't really matter
maccam94: i've got textbooks that weigh more haha
maligor: I doubt I'll ever buy a laptop as such
maligor: I'd get a snapdragon based arm netbook if I could grab one tho
vadi2: well, finished the game
vadi2: its at 71C
vadi2: I'll see if I can up it to where it still is around that
maccam94: you really can't OC laptops very easily
maligor: a bit more and you could use it for cooking :P
maccam94: you have to buy special unlocked laptops that cost thousands of dollars in order to be able to OC the cpu
vadi2: how can I tell if mine is locked?
maccam94: did you by a core 2 extreme processor?
maccam94: if not, then you probably can't OC it
maccam94: those processors alone cost $1000
vadi2: I bought an CPU since I bought the laptop and upgraded it like that
vadi2: and I was thinking about OCing the video card first
maccam94: what laptop do you have?
vadi2: system76 serval 3
maccam94: i'm pretty sure you can't oc it
maccam94: and while nvidia has CoolBits overclocking in the nvidia-settings tool, mobile cards don't listen to it
maligor: nothing is impossible ;)
vadi2: I actually dot see the coolbits in nvidia-settings
vadi2: only found it when I installed the nv-clock thing
vadi2: told me to add the option in xorg.conf... need to restart to see if it'll take any effect. sec
maccam94: you need to add coolbits to xorg.conf
maccam94: then the option will show
maccam94: but it won't work
maccam94: i don't think there's anything you can do short of VBIOS hacking
maccam94: which you really don't want to do
maccam94: it's a good way to brick your system
maccam94: you could possibly check if your graphics card is user-upgradeable
maccam94: mine is, and your laptop looks like another sager model
vadi2: no its a compal fl90
maccam94: hmph interesting
maccam94: they use sager for the bonobo, i wonder why they use multiple suppliers...
maccam94: oh well
maccam94: you could still check
maccam94: though i'm not sure if you need special bios updates to use some cards...
vadi2: I think it's buggy on overclocking
vadi2: I kept trying and trying, and now its at 513Mhz
vadi2: with the standard apparently 475
ZeXx86: 'lo .. still no news about UT3 Linux client ?
monreal: hahaha.... good one
JayF: hi, any gamer with an ati card here?
sandeen: michaellarabel, which version of PTS was used for the "OpenSuSE, Ubuntu, Fedora, Mandriva Benchmarks" ?
michaellarabel: PTS 2.0.0b2 or something. Should show on the graphs, sandeen.
sandeen: ah, sure does, thanks
sandeen: and is that still available for d/l?
sandeen: I wanted to look over some of the weird fedora results
sandeen: ok looks like it is
Kano: compile benchmarks or something else?
sandeen: michaellarabel, is it correct that if I set PTS_USER_DIR to some path, then the tests will build & run there, and fs tests will also use that dir?
sandeen: Kano, more interested in the fs tests, dbench & postmark for starters
sandeen: michaellarabel, did you ever look into whether the apache test was writing to error logs for you?
michaellarabel: sandeen: PTS_USER_DIR should still I believe, I usually toggle it through ~/.phoronix-test-suite/user-config.xml with the EnvironmentDirectory. And yes, all PTS is available for download
michaellarabel: That's still on my TODO list with regard to apache
sandeen: michaellarabel, one thing I was going to suggest is that in your summary of the systems tested, in addition to fs type, it might be nice to say what the block device is (lvm, or plain partition, or what) just so that for the IO sorts of tests, there's some idea of what the IO stack is
sandeen: 'cause for mandriva, for example, I have no idea :)
michaellarabel: Yeah, sorry about that, I should add that to future tests.
sandeen: it's hard to know what might matter
sandeen: upstream recently changed default journalling mode for ext3 to something that very fast and very unsafe on a crash
sandeen: I bet that explains the mandriva numbers (though how dbench claim to go 163MB/s on a single spindle I still have no idea)
sandeen: michaellarabel, hrm, making EnvironmentDirectory "/mnt/test" isn't doing any IO to that dir
sandeen: oh, crap.
sandeen: adds trailing /
michaellarabel: hmm, was it not working right without trailing /?
sandeen: nope, I got stuff in /mnt/testdbench instead of /mnt/test/dbench
Kano: michaellarabel: thats a common error, you always have to use ending / in your settings
Kano: well i know it
Kano: as thats the same for dl caceh
michaellarabel: that's an easy bug to fix... okay.
sandeen: michaellarabel, hm maybe rather than asking for things like the IO device, it'd be cool if phoronix had some sort of system-state-capture script
sandeen: so for anything you publish, you can link to things like lists of packages installed & versions, root device, etc
michaellarabel: The trailing slash issue is fixed in Git.
sandeen: woo ;)
michaellarabel: sandeen: There is already that sort of.... Well, besides not being linked to from article, need to wait till there is Phoronix Global support for uploading system states.
sandeen: michaellarabel, well, for user-contributed things sure, but maybe just for your officially published articles?
michaellarabel: But depending upon your settings, PTS can log the glxinfo, Xorg.0.log, dmesg, lspci, etc for each test system.
sandeen: "sosreport" grabs a crapton of things ,for example
sandeen: part of the "sos" rpm on fedora anyway. just a thought
sandeen: if something grabbed the kernel .config under test, that'd be awesome.
sandeen: same for fs mount options, selinux state, running daemons, etc
sandeen: just to put results in context for people who want to dig deeper
sandeen: or who want to reproduce the thing exactly :)
michaellarabel: sandeen: If you would like to come up with a list of everything you would like to see, I would be happy to implement it.
sandeen: i'll try to compile a list as I look through stuff :) How space-constrained would you be?
sandeen: sosreport can be pretty big; maybe not something you'd want for submitted benchmarks, but maybe worthwhile for your official ones
michaellarabel: For Phoronix Global storage of it? Umm, I would prefer say like 50k or less for user-contributted submissions on average.
sandeen: ok, that'd be pretty small; maybe there can be different sets for submitted systems vs. your own testing
michaellarabel: or 100k I suppose for users
sandeen: wonders how many of the OSs regularly under test provide sosreport
sandeen: I guess you could d/l and build it like everything else :D
sandeen: ok, gotta bike to the pub :) thanks for the info.
michaellarabel: np cya