Page 2 of 6 FirstFirst 1234 ... LastLast
Results 11 to 20 of 52

Thread: AMD Phenom 9500 Linux Performance

  1. #11
    Join Date
    Oct 2007
    Location
    Moscow
    Posts
    3

    Default

    What should be added:
    1. Comparison with other quad core CPUs
    2. Truly much threaded applications
    3. Java benchmarks. Why java? It can be any vm. Mono for example.
    4. Mathematica benchmark

  2. #12
    Join Date
    Oct 2007
    Posts
    52

    Default

    Quote Originally Posted by tduglas View Post
    What should be added:
    1. Comparison with other quad core CPUs
    2. Truly much threaded applications
    3. Java benchmarks. Why java? It can be any vm. Mono for example.
    4. Mathematica benchmark
    +1

    and i would also like if phoronix could do a run without the tlb fix since phenom would show its real power then

  3. #13
    Join Date
    Aug 2007
    Posts
    6,631

    Default

    Or just compile a kernel with the AMD hotfix.

  4. #14
    Join Date
    May 2007
    Location
    Third Rock from the Sun
    Posts
    6,584

    Default

    Quote Originally Posted by Kano View Post
    Or just compile a kernel with the AMD hotfix.
    Yup I've got two Phenom based systems, both on Asus boards M2N-E and M2A-VM HDMI. I haven't had any of the issues described in the article. Both were drop in CPU replacements on existing X2 systems running openSUSE. I just popped them in and away they went, no reinstall needed. I did later though compile the kernels with the hotfix to workaround the TLB bug with no performance hit.

  5. #15
    Join Date
    Jan 2008
    Posts
    1

    Default

    Yep got a 9500 tried to startup for the first time with an old linux disk... like ubuntu 6.10 or it may of been the next version but it didn't work... that whole kernal stuff happened.

  6. #16

    Default

    We tried it with and without the TLB fix enabled -- in both cases it exhibited the same problems. I'll be delivering more benchmarks, etc next week once I return from Las Vegas for CES.


    I'll also be asking AMD-Legacy about this on Monday and see what those on the show floor have to say.

  7. #17
    Join Date
    Jan 2008
    Posts
    2

    Thumbs up Installs and runs FLAWLESSLY for me--CentOS 5.1

    Hey Guys. Just wanted to add my thoughts and experiences. In the last week I setup a new Phenom system. Phenom 9500, ECS - AMD 770 motherboard running CentOS 5.1 x86_64. I must say, I have had absolutely NO problems. No kernel panics, no stability problems whatsoever (I stress tested it for some time to see what would happen with stability and temperature). Maxing out all 4 cores I can't get the chip to run hotter than 64 degrees Celsius. Everything runs perfectly smooth. In fact, of the 4 linux servers I've built in the last year this one was the most painless. My benchmarks are incredible. My RAID 5 setup (4 SATA2 drives) is giving me 200 MB/sec read speeds, the processor is more than powerful enough, and everything is running as planned. Maybe some of these issues you are talking about are specific to Ubuntu 7.10? Or possibly a problem relating to older kernels? CentOS 5.1 just recently came out so maybe these problems have been fixed in the c5.1 kernel.

    Cheers and good lucking running the Phenom. I recommend it to EVERYONE out there! I was able to built a $3000 server for just $900 thanks to the Phenom 9500 and Fry's electronics great holiday deals!

    Devin

  8. #18
    Join Date
    Jan 2008
    Posts
    187

    Default

    You would be better off using a parallelized compressor for these tests, like parallel bzip http://compression.ca/pbzip2/

    (I've also used parallel gzip, but that doesn't show as dramatic gains.) You should also use multithreaded LAME.

    I've posted results from testing OpenLDAP with multiple cores/threads across Opteron 875, Opteron 2347, and Xeon 5345 here
    http://connexitor.com/blog/pivot/entry.php?id=191
    With code that's explicitly written for threading and concurrency, the AMD Barcelona design really shines.

    I think you need to state more clearly what your goal is in these benchmarks. If you really want to show the performance potential of a new platform, you should actually use test software that takes advantage of that platform. Probably that means not using any binary-based distros at all, and use gentoo instead. At least, for these SSE/media codec packages you ought to use a gcc with -march=amdfam10 support (AMD's 4.2 distro, or gcc 4.3 from CVS). The only test you've conducted here that shows any big advantage for the Phenom processor is compiling a kernel in parallel with make -j. All this test proves is that GNUmake works as designed. (By the way, I designed the parallel make support in GNUmake, over 16 years ago. So this test shows nothing new at all; it's been working fine for over a decade.)

  9. #19
    Join Date
    Dec 2006
    Posts
    222

    Default

    Just wanted to make a couple of comments. The syncing problems are probably due to the still shaky bios support and are not related to the TLB bug. In fact a lot of the early adopters had trouble running their systems at stock settings, although the situation seems to be improving. In fact at another forum several Phenom users are trying to get the bug appear so far with no success. This bug is not something that should concern desktop users and benchmarks should probably be done without the fix enabled as it can seriously hamper performance. Also to achieve a higher overclock try dropping the northbridge multiplier as it seems it cant clock much above 2GHz on most Phenoms while the cores can often hit 2.8GHz. Im curious about the kernel fix though. Is it it stable?

  10. #20
    Join Date
    Oct 2007
    Posts
    52

    Default

    Quote Originally Posted by highlandsun View Post

    I've posted results from testing OpenLDAP with multiple cores/threads across Opteron 875, Opteron 2347, and Xeon 5345 here
    http://connexitor.com/blog/pivot/entry.php?id=191
    With code that's explicitly written for threading and concurrency, the AMD Barcelona design really shines.
    Wow that seems very impressive .
    Code:
    Clients	Opteron 875	Opteron 2347	Xeon 5345
    1	3.1s	         2.9s	         1.6s
    2	6.7s	         6.5s	         12.6s
    3	12.1s	        10.5s	         20.5s
    4	18.1s	        14.9s	         44.6s




    Ive gonna be happy with my phenom thats fort sure

    Its also seems you need to run 4 vm to get the tlb bug appear, so i can safely run xen virtualization with atleast 2 vm system and the tlb fix disabled . Im considering a 64 bit Ubuntu 8.04 with a 32 bit linux and 32 bit windows vm:=)
    After all it would be a shame not to use the virtualization functions of the Phenom.
    Last edited by teroedni; 01-06-2008 at 03:36 PM.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •