Advertisement
How to add spoiler tags, edit posts, add images etc. How to - a user's guide to the new version of Boards
Mods please check the Moderators Group for an important update on Mod tools. If you do not have access to the group, please PM Niamh. Thanks!

[OC log] (\/)0n0L17(-) II

  • #1
    Registered Users Posts: 8,067 L31mr0d


    Ok well I finally got my RAM replaced from HWV. Funny thing is they didn't return any of my emails for 3 weeks then they sent my out a kit of RAM thats actually better than the one I sent them?

    I used to have a 4GB kit of Mushkin HP2-6400, they've sent me out a 4GB kit of XP2-6400. Basically the same but with tighter timings.

    So I thought I wouldn't be stupid like last time and just start ramping up voltages and clocks and seeing if they work, i'm going to try and OC this rig methodically and see what my results are. My rig is a fairly common high/mid end machine so hopefully someone will find this usefull.

    The rig i'm overclocking is:

    (\/)0n0L17(-) II
    CPU: QX6700 w/ Tuniq Tower lapped
    RAM:4GB Mushkin XP2-6400 @ 4-4-4-12 2T
    Mobo: eVGA 680i-A1
    GPU: eVGA 8800 GTX @ 620/1100
    HDD1: 2x320GB WD SE16 RAID0
    HDD2: 500GB Samsung Spinpoint
    Soundcard: X-FI Fatal1ty
    PSU: OCZ 700w GX
    Optical: SATA DVD-ROM 18x
    Case: P180b

    Here is my process for overclocking. I'll be leaving the FSB unlinked for this

    1: Find the limits of the RAM:
    This involves setting the timings to their rated settings and stepping up the Mhz of the FSB 10 at a time until I fail memtest86 before 8 hours. Once I fail i'll up the voltages and continue. I'll only be going up as far as 2.0v

    2: Find the limits of the CPU:
    This involves loosening the RAM timings and dropping the RAM clocks down, then uping the CPU's FSB 10Mhz at a time and testing for 8 hours in OCCT. Once it fails i'll up the voltages and test again, i'll do this up as far as 1.4v

    3: Find the best balance between RAM and CPU:
    This area is always fiddly, i'll try testing the RAM and CPU at their limits together and see if i'm stable. If not i'll increase the SPP, MCP, FSB and SPP<->MCP volts and retest. If I can't get it stable i'll drop both the cpu and ram clocks by 10mhz each until i'm stable.

    4: As a final test i'll run an 8hour run of rthdribl (or maybe the crysis benchmark) and OCCT to make sure everything is stable including my GPU overclocks.

    Here are some screenshots of the BIOS, feel free to give me advice on settings which are important and onces I should ignore.

    1.JPG

    2.JPG

    3.JPG

    4.JPG

    5.JPG


«13

Comments



  • Just 2.0v with D9's? You won't be going to far at that voltage.. Something in the 2.2 range would be better, and there shouldn't be any danger of killing them at that, ballistix are rated 2.2v stock using the same ic's.




  • CPU Multiplier: 10
    FSB – Memory Clock Mode: Unlinked
    CPU FSB Mhz: 1066
    RAM FSB Mhz: 810
    FSB Memory Ratio: Auto

    Timings
    TCL: 4
    TRCD: 4
    TRP: 4
    TRAS: 12
    CMD: 2
    TRRD: 3
    TRC: 24
    TWR: 5
    TWTR: 14
    TREF: Auto

    Volts
    CPU Core: 1.29
    CPU FSB: 1.2
    Memory: 1.9
    Nforce SPP: 1.25
    Nforce MCP: 1.5
    HT nForce SPP <-> MCP: 1.2

    Time Ran (hrs): 9.3
    Memtest86 PASS




  • :mad: ok a weird thing is happening now, I set the RAM FSB to 820Mhz and ran memtest86, it passed 8 hours BUT... I don't think my FSB settings are being applied? CPU-Z and Everest are still picking up my RAM FSB as only 800Mhz?

    So I set the Memory Clock Mode to "linked", set the memory ratio to 1:1 and dropped the FSB to 820. I then rebooted and in post I now get "System has entered safe mode"... wth is going on???

    Anyone any ideas? :confused:




  • CPU Multiplier: 10
    FSB – Memory Clock Mode: Unlinked
    CPU FSB Mhz: 1066
    RAM FSB Mhz: 830
    FSB Memory Ratio: Auto

    Timings
    TCL: 4
    TRCD: 4
    TRP: 4
    TRAS: 12
    CMD: 2
    TRRD: 3
    TRC: 24
    TWR: 5
    TWTR: 14
    TREF: Auto

    Volts
    CPU Core: 1.29
    CPU FSB: 1.2
    Memory: 2.0
    Nforce SPP: 1.25
    Nforce MCP: 1.5
    HT nForce SPP <-> MCP: 1.2

    Time Ran (hrs): 11.43
    Memtest86: PASS


    Wow, these forums are really buzzing with life :rolleyes: well anyway, I got some help over at the eVGA forums about why the increase in the RAM FSB wasn't being shown in the POST.

    Unlinked doesn't necessarily mean the RAM isn't still hindered by the divider (which can't be changed in this mode) So the ACTUAL RAM Mhz didn't go up until I changed it to 830Mhz. I had to up the voltage to 2.0v to stabilize this though. I don't really want to go much higher than this but i'd like to at least get to 900Mhz at these timings.

    Alternatively i'm going to try and leave them at 400Mhz and lower the timings.

    EDIT: Oh, I meant to ask... how does the HT multiplier work for Intel? When should I be dropping it to say, x4 and x3? Does it read the FSB for the CPU or the RAM? Does lowering it affect performance much?




  • :rolleyes: ...

    ok having a REALLY weird problem now. I set the RAM FSB to 840Mhz and the VRAM to 2.0v and booted into Vista. i then played a few hours of Crysis, and when I exited the game my RAM had dropped back down to 800Mhz and the voltage had dropped to 1.9v???

    WTH???

    Anyone any ideas...


  • Advertisement


  • CPU Multiplier: 10
    FSB – Memory Clock Mode: Unlinked
    CPU FSB Mhz: 1066
    RAM FSB Mhz: 840
    FSB Memory Ratio: Auto

    Timings
    TCL: 4
    TRCD: 4
    TRP: 4
    TRAS: 12
    CMD: 2
    TRRD: 3
    TRC: 24
    TWR: 5
    TWTR: 14
    TREF: Auto

    Volts
    CPU Core: 1.29
    CPU FSB: 1.2
    Memory: 2
    Nforce SPP: 1.25
    Nforce MCP: 1.5
    HT nForce SPP <-> MCP: 1.2

    Time Ran (hrs): 9.4
    Memtest86: PASS


    Ok... well I figured out the issue I was having with my voltages and clocks resetting. I have ntune installed and had saved a profile with my GPU overclocked values in it. Before loading up Crysis I had set ntune to load that profile, unknown to me, ntune also saves whatever your BIOS settings where when you saved that profile and then loads them. There is also no way to stop it from doing this :mad: Looks like i'll be getting another program to overclock my GPU.




  • Ah only just seen this now. And i was about to blame ntune. Had the exact same problem myself. Use riva tuner, atleast you can overclock the shaders too now :)




  • Anti wrote: »
    Ah only just seen this now. And i was about to blame ntune. Had the exact same problem myself. Use riva tuner, atleast you can overclock the shaders too now :)

    Yeah just got the latest version. Luckily there are some released signed drivers for it so no need to turn driver chechking off when rebooting.

    Having some problems now with the RAM. Its failing after 7 hours at 854Mhz :( I've lowered the timings a bit to see if I can get the clocks higher, although I'd be surprised if I can hit 900Mhz on this kit at 2.0v. I'd go to 2.1v but I haven't tracked down what killed my last set and I don't really want to kill this set right before the xmas holidays :rolleyes:

    CPU Multiplier: 10
    FSB – Memory Clock Mode: Unlinked
    CPU FSB Mhz: 1066
    RAM FSB Mhz: 854
    FSB Memory Ratio: Auto

    Timings
    TCL: 4
    TRCD: 4
    TRP: 4
    TRAS: 12
    CMD: 2
    TRRD: 3
    TRC: 24
    TWR: 5
    TWTR: 14
    TREF: Auto

    Volts
    CPU Core: 1.29
    CPU FSB: 1.2
    Memory: 2
    Nforce SPP: 1.25
    Nforce MCP: 1.5
    HT nForce SPP <-> MCP: 1.2

    Time Ran (hrs): 7.2
    Memtest86: FAIL




  • Ok well lowering the timings seems to have helped. I'm going to see how far I can go at these timings.


    CPU Multiplier: 10
    FSB – Memory Clock Mode: Unlinked
    CPU FSB Mhz: 1066
    RAM FSB Mhz: 854
    FSB Memory Ratio: Auto

    Timings
    TCL: 5
    TRCD: 5
    TRP: 5
    TRAS: 12
    CMD: 2
    TRRD: 3
    TRC: 24
    TWR: 5
    TWTR: 14
    TREF: Auto

    Volts
    CPU Core: 1.29
    CPU FSB: 1.2
    Memory: 2
    Nforce SPP: 1.25
    Nforce MCP: 1.5
    HT nForce SPP <-> MCP: 1.2

    Time Ran (hrs): 9.5
    Memtest86: PASS




  • CPU Multiplier: 10
    FSB – Memory Clock Mode: Unlinked
    CPU FSB Mhz: 1066
    RAM FSB Mhz: 867
    FSB Memory Ratio: Auto

    Timings
    TCL: 5
    TRCD: 5
    TRP: 5
    TRAS: 12
    CMD: 2
    TRRD: 3
    TRC: 24
    TWR: 5
    TWTR: 14
    TREF: Auto

    Volts
    CPU Core: 1.29
    CPU FSB: 1.2
    Memory: 2
    Nforce SPP: 1.25
    Nforce MCP: 1.5
    HT nForce SPP <-> MCP: 1.2

    Time Ran (hrs): 9.3
    Memtest86: PASS


    looks like the lowered timings are helping with the overclock.


  • Advertisement


  • CPU Multiplier: 10
    FSB – Memory Clock Mode: Unlinked
    CPU FSB Mhz: 1066
    RAM FSB Mhz: 889
    FSB Memory Ratio: Auto

    Timings
    TCL: 5
    TRCD: 5
    TRP: 5
    TRAS: 12
    CMD: 2
    TRRD: 3
    TRC: 24
    TWR: 5
    TWTR: 14
    TREF: Auto

    Volts
    CPU Core: 1.29
    CPU FSB: 1.2
    Memory: 2
    Nforce SPP: 1.25
    Nforce MCP: 1.5
    HT nForce SPP <-> MCP: 1.2

    Time Ran (hrs): 9.0
    Memtest86: PASS




  • Ok getting a REALLY weird issue with overclocking the RAM (although nobody will probably read this :rolleyes:) My performance is DROPPING to my INCREASES in FSB (using EVEREST to bench)??? But the sticks are still 100% stable after 8 hours of memtest86???

    i.e.

    @ 427Mhz -
    * mem read: 7600 MB/s
    * mem write: 4830 MB/s
    * mem copy: 5368 MB/s

    @ 444Mhz -
    * mem read: 7411 MB/s
    * mem write: 4828 MB/s
    * mem copy: 5344 MB/s

    @ 457Mhz -
    * mem read: 7358 MB/s
    * mem write: 4852 MB/s
    * mem copy: 5400 MB/s

    Very very odd??? :confused: anyone any ideas, i've never seen this issue before when overclocking.



    CPU Multiplier: 10
    FSB – Memory Clock Mode: Unlinked
    CPU FSB Mhz: 1066
    RAM FSB Mhz: 915
    FSB Memory Ratio: Auto

    Timings
    TCL: 5
    TRCD: 5
    TRP: 5
    TRAS: 12
    CMD: 2
    TRRD: 3
    TRC: 24
    TWR: 5
    TWTR: 14
    TREF: Auto

    Volts
    CPU Core: 1.29
    CPU FSB: 1.2
    Memory: 2
    Nforce SPP: 1.25
    Nforce MCP: 1.5
    HT nForce SPP <-> MCP: 1.2

    Time Ran (hrs): 8.0
    Memtest86: PASS




  • try giving it some more voltage just a small bit and see what heppens.




  • Anti wrote: »
    try giving it some more voltage just a small bit and see what heppens.

    I don't think thats it, i've been reading around and supposedly this drop in performance happens when the northbridge strap increases from 1066 to 1333. Supposedly this happens when the FSB goes over a certain value.

    Heres a graph I found on this particular problem.

    attachment.php?attachmentid=121&stc=1&d=1160346346

    CPU Multiplier: 10
    FSB – Memory Clock Mode: Unlinked
    CPU FSB Mhz: 1066
    RAM FSB Mhz: 934
    FSB Memory Ratio: Auto

    Timings
    TCL: 5
    TRCD: 5
    TRP: 5
    TRAS: 12
    CMD: 2
    TRRD: 3
    TRC: 24
    TWR: 5
    TWTR: 14
    TREF: Auto

    Volts
    CPU Core: 1.29
    CPU FSB: 1.2
    Memory: 2
    Nforce SPP: 1.25
    Nforce MCP: 1.5
    HT nForce SPP <-> MCP: 1.2

    Time Ran (hrs): 9.0
    Memtest86: PASS


    EVEREST Bench results: (OMG :mad: stop going down!!!)

    *mem read: 7227 MB/s
    *mem write: 4827 MB/s
    *mem copy: 5363 MB/s




  • Ok, stable at 948Mhz after 8 hours... plus my bench scores seem to be rising again. So I have a window of DDR 100Mhz (i.e. 50Mhz) where my performance drops, now i've got to see how much further I have to push the FSB to get back to where I was before? Seems kinda fruitless overclocking when there are such blatant flaws on the motherboard. It might prove better leaving the DDR FSB low and just tightening the timings.

    @ 474Mhz
    *mem read: 7342 MB/s
    *mem write: 4850 MB/s
    *mem copy: 5436 MB/s

    CPU Multiplier: 10
    FSB – Memory Clock Mode: Unlinked
    CPU FSB Mhz: 1066
    RAM FSB Mhz: 948
    FSB Memory Ratio: Auto

    Timings
    TCL: 5
    TRCD: 5
    TRP: 5
    TRAS: 12
    CMD: 2
    TRRD: 3
    TRC: 24
    TWR: 5
    TWTR: 14
    TREF: Auto

    Volts
    CPU Core: 1.29
    CPU FSB: 1.2
    Memory: 2
    Nforce SPP: 1.25
    Nforce MCP: 1.5
    HT nForce SPP <-> MCP: 1.2

    Time Ran (hrs): 8.0
    Memtest86: PASS




  • jesus, thats a bit mad isnt it. And yeah i think i agree with you.




  • Anti wrote: »
    jesus, thats a bit mad isnt it. And yeah i think i agree with you.

    I'd say i'll be up to around 500Mhz before i'm back to the same performance I was at at 427Mhz. Supposedly this is a common problem but because people don't bench, but just stress test at each clock increase they just assume that the performance is rising proportional to the increase im Mhz, which it isn't.

    The only way around this is to overclock the RAM from within Windows using something like ntune or clockgen as the strap won't change from within windows. This is also the reason why overclocking within windows will lead to stability problems a lot earlier than overclocking from the BIOS.




  • Ok WTH??? My scores are dropping again. This is very weird now.


    @ 480Mhz
    *mem read: 7273 MB/s
    *mem write: 4828 MB/s
    *mem copy: 5406 MB/s

    CPU Multiplier: 10
    FSB – Memory Clock Mode: Unlinked
    CPU FSB Mhz: 1066
    RAM FSB Mhz: 960
    FSB Memory Ratio: Auto

    Timings
    TCL: 5
    TRCD: 5
    TRP: 5
    TRAS: 12
    CMD: 2
    TRRD: 3
    TRC: 24
    TWR: 5
    TWTR: 14
    TREF: Auto

    Volts
    CPU Core: 1.29
    CPU FSB: 1.2
    Memory: 2
    Nforce SPP: 1.25
    Nforce MCP: 1.5
    HT nForce SPP <-> MCP: 1.2

    Time Ran (hrs): 9.3
    Memtest86: PASS




  • BTW, what voltage is your ram warranty covered to? I wouldn't think twice bout giving them 2.2V




  • he's supposed to give them 2.2v there d9 chips.


  • Advertisement


  • awhir wrote: »
    he's supposed to give them 2.2v there d9 chips.

    I've been on the Mushkin forums and these sticks are warrentied to 2.1v. But I doubt that's whats causing this issue. If I start failing memtest86 then i'll increase the volts. No point giving them more volts than they need.

    I've a feeling as i'm increasing the FSB the Northbridge strap is changing the memory latency proportionally (a memory setting 680i boards don't let you control) this means that what ever extra clock I give them is being cancelled out by this latency. They seem to have settled at around 7300 MB/s Read and aren't really increasing anymore.




  • L31mr0d wrote: »
    Ok, stable at 948Mhz after 8 hours... plus my bench scores seem to be rising again. So I have a window of DDR 100Mhz (i.e. 50Mhz) where my performance drops, now i've got to see how much further I have to push the FSB to get back to where I was before? Seems kinda fruitless overclocking when there are such blatant flaws on the motherboard. It might prove better leaving the DDR FSB low and just tightening the timings.

    @ 474Mhz
    *mem read: 7342 MB/s
    *mem write: 4850 MB/s
    *mem copy: 5436 MB/s

    CPU Multiplier: 10
    FSB – Memory Clock Mode: Unlinked
    CPU FSB Mhz: 1066
    RAM FSB Mhz: 948
    FSB Memory Ratio: Auto

    Timings
    TCL: 5
    TRCD: 5
    TRP: 5
    TRAS: 12
    CMD: 2
    TRRD: 3
    TRC: 24
    TWR: 5
    TWTR: 14
    TREF: Auto

    Volts
    CPU Core: 1.29
    CPU FSB: 1.2
    Memory: 2
    Nforce SPP: 1.25
    Nforce MCP: 1.5
    HT nForce SPP <-> MCP: 1.2

    Time Ran (hrs): 8.0
    Memtest86: PASS

    It's really not a flaw, just the nature of how the thing works. The NB-RAM and NB-CPU interconnects have latencies too, just like the ram itself does. Increasing the clock speed is obviously going to mean the board will need to increase its internal latencies to allow for the higher speeds.. just like your ram does. Also really give that RAM 2.2v man, it needs it!

    Pure bandwidth will usually prove better than lower FSB with tight latencies.. Sorta depends on the app but in general it'll be better.




  • Ok I give up :confused: it seems some other setting on the motherboard that isn't visable in the BIOS is fighting against me and defeating whatever overclock I give these sticks. I think the reason i'm not failing memtest86 at these speeds is that they are still running way below their rated bandwidth. At DDR1000 i'm still running under the bandwidth I was at at DDR850.

    I'm going to cut my losses and start testing these sticks linked and with the divider set to sync mode (i.e. DDR for RAM, QDR for CPU)

    Also nibble, your comment makes no sense to me. Tight timings will give me more bandwidth so i'm not sacrificing one for the other. If you mean i'll get more bandwidth from a higher clock I think i've already proved that with my board (at least with the clocked mode set to unlinked) this isn't so.

    Also why do people keep telling me to up my volts??? Volts are only going to affect stability, as my sticks are stable I don't need to give them more volts yet. Upping the voltages needlessly is just going to create more heat.


    @ 500Mhz
    *mem read: 7356 MB/s
    *mem write: 4834 MB/s
    *mem copy: 5440 MB/s



    CPU Multiplier: 10
    FSB – Memory Clock Mode: Unlinked
    CPU FSB Mhz: 1066
    RAM FSB Mhz: 1000
    FSB Memory Ratio: Auto

    Timings
    TCL: 5
    TRCD: 5
    TRP: 5
    TRAS: 12
    CMD: 2
    TRRD: 3
    TRC: 24
    TWR: 5
    TWTR: 14
    TREF: Auto

    Volts
    CPU Core: 1.29
    CPU FSB: 1.2
    Memory: 2.0
    Nforce SPP: 1.25
    Nforce MCP: 1.5
    HT nForce SPP <-> MCP: 1.2

    Time Ran (hrs): 11.2
    Memtest86: PASS




  • CPU Multiplier: 6
    FSB – Memory Clock Mode: linked
    CPU FSB Mhz: 1600
    RAM FSB Mhz: 800
    FSB Memory Ratio: Synced mode

    Timings
    TCL: 4
    TRCD: 4
    TRP: 3
    TRAS: 12
    CMD: 1
    TRRD: 3
    TRC: 24
    TWR: 5
    TWTR: 14
    TREF: Auto

    Volts
    CPU Core: 1.29
    CPU FSB: 1.2
    Memory: 2.1
    Nforce SPP: 1.25
    Nforce MCP: 1.5
    HT nForce SPP <-> MCP: 1.2

    Time Ran (hrs): 4.3
    Memtest86: FAIL


    Ok tried tightening the timings today, i've been trying to get the timings above stable. At 2.1v they are stable for 4.3 hours. Still anything less than 8 hours is not good enough. I'll try 2.15v tonight and see how it goes




  • CPU Multiplier: 6
    FSB – Memory Clock Mode: linked
    CPU FSB Mhz: 1600
    RAM FSB Mhz: 800
    FSB Memory Ratio: Sync mode

    Timings
    TCL: 4
    TRCD: 4
    TRP: 3
    TRAS: 12
    CMD: 1
    TRRD: 3
    TRC: 24
    TWR: 5
    TWTR: 14
    TREF: Auto

    Volts
    CPU Core: 1.29
    CPU FSB: 1.2
    Memory: 2.15
    Nforce SPP: 1.25
    Nforce MCP: 1.5
    HT nForce SPP <-> MCP: 1.2

    Time Ran (hrs): 9.3
    Memtest86: FAIL


    hmm... ok they failed again. Just to be sure i'm going to loosening the timings back to stock today to make sure I haven't broken these sticks as the error is always coming up at the same location.




  • For a quick memory test, run a 15x loop of test 5 & 10x of test 6, will take about 1.5hrs. These are the most intensive tests that will produce the most fails quickly. Ya must have great patience running them all night :D

    Have ya tries 2.2V with the tighter timings? They're d9 chips & do actually need it & can cope with it




  • PogMoThoin wrote: »
    For a quick memory test, run a 15x loop of test 5 & 10x of test 6, will take about 1.5hrs. These are the most intensive tests that will produce the most fails quickly. Ya must have great patience running them all night :D

    Have ya tries 2.2V with the tighter timings? They're d9 chips & do actually need it & can cope with it

    i'm in no rush, i'd rather get it right then end up breaking them. I run the testing overnight, and then again when i'm in work, so it really isn't eating into my time at all. Plus it would only bother me if I needed the performance for some game i'm playing which I don't

    Also, I've found the failing is happening in test 4

    I'm going to have to rethink how i'm going to overclock this board as you can't really ever seperate the ram from the cpu. The strap changes affect both of them, plus the 680i boards suffer from having FSB holes where the system just won't boot at one clock, yet a higher clock will. Its the oddest board i've ever had to overclock on.

    Plus for some reason, my system enters safe mode when I enable linked mode with a divider of 1:1... it really does not like it, i've tried a few different clock settings and always the same.




  • CPU Multiplier: 10
    FSB – Memory Clock Mode: unlinked
    CPU FSB Mhz: 1067
    RAM FSB Mhz: 800
    FSB Memory Ratio: Sync mode

    Timings
    TCL: 4
    TRCD: 4
    TRP: 3
    TRAS: 12
    CMD: 2
    TRRD: 3
    TRC: 24
    TWR: 5
    TWTR: 14
    TREF: Auto

    Volts
    CPU Core: 1.29
    CPU FSB: 1.2
    Memory: 2.2
    Nforce SPP: 1.25
    Nforce MCP: 1.5
    HT nForce SPP <-> MCP: 1.2

    Time Ran (hrs): 8.4
    Memtest86: PASS




  • Told ya :p

    Have a try at 4-4-3-10 @ 2.2v. Or use memset & test with orthos.


  • Advertisement


  • PogMoThoin wrote: »
    Told ya :p

    Have a try at 4-4-3-10 @ 2.2v. Or use memset & test with orthos.

    well it runs stock 4-4-4-12 @ 1.9v, it actually runs stable at 4-4-3-12 @ 2.05v but i'm sick of voltage being an issue so i've just upped to 2.2v as I ran memtest overnight at this voltage and I wasn't getting any errors.

    The next timings I'll be trying are:

    4-4-3-10
    4-3-3-10
    4-3-3-9

    then if i'm lucky 1T, but I'm not holding my breath for that as there isn't many 4GB kits that can handle 1T.


Advertisement