Page 5 of 6 FirstFirst 123456 LastLast
Results 61 to 75 of 90
  1. #61
    Joined
    Jan 2005
    Location
    Marietta, GA
    Age
    30
    Posts
    576

    Re: New CPU Benchmark

    on version 1.3 the total progress bar goes from 0% to 50% to 100% but doesn't show anything inbetween like the previous versions. also it says you added more reference scores but i see a lot less.
    (\__/)
    (='.'=)
    (")_(")

  2. #62
    Joined
    Jan 2002
    Location
    Joliet, Illinois
    Age
    34
    Posts
    15,846

    Re: New CPU Benchmark

    Quote Originally Posted by morerandomsn View Post
    on version 1.3 the total progress bar goes from 0% to 50% to 100% but doesn't show anything inbetween like the previous versions. also it says you added more reference scores but i see a lot less.
    maybe the updating the progress bar real time requires too much cpu usage possibly skewing results?

    practically every sisoft sandra benchmark does this so i assume it's normal for this type of benchmark.
    Corsair 800D / Corsair TX750 PSU
    ASUS Sabertooth P67 / Win7 Pro
    Intel i7 2600k @ 4.4ghz (44 x 100) + Corsair H100
    Corsair Vengeance DDR3 1600 x 16GB
    2 x XFX R9 280x + 3 x Asus VS24AH-P 24" IPS 1920x1200
    Samsung 840 Evo 250GB OS + 2 x 1TB WD Caviar Black




    The member formerly known as SuBX3r0 HEAT

  3. #63
    Joined
    May 2002
    Location
    Kansas
    Age
    33
    Posts
    1,088

    Re: New CPU Benchmark

    I noted that I "removed 'fine' progress reporting due to overhead" in the release notes. It was eating up way too much CPU time. Now I only report the progress of the steps in the benchmark. Fifty percent is when the test data has finished generating. One hundred percent is when the encryption stage is done. It was responsible for about 40% of the total work load as it was before, so I removed it. That's one reason scores jumped so much.

    Version 1.3 Testing had only two reference scores. I had to clear out all the ones prior due to changes in the program leading to non-comparable scores. The progress reporting thing was partly responsible. The multi-threaded capable version of the Aleris algorithm was the other thing. Version 1.3b Testing has five reference scores. To me, this is an increase.
    Sometimes I feel like I'm becoming a dinosaur.

  4. #64
    Joined
    Jan 2002
    Location
    Joliet, Illinois
    Age
    34
    Posts
    15,846

    Re: New CPU Benchmark

    Quote Originally Posted by TBird761 View Post
    I noted that I "removed 'fine' progress reporting due to overhead" in the release notes. It was eating up way too much CPU time. Now I only report the progress of the steps in the benchmark. Fifty percent is when the test data has finished generating. One hundred percent is when the encryption stage is done. It was responsible for about 40% of the total work load as it was before, so I removed it. That's one reason scores jumped so much.

    Version 1.3 Testing had only two reference scores. I had to clear out all the ones prior due to changes in the program leading to non-comparable scores. The progress reporting thing was partly responsible. The multi-threaded capable version of the Aleris algorithm was the other thing. Version 1.3b Testing has five reference scores. To me, this is an increase.
    so tell me, are there any "optimizations" is there going to be a scandle ala futuremarks ati vs nv?

    kidding of course, but i totally called it on the progress bar thing, i know a thing or two about a thibng or two.
    Corsair 800D / Corsair TX750 PSU
    ASUS Sabertooth P67 / Win7 Pro
    Intel i7 2600k @ 4.4ghz (44 x 100) + Corsair H100
    Corsair Vengeance DDR3 1600 x 16GB
    2 x XFX R9 280x + 3 x Asus VS24AH-P 24" IPS 1920x1200
    Samsung 840 Evo 250GB OS + 2 x 1TB WD Caviar Black




    The member formerly known as SuBX3r0 HEAT

  5. #65
    Joined
    May 2002
    Location
    Kansas
    Age
    33
    Posts
    1,088

    Re: New CPU Benchmark

    Heh. I know it seems like my benchmark favors Intel pretty heavily, but I can't help it. It's a purely honest implementation of the algorithm and it's almost entire straight integer math. It just seems that the Intel chips are better at it. I've always been an AMD fan, myself.
    Sometimes I feel like I'm becoming a dinosaur.

  6. #66
    Joined
    Jan 2005
    Location
    Marietta, GA
    Age
    30
    Posts
    576

    Re: New CPU Benchmark

    perhaps the reference scores shown should change depending on what settings you have picked. a dual core processor gets higher scores with two threads than one so when you pick two threads it will show the score that the dual core got with those settings and so forth with the other settings. none of the reference scores show the settings that they were tested at.
    (\__/)
    (='.'=)
    (")_(")

  7. #67
    Joined
    Jan 2005
    Location
    Marietta, GA
    Age
    30
    Posts
    576

    Re: New CPU Benchmark

    scores for all 15 possible settings at stock for the athlon 64 3000.
    1 thread:
    128 - 14,098
    256 - 14,004
    512 - 14,063

    2 threads:
    128 - 14,003
    256 - 13,508
    512 - 13,640

    4 threads:
    128 - 13,752
    256 - 13,685
    512 - 13,541

    8 threads:
    128 - 13,574
    256 - 13,729
    512 - 13,476

    16 threads:
    128 - 13,662
    256 - 13,673
    512 - 13,470

    i noticed even when it cleared out the RAM after each test i still got the best scores by shutting down the program between each test. all of these scores where with no other programs running. i'll try and post all the scores at OCed settings tomorow. and i still haven't gotten around to testing the Pentium 4 1.8Ghz machine but i'm planning on doing it soon. maybe tomorow but i don't thinks so. hopefully before the week is over.
    (\__/)
    (='.'=)
    (")_(")

  8. #68
    Joined
    May 2002
    Location
    Kansas
    Age
    33
    Posts
    1,088

    Re: New CPU Benchmark

    Cool. Thanks for the scores like that.

    As for reporting settings, 1.3b does show how many threads and special options (+HT = "With Hyperthreading"), etc, but not test size. I think I will add some filtering abilities like you suggested, however.

    The next version will also see multi-threaded test data generation and a totally revamped "stress test" mode. Currently, it works like this:

    Program's thread:
    - Generate 128MB test data
    - Launch testing threads

    Testing thread(s):
    - Encrypt data
    - Check if all threads are done when one finishes (if so, launch decode threads)

    Decode thread(s):
    - Decrypt data
    - Check if all threads are done when one finishes (if so, launch verify thread)

    Verify thread:
    - Check to see if the original data and the encrypted/decrypted data match up
    - Launch the test over again from beginning

    As you can tell, CPU loading is less than optimal. The new model will keep the CPU at 100% all the time and not switch between threads so much. Each worker thread will simply have its own data set instead of feeding out of a single pool of test data. This isn't efficient or logical for the benchmark portion, but for the stress test it doesn't matter if one thread finishes before another one. All that matters is if an error is found at any point along the line.
    Last edited by TBird761; 11-08-2006 at 12:34 AM.
    Sometimes I feel like I'm becoming a dinosaur.

  9. #69
    Joined
    Jan 2005
    Location
    Marietta, GA
    Age
    30
    Posts
    576

    Re: New CPU Benchmark

    scores for Athlon 64 3000 OCed to 2.7Ghz
    1 thread:
    128 - 19,971
    256 - 20,738
    512 - 20,447

    2 threads:
    128 - 21,076
    256 - 20,789
    512 - 20,410

    4threads:
    128 - 20,713
    256 - 21,130
    512 - 20,324

    8 threads:
    128 - 20,312
    256 - 20,610
    512 - 21,264

    16 threads:
    128 - 20,362
    256 - 20,585
    512 - 20,522

    realized that the pentium 4 rig that i was planning on running the test on only has 512MBs of RAM in it so do you think it would still be worth it to run the test on it?
    (\__/)
    (='.'=)
    (")_(")

  10. #70
    Joined
    May 2002
    Location
    Kansas
    Age
    33
    Posts
    1,088

    Re: New CPU Benchmark

    Yeah. Slower machines should still give accurate results at lower test sizes.
    Sometimes I feel like I'm becoming a dinosaur.

  11. #71
    Joined
    Dec 2002
    Location
    USA
    Posts
    70

    Re: New CPU Benchmark

    Opteron 165 OC @ 2900 or 9x322/263

    1 thread:
    128 - 22,919
    256 - 22,795
    512 - 22,826

    2 threads:
    128 - 46,103
    256 - 45,104
    512 - 45,777

    4 threads:
    128 - 46,348
    256 - 45,853
    512 - 44,620

    8 threads:
    128 - 46,348
    256 - 44,865
    512 - 45,225

    16 threads:
    128 - 45,354
    256 - 45,837
    512 - 46,282

    Didn't experience any problems with the memory not releasing in between runs.
    DFI Expert Bios 04/06; Opty 165 (CCBBE 0615) 9x323/264@1.425Vcore 24/7; G.Skill F1-4000USU2-2GBHZ @ 2.6V; Sapphire X800GTO^2 (16p-stock); OCZ Powerstream 520ADJ; Air-cooled SI-120_Panaflo FBA12G12U 120X38MM @ 115CFM;

  12. #72
    Joined
    Jan 2005
    Location
    Marietta, GA
    Age
    30
    Posts
    576

    Re: New CPU Benchmark

    1 thread 128MB test on pentium 4 2.4GHz computer - 14,266 computer only has 512MBs of RAM so i can't do any other tests.
    would you like me to overclock my computer to other settings to simulate athlon 64 3200, 3500, and 3800s?
    (\__/)
    (='.'=)
    (")_(")

  13. #73
    Joined
    Jan 2005
    Location
    Marietta, GA
    Age
    30
    Posts
    576

    Re: New CPU Benchmark

    Bump. any updates to the benchmark yet?
    (\__/)
    (='.'=)
    (")_(")

  14. #74
    Joined
    Jul 2003
    Age
    37
    Posts
    209

    Re: New CPU Benchmark

    Just tried it on my Athlon xp 2500+ (sempron) @ stock

    256mb test size w/ 1 thread I got 10,951 (in v1.3..with a lot of junk running in the background )

  15. #75
    Joined
    Jul 2005
    Location
    College Station, TX
    Age
    38
    Posts
    1,049

    Re: New CPU Benchmark

    Getting pretty standard scores for my CPU/Memory setups.

    AMD Laptop Turion with about 768mb of ram running at 2gHz hits 16k easily regardless of test.

    C2D I have in my system specs hits 60k.


    CPU-Z My Heat
    Request Granted. Safe Passage Guaranteed via JimZ

Posting Permissions

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