Frequency vs timings *update* results in OP

Discussion in 'Memory' started by DavLe.BeRRy, Sep 10, 2007.

  1. DavLe.BeRRy

    DavLe.BeRRy Member

    Joined:
    Jun 11, 2006
    Messages:
    1,989
    Location:
    Perth, WA
    Hi guys,

    I have very limited knowledge of exactly how timings work but the main understanding i have is that people try to achieve the "tightest" timings (lowest) @ the highest frequency possible.

    Would a 667Mhz kit @ 4-4-4-12, be better than 800MHz 5-5-5-15?
    When adjusting timings what does the last number represent and why is it higher than the others?

    4+4+4 = 12
    5+5+5 = 15

    ^ does this mean anything or is it a coincidence.


    i've been reading this articles but it seems a little oblivious to me :confused: and their conclusions haven't really summed up what "optimal timings" are.. which is what i'm trying to work out.

    Can anybody help me out here?

    Update

    I have added some basic testing and here are my results as follows:

    Test Setup
    • Intel Core 2 Quad Q6600 @ 2.40GHz
    • 2x1GB PC6400 ADATA DDR2-800MHz 1.83v 5-5-5-18
    • Seagate 320GB SATAII 7200RPM 16mb cache
    • Albatron Geforce 8800GTX 768mb
    • EVGA 650i Ultra A1
    • ASUS SATA DVDRW 18x
    • Windows Vista Home premium 32bit

    I chose to compare the sticks at their stocks speeds and then @ 4-4-4-10 2.0v 667MHz

    CPU-Z defaults
    [​IMG]

    CPU-Z overclocked
    [​IMG]

    Testing
    For testing i have used Passmark Performance test 6.1 [Basic and advanced memory testing read/write]

    Passmark Defaults
    [​IMG]
    Click for full-sized image

    Passmark Overclocked
    [​IMG]
    Click for full-sized image!

    Seems Write speed is slightly down, what would be the cause of this?

    3DMark 2006 1.1.0 Direct comparison
    [​IMG]

    Gaming Counter-Strike: Source [Video Stress Test - Maximum graphics settings]

    settings
    [​IMG]
    results
    [​IMG]

    Conclusion
    Well there you have it, results pretty much speak for them self although I have lost 133MHz on my RAM the tighter timings have given me a very slight gain in performance, both benchmarking and gaming :) I would have liked to have included results for the time demo featured in F.E.A.R but i have yet to install the game and will need the massive patch for solid results.

    Some of you may ask, why didn't u set it to 800MHz @ 4-4-4-10? well even at 2.10v (max on my board) this memory won't budge, it wont even overclock by 1MHz on its stock timings.

    Very interesting learning experience there :) hope someone else finds it useful.

    edit: just noticed i did the read test twice for the OC'd values.. i'll update that now.
     
    Last edited: Sep 10, 2007
  2. Jazper

    Jazper Member

    Joined:
    Jul 28, 2001
    Messages:
    2,475
    Location:
    Melbourne Australia
    More mhz gives more performance than tighter timings on the whole. That is all. Sandra has a memory benchmark.
     
  3. antari

    antari Member

    Joined:
    Aug 27, 2006
    Messages:
    1,338
    Location:
    Bendigo
    The last number is called tRAS, and I don't actually know the actual definition of what it does but I've found it makes little to no difference to performance at all. Certain types of memory IC are pretty much oblivious to it (in that you can pretty much set it down to the same as the others) whereas others need it to be in roughly that pattern in those ones you've listed.

    Basically, "optimal" timings are the lowest timings you can set at a certain speed and voltage. You do have to be careful not to use too high of a voltage (for example Micron D9 I wouldn't do more than 2.15V 24/7). The one exception to this "lowest = best" rule that I know is on Asus P5B boards, as some of them didn't like x-3-x-x timings IIRC.

    Depending on your platform (CPU/chipset), low timings, high clocks or a compromise will be the best option, and for this you need to experiment... for example:

    SuperPi 1M times at 2.8GHz (7x400):
    RAM @ 800 3-3-3-3: 19.687
    RAM @ 1000 4-4-4-4: 19.453
    RAM @ 1200 5-5-5-5: 19.407
    For each I ran three runs and picked the best one... subtimings in BIOS were all 3 except for tRFC, I used 20 for the CAS 3 run, 25 for the CAS 4 run and 30 for the CAS 5 run.

    Memory benchmarking I use tools like Everest and SiSoft Sandra as well as SuperPi.
     
    Last edited: Sep 10, 2007
  4. impact_player

    impact_player Member

    Joined:
    Jul 12, 2006
    Messages:
    1,168
    Wow, this exact question has been asked in the thread below here. Anyone got an answer?
     
  5. OP
    OP
    DavLe.BeRRy

    DavLe.BeRRy Member

    Joined:
    Jun 11, 2006
    Messages:
    1,989
    Location:
    Perth, WA
    Results have been added to OP :)

    My only question that remains is: will this slight performance gain remain and hold is truth when the CPU is overclocked with it?
     
    Last edited: Sep 10, 2007
  6. cbjaust

    cbjaust Member

    Joined:
    Sep 11, 2003
    Messages:
    1,015
    Location:
    Melbourne
    Have aread of this post over at the Tech Repository.

    It is very interesting.

    Cheers
     
  7. OP
    OP
    DavLe.BeRRy

    DavLe.BeRRy Member

    Joined:
    Jun 11, 2006
    Messages:
    1,989
    Location:
    Perth, WA
    Thanks CBJaust, i'll look into that :)
    i have reverted to stock 800MHz, 5-5-5-18 2T for the time being.

    I am still waiting on all my watercooling gear. Once it arrives i shall begin overclocking my CPU and perform more tests on my RAM to see the affect in an overclocked environment..
     
  8. cbjaust

    cbjaust Member

    Joined:
    Sep 11, 2003
    Messages:
    1,015
    Location:
    Melbourne
    here is a small comparison I have made with SuperPI mod/1.5 XS with my Opteron at 2830MHz and the RAM at 201 and 235MHz with different main timings. The 201MHz with tighter timings is slightly but definately faster. the Sub timings were the same in both cases. In fact, I always run the same sub timings regardless of the RAM frequency and what tCL, tRCD, tRP and tRAS are set to.

    Code:
    CPU		AMD Opteron 170 Dual Core	AMD Opteron 170 Dual Core
    OPN		OSA170DAA6CD			OSA170DAA6CD
    Code		1461398E60656			1461398E60656
    Stepping	CCBBE 0617FPMW			CCBBE 0617FPMW
    Core		Denmark				Denmark
    L2		2 x 1024 kiB			2 x 1024 kiB
    digits v/Speed>	Opteron 170 @ 2830/202MHz	Opteron 170 @ 2830/235MHz
    16		0:00.187			0:00.187
    32		0:00.453			0:00.453
    64		0:01.015			0:01.015
    128		0:02.282			0:02.297
    256		0:05.359			0:05.406
    512		0:13.078			0:13.141
    1024		0:30.328			0:30.469
    2048		1:09.969			1:10.390
    4096		2:36.031			not stable
    8192		5:39.672			not run
    16384		12:31.656			not run
    32768		27:17.047			not run
    Mainboard	Abit AN8 32X			Abit AN8 32X
    VGA Card	nvidia GeForce 8800GTS		nvidia GeForce 8800GTS
    Memory		TwinX2048-3200C2		TwinX2048-3200C2
    Memory Timings	2.0-2-2-7 1T			2.5-3-3-7 1T
    Memory Speed	202 MHz				235 MHz
    CPU FSB/HTT	283 MHz				283 MHz
     
  9. Sub.Zer0

    Sub.Zer0 (Banned or Deleted)

    Joined:
    Aug 7, 2007
    Messages:
    3,521
    Location:
    Melbourne
    Don't forget that AMD and Intel are completely different kettle of fish when it comes to RAM performance. What works best for one will most likely not be optimal for the other. AMD have on on-chip mem controller whereas Intel's is on the mobo.
     
  10. ANZACATTACK

    ANZACATTACK Member

    Joined:
    Jan 11, 2007
    Messages:
    448
    Location:
    Sydney
    So next question is how much better would it be with the RAM at stock but with the CPU OC'ed to 3.2Ghz.

    I've been wanting to do the same to my rig, as it's currently running a stock

    Will bumping the FSB up to 400 and setting the multiplier at 8 give my rig (in sig) a 1:1 ratio with my RAM? ( I guess what I'm asking is what is the benifits of running at 1:1? )
     

Share This Page