Page 4 of 8 FirstFirst 1234567 ... LastLast
Results 76 to 100 of 177

Thread: OC Report - Cellshock DDR3 1866 BLUE D9JNL

  1. #76
    silver wall jumper X
    Join Date
    Apr 2005
    Location
    Las Vegas
    Posts
    1,579
    Quote Originally Posted by Gautam View Post
    Heh, all this talk about clock skew is giving me an idea for that Foxconn article contest.
    Something along those lines?



    Mike's quick and dirty guide to tuning your Clock Skew settings:


    1. Setup BIOS

    Setup your BIOS/System for your max stable settings in the particular benchmark you want to tweak your memory for. Make sure that the system can handle at least an extra 20mhz FSB 100% stable - you can check this by just using a lower memory divider, while keeping CPU, FSB clocks, voltages, etc. intact. Have the clock skew set to “Normal” on channel A+B.


    2. Find Max Stable Setting with “Normal” Clock Skew

    Run the benchmark/test at the max stable mem clocks. Now you have your maximum stable memory settings before the clock skew tune.

    3. Find best Clock Skew Difference Value


    Now check first to see if a different clock skew between channel A and B helps. First only advance channel B by the minimum possible value, check results with your benchmark - Next only delay Channel A with Channel B on Normal, check results, and then reverse delays with advances ...
    example:
    Test 1: Channel A = Normal / Channel B = Advance 25ps
    Test 2: Channel A = Delay 25ps / Channel B = Normal
    Test 3: Channel A = Advance 25ps / Channel B = Normal
    Test 4: Channel A = Normal / Channel B = Delay 25ps

    might only be necessary to run Test 1 + Test 2

    After four quick tests you should be able to tell if different values for channel A and B help to stabilize your clocks, or if not you'll know that both channels require an equal value. If different values are beneficial - see if extending the difference shows even more improvement.
    You can try advancing and delaying simultaneously - or keeping one channel normal while further extending the difference on the other channel until you find the best difference value.

    example
    Test 5: Channel A = Delay 25ps / Channel B = Advance 25ps
    Test 6: Channel A = Delay 50ps / Channel B = Normal
    Test 7: Channel A = Normal / Channel B = Advance 50ps

    might only be necessary to run one type of the three variations

    4. Find best Combined Clock Skew Value

    Now that you have best difference (or know that an equal value is best), check to see if advancing or delaying both channels will give you any further gains, while keeping the difference from Step 3. between channel A + B ( or keeping equal value) . You should be able to see a trend quickly if you need to delay both channels or advance both channels and be able to zoom in on the best combination for your clock skews.


    Examples ( with Clock Skew Difference from step 3 e.g. = Channel A delay 25ps) :

    Test 8: Channel A = delay 150ps / Channel B = delay 125ps
    Test 9: Channel A = advance 125ps / Channel B = Advance 150ps

    5. Re-adjust Clock Skew Difference Value

    You can retest the Clock Skew Difference Value now that you have advanced or delayed by reducing or extending the difference again – but most likely the value will remain identical.

    6. Enjoy your improved clocks!


    Clock Skews might need to be re-adjusted for different latencies, voltages, dividers and even benchmarks. They should nevertheless be not too far off from what you get from your first clock skew tune.

    Clock Skews settings will be different for each different motherboard - even if it is the same model. Don't copy other peoples clock skew settings - they might make things a lot worse for you.

    Make sure you remember which order you memory sticks were in the motherboard, as well which slots were used – you might need adjustment if the order is reversed.

  2. #77
    Xtreme X.I.P.
    Join Date
    Nov 2002
    Location
    Shipai
    Posts
    31,147
    sweet, so the extra vdimm scaling on that board makes up for the slighlty worse mhz per vdimm right?

    in theory, the better and cleaner and more stable vdimm is, the less of it you need to get the same clocks stable, right? so then the other board should have "better" vdimm than this one? maybe this one just clocks better cause of a better nb but vdimm quality is actually worse? just speculating here...
    so your pushing for max cas8 clocks now?
    does it actually help perf wise? I always thought the few extra 100mhz or 200mhz wasnt really worth dropping to cas8.
    but then again, with ddr3 mhz really seems to have a much bigger impact than timings.

    about clock skew, so there are two ways it actually helps?
    advancing or delaying both channels, and then finding the right delay betweeen channel a and channel b? interesting...

    thanks for the details mike! great work

  3. #78
    Xtreme 3DTeam Member
    Join Date
    Aug 2006
    Location
    Where You don't get nothin' for $
    Posts
    1,001
    Mike - do You still have some BIOS settings from the MAXIMUS EXTREME?

    I received my kit yesterday and played for a few hours now, screen will follow later.

    I can do 8x500 strap 266 at 1000MHz 8-8-8-24 with only 1.87 Vdimm, thats nice . My DOMINATORS were not able to boot into WIN XP at 1000MHz on this board, no matter of volts or timings.

    Sadly I can not run 7-7-7-x with the Cellshocks, system posts at 2.06 Vdimm but no boot into WIN XP, either I get a BIOS failure message to recover it or XP gives the message of no starting caused by a broken file. Tried also higher volts, no way. But when I jump back to 8-8-8-x all is fine again, of course no broken XP or corrupt BIOS. Very strange behaviour......... - never had this before.
    If nothing works nomore......:


  4. #79
    silver wall jumper X
    Join Date
    Apr 2005
    Location
    Las Vegas
    Posts
    1,579
    Quote Originally Posted by saaya View Post
    sweet, so the extra vdimm scaling on that board makes up for the slighlty worse mhz per vdimm right?

    in theory, the better and cleaner and more stable vdimm is, the less of it you need to get the same clocks stable, right? so then the other board should have "better" vdimm than this one? maybe this one just clocks better cause of a better nb but vdimm quality is actually worse? just speculating here...

    Yes the extra scaling makes up for the lost clocks - on one hand the theory does make sense; on the other hand maybe this board has cleaner vdimm - maybe because of that it can handle a higher vdimm as compared to other boards. Or this boards droops vdimm extremely. I would assume if there is a vdimm wall for max vdimm - it should remain that wall - and "uncleaner" voltage should actually make things worse. But since the memory can handle vdimm better - I tend to suspect that the vdimm is cleaner. Also just speculating. But I was at first shocked to see how much I lost. Shows again how much we are dependant on a good motherboard.

    It is possible that the first board could have scaled even better - I was on that board too short to be sure it might have not scaled just like this board. Then the dirtier vdimm makes sense yet again For sure on the first board I running 7/7/6 - so this new P5E3 def. is not as good.
    Also on this P5E3 I am doing a lot worse on Cas 6, then I was before. All confusing...

    Quote Originally Posted by saaya View Post

    so your pushing for max cas8 clocks now?
    does it actually help perf wise? I always thought the few extra 100mhz or 200mhz wasnt really worth dropping to cas8.
    but then again, with ddr3 mhz really seems to have a much bigger impact than timings.
    Was just trying to see if I could somehow get to DDR3 2300mhz for the screeny - but the CPU has has me shut down at 2260mhz. But I assume perfromance wouldn't really improve unless I was running DDR3 2150
    MHZ Cas8 1T ( which this set cannot - 2080 1T is the wall for this set )

    Quote Originally Posted by saaya View Post
    about clock skew, so there are two ways it actually helps?
    advancing or delaying both channels, and then finding the right delay betweeen channel a and channel b? interesting...

    thanks for the details mike! great work
    Yes correct - finding the difference between channel A + B might help you to tune differences between your channels - for example due to different trace lengths; eg. Channel A has shorter traces and Channel B has longer traces - hence it takes a bit longer for the signals to travel and syncing them could help.

    I have a feeling that auto settings start increasing those Clock Skew Differences at high memory speeds, which in actually might not be needed, and hence harm the perfromance.
    I have noticed that AUTO CPU gTL settings often follow this logic - and assume that BIOS designers do the same for memory.


    I dunno why why advancing / delaying both channels actually helps - but I have gotten the biggest gains from doing so.
    Last edited by mike; 04-25-2008 at 12:11 PM.

  5. #80
    silver wall jumper X
    Join Date
    Apr 2005
    Location
    Las Vegas
    Posts
    1,579
    Quote Originally Posted by camouflage View Post
    Mike - do You still have some BIOS settings from the MAXIMUS EXTREME?

    I received my kit yesterday and played for a few hours now, screen will follow later.

    I can do 8x500 strap 266 at 1000MHz 8-8-8-24 with only 1.87 Vdimm, thats nice . My DOMINATORS were not able to boot into WIN XP at 1000MHz on this board, no matter of volts or timings.

    Sadly I can not run 7-7-7-x with the Cellshocks, system posts at 2.06 Vdimm but no boot into WIN XP, either I get a BIOS failure message to recover it or XP gives the message of no starting caused by a broken file. Tried also higher volts, no way. But when I jump back to 8-8-8-x all is fine again, of course no broken XP or corrupt BIOS. Very strange behaviour......... - never had this before.

    I started out with around 1950mhz Cas 7 at first myself, and then tuned/tweaked my way up to 2000mhz Cas7 - don't despair yet. The Boot Failure Message is a normal behaviour when your memory clocks are overboard - I have seen this message a lot in my life

    Settings, I will try to take some pics - actually you might be able to extract them from the 1m youtube video - I don't have the board set up at moment and cannot quickly get them.
    http://www.youtube.com/watch?v=gWE3k8EpaNo


    I have not been able to test the BIOS 9.04 yet are you 100% sure that memory clocks don't differ between the 2 BIOSs? With the P5E3 now with the newer BIOS I cannot run Strap 266 - only 333 and 400 straps ; yet with the older BIOS I was ONLY able to run 266. I honestly wouldn't be surprised that some of the latest Maximus Extreme BIOS might be interesting for memory clocking.

    I think Vfsb helped a lot - I was quiet high on that one ( around 1.4+ iirc) . expect to shoot up to 1.85V in the NB - make sure it is not running too hot. If you are 50C + you're about to enter NB instabilty territory

    Also you might be able to go higher vdimm on your set of cellshocks - try quickly 2.08V and 2.10V ; defenitely try to play with clock skews - I have a feeling delaying the skews might be helpful on the Maximus if it behaves similar to mine. I make Clock SKews responsible for at least 30mhz in reaching the 2000mhz marker.

    I my case I can run performance level 6 at CAS7, at CAS8 Performance levels are very "itchy" and I am forced to lossen up . Try higher perfromance levels - just for a quick go - also tRFC are always good to be loose at first.

    Your sticks might be CAS8 lovers nevertheless - mine do crap out at DDR3 2080 after all, regardless of me being able to post that silly 1m at CAS8 DDR3 2250mhz - not stable whatsoever.


    Please make sure not to jump back and forth from Dominators to Cellshocks - or if you do to clear BIOS and NOT use previously svaed OC profiles - they might have Dominator specfic settings that might conflict with the Cellshock settings. I made that error much too often myself.
    Last edited by mike; 04-25-2008 at 12:04 PM.

  6. #81
    Xtreme X.I.P.
    Join Date
    Aug 2004
    Location
    Chile
    Posts
    4,151
    Dude mike your NB is so hiiiiiiiigh i'm only at 1.51v, maybe that is what i need.

  7. #82
    Xtreme 3DTeam Member
    Join Date
    Aug 2006
    Location
    Where You don't get nothin' for $
    Posts
    1,001
    Thanks a lot Mike - will keep on trying. My Vfsb is 1.44 - the board needs that much.

    NB is under chilled water - no temp probs at all, ~30°C @1.66 Vmch.

    Your last comment is funny - it just happened to me. OCed Dominators removed and Cellshocks plugged in = no boot. Then I did CMOS clear, = no boot. tried this 5 times, = no boot. Has to install back the Corsairs again and set them down to default settings, = boot now !

    Then I installed the Cellshocks again and all was fine....... - took me about one hour....

    BTW, here's my screen:



    12.5k read at 4GHz isn't bad - though it is only Cas 8.

    If nothing works nomore......:


  8. #83
    Xtreme 3DTeam Member
    Join Date
    Aug 2006
    Location
    Where You don't get nothin' for $
    Posts
    1,001
    But my E8500 can do better :



    If nothing works nomore......:


  9. #84
    Xtreme X.I.P.
    Join Date
    Nov 2002
    Location
    Shipai
    Posts
    31,147
    Quote Originally Posted by Rob Halford View Post
    @ saaya; Can you comment upon any 2 * 2 GB kits (14400) coming our way soon? Extremely pleased with today’s sticks but it seems that I need a couple of GB’s more. Vista Ultimate is somewhat demanding and I don’t want to go with 4 * 1GB...
    yepp, 2x2gb is coming soon

    camouflage, what chipset voltage are you using?

  10. #85
    Xtreme 3DTeam Member
    Join Date
    Aug 2006
    Location
    Where You don't get nothin' for $
    Posts
    1,001
    Quote Originally Posted by saaya View Post
    yepp, 2x2gb is coming soon

    camouflage, what chipset voltage are you using?
    1.7 Vmch - now I'm running 8-8-8-20 @1.89 Vdimm. 7-7-7-x is not stable at all 1T & PL6 on my Maximus Extreme.

    Oh, and I had to set 1.44 Vfsb for stability even for 8-8-8-x @ 1000MHz...........

    NB @33C with chilled water...............
    If nothing works nomore......:


  11. #86
    Xtreme X.I.P.
    Join Date
    Nov 2002
    Location
    Shipai
    Posts
    31,147
    i wouldnt call 33 celsius chilled heheheh but as long as its below 40 all chips should be having a good time

    so whats the max you can get with 777?
    try PL7 and dont forget, the max voltage the sticks will accept are different depending on what timings you use. with 888 they usually like less vdimm than with 777

    with D9GTR it was around 0.15v more they liked if you reduce cas, trcd and trp by 1 clock, with D9JNL its only 0.05v more or less from what ive seen, but still, it changes depending on the timings you use, so with cas7 you can try higher vdimm.

  12. #87
    Xtreme 3DTeam Member
    Join Date
    Aug 2006
    Location
    Where You don't get nothin' for $
    Posts
    1,001
    Quote Originally Posted by saaya View Post
    i wouldnt call 33 celsius chilled heheheh but as long as its below 40 all chips should be having a good time

    so whats the max you can get with 777?
    try PL7 and dont forget, the max voltage the sticks will accept are different depending on what timings you use. with 888 they usually like less vdimm than with 777

    with D9GTR it was around 0.15v more they liked if you reduce cas, trcd and trp by 1 clock, with D9JNL its only 0.05v more or less from what ive seen, but still, it changes depending on the timings you use, so with cas7 you can try higher vdimm.
    Hm, I think on air I'd go above 50°C.............
    If nothing works nomore......:


  13. #88
    Xtreme Member
    Join Date
    Feb 2008
    Location
    Italy
    Posts
    423
    well
    i wanna get ddr3, i use liquid nitrogen , phase change ecc.
    whats better for me to use with e8xxx and qx9650, this new blue 1866 or old 1800 ??

  14. #89
    Aussie God
    Join Date
    Feb 2005
    Location
    Copenhagen, Denmark
    Posts
    4,596
    Quote Originally Posted by aerie View Post
    well
    i wanna get ddr3, i use liquid nitrogen , phase change ecc.
    whats better for me to use with e8xxx and qx9650, this new blue 1866 or old 1800 ??
    This is my oppinion...
    First off, using liquid nitrogen on ram will give you some problems, as the PCB will get very cold, not only on sticks, but mobo as well, and thereby give a headeech insulating....
    Second... QX9650 will easen the requirements on your motherboard as for unlocked multiplier, however, hard for NB.. (I dont know exactly why, even after reading Tony's post in this section)
    As for ram, when using great cooling, I would deff go for ram that scales with voltage.... Im not aware what chips the "old" 1800 uses, however, if before JNL, its highly likely that they used GTR or GTS - saaya will probably be able to answer this....
    For extreme cooling, and high voltage, I would deff' say one of them chips...
    Competition ranking;
    2005; Netbyte, Karise/Denmark #1 @ PiFast
    2008; AOCM II, Minfeld/Germany #2 @ 01SE/AM3/8M (w. Oliver)
    2009; AMD-OC, Viborg/Denmark #2 @ max freq Gigabyte TweaKING, Paris/France #4 @ 32M/01SE (w. Vanovich)
    2010: Gigabyte P55, Hamburg/Germany #6 @ wprime 1024/SPI 1M (w. THC) AOCM III, Minfeld/Germany #6 @ 01SE/AM3/1M/8M (w. NeoForce)

    Spectating;
    2010; GOOC 2010 Many thanks to Gigabyte!


  15. #90
    Xtreme Member
    Join Date
    Feb 2008
    Location
    Italy
    Posts
    423
    .... i dont use nitrogen on rams............................
    only cpu and gpu....
    i only wanna know for extreme clock what is better , blue ore black?

  16. #91
    silver wall jumper X
    Join Date
    Apr 2005
    Location
    Las Vegas
    Posts
    1,579
    Quote Originally Posted by aerie View Post
    .... i dont use nitrogen on rams............................
    only cpu and gpu....
    i only wanna know for extreme clock what is better , blue ore black?


    For wolfsdale (8500) go D9JNL ( cas7/cas8) - for yorkfield (9650) go D9GTR ( cas 6 )

  17. #92
    Xtreme X.I.P.
    Join Date
    Nov 2002
    Location
    Shipai
    Posts
    31,147
    Quote Originally Posted by aerie View Post
    well
    i wanna get ddr3, i use liquid nitrogen , phase change ecc.
    whats better for me to use with e8xxx and qx9650, this new blue 1866 or old 1800 ??
    i think mike is right about this:
    Quote Originally Posted by mikeguava View Post
    For wolfsdale (8500) go D9JNL ( cas7/cas8) - for yorkfield (9650) go D9GTR ( cas 6 )
    id say for dualcores D9JNL is better and for quads D9GTR cause quads cant get to as high fsb and hence mem speeds, which is where d9jnl really shines.
    i think the difference is tiny though, for benching it makes a difference if you use gtr or jnl, but for 24/7... no big difference id say...

  18. #93
    Xtreme Enthusiast
    Join Date
    Jul 2007
    Posts
    668
    Quote Originally Posted by saaya View Post
    i think mike is right about this:


    id say for dualcores D9JNL is better and for quads D9GTR cause quads cant get to as high fsb and hence mem speeds, which is where d9jnl really shines.
    i think the difference is tiny though, for benching it makes a difference if you use gtr or jnl, but for 24/7... no big difference id say...
    But who uses Ln2 for 24/7?

  19. #94
    Xtreme Member
    Join Date
    Feb 2008
    Location
    Italy
    Posts
    423
    d9GTR can get best frequences for cas6 ??

  20. #95
    Xtreme X.I.P.
    Join Date
    Nov 2002
    Location
    Shipai
    Posts
    31,147
    Quote Originally Posted by DMH View Post
    But who uses Ln2 for 24/7?
    kingpin, hipro5, shamino, kinc... ^^
    i think kingpin has a 300L tank in his basement iirc, and i know some benchers in northern europe have stationary tanks installed
    At cellshock we have a 1000L stationary tank installed
    nordichardware is actually thinking of getting a ln2 pipeline into their main lab for unlimited ln2 supply
    But yes, lets get back to reality and what most people do

    if you bench a lot and want to tweak the last 10mb/s bandwidth out of your mem/system, then yeah, get some d9gtr, if you only bench every now and then theres no big difference to d9jnl... well, lets put it this way, really good d9jnl is pretty much as good as really good gtr, maybe slightly worse with cas6. normal d9jnl is better than normal d9gtr...

    Quote Originally Posted by aerie View Post
    d9GTR can get best frequences for cas6 ??
    yes, the average d9gtr is good for ~1650 cas6, the average d9jnl is good for ~1600 cas6. The best GTR can do 1800+ cas6, the best jnl can do 1750+ cas6. and gtr can run 665 or 655, while jnl can only run 666 or 665

    So as i said, its a very small difference...

    and when it comes to cas7 and cas8, from what ive seen there is only very few REALLY good GTR that can keep up and surpass JNL... so there good JNL is actually beating good GTR...
    Last edited by saaya; 05-04-2008 at 04:13 AM.

  21. #96
    Memory Addict
    Join Date
    Aug 2002
    Location
    Brisbane, Australia
    Posts
    11,651
    never used DI/LN2 myself heh

    nordichardware LN2 pipeline LOL

    These 1866 are looking pretty good if you can match it with right motherboard it seems from these results
    ---

  22. #97
    Xtreme Enthusiast
    Join Date
    Jun 2005
    Posts
    601
    Quote Originally Posted by camouflage View Post
    Thanks a lot Mike - will keep on trying. My Vfsb is 1.44 - the board needs that much.

    NB is under chilled water - no temp probs at all, ~30°C @1.66 Vmch.

    Your last comment is funny - it just happened to me. OCed Dominators removed and Cellshocks plugged in = no boot. Then I did CMOS clear, = no boot. tried this 5 times, = no boot. Has to install back the Corsairs again and set them down to default settings, = boot now !

    Then I installed the Cellshocks again and all was fine....... - took me about one hour....

    BTW, here's my screen:



    12.5k read at 4GHz isn't bad - though it is only Cas 8.

    how come that we have almost the same result i am using DDr2 @ 500 Mhz with 5-5-5-15
    maybe the 0.1 Mhz make the deference
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	untitled.jpg 
Views:	785 
Size:	186.9 KB 
ID:	77872  
    2600K working in 4.8 GHZ so far

    2600k @(4600 Ghz) 1.42v : (under water)
    Asus B2 mobo
    4 GiG DDR 2400 MHZ
    GTX 570 @ 1.063v ( 910/1820/2001 Mhz) .(under water).
    1020 W PSU

    Hobie's : Overclocking .... Overclocking .... Overclocking

  23. #98
    Memory Addict
    Join Date
    Aug 2002
    Location
    Brisbane, Australia
    Posts
    11,651
    Quote Originally Posted by makaka View Post
    how come that we have almost the same result i am using DDr2 @ 500 Mhz with 5-5-5-15
    maybe the 0.1 Mhz make the deference
    1M won't be able to show the differences.. you need something like super pi 32M
    ---

  24. #99
    Xtreme 3DTeam Member
    Join Date
    Aug 2006
    Location
    Where You don't get nothin' for $
    Posts
    1,001
    Here's a 16M - sadly I need tRFC 82 for full stability:



    If nothing works nomore......:


  25. #100
    Xtreme Enthusiast
    Join Date
    Jul 2007
    Posts
    668
    Quote Originally Posted by saaya View Post
    kingpin, hipro5, shamino, kinc... ^^
    i think kingpin has a 300L tank in his basement iirc, and i know some benchers in northern europe have stationary tanks installed
    At cellshock we have a 1000L stationary tank installed
    nordichardware is actually thinking of getting a ln2 pipeline into their main lab for unlimited ln2 supply

    But yes, lets get back to reality and what most people do

    if you bench a lot and want to tweak the last 10mb/s bandwidth out of your mem/system, then yeah, get some d9gtr, if you only bench every now and then theres no big difference to d9jnl... well, lets put it this way, really good d9jnl is pretty much as good as really good gtr, maybe slightly worse with cas6. normal d9jnl is better than normal d9gtr...


    yes, the average d9gtr is good for ~1650 cas6, the average d9jnl is good for ~1600 cas6. The best GTR can do 1800+ cas6, the best jnl can do 1750+ cas6. and gtr can run 665 or 655, while jnl can only run 666 or 665

    So as i said, its a very small difference...

    and when it comes to cas7 and cas8, from what ive seen there is only very few REALLY good GTR that can keep up and surpass JNL... so there good JNL is actually beating good GTR...
    LN2 addict's,I knew some bencher's had acess to alot fo LN2 but never though of them having stationary tanks for just them

Page 4 of 8 FirstFirst 1234567 ... 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
  •