x264 video encoding benchmark

[directhex]

you have a 148 core machine at work? that's insane.

152 core, but i can't span jobs across the "boot" cores and "work" cores easily

[graysky]

run it!

[mroz]

---------- RUN1PASS1.LOG
encoded 1749 frames, 20.26 fps, 1854.10 kb/s

---------- RUN2PASS1.LOG
encoded 1749 frames, 20.00 fps, 1854.10 kb/s

---------- RUN3PASS1.LOG
encoded 1749 frames, 19.92 fps, 1854.10 kb/s

---------- RUN4PASS1.LOG
encoded 1749 frames, 19.95 fps, 1854.10 kb/s

---------- RUN5PASS1.LOG
encoded 1749 frames, 19.98 fps, 1854.10 kb/s

---------- RUN1PASS2.LOG
encoded 1749 frames, 5.28 fps, 1825.89 kb/s

---------- RUN2PASS2.LOG
encoded 1749 frames, 5.29 fps, 1825.89 kb/s

---------- RUN3PASS2.LOG
encoded 1749 frames, 5.30 fps, 1825.89 kb/s

---------- RUN4PASS2.LOG
encoded 1749 frames, 5.30 fps, 1825.89 kb/s

---------- RUN5PASS2.LOG
encoded 1749 frames, 5.26 fps, 1825.89 kb/s

Athlon XP 3200+ @ stock 202x11, sis 741gx, 1GiB PC3200 DDR stock 200Hz 2.5-3-3-6, XP SP2

Will post results for new Q6600 rig when finished fixing on OC

[directhex]

run it!

ran it across 116 cores. like i said, the lack of hand-optimised assembly routines for itanium kill it in this benchmark

[aidanjt]

You mean there is another Intel CPU other than the Pentium 4 that sucked at FPU performance?

[graysky]

Athlon XP 3200+ @ stock 202x11, sis 741gx, 1GiB PC3200 DDR stock 200Hz 2.5-3-3-6, XP SP2

Will post results for new Q6600 rig when finished fixing on OC

Thanks for the results!

[directhex]

You mean there is another Intel CPU other than the Pentium 4 that sucked at FPU performance?

no, itanium beats a lot of the competition into a pulp, clock for clock. however, it's enormously compiler-dependent, and gcc simply doesn't do a very good job at optimizing code in general, let alone a chip like IA64. the solution taken by performance-critical gcc-centric codes like mplayer or x264 is to write some sections in assembly, but assembly is per-arch, and there's no equivalent section for IA64

before you ask, no, I couldn't get it to compile with icc

[directhex]

okay, i absolutely cannot for the life of me work out how to get the memory timings out of the altix. sorry, looks like you wont' be able to include that one on your list

that said, the concept of memory latency in the conventional sense is broadly irrelevant for a machine like the altix, since memory can be spread over a rather large number of individual "motherboards"

tidying up other entries & adding extra data like mobo chipset, though

[directhex]

Intel Pentium M 730, stock 133 x 12, i915GMS/i910GML, 4-4-4-11 @ 266MHz, Ubuntu GNU/Linux 7.10 i386

encoded 1771 frames, 9.06 fps, 1853.53 kb/s
encoded 1771 frames, 9.08 fps, 1853.53 kb/s
encoded 1771 frames, 9.09 fps, 1853.53 kb/s
encoded 1771 frames, 9.08 fps, 1853.53 kb/s
encoded 1771 frames, 9.12 fps, 1853.53 kb/s
encoded 1771 frames, 1.80 fps, 1825.71 kb/s
encoded 1771 frames, 1.80 fps, 1825.71 kb/s
encoded 1771 frames, 1.80 fps, 1825.71 kb/s
encoded 1771 frames, 1.80 fps, 1825.71 kb/s
encoded 1771 frames, 1.80 fps, 1825.71 kb/s

this craptop's a bit slow, really

[aidanjt]

no, itanium beats a lot of the competition into a pulp, clock for clock. however, it's enormously compiler-dependent, and gcc simply doesn't do a very good job at optimizing code in general, let alone a chip like IA64. the solution taken by performance-critical gcc-centric codes like mplayer or x264 is to write some sections in assembly, but assembly is per-arch, and there's no equivalent section for IA64

Aye, that's true. IA64 isn't as well supported as it should be. Even x86_64 support is a little weak compared to x86.

before you ask, no, I couldn't get it to compile with icc

I wasn't going to say anything about icc, can't say I'm supprised icc didn't compile them.

[kungpo]

e4300 stock 9x200, RAM 400, 4-4-4-12, i965

encoded 1749 frames, 48.62 fps, 1850.89 kb/s
encoded 1749 frames, 50.82 fps, 1850.89 kb/s
encoded 1749 frames, 51.02 fps, 1850.89 kb/s
encoded 1749 frames, 50.69 fps, 1850.89 kb/s
encoded 1749 frames, 50.87 fps, 1850.89 kb/s
encoded 1749 frames, 12.53 fps, 1826.20 kb/s
encoded 1749 frames, 12.51 fps, 1826.37 kb/s
encoded 1749 frames, 12.42 fps, 1826.37 kb/s
encoded 1749 frames, 12.58 fps, 1826.37 kb/s
encoded 1749 frames, 12.57 fps, 1826.37 kb/s

e4300 oc 9x340, RAM 680, 4-4-4-12, i965

encoded 1749 frames, 83.71 fps, 1850.89 kb/s
encoded 1749 frames, 84.23 fps, 1850.89 kb/s
encoded 1749 frames, 84.68 fps, 1850.89 kb/s
encoded 1749 frames, 84.57 fps, 1850.89 kb/s
encoded 1749 frames, 84.73 fps, 1850.89 kb/s
encoded 1749 frames, 20.88 fps, 1826.38 kb/s
encoded 1749 frames, 21.08 fps, 1826.38 kb/s
encoded 1749 frames, 21.13 fps, 1826.38 kb/s
encoded 1749 frames, 21.08 fps, 1826.38 kb/s
encoded 1749 frames, 21.05 fps, 1826.37 kb/s

Both clean boots, all necessary startup items stopped. Big difference between the two! Vista Business 32Bit with all updates and VirusScan 8.5 Enterprise.

[directhex]

okay, scroll up for 4 cleaned up results (everything but the Altix)

[graysky]

Thanks for the results, guys!

[kungpo]

You're more than welcome. Looking forward to seeing the overall results!

[directhex]

i should rerun a test on windows, for curiosity's sake

[mroz]

i should rerun a test on windows, for curiosity's sake

I'd be very interested to know how the two compare on the same machine