View Poll Results: What do you want to see in the next guide?

Voters
97. You may not vote on this poll
  • Samsung UCCC 2GB stability guide

    14 14.43%
  • Winbond BH-UTT low latency stability guide

    22 22.68%
  • TCCD DDR600 2.5-3-3-7 24/7 guide

    59 60.82%
  • No guides anymore, figure it out myself!

    2 2.06%
Page 1 of 20 123411 ... LastLast
Results 1 to 25 of 481

Thread: TCCD DDR600+ DFI nF4 Ultra/SLI guide

  1. #1
    Bandwidth Freak
    Join Date
    Dec 2003
    Posts
    967

    TCCD DDR600+ DFI nF4 Ultra/SLI guide

    This guide will be updated as long as there is question to it

    Introduction

    Why the need for such a guide? On several web logs there are a lot of people who upgraded their S754 platform to the new S939 with PCI Express.
    This is because the FX processor is not the only CPU anymore with the S939 socket. Most of these people want high HTT speeds to overclock their processor and memory so they buy Samsung TCCD or low latency memory like Winbond BH5/UTT.

    Recently G.Skill PC4400/4800 became the hottest memory on the market together with OCZ PC4800 based on Samsung TCCD. This memory requires a high HTT speed, so the system must be overclocked.

    A big problem that occurs when trying to reach high HTT speed is the memory that doesn’t want to run that speed although it is rated at speeds 275 or 300MHz. This results in many errors showing up in Memtest86+ and sudden BSOD in Windows. Some people might think this is caused by faulty memory, but nothing could be more untrue.

    There is another culprit here, the processor on-die memory controller. The overclocking potential of every memory controller is different, there is no doubt that there are bad and good ones. In the low- and midrange processors from AMD the quality of the controller is not always consistent, this is a problem for most of the users. This is not so much an issue if you plan to buy a high end FX processor.

    The cores available of the Athlon 64 / X2 series at this moment are:

    - Clawhammer 130nm
    - Winchester 90nm
    - Newcastle 130nm
    - San-diego 90nm
    - Venice 90nm
    - Italy 90nm
    - Paris 90nm
    - Manchester 90nm (Dual core)
    - Toledo 90nm (Dual core)

    AMD has launched two new cores named “Venice and San Diego” (revision E3 and E4). These cores have an upgraded instruction set (SSE3) and a better memory controller which is capable of addressing four memory modules in Dual Channel. (Please read the extra section SANDIEGO tweaks)

    Unfortunately the only way to find out if your CPU's memory controller is by testing it, the core revision or stepping does not tell you enough. All I can is recommend is; buy the most recent processor. This will increase the chance that you have a new memory controller.

    The chance that you will have a “bad” memory controller on the Winchester core is 6 of 10. When you have a good controller, this does not mean you will reach those speeds 100% around DDR500~600. The RAM must be tuned, that’s why I've created this guide.

    In this guide I will help you to get your TCCD memory running stable at your stock rating and above.

    I have used the following modules for testing:

    - G.Skill PC4400 1GB LE (Samsung TCCD 431) (2x512MB)


    - G.Skill PC4800 1GB FR (Samsung TCCD 440) (2x512MB)


    - G.Skill PC4800 512 LA (Samsung TCCD 440) (2x256MB)



    If you use other TCCD based memory products like OCZ, PQI, Kingston, Corsair etc. you also can use this guide, but some results may vary from what i saw with this gskill memory.


    The Goals:

    best timings=performence for

    2x 512MB at:

    275MHz (DDR550) (CL2.5-3-3-6 1T)
    300MHz (DDR600) (CL2.5-3-3-6 1T)
    310Mhz (DDR620) (CL2.5-4-4-8 1T)

    2x 256MB at:

    320MHz (DDR640) (CL2.5-4-3-6 1T)

    Clear I think

    Users with a DFI nF4 Ultra/SLI motherboard would have known that this board has a very detailed memory setting menu in the BIOS. These settings are same on the DFI nF3 UT 250Gb board. If you don’t know what all those settings mean, I recommend you to read this guide first

    All those settings are important to tune the memory to be able to run at DDR600 and beyond. For users who do not own a DFI nF4, they can use A64tweaker 0.5 for windows, this program allows you to change advanced memory timing related settings straight from within Windows.

    The Method

    To test stability, I will use a well-known program called Memtest86+. This program is implemented in the BIOS on the DFI nF4. You can download the new bios from the DFI website. When you use another board then DFI, use google to find memtest86++.

    The steps below are the steps I use to find the sweetest point of the memory.

    1. Test which module combination works best on your motherboard. Sometimes two modules works better if you switched them from place. Why? I don't know, but somehow modules can react different. For TCCD always use the orange banks. (2 and 4), for Winbond "voltages-freaks" use the yellow banks.

    2. At what speed will TCCD show her true speed? Yes DDR600+, this is somekind of must for this memory due the looser timings (2.5-3(4)-3(4)-7(8) Test the memory on e.g. DDR600 with minimum voltage. (2.6~2.7v). If you get let say 20 to 50 errors in Memtest86+ test 5,8 you can tune the memory from there. If you get plenty of errors you have to increase the voltage or lower the clock speeds.

    Use test 5 for testing the memory. This procedure will stress the memory with one patern. This test is the fastest but less error detection.

    Use test 8 for more memory testing. This procedure uses a different patern then test 5 and has a better error detection, but takes longer.

    You can run all the test by running the STD procedure, then the program will run all the test. The best and complete procedure in my opionion.

    3. If you are done with tuning on a specific speed, try to clock further with more VDDR or signal strength (Drive/Data strength).
    Or if you think it is enough you can tune that specific speed. (read the article how you can do this)

    4. When you are done with testing in memtest86++, you can try booting into windows and bench from there. NOTE Memtest86++ stable does not always mean Windows stable. There are some users having this problem like unstable windows (BSOD's) while memtest86++ is just fine.

    When you get this problem, you can add some vDDR trough your memory and/or increasing the drive and data strength.

    In this article you have seen me writing about stable speeds. My criteria of stable are a minimum pass of 5 times in Memtest86+ test 8 without errors.
    Most of the users use test 5 for stability test but I recommend test 8.

    Piece of cake you might think, but this won’t be easy especially when you are trying to get high speeds stable with 1GB ram.

    For guaranteed "Windows" stability always test with Prime95 (long period stressing, Memtest 3.2, RightMark Memory test) or houres of gaming. Just Memtest86++ is just not enough!

    Bios Version

    As you might know, DFI is a motherboard manufacturer that often releases new BIOS updates which is very great. I recommend using the latest beta bioses for the best support and performance for TCCD.

    Which TCCD is the best?

    A difficult question to answer is; what revision if TCCD is the best? At this moment there are five revisions: 425, 431, 437, 440, 449, 504, 507, 519 and 525. According to some users experience, newer TCCD are performing better at a higher voltage.

    Test Setup

    I used a competitively priced A64 Winchester CPU in test setup, rather then an overly expensive A64 FX-?? one.

    - Athlon 64 3000+ (CBBHD 0448 RPMW)
    - DFI nF4 Ultra-D rev. A02
    - G.skill PC4400 & 4800 LE/LA
    - Tagan 480Watt-U22

    I had quite a bit of luck with the on-die memory controller of the A64 3000+ as it turned it runs at DDR600 speeds flawlessly.

    Cooling: I can’t stress this enough; use an 80mm fan or bigger to cool down your memory as it has a very large impact on the result. At lower speeds and voltage the TCCD chips hardly get hot, but when pushing the envelope they start cooking and this will deter the overclocking potential.

    2x 512MB DDR550 2.5-3-3-6 1T Dual Channel





    BIOS Settings

    Vddr: 2.7v
    Tcl: 2.5
    Trcd: 3
    Tras: 6
    Trp: 3
    Trc: 9
    Trfc: 12
    Trrd: 2
    Twr: 2
    Twrt: 2
    Trwt: 2
    Tref: 3072 (200MHz 3.9us)
    Twcl: 1

    Bank int.: Enable
    Skew control: 255 +
    Drive strength: 7
    Data strength: 4
    Max. Async: 7ns
    Read Preamble: 5ns
    Idle Cycle: AUTO
    Dyn. Counter: Enable
    R/W Bypass: 16x
    Bypass Max: 7x
    32Bit Gran.: Disable

    With these settings I got high performance and stability. The Tref (Refresh interval) can set to 4078 (166MHz 1.95us) for better performance.

    If the settings mentioned above prove not to be stable, you can change the following:

    - Max. Async. to 8ns
    - Lower the refresh interval (Tref)
    - Trrd to 3
    - Idle Cycle at 128/256


    These settings will increase stability and keep the performance at a descent level. If those settings are still unstable, there is something wrong with your memory or memory controller. You make sure your memory is not the issue; test the memory modules separately in bank 1 or 2.

    When the memory works fine, then the problem must be the memory controller. You should disable the Command per Clock (CPC). If the problem is still there, you should consider lowering the speeds or, as a last resort, buy another processor.

    2x 512MB DDR600 2.5-3-3-6 1T Dual Channel





    BIOS Settings

    Vddr: 2.7v
    Tcl: 2.5
    Trcd: 3
    Tras: 6
    Trp: 3
    Trc: 9 (12)
    Trfc: 14
    Trrd: 2
    Twr: 2
    Twrt: 2
    Trwt: 2
    Tref: 3072 (200MHz 3.9us)
    Twcl: 1

    Bank int.: Enable
    Skew control: 255 +
    Drive strength: 7
    Data strength: 2 (4)
    Max. Async: 8ns
    Read Preamble: 5ns
    Idle Cycle: AUTO
    Dyn. Counter: Enable
    R/W Bypass: 16x
    Bypass Max: 7x
    32Bit Gran.: Disable


    The settings above are set for high performance and stability. The Tref can be tuned higher to e.g. 1.95us for better performance. Tref=3072 is 200MHz 3.9us refresh interval. You can also set Trfc to 12.

    When those settings are not stable, you can change the following:

    - Max. Async. to 8ns
    - Lower the refresh interval (Tref)
    - Trrd to 3
    - Twr to 3
    - Trwt to 3
    - Idle Cycle at 128/256


    If those settings are still unstable, be careful working with a higher voltage. TCCD does not like high voltage compare to Winbond BH5. Maybe the newer revision TCCD like higher voltage but this one doesn’t. Always start with a lower voltage and increase the voltage step by step to a satisfied value. As you see I only used 2.7v. higher voltage 2.8 ~ 2.9v only gives me errors in Memtest86+!

    Always try to run at CL2.5-3-3-6 1T to keep the performance at a high level. But if there is no other way to get this stable you can raise the Trcd to 4. This will decrease the bandwidth by 100-150Mb.

    Test known-how’s

    - Data strength 2 is the best value for stability, I recommend not increasing or decreasing it. But always test which setting is best for you.

    - Drive strength 7 is the best value for stability, I recommend not increasing or decreasing it.

    - Always set the bank interleave to enable when you are using two modules. This will increase the performance and stability.

    2x 512MB DDR620 2.5-4-4-8 1T Dual Channel





    BIOS Settings

    Vddr: 2.7v
    Tcl: 2.5
    Trcd: 4
    Tras: 8
    Trp: 4
    Trc: 7
    Trfc: 14
    Trrd: 2
    Twr: 2
    Twrt: 2
    Trwt: 2
    Tref: 3072 (200MHz 3.9us)
    Twcl: 1

    Bank int.: Enable
    Skew control: 0 (off)
    Drive strength: 7
    Data strength: 2
    Max. Async: 9ns
    Read Preamble: 5ns
    Idle Cycle: AUTO (16)
    Dyn. Counter: Enable
    R/W Bypass: 16x
    Bypass Max: 7x
    32Bit Gran.: Disable

    The settings above are set for high performance and stability.

    When those settings are not stable, you can change the following:

    - Read Write Bypass to 8x
    - Bypass Max to 4x
    - Idle Cycle at 128/256
    - Set DQS Skew to 128, Decrease


    If those settings are still unstable, be careful working with a higher voltage. TCCD does not like high voltage compare to Winbond BH5. Maybe the newer revision TCCD like higher voltage but this one doesn’t. Always start with a lower voltage and increase the voltage step by step to a satisfied value. As you see I only used 2.7v. higher voltage 2.8 ~ 2.9v only gives me errors in Memtest86+!

    Test known-how’s

    - Decreasing the DQS Strobe timing gives the memory a huge stability boost, and little performance loss.

    2x 256MB DDR640 2.5-4-3-6 1T Dual Channel

    Unfortunately I cannot do 10 pass stable with Memtest86+ at 320MHz CL2.5-3-3-6 1T. I only get 5 pass stable. The problem is my memory, one module being not quite as good as the other.



    BIOS Settings

    Vddr: 2.9v
    Tcl: 2.5
    Trcd: 4
    Tras: 6
    Trp: 3
    Trc: 9
    Trfc: 16
    Trrd: 3
    Twr: 3
    Twrt: 3
    Trwt: 2
    Tref: 3072 (200MHz 3.9us)
    Twcl: 1

    Bank int.: Enable
    Skew control: 128+
    Drive strength: 7
    Data strength: 1
    Max. Async: 9ns
    Read Preamble: 5ns
    Idle Cycle: AUTO
    Dyn. Counter: Enable
    R/W Bypass: 16x
    Bypass Max: 7x
    32Bit Gran.: Disable

    The settings above are set for high stability. These settings are stable for better modules. So you have to test it by yourself. If these settings are stable with your ram, you can lower the Max. Async. to 8ns for even better performance. You also can lower the Trrd and Twr to 2.

    Tref is set 3072 for stable refreshrates. For benching I recommend a faster refresh rated 1.95us.

    For better stability you can set the Idle Cycle time to 128 or 256.

    Increasing the Trfc by 1 also will increase some stability to your memory.


    Test known-how’s

    - When you use 2x256MB TCCD modules, I recommend you not to use to much signal strength.

    Using 4x512MB of RAM, The Tweaks

    Nowadays, 2 Gigabytes of RAM is not unpossible. It is affordable, and not a overkill anymore. Games and mostly graphics-based sofware need that much of RAM. Should you upgrade to 4x512MB RAM if you play alot of games?
    The answer is YES, it really helps in shorting loading times and reduce alot of frames drops.
    Using 2GB of RAM, the speed /timings of the RAM will be looser and the memorycontroller stress will be increased. Sometimes a little more Vcore is needed. Running at lower speeds like 400MHz-500MHz is not a very big problem, but running at high speeds especially with TCCD DDR550~600 is a hard goal to achieve.

    The most important setting change when using 4x512MB is that the system must run 2T command rate. 1T will fail with the nowadays memorycontroller even for the Sandiego/Venice memory-controller.
    Further changes are longer asynchronic latency and preamble read latency.

    Here I have made some test to achieve the maximum stable speeds with 4x 512MB PC4800 DDR600 RAM. Unfortunately my controller cannot do DDR600, even with looser timings (2.5-4-4-8 2T). The maximum stable lies around 270-280MHz 2.5-3-3-7 1T and 290Mhz 2.5-4-4-8.



    BIOS Settings

    Vddr: 2.8v
    Tcl: 2.5
    Trcd: 3
    Tras: 6
    Trp: 3
    Trc: 7
    Trfc: 14
    Trrd: 0
    Twr: 2
    Twrt: 1
    Trwt: 2
    Tref: 3072 (200MHz 3.9us)
    Twcl: 1

    Bank int.: Disable
    Skew control: AUTO
    Drive strength: 8
    Data strength: 3
    Max. Async: 11ns
    Read Preamble: 8.5ns

    Idle Cycle: AUTO (16)
    Dyn. Counter: Enable
    R/W Bypass: 16x
    Bypass Max: 7x
    32Bit Gran.: Disable

    As you can see, the major changes are stated bold. These are required when cranking up the speeds. I can tell you, tweaking 2048M of RAM is hard .

    When you don't know at what value you should set the read preamble and Max. async, just set AUTO and check A64tweaker in windows. The BIOS automatically set the right setting. Most of time enough to make it stable. This "feature" makes it easier to tweak. Let BIOS do the first step and you do derest.

    Overall, the timings are slower. I know. 2T, longer latency timings. BUT, what you get is must better that is for sure!

    Is more better? yes only when the software ask for it

    SAN DIEGO Memory Tweaks

    As now the Venice and Sandiego 90nm cores are fully available, these controllers are quiet different to the Newcastle/Clawhammer controllers.

    There are changes in the bandwidth, the new cores especialy the Sandiego has a lower bandwidth then the older cores. How this happen is still unknown, soon some clear information will be released. The latency are still the same though.

    I have tested a Sandiego 3700+ 1MB processor with memorymodules like PC4800 2.5-4-4-8 1T and PC4800 2.5-4-4-8 2T.

    Definitely this core needs some more tweaking then a clawhammer core. My memorycontroller can be lacky but I post my experiences for those who have troubles getting their memory stable at STOCK clocks.

    When you run DDR600 2.5-4-4-8 or 2.5-3-3-6 1T you can turn the Bank interleave modus to disable. This might improve the stability on some memorycontrollers.

    Second, Decrease the DQS with value 128. This stabelize the Memtest86++ TEST 8 procedure which is more accurate then TEST5.
    Please use TEST 8 for memory test procedures instead of TEST 5, plus the errors most of the time occures at 10% , 30% and 60-70% procedure.

    Also make sure these settings are at:

    - Drive strength 7
    - Data strength AUTO or 2

    I hope this guide will help you on your road to DDR600 speeds with your brand new S939 setup!

    Kakaroto,

    Thnx [M] for improving the translation work!

    Very Handy Articles related to Tweaking TCCD

    - What is DQS? Discuss it! ~by OPB!

    updates

    1 april 2005 - added new TCCD series.
    12 april 2005 - small change Tref setting.
    25 may 2005 - San Diego memory tweaks section added.
    27 may 2005 - Add DDR620 2.5-4-4-8 1T stability guide
    11 juli 2005 - Add 2048MB RAM stability guide
    7 januari 2006 - some minor changes
    Last edited by kakaroto; 03-06-2006 at 02:20 PM.

  2. #2
    Xtreme Addict
    Join Date
    Oct 2004
    Location
    Boston, MA
    Posts
    1,448
    Thanks for the great guide, its gonna help me a lot once I finally get this rig running

    File Server:
    Super Micro X8DTi
    2x E5620 2.4Ghz Westmere
    12GB DDR3 ECC Registered
    50GB OCZ Vertex 2
    RocketRaid 3520
    6x 1.5TB RAID5
    Zotac GT 220
    Zippy 600W

    3DMark05: 12308
    3DMark03: 25820

  3. #3
    XIP
    Join Date
    Nov 2004
    Location
    NYC
    Posts
    5,523
    Quote Originally Posted by HiJon89
    Thanks for the great guide, its gonna help me a lot once I finally get this rig running
    Thats your LE....431 Kakaroto's #892

  4. #4
    Xtreme Addict
    Join Date
    May 2004
    Posts
    1,755
    Very nice, will probably help tons when i'll finaly be able to put together the 4x512 TCCD sticks that should be there next week with one of those Venices and a DFI if the Gruper is late to the party.
    You forgot the rev 504 TCCD in your list

  5. #5
    Registered User
    Join Date
    Jan 2005
    Location
    Texas
    Posts
    334
    Yep, great guide by [M]


    Got news? Post it HERE

  6. #6
    Bandwidth Freak
    Join Date
    Dec 2003
    Posts
    967
    Quote Originally Posted by SLaY3r07
    Yep, great guide by [M]
    I wrote this guide Gave permission to post on [M] and to improve my translation.

  7. #7
    Registered User
    Join Date
    Jan 2005
    Location
    Texas
    Posts
    334
    Quote Originally Posted by kakaroto
    I wrote this guide Gave permission to post on [M] and to improve my translation.
    Yeah I just saw this guide on [M]. Great guide kakaroto


    Got news? Post it HERE

  8. #8
    Bandwidth Freak
    Join Date
    Dec 2003
    Posts
    967
    If anyone have questions or other related problem with the guide you can post here.

    For G.Skill memory related problems you can post here: http://xtremesystems.org/forums/showthread.php?t=55229

    Any suggestions are most welcome!

  9. #9
    Xtreme X.I.P.
    Join Date
    Aug 2004
    Location
    Chile
    Posts
    4,151
    awsome kakaroto really nice i linked it a chilean forum, i hope you dont mind

  10. #10
    Registered User
    Join Date
    Dec 2004
    Location
    Worcester UK
    Posts
    14
    Very Nice! I'm sure this will help me. Thanks very much.

  11. #11
    Xtreme Member
    Join Date
    Feb 2005
    Location
    On a celestial body that revolves around a fusion ball of Hydrogen
    Posts
    111
    Kakaroto, this guide will be xtremely helpful for me and Im sure many others, great job, thanks.
    ASUS A7N8X Deluxe Rev II ~ Barton 2500 AQXEA0323 @ 2420MHZ
    2x512MB Corsair 3200LLPT BH-5 ~ BFG AGP 6800GT OC
    2x80GB Seagate SATA in RAID 0 ~ SB Live Plat 5.1 ~ Live Drive
    Princeton LCD19D ~ Enermax EG651P-VE FMA ~ Enermax EG465P-VE(D)
    DD M4 ~ M4 A ~ BIX II ~ Eheim 1260

    DFI NForce4 SLI-DR ~ Winnie 3200 CBBHD 0449 MPM @ 2700MHZ
    2x512MB G.Skill 4800LA ~ 2xBFG 8800GTX
    2x74 GB Raptors in RAID 0 ~ Audigy II ZS
    ViewSonic VP201s ~ PCP&C 510 SLI ~ Enermax EG465P-VE FMA
    DD TDX ~ 2xM4 A ~ Weapon ~ Iwaki MD20

  12. #12
    xtreme energy
    Join Date
    Oct 2004
    Location
    Europe, Latvia
    Posts
    4,145
    Thanks.
    I still have to find the best timings and speed for my pair of sticks but this guide saves a lot of time.
    ...

  13. #13
    Xtreme Guru
    Join Date
    Dec 2003
    Location
    Vancouver, Canada
    Posts
    3,858

    Cool

    Thanks for sharing your experience Kakaroto. I will be sure to look at this when I am building my new system
    i5 750 4.20GHz @ NH-D14 | 8GB | P7P55DLE | 8800U | Indilinx SSD + Samsung F3 | HAF922 + CM750W
    Past: Q6600 @ 3.60 E6400 @ 3.60 | E6300 @ 3.40 | O165 @ 2.90 | X2 4400+ @ 2.80 | X2 3800+ @ 2.70 | VE 3200+ @ 2.80 | WI 3200+ @ 2.75 | WI 3000+ no IHS @ 2.72 | TBB 1700+ @ 2.60 | XP-M 2500+ @ 2.63 | NC 2800+ @ 2.40 | AB 1.60GHz @ 2.60
    Quote Originally Posted by CompGeek
    The US is the only country that doesn't use [nuclear weapons] to terrorize other countries. The US is based on Christian values, unlike any other country in the world. Granted we are straying from our Christian heritage, but we still have a freedom aimed diplomatic stance.

  14. #14
    Registered User
    Join Date
    Feb 2005
    Posts
    7

    Thumbs up

    This guide is awesome, thank you very much for your time kakaroto .
    It helped me push my 1GB LA past the DDR620 territory, memtest stable for hours.
    The main thing I never considered was enabling the dynamic counter which seems to help a lot. Setting the drive strength to 7 was also a great tip (it turned out to be much more stable than 8 which I always used).
    Once again, thank you !
    Athlon 64 3200+@2700 (1.45Vcore)| DFI LP nF4-Ultra D (217 Bios)| 1 GB (2x512mb) G.Skill PC4800 LA (2.5-3-3-6 @ 270 , 3.0Vdimm)| Tul Radeon X850XT | Enermax Noisetaker V2.0 485w PSU | Zalman 7700 AlCu | Seagate 160GB Sata NCQ

  15. #15
    Xtreme Member
    Join Date
    Dec 2004
    Posts
    455
    this is an awesome, very CLEAR and detailed guide !

    I am looking for a LONG time already on many forums for TCCD optimisations around 300 HTT or so....problem is mostly you find "general" advice for people having problems at stock.....RARELY a complete and detailed advice HOW (!) exactly and with which settings to go 300 etc.


    I am trying for WEEKS already to run my OCZ TCCD REV2 at 2.5 3 3 7 (instead of 2.5 4 3 7) running at 289.....and i will try your tips tomorrow starting with decreasing my voltage etc...

    VERY NICE and thanx again !

  16. #16
    Xtreme Member
    Join Date
    Dec 2004
    Posts
    455
    hey, i know this is mainly for Gskill....anyway i used this guide for my OCZ 3200 el rev2 plats and i can tell with CERTAINITY that they dont do 289 @ 2.5 3 3 7.
    [i am running a 3500 winchester at 9x289/288, i am positive my CPU does 2600mhz)

    The funny thing is ONE freaking mhz less (ie. 288) and i can run them finally 2.5 3 3 7 at 288. I dont know how many *hours* i tweaked and memtested.....but this ONE mhz was which almost drove me into insanity.

    Eg. w/ 289 i was so on the edge that i passed 9 runs memtest #5 and got some errors on the 10th run ---> instable. My criteria is 15 runs of memtest #5 to consider them stable right now.

    I can also say with certainity that *my* OCZ sticks run best at 2.9V....started out 2.7/2.8 but 2.9 is the golden sweet spot and 3.0 is the same...so 2.9V is it, no more, no less. memtest #5 is a nice indicator for needed voltage.

    One finding:

    I think my efforts (and also the recommendations to IF ANY POSSIBLE run 2.5 3 3 7 instead of 2.5 4 3 7) are more "theoretical" than doing anything in real life.

    My mem bandwidth (after tweaking) in memtest is always around 3176MB/s...and after i finally managed to run the sticks 2.5 3 3 7 (with 9mhz less on the CPU since i changed from 9x289 to 9x288 but with TRCD@3) i did my benchies.

    Eg. Farcry 800x600 Hardware OC benchmark (which is good for testing latency/mem bandwidth impact etc.) has the same results as i have with the former TRCD @4 - and everest/memtest etc. is all about the same.

    I cannot confirm that TRCD at 3 is some magic goal to achieve for great improvement...at least i dont see ANYTHING which would make me conclude that in any benchmark results !
    In other words:
    If you run 2.5 4 3 7 in the 290-3xx HTT range...and you get ~3200MB/S in memtest then you might be well off and can save yourself the time to go to 2.5 3 3 7 since it doesnt do ANYTHING. Zero !

    Feel free to prove me wrong by showing me a benchmark where i can see a huge impact by changing TRCD 4-->3. At least not in my case and at speeds in the HTT range 290-3xx.



    NOTES:

    Beware of "async latency"....eg. set it to 9ns and it passes memtest.....but i get bad crashes in windows. So leave at 8ns or 7ns.

    Beware of every memory benchmark in windows...especially everest since the results can change GREATLY - depending when you run it.
    GREAT results right after boot - do some stuff, run everest again.....and the values hit rock bottom. This is hardly reliable and usable for comparison w/ other people.
    IMHO the only halfway reliable measurement for bandwidth is from within memtest.

  17. #17
    Xtreme Addict
    Join Date
    Oct 2004
    Location
    Boston, MA
    Posts
    1,448
    I'm at 8x300 now thanks to this guide, thanks a ton

    File Server:
    Super Micro X8DTi
    2x E5620 2.4Ghz Westmere
    12GB DDR3 ECC Registered
    50GB OCZ Vertex 2
    RocketRaid 3520
    6x 1.5TB RAID5
    Zotac GT 220
    Zippy 600W

    3DMark05: 12308
    3DMark03: 25820

  18. #18
    Xtreme Enthusiast
    Join Date
    Jan 2005
    Location
    Canada, Qubec
    Posts
    648
    Great guide!!! Thanks a lot it helped me with my ocz plat rev2. Nevetheless, I was unable to run them 2.5-3-3-7 @ 300 htt but it was close! I always get some 6-8 errors each 10 pass of test 5 so I run them 2.5-4-3-7.

  19. #19
    Banned
    Join Date
    Jan 2004
    Location
    Land of Buckeye
    Posts
    2,881
    Definitely the TCCD bibble here.

  20. #20
    Xtreme Member
    Join Date
    Dec 2004
    Posts
    455
    Quote Originally Posted by Charloz24
    Great guide!!! Thanks a lot it helped me with my ocz plat rev2. Nevetheless, I was unable to run them 2.5-3-3-7 @ 300 htt but it was close! I always get some 6-8 errors each 10 pass of test 5 so I run them 2.5-4-3-7.
    see my reply

    maybe ONE reason is that the OCZ need more voltage, at least mine do..they love 2.9V. And they get HOOOTTT Thats probably the reason we can run 9 or so passes fine...and then slowly the errors are popping up because of heat issues.


    In the guide he uses 2.7V for his Gskills btw.

  21. #21
    Xtreme Addict
    Join Date
    Dec 2004
    Location
    The Holy Land
    Posts
    1,205
    Quote Originally Posted by Onepagebook
    Definitely the TCCD bibble here.

    and you right my friend
    Asus P5-E X38 Intel QX9650 INOO3D 9800 Gx2 4x Super Talent's T800UX2GC4
    Creative X-Fi Raid 0 Raptors 36Gb data 1Tb Seasonic M12 700w
    Dell 2407WFP Lian-Li Pc-65 mod Custom Built Phase Change Unit |Logitech G25|

    RIG#2: DFI LANPARTY LT X48-T2R|Q9550| INOO3D 9800GTX|
    2xCellshok 1000| PC Power and cooling 750w Thermaltake v1
    Wife: DFI LANPARTY UT P35-T2R|Q930|BFG 8800GTS 512MB|
    Samsung ddrII 667| PC Power and cooling 610eps12v Thermaltake v1

    Htpc:Dfi Sli-Dr Venus|3700+|2x1gb Team Group ddr500|Thermaltake Mozart TX


  22. #22
    Registered User
    Join Date
    Jan 2005
    Location
    Catalonia
    Posts
    70
    nice guide, i can reach 305x9 (2.5-4-3-7) but unstable , maybe for memory controller.
    ASUS P5Q Deluxe / Q6600 G0 3.2Ghz (400x8) @ 1.31V / 2x2Gb G.Skill F2-8800 PC2-1100 1.90V / Corsair HX620 / Noctua NH-U12P / 1x7200.11 640 Gb / BenQ G2400W / Asus DRW-2014L1T / Windows XP uE 2008.1

  23. #23
    Bandwidth Freak
    Join Date
    Dec 2003
    Posts
    967
    Quote Originally Posted by flexy
    hey, i know this is mainly for Gskill....anyway i used this guide for my OCZ 3200 el rev2 plats and i can tell with CERTAINITY that they dont do 289 @ 2.5 3 3 7.
    [i am running a 3500 winchester at 9x289/288, i am positive my CPU does 2600mhz)

    The funny thing is ONE freaking mhz less (ie. 288) and i can run them finally 2.5 3 3 7 at 288. I dont know how many *hours* i tweaked and memtested.....but this ONE mhz was which almost drove me into insanity.

    Eg. w/ 289 i was so on the edge that i passed 9 runs memtest #5 and got some errors on the 10th run ---> instable. My criteria is 15 runs of memtest #5 to consider them stable right now.

    I can also say with certainity that *my* OCZ sticks run best at 2.9V....started out 2.7/2.8 but 2.9 is the golden sweet spot and 3.0 is the same...so 2.9V is it, no more, no less. memtest #5 is a nice indicator for needed voltage.

    One finding:

    I think my efforts (and also the recommendations to IF ANY POSSIBLE run 2.5 3 3 7 instead of 2.5 4 3 7) are more "theoretical" than doing anything in real life.

    My mem bandwidth (after tweaking) in memtest is always around 3176MB/s...and after i finally managed to run the sticks 2.5 3 3 7 (with 9mhz less on the CPU since i changed from 9x289 to 9x288 but with TRCD@3) i did my benchies.

    Eg. Farcry 800x600 Hardware OC benchmark (which is good for testing latency/mem bandwidth impact etc.) has the same results as i have with the former TRCD @4 - and everest/memtest etc. is all about the same.

    I cannot confirm that TRCD at 3 is some magic goal to achieve for great improvement...at least i dont see ANYTHING which would make me conclude that in any benchmark results !
    In other words:
    If you run 2.5 4 3 7 in the 290-3xx HTT range...and you get ~3200MB/S in memtest then you might be well off and can save yourself the time to go to 2.5 3 3 7 since it doesnt do ANYTHING. Zero !

    Feel free to prove me wrong by showing me a benchmark where i can see a huge impact by changing TRCD 4-->3. At least not in my case and at speeds in the HTT range 290-3xx.



    NOTES:

    Beware of "async latency"....eg. set it to 9ns and it passes memtest.....but i get bad crashes in windows. So leave at 8ns or 7ns.

    Beware of every memory benchmark in windows...especially everest since the results can change GREATLY - depending when you run it.
    GREAT results right after boot - do some stuff, run everest again.....and the values hit rock bottom. This is hardly reliable and usable for comparison w/ other people.
    IMHO the only halfway reliable measurement for bandwidth is from within memtest.
    First, thank you for sharing your experience!

    The difference between Trcd 4 to 3 can only be seen in benchmarks.

    If you test Sandra 2004/2005 you can see some bandwidth loss when running trcd 4 instead of 3.

    Here a sample with Pifast calculating 1M pi:



    MEMTEST86++ bandwidth is not accurate. I have experienced heavy bandwidth difference when changing Trcd, Tref, Twr and Trrd, sometimes 100-200MB difference.
    The bandwidth is also related to your processor speed, not constant like on the S754 platform.
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	trcd_Test.JPG 
Views:	89155 
Size:	136.6 KB 
ID:	27598  
    Last edited by kakaroto; 03-28-2005 at 01:23 PM.

  24. #24
    Registered User
    Join Date
    Jan 2005
    Location
    Catalonia
    Posts
    70
    nice guide for gskill, in my ocz plat rev2 i need to put several settings to auto to run memtest without errors. So settings for gskill not for ocz tccd
    ASUS P5Q Deluxe / Q6600 G0 3.2Ghz (400x8) @ 1.31V / 2x2Gb G.Skill F2-8800 PC2-1100 1.90V / Corsair HX620 / Noctua NH-U12P / 1x7200.11 640 Gb / BenQ G2400W / Asus DRW-2014L1T / Windows XP uE 2008.1

  25. #25
    Xtreme Member
    Join Date
    Dec 2004
    Posts
    455
    hi kakaroto,

    thanks for the posting and the screenshots.


    i have more information coming after a weekend and today tweaking and benching and priming without end Btw. i have a winnie week 42 3500+, dfi ultra-d, 310p bios, the two OCZ 3200 rev2s...X-120 heatsink

    Ok...first...you are right, cooling is EVERYTHING !

    I put a 80mm Sunon Fan over my sticks and BINGO ! - i could tighten my timings a lot and finally could run them 289 2.5 3 3 7 1T with the tightest timings possible.

    I tested the new setting (now with fan blowing) 15 runs memtest #5, pass memtest #8 and one complete memtest run....and i did not HAVE one error.

    Also...after priming (better OCCT-ing since OCCT is even more sensitive than prime) i *know* that my CPU is good for 2600.

    So...i took the new memory timings (which were actually the SAME like my 2.5 4 3 7 timings except the "4").....and as said i pass memtest like a champ....but then disaster struck in Windows !

    OCCT/Prime...i couldnt even run 5 secs with the VERY SAME settings which passed memtest..and at 2601mhz and same voltage as i had before.

    So...i took a64tweaker and started to tweak back (relax) many timings...kept the "3" TRCD at any cost but tweaked all the usual stuff down....but for nothing on earth i could get prime/occt running more than 7mins without it complaining - and then i already tweaked down all the nice settings and lost bandwidth.....but prime STILL complaining.

    After hours and hours and experimenting i am now back to my old 2.5 4 3 7 settings..and i am priming right now (just to double check) and of course everything is fine.

    What i want to say:

    If you tweak the memory - well we all know the memory controller is on the CPU. Change TRCD (or anything) from 4 to 3 - and your CPU will have a totally different overclocking behaviour (which makes sense - kind of).
    The relationship CPU/memcontroller/memory is that way that tweaking the one thing will influence the other. Maybe this is often overseen.

    The result: You might run memtest for hours...but you might get a surprise in Windows...and especially once you start testing with prime/occt.

    I am really still kind of surprised since i did not expect that GREAT of problems, eg. not even 5 secs stable in OCCT is *serious* instable - i also had errors like page faults and BSODS whatever - this just to show you that memtest...well....i dont know. What does it mean when i can pass memtest for hours but cant pass 5 secs in OCCT ?

    It MIGHT be that these highly tweaked settings are only good for CPUs at stock speed....eg. you can run your memory at those speeds and the CPU can handle the tight latency if its not overclocked....and if the CPU is already on the edge (like mine w/ 2600mhz)...it just cannot handle the latency ?

Page 1 of 20 123411 ... LastLast

Bookmarks

Bookmarks

Posting Permissions

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