Phoronix IRC Log: 2008-06-13
starkmjolk: Someone more than me was interested in getting a nice new MID right?
copper: Deanjo: overclocked?
starkmjolk: must be, but a 3GHz phenom, that's nice... :)
starkmjolk: I'm running 3GHz X2 and I think this machine is blazing fast
copper: "Polish guy clocks Phenom X4 to 3.7GHz" http://www.fudzilla.com/index.php?option=com_content&task=view&id=7868&Itemid=35
copper: I'm basically stuck at 2.3GHz with my 9600BE
starkmjolk: I saw that link earlier
copper: not that I really mind
starkmjolk: running with TLB fix I suppose?
copper: I don't use virtualization, my system is stable
starkmjolk: fair enough I suppose :)
copper: with virtualization enabled however (not even in use), I ran into the TLB bug within hours
HM: So does the Phenom X4 clocked to 3.7Ghz built the nearest Core2 Quad offering?
HM: or does it still fail
starkmjolk: I thought about it, but probably won't swap in a phenom until the 45 nm shrink
copper: yeah you can wait
starkmjolk: I've put a hold on my FPGA-playing, otherwise I'd probably upgrade in an instant
starkmjolk: so easy to get build times of hours and even days when you have some complexity :)
copper: AM2 socket seems to incur a slight speed penalty with the Phenom though, just so you know
starkmjolk: I have a 780G board, so my only worry is power delivery
copper: oh, cool
copper: mine is an AM2 board
copper: and I suspect LAME is one of those apps that runs slower because of it
copper: too many Phenom owners run it faster
copper: with other apps there's no difference
copper: HM: what did you mean?
HM: The Phenom X4 clocked to 3.7Ghz
HM: does it kick Core 2 Quad arse yet
copper: the polish dude in question didn't provide benchmarks AFAIK
copper: it's not a pissing contest though
copper: it's a penis enlargement contest
HM: Oh I see
starkmjolk: I'd recon a OCed Core 2 Quad probably has higher integer performance, but it depends on the application
starkmjolk: I always get put off by bad architectures, and even if the Core2s are very fast there are many stupid and bad architecture decitions in em
starkmjolk: the FSB is old and stupid, the off chip memory controller brings unneccesary complexity and less performance
copper: Nehalem will fix that, although it'll be a while before it gets mainstream (i.e. affordable)
starkmjolk: dual-dies that can't communicate over the package is stupid
starkmjolk: I know the FSB is about to go now, but it should've gone a few years ago (or really about year 2000, but HP ruined that)
starkmjolk: and then the big one, x86 is not a good architecture. CISC isn't as clever as once thought
Exopaladin: won't be getting rid of x86 any time soon unless MS dies, probably :P
starkmjolk: Exopaladin: a shame :)
starkmjolk: If I had a few spare mils and some venture capital I'd take a go at it, atleast challange atom/celeron/geode with a very low power HP RISC
starkmjolk: ARM is on the right track, but take that to a 45 or 32 nm node with some fresh thinking and you got a winner
copper: finance ARM then :P
starkmjolk: this time HP was high performance, not hewlett :P
Exopaladin: I don't really get the point of atom, aside from the silly people wanting to run vista on mobile devices :/
copper: x86 compatibility
starkmjolk: Exopaladin: Me neither, but I do get the point of moblin and MIDs, so I still want one :)
starkmjolk: copper: as he said, vista on mobile devices :P
copper: VIA's Nano will probably be better
Exopaladin: x86 compatability isn't really needed on mobile devices unless you're aiming to run MS shit :P
copper: there's much more to x86 compatibility than Vista
starkmjolk: XP? :)
copper: there's other closed-source vendors to consider, like NVIDIA
starkmjolk: they adapt to the hardware
copper: anyway, NVIDIA is working with VIA, that should be a winner
starkmjolk: perhaps, but VIAs size is starting to hurt them, even more so than AMD :(
copper: they managed to come up with the Nano, which is more powerful than Intel's Atom
starkmjolk: and higher wattage too. And they could've got the nano out of the door like two years ago
starkmjolk: Just as phenom the delay made it less of a revolution and more of an evolution, not that it made the cpu less good, it's still a nice peace of engineering
GNU\colossus: starkmjolk: modern x86 CPUs aren't CISC on the level below the original ISA
copper: uh, it's only twice as fast as the C7
starkmjolk: GNU\colossus: I'd call em CISC, very many, seldom used instructions
GNU\colossus: well, the x86 ISA is merely an emulated layer above the "real" core nowadays
starkmjolk: GNU\colossus: true, but I want to get rid of that layer :)
GNU\colossus: I doubt that wouldoffer much benefit, actually
Ivanovic: GNU\colossus: it would make life for coders easier
GNU\colossus: there are other arches which do not suffer from the supposed cruft x86 has accumulated, yet they don't outperform x86 by that vast a margin, if at all
Ivanovic: that is allowing better specific optimisations and simpler compiler
GNU\colossus: Ivanovic: compiler developers maybe
starkmjolk: GNU\colossus: check atom vs TI OMAP with performance / watts
starkmjolk: try making <1W decent X86 performance
Ivanovic: the arm wins easilyy
starkmjolk: I want to see a future for consumer devices that aren't faster, just smarter and smaller. We already have too much performance on our hands :)
GNU\colossus: yeah, granted. but there's no arm core that can compete with a Core or Nehalem on a pure performance basis
starkmjolk: true for desktop
starkmjolk: but I think both Niagara2 and Power6 have something to show when it comes to multithreaded server environments
GNU\colossus: I've seen benchmarks of the niagara platform, the results where outright disappointing
starkmjolk: for som,e yes, for others no
GNU\colossus: I'm not arguing x86 is the best architecture, yet it proved fit enough to dominate the market... that's not necessarily a good thing, but most probably a good-enough thing :)
copper: windows dominates the market
Ivanovic: GNU\colossus: it dominates due to binary compatibility
Ivanovic: that is the only real reason
starkmjolk: GNU\colossus: It has proven what big truckloads of cash can do to keep a bad architecture on top. And that that capitalism doesn't always promote the smartest choises
GNU\colossus: yeah, I fully agree
Ivanovic: if the same amount of work was put into arm/ppc/whatever, those would be as fast as x86
GNU\colossus: I bet they would be faster
GNU\colossus: but noone has invested that amount of work
GNU\colossus: and I doubt anyone will in the decades to come
starkmjolk: time to change our ways asap then :)
starkmjolk: I think they will
starkmjolk: the way consumers use PCs isn't changing much, mail, office, web
starkmjolk: now they have performance from the skies that isn't used, and big fans and power bills
starkmjolk: when the smaller CPUs with good software can match the Vista power horse for 1/3 or 1/2 the price, things will change
GNU\colossus: and Aero Glass and Superfetch and four trojan horses in parallel despite UAC to eat it all up
starkmjolk: Nvidia is launching new ARMs too, with more dedicated hardware accelerators than earlier seen in CPUs
starkmjolk: Showing that performance can be high without X86, and with a much smaller budget
GNU\colossus: well, we'll see what all this GPGPU craze is going to do to computing
starkmjolk: get us back 20 years? They'll end up with what Cray had in the 70's and 80's
starkmjolk: but that isn't a bad thing, moores law has made it possible
starkmjolk: taco lunch!
m-c: michaellarabel: seen this yet? >> http://www.tgdaily.com/html_tmp/content-view-37925-113.html
GNU\colossus: the demonstration ran on a 16-core (4 socket, 4 core) Tigerton system running at 2.93 GHz.
Deanjo: http://global.phoronix-test-suite.com/index.php?k=profile&u=deanjo-19993-24094-25377 <--- well she looks stable @ 3Ghz time to start upping voltages tonight
starkmjolk: following the transistor increese with mores law for a few more years and that's in your desktop, not bad, not bad.
schestowitz: : seen my pointer on gaming the other day?
GNU\colossus: michaellarabel: nice article on the tagan nas there. if it really works out with the tagan/phoronix joint venture, you might want to let the nice guys over at www.nas-central.org know about it.
michaellarabel: schestowitz: I don't believe so
vadi2: intel ray-traced etqw: http://www.tgdaily.com/html_tmp/content-view-37925-113.html
Ivanovic: michaellarabel: hmm, how noisy is this device?
Ivanovic: those small 40mm fans tend to be really annoying
michaellarabel: very quiet
schestowitz: I wrote: " : article idea (one that would be a handy ref) is an overview of Linux games you've covered+screenshots." It would be good 'advocacy ammo'.
Ivanovic: and how much power does it drain?
vadi2: looks very cool, but not exactly feasible - it was running on a 16core 2.93ghz with fps under 30..
Ivanovic: that is my current solution using an nslu2 with a 2.5" usb hdd attached needs <10W
vadi2: And the best part - 'The icing on the cake was that the game was actually demonstrated running on a 64-bit Linux operating system. Intel stated that with ray-tracing, the company now supports 32-bit and 64-bit versions of Linux and Windows operating systems. We'll see what will happen with Mac OS X support, but that should be on the cards as well.'
vadi2: So at least we aren't going to be left out there.
Ivanovic: and: do you have shell access (via ssh) onto the box, or is there only the webinterface
Ivanovic: which distri is the system built onto, ...
michaellarabel: GeForce 8200 vs. Radeon HD 3200 coming Sunday or Monday...
vadi2: What's the ati equivalent of an 8600gt?
michaellarabel: Equivalent for performance or price?
michaellarabel: HD 3650 or so
michaellarabel: ( The 8200 and 3200 are the IGP parts )
copper: Ivanovic: what were you saying about ET: QW? Threading not working?
copper: the threaded binary from the demo doesn't seem to help much on my phenom
uncle_fungus: copper: its worse than "doesn't help" on my q6600
uncle_fungus: using the rthread binary knocks 26% off the unthreaded frame rate
copper: so what's the deal
uncle_fungus: their "render threading" doesn't work
copper: do they know about it?
uncle_fungus: i don't know. I've only just checked it
michaellarabel: The threaded binary will cause problems with ATI especially.
michaellarabel: I have discussed it before with both id Software and AMD.
copper: all that processing power wasted
michaellarabel: Basic thing.... id Software doesn't really test with AMD on Linux... so it was news to them when I broke it to them during the closed testing process. AMD says that the performance is dropping because the driver is multi-threaded and when running the multi-threaded ETQW, it's essentially fighting with one another.
vadi2: So isn't that even better? non-multithreaded games (which are majority) perform better?
copper: uh, no
copper: how would that be better
copper: 4 cores > 1 core
copper: if only one core is used, three cores are wasted
vadi2: "driver is multi-threaded"
copper: ah, sorry, I misunderstood
vadi2: Oh wow. Suddenly my eve game got corrupted. I'm in a box now with the pidgin screen on one side...
maligor: there's a difference between a multi-threaded renderer and a multi-threaded game also
vadi2: it fixed itself now. but this was kinda fun: http://img266.imageshack.us/img266/1843/screenshotnj3.png http://img266.imageshack.us/img266/9350/screenshot1yz8.png http://img238.imageshack.us/img238/451/screenshot2xz0.png
copper: gonna try booting Arch Linux with a custom kernel, bbl
copper: whadoyouknow, it works
copper: oggenc -q10 -o /dev/null pts-trondheim-3.wav: 20 seconds
copper: oggenc2_sse3_mt -q10 -o /dev/null pts-trondheim-3.wav: 8 seconds
copper: oggenc2_sse3_st -q10 -o /dev/null pts-trondheim-3.wav: 12 seconds
copper: someone needs to port http://homepage3.nifty.com/blacksword/ :(
michaellarabel: deanjo: Have you looked at batch-benchmark?
m-c: Hello again - my new computer kit showed up today. First time I am doing a build-it-myself, in a long while. The kit did not come with CPU Thermal Gel. Any recommendations where I can find some locally? I just need a squirt, and I do not want to wait around another week for it.
maligor: the cpu fansink didn't include one?
m-c: No, and I bought a retail version
copper: third-party sink?
m-c: Called Tiger Direct and they said no one *ever* asks why there is no gel in the packaging.
copper: Boxed CPUs include a fan sink with thermal paste pre-applied
maligor: my massive copper monstrosity had it included
maligor: I used my old store of arctic silver anyway tho
copper: you callin' me a monster?
copper: m-c: did you buy a sink separately from the CPU, or did it come with the CPU?
maligor: http://www.frostytech.com/articleview.cfm?articleID=2045 rather
maligor: m-c, any well equipped computer store would have the stuff
copper: sadly there's a difference between "computer store" and "computing hardware store"
maligor: perhaps in the US :P
copper: in France too
maligor: only apple stores here are like that
copper: I have to buy my hardware in Paris
maligor: the rest are the geeky sort
maligor: unless you count in general household electronics shops, but those don't qualify
m-c: Hmm - there is an apple store nearby
copper: m-c: you didn't answer my question
maligor: I really really doubt they'd have it
m-c: Oh, okay
maligor: apple people don't install processors
maligor: they buy their fix from apple straight
michaellarabel: m-c: Did you talk to Mario yet
copper: Is he /ignoring me?
m-c: Didn't Dell , Best Buy , and Circuit City drive all those little computer stores out of business? ;)
maligor: m-c, we still have those here
maligor: as well as a big computer stores that sell a large selection of stuff
m-c: michaellarabel: Honestly, I looked for his name and I did not find him listed in the project
copper: why do I even bother
m-c: Shame to have to purchase a tube of it when I need such a tiny bit
maligor: yeah, the tubes last forever
michaellarabel: superm1 @ ubuntu.com
maligor: copper, I think the answer is self evident :P
m-c: Thanks michaellarabel - I will contact him this weekend
maligor: copper, boxed sinks always have those grease pads
michaellarabel: I think superm1 might also be his IRC handle, not sure
copper: maligor: it's his first time, maybe he doesn't realize what it is, or maybe a friend strongly suggested to use thermal paste
copper: in any case he's not listening
maligor: m-c, might be a good idea to check that the sink doesn't have thermal grease pads on it btw
maligor: never seen those except in cpu bundled sinks tho
m-c: maligor: I actually just got off the phone with someone who pointed that out to me. You're right. It's on the bottom, with a plastic cover keeping it from drying.