Phoronix IRC Log: 2008-10-20
maligor: nice performance these new CF cards give.. 30MB/s read on a usb reader
GNU\colossus: mine did 45mb/s read via IDE
GNU\colossus: and about 30mb/s write
GNU\colossus: A-DATA 266x
maligor: this is trancend MLC 133x 16GB
maligor: and this sandisk ultra ii 8GB (15MB/s version) SDHC gives 17MB/s
maligor: pretty decent for disk use already
Milyardo_: Good morning #phoronix
Deanjo: checks to see if Tom's fixed their openssl benches yet now that they are looking into it
Milyardo_: Of course not
Milyardo_: That would mean that they care about presenting unbiased and fair information
Deanjo: It's funny how I didn't get a reply until I posted real benchmark results
Deanjo: Hell the only way I could actually get close to reproducing their results was by compiling it for a plain jane 386 with no optimization flags whatsoever.
Deanjo: Even with that I was still pulling in faster results
Deanjo: Even utilizing the precompiled rpm from the openSUSE distro I was still getting 180 signs
Milyardo_: I don't know how they did it
Milyardo_: Did you ask for steps to reproduce the results?
Deanjo: copper was saying that he got a reply back and they said that they used the ubuntu 32 openssl deb
Milyardo_: I thought they were using pts though...
Deanjo: they are, so they had to have modified it
Milyardo_: Complete incompetence or are they just lying to cover somethinf up?
Deanjo: The only other thing I can think of is that they compiled it on a P4 and it's trying to use SSE3
Deanjo: and SSE3 on anything but a P4 is totally useless
Milyardo_: There were no P4s in the test though
Milyardo_: (far as I saw)
Deanjo: Ya but who knows how their lab functions
Deanjo: Probably compiled it on a P4 server in a VM or something then used the disk image across all the machines
Milyardo_: But they claimed to use all of the optimazations availible for each CPU
Milyardo_: So that doesn't make sense either
Deanjo: Where do they claim that?
Milyardo_: hrrmmm maybe I was mistaken... though I could have sworn copper said something that made me think that
Milyardo_: still going through IRC logs for the past month
Milyardo_: It was this
Milyardo_: Copper linked me this
Milyardo_: its in French btw
Deanjo: Oct 16 10:21:44 you were fucking right
Deanjo: Oct 16 10:22:04 they used a single binary for all CPUs, with *no* optimization, and the same environment
Deanjo: Oct 16 10:22:43 lol geee that sort of defeats the whole purpose now doesn't it?
Deanjo: Oct 16 10:23:32 they justify it by saying that Joe six-pack will use apt-get and not a self-compiled binary, that it allows them to compare CPUs accurately, and that their results for the C2D E7200, E6850 and E6600 are close to those reported on PTS Global
Milyardo_: Well in that page apaige says the oppisite
Milyardo_: to quote
Milyardo_: "Sans la moindre optimization, ce n'est pas le CPU dont vous faites le benchmark, ni même de l'application. "
Milyardo_: nerver mind
Milyardo_: my french isn't as good as I thought maybe
Milyardo_: Google translates "Sans la moindre optimization" to "Without any optimization", although I don't why. When I read that I see "With all the optimazations"
Deanjo: heh, translators still leave alot to be desired
Lightkey: because sans = without, that word exists in english too with that meaning
Deanjo: There are discrepancies with some of the other linux benchmarks as well but nowhere near the ssl test.
Deanjo: The other thing that bugs me about their charts is that they don't report what other hardware is being used. ie amount of ram etc
maligor: maybe they have ddr400 on the phenoms because joe sixpack will use ddr400
Deanjo: lol, that would be impossible
Deanjo: maybe they disabled 3 out of 4 cores on the phenoms
maligor: because joe sixpack doesn't know what a core is
Deanjo: Well joe only uses 1 core. The others are just spares
maligor: did the benchmark use x86_64 mode?
maligor: considering that would enable some stuff
Deanjo: 8.04 32-bit
maligor: i386 then
maligor: maybe someone should bug them to do a x86_64 bench too
Deanjo: Well they should considering every single proc that they benched supported 64-bit
maligor: and maybe they should benchmark graphics cards on directx 3d 3.0
Deanjo: Like I mentioned before, even just using the base 64-bit rpm from openSUSE to run the openSSL bench without openSSL being recompiled gets nearly 3x the results of theirs
Deanjo: I don't mind a 5% discrepancy in benchmarks but 150% is way to much
Deanjo: 5% allows for "luck of the draw" quirks
Deanjo: even that though is being generous
maligor: openssl likes x86_64?
Deanjo: ya, a bit, the instruction set makes more of a dent though
Deanjo: 32-bit vs 64-bit can make some nice gains depending on the application. MySQL for example loves 64-bit
Lightkey: some like the doubled registers much :-)
Deanjo: Yup, it varies alot.
Deanjo: On a server there is no excuse to be running 32-bit if the proc supports 64
maligor: unless it has 1kb of memory
maligor: 64bit uses more memory
maligor: don't see any reason to run 32bit once there's a working flash
Deanjo: if they would at least fix the fuckin wmode when using the wrapper it wouldn't be bad
Deanjo: I've only been bitchin to adobe about that damn bug for 5 years now
maligor: yeah, and debian doesn't ship lib32nss3
maligor: which flash 10 depends on
maligor: running swfdec now
Deanjo: Well debian has to get with the game
maligor: swfdec actually allows for floating elements ontop of the flash elements
Deanjo: too bad you couldn't dictate what plug-in handles what
maligor: it still has quite a lot of bugs tho
maligor: but it hasn't actually crashed firefox a single time
maligor: (unlike gnash)
Deanjo: Adobe makes a mess of any application it did not initially develop
maligor: they seem to have a lack of innovation too
maligor: did you check CS4 improvements?
Deanjo: No, never use CS
maligor: it was like a list of new inventions unveiled in siggraph :P
michaellarabel: Is CS 4 the new one with GPU acceleration?
maligor: cool features but still
Deanjo: I did see the GPGPU improvements
maligor: oh, never noticed those
michaellarabel: There was some development build of CS at the RV770 launch event in June and they had a very nice demo
maligor: that reminds me, I haven't tested if my gpgpu app still crashes X
maligor: or rather crashes the kernel
michaellarabel: What gpgpu app is that?
maligor: raw interpolator
michaellarabel: Can it be made into a benchmark for PTS? :P
maligor: well, if it worked.. maybe
maligor: brook+ compiler turns it into a 1MB header file
maligor: last time I tried it, it was catalyst 8.9
Deanjo: until openCL hits (or at least until folding@home comes out with a linux client that can be benchmarked for both ATI and Nvidia), GPGPU benching isn't going to be all that meaningful
maligor: let's see if it still crashes :P
maligor: oh well
maligor: it was pretty bad code anyway
maligor: been meaning to test it out in windows but never got around to it
maligor: I see I remembered it wrong too: 1948394 2008-10-20 21:44 filter_gpu.hpp :P
Deanjo: OMG now there is a OEM that makes a oil submerged PC system http://www.maximumpc.com/article/features/hardcorepc_reactor?page=0%2C0
Deanjo: You should see if they will send you an eval system michaellarabel
maligor: next I expect we'll see liquid metal cooling
Deanjo: That case is damn sexy
maligor: hard to service yourself tho I guess
Deanjo: It would be messy as all hell
maligor: it looks like non-expandable stuff tho
maligor: atleast can't see any standard pci bracket holes :P
Deanjo: It comes pretty loaded as is
Deanjo: I would love to see a aircooled version of that case
maligor: I wonder where it extracts the heat
maligor: no good pictures
maligor: oh, it has more pages
Deanjo: Ya there is a radiator on it
Deanjo: Great choice for a MB manufacturer on it
maligor: meh.. two options, Windows Ultimate 32 and 64, both for $190
maligor: where's the FreeDOS or Linux option
Deanjo: wonders when the kernel will have oil pump monitoring support :P
maligor: expensive things
Deanjo: also wonders if a block heater is needed for northern Canada :D
maligor: $4500 at cheapest
maligor: you mean the oil freezes?
maligor: might be pretty bad if it expands or contracts
Deanjo: You better believe that oil gets thick as molasses at -30
Deanjo: I can just see the headlines "PC's now using up world's oil supply, oil prices rise $200 a barrel"
maligor: "Due to unprecedented changes in the oil industry, oil is now sold by the case, the old barrel pricing becoming obsolete"
Deanjo: "Greenpeace now targets hardcorecomputer's as public enemy #1, Apple gets moved down the charts."
maligor: ext4 is definitely looks tempting
maligor: like nanosecond timestamps.. I wonder who needs that
Deanjo: heh, I disable that crap
Deanjo: "I find if I use synthetic oil my super-pi results improve....."
maligor: you mean natural oil
maligor: synthetic oil is for the joe sixpack massmarket
Deanjo: I'm just picturing all types of new forum arguments lol
maligor: natural oil shows you care for your computer
maligor: wouldn't feed my computer any of that synthetic crap
maligor: all the ext4 benchmarks seem to say it's a faster version of ext3 with more features, pretty impressive
Deanjo: Ya, which will then be eclipsed shortly after by btrfs
maligor: btrfs just seems a tad complex
maligor: chunkfs looks promising
maligor: linux kernel must have some sort of world record on amount of filesystems supported natively