Asus P5E-VM HDMI: best matx ever?

Discussion in 'Small Form Factor Systems' started by pvhk, Oct 20, 2007.

  1. NichoTL

    NichoTL Gawd

    Messages:
    772
    Joined:
    Apr 15, 2008
    About --DRAM RAS# Activate to Precha: 18 DRAM Clocks:
    on my BIOS I can't set it higher than 15. Can you?
    Also don't forget FSB, say, 1.36V.
     
  2. Phuncz

    Phuncz 2[H]4U

    Messages:
    2,630
    Joined:
    Apr 12, 2009
    Yeah I see your point :) Higher multipliers only are useful at standard FSB speeds, as the 333 x 9 already gives me 3.0GHz. If I would put the FSB at 400, I'd have 3.7GHz which is already getting close to the limit. I got this CPU on the cheap, I didn't buy it specifically for the unlocked multiplier.

    My opinion is that without putting too much voltage on the components (FSB +0.08v, NB +0.10v) I get a faster system all-round, where as getting a higher overclocked CPU leaves me with a high voltage increase and thus bigger heat and power usage all the time but only a speed increase some of the time. My system is used for 60% browsing, 20% multitasking, 10% gaming and 10% Photoshop. Since I need a lot of juice for a 17% clock speed increase, my guess is a 500MHz FSB running 1:1 with the memory seems a better trade-off. That's my opinion ofcourse, if you think otherwise I'd like to hear it :) I'll be sure to test how far this processor clocks after the weekend.
     
  3. dosu23

    dosu23 [H]Lite

    Messages:
    70
    Joined:
    Jan 3, 2009
    to NichoTL

    I have speedstep disabled but cpuz keeps showing me at 2400mhz x6 insted of my 3200mhz 8x, there a fix for that or is there something else in the bios i should fix? im not sure what to do.


    edit: also prime95 for 20min cpuz shows 1.36v cpu at 3.2ghz and no core over 55c during prime.
    what shall i try next? my goal is i guess 3.4ghz and pretty much maxing out my ram as much as possible

    edit: one core had a error, should i give more cpu voltage or?
     
    Last edited: May 2, 2009
  4. NichoTL

    NichoTL Gawd

    Messages:
    772
    Joined:
    Apr 15, 2008
    Hmm whether you get 4GHz (as an example) by going 8*500 or 9*445, you still need to apply more or less the same amount of voltage to the CPU. In the first case, you have to choose the 1:1 divider (5:6 would put your RAM at 1200) and then your RAM is running at 1000MHz. Second case you can choose the 5:6 divider and have your RAM at 1068. Now you seem to think (from what I understand) that the higher FSB will give better performance, but I'd say that the difference (if there is any) will be marginal at best and more than made up for by the fact that the mem runs faster in the lower FSB scenario. But I'm still pretty new at this so I'd like to hear from an experienced OCer on this topic. On top of that, the higher FSB will likely require higher FSB, NB and PLL voltages to run smoothly, which might become a problem in the long run.
    I personally would choose the lower FSB scenario in this situation, because both FSB are pretty close. If it was a choice between, say, 200 and 400 as FSB, I'd pick 400 (all things being equal).
    Now if and when you get a stable overclock running both scenarios, it would be interesting to see the results of, for example, Sandra memory benchmarks for both situations.
     
  5. dosu23

    dosu23 [H]Lite

    Messages:
    70
    Joined:
    Jan 3, 2009
    To NichoTL
    Oh also, what volts should i lower slowly other than ram and cpu after i find a good stable point(if i do)
     
  6. NichoTL

    NichoTL Gawd

    Messages:
    772
    Joined:
    Apr 15, 2008
    Don't worry about speedstep. As soon as you put load onto your CPU, it'll go to 8* multiplier. About the voltages to lower when you are 100% sure you are stable? Most important is CPU, of course. Then you can try lowering NB, FSB and PLL. Don't really have a recipe for this, sorry. You've just got to try.
     
  7. dosu23

    dosu23 [H]Lite

    Messages:
    70
    Joined:
    Jan 3, 2009
    okay, is the ratio for like 2:1 ram the same as just setting like 9 multi with 266 then selecting for dram ddr2 400, 533 ect? or? like my 8x400 i can select ddr2 833 and 1000? whats the highest my ram can go? 1000mhz? still a little confused with that, i guess i want 3.4ghz and the highest my ram can go. so if i want 8x430 then select the highest dram setting? or would like 7x500 be faster overall?


    also core #3 always fails and runs cooler on idle than the others? faulty core or just not enough volts? i put 1.50 in bios but cpuz shows 1.39v

    edit: i dont think the pencil mod was working so i tried again and now no fails in prime
    72c in prime full load is fine? idles around 52c or so says realtemp with 26-37 tj max
     
    Last edited: May 2, 2009
  8. NichoTL

    NichoTL Gawd

    Messages:
    772
    Joined:
    Apr 15, 2008
    Multiplier has nothing to do with your memory.
    The strap you choose sets what RAM speeds are available to you for a certain FSB:
    example: STRAP 333 imagine your FSB is also 333
    support for DDR2 667 gives you the 1:1 divider (333:333 which becomes 667 because its ddr2) and support for DDR2 800 gives you the 5:6 divider (333:400)
    so if you're running FSB at 400, you have 2 choices for RAM: 800 (1:1) and 960 (5:6)
    on STRAP 266: you have 3 choices: 1:1 (which would be support for DDR2 533), 4:5 (DDR2 667) and 2:3 (DDR2 800), etc...

    Now an important part of choosing which kind of setting you want to run is determining at what speed you want to run your RAM:
    your RAM is rated 1000MHz so you could choose FSB 500 and divider 1:1 on strap 333 (possible also on strap 266 but you want to pick the highest strap available that's still lower than the FSB) or FSB 416-ish with 5:6 divider. Of course you don't have to aim for your RAM at 1000MHz exactly so that gives you some flexibility.
     
    Last edited: May 2, 2009
  9. NichoTL

    NichoTL Gawd

    Messages:
    772
    Joined:
    Apr 15, 2008
    My guess is that you can start lowering your voltage. The temp is acceptable but better play safe...
     
  10. dosu23

    dosu23 [H]Lite

    Messages:
    70
    Joined:
    Jan 3, 2009
    cpuz shows 8x400 3200mhz 1.32 volts, 71c in prime and dram freq 480mhz, is that pretty good or what should i change to get more out of my setup seeing how i dont know too much
     
  11. NichoTL

    NichoTL Gawd

    Messages:
    772
    Joined:
    Apr 15, 2008
    Which stepping is your Q6600? (you can check on cpuz)

    If it's a G0 it's probably capable of more.

    If your setup is stable, then that's not too bad. You've got 2 possibilities now:
    - either you try and set the multiplier to 9 and find the CPU voltage that will make it stable (that's IF it is a G0. If it's a C0/C1, forget it).
    - or you leave the multiplier at 8 and push the FSB to, say, 425. But in this case you will possibly need to modify all other voltage settings: NB, FSB, PLL and DRAM, which could be a chore (to take the CPU out of the equation, you first drop the multiplier to 6 or 7, push the FSB to 425 and then do whatever is needed to stabilize everything. And then put the multiplier back to 8 and adjust the CPU voltage).
     
  12. dosu23

    dosu23 [H]Lite

    Messages:
    70
    Joined:
    Jan 3, 2009
    okay ill try the second. im using coretemp that has tjmax at 90c and real temp that the tj max is always different like 30-70c always and temps are like 20c different between progs, im not sure what the real temp is

    and lets say insted of 8 x 425 i do like 9 x 390, what would that benefit?


     
    Last edited: May 3, 2009
  13. antaholics

    antaholics [H]Lite

    Messages:
    106
    Joined:
    Jan 9, 2009
    if i rmbr correctly realtemp is crap... and core temp agrees with a few other apps (everest, etc), so my bet is on core temp
     
  14. NichoTL

    NichoTL Gawd

    Messages:
    772
    Joined:
    Apr 15, 2008
    Can't help about the temp, I usually use hwmonitor.
    If you go, for example, 8.5*400, then you ONLY have to readjust the CPU voltage (because you say that you have 8*400 stable). But your RAM is running at 480*2=960.
    If you go, 8*425, then obviously you have to readjust all the voltages, but then your RAM is running at 425*1.2*2=1020 slightly better (your RAM should be able to handle it, hopefully). If you get that stable, that's pretty good: 3.4GHz on your CPU with 1020MHz on your RAM.
     
  15. dosu23

    dosu23 [H]Lite

    Messages:
    70
    Joined:
    Jan 3, 2009
    I'd like to try for 3.4 but at 3.2 my prime im at 71c, isnt that a little high?
    how that i have the pencil mod can/should i lower my pll or anything its at PLL Voltage: 1.58v
    or does that not have anything to do with vdroop?


     
  16. NichoTL

    NichoTL Gawd

    Messages:
    772
    Joined:
    Apr 15, 2008
    Yeah it is indeed high. I doubt the pencil mod has an influence on other voltages, but I can't be sure. You could try lowering your CPU voltage and see what's the lowest that keeps your 3.2 setup stable. Temps will go down.
     
  17. dosu23

    dosu23 [H]Lite

    Messages:
    70
    Joined:
    Jan 3, 2009
    Will do, thanks
    1 hour of prime is good enough I suppose to see whats a stable volt?
    im pretty sure the pencil mod worked, but i set in bios 1.35 and in cpuz it shows 1.296v, it doesnt seem to raise when i raise the volts to get a stable v




     
    Last edited: May 4, 2009
  18. Phuncz

    Phuncz 2[H]4U

    Messages:
    2,630
    Joined:
    Apr 12, 2009
    Today I tried the CPU overclocking thing and I'm not happy about the result... This mobo seems to be picky or I got a really strict CPU. I also tried to set the RAM to lower rates since I was going to test standard FSB anyway. Results:

    STABLE = about 5 runs of Prime95 on all cores without error
    BOOT = POST ok, Windows bootup ok, crash/errors with Prime95 or Windows
    NO POST = doesn't POST, obviously
    Auto volt seems to go to 1.4v if needed, to really get anywhere I need to put in 1.5v or higher

    STABLE FSB 333 x 9 = 3.0GHz (auto volt) - RAM: 333MHz 5-5-5-15 @ 2.1v
    BOOT FSB 333 x 10 = 3.33GHz (auto volt) - RAM: 333MHz 4-4-4-12 @ 2.1v
    BOOT FSB 333 x 10 = 3.33GHz (auto volt) - RAM: 333MHz 5-5-5-15 @ 2.1v
    NO POST FSB 333 x 11 = 3.67GHz (auto volt) - RAM: 333MHz 4-4-4-12 @ 2.1v
    NO POST FSB 333 x 11 = 3.67GHz (1.5v or 1.6v) - RAM: 333MHz 4-4-4-12 @ 2.1v
    STABLE FSB 333 x 9 = 3.0GHz (auto volt) - RAM: 333MHz 3-3-3-9 @ 2.1v
    NO POST FSB 400 x 8 = 3.2GHz (auto volt) - RAM: 400MHz 3-3-3-9 @ 2.1v
    STABLE FSB 400 x 8 = 3.2GHz (auto volt) - RAM: 400MHz 4-4-4-12 @ 2.1v
    STABLE FSB 433 x 7 = 3.03GHz (auto volt) - RAM: 433MHz 4-4-4-12 @ 2.1v
    (testing that last one right now)

    So I'm trying to find the middle road here in lowest memory latency and highest RAM clock speed, in between finding a stable higher CPU speed. But it seems I have to go 1.6v and higher to get it stable above 3.3GHz.

    Do I need to adjust voltages on other components for in increase in RAM and CPU speed ? Until now I've figured out I can run 500MHz FSB stable without too much extra volt on NB and FSB, the RAM can handle better timings on 433MHz (maybe more) on the 533MHz-rated 2.1v and the maximum stable CPU clock speed on the stock Intel cooler is 3.5GHz with 1.6v on the vCore.

    As it seems I can go any direction I want: faster CPU, faster RAM, faster interconnect (NB, FSB). I'm still doubting which ones are the most gain in performance. I could run Sisoft Sandra all week to find out exactly but those are synthetic and not real. At the moment I'm hoping to get the RAM at higher frequencies on the 4-4-4 latencies and thus increasing the FSB along, making all components faster.
     
  19. antaholics

    antaholics [H]Lite

    Messages:
    106
    Joined:
    Jan 9, 2009
    i dont have time to read anything right now, but a quick advice: DONT use auto volt for CPU if you're OC's arent stable/posting
     
  20. Phuncz

    Phuncz 2[H]4U

    Messages:
    2,630
    Joined:
    Apr 12, 2009
    Well I found out it rarely increases stability if I put it on 1.4v, I always need to put 1.45v or higher to actually improve. At the moment I'm trying not to increase voltage since I'm running the stock Intel cooler which has a hard time keeping up.

    Further testing:

    STABLE = about 5 runs of Prime95 on all cores without error
    BOOT = POST ok, Windows bootup ok, crash/errors with Prime95 or Windows
    NO BOOT = doesn't successfully boot into Windows
    NO POST = doesn't POST, obviously
    Auto volt seems to go to 1.4v if needed, to really get anywhere I need to put in 1.5v or higher

    BOOT FSB 433 x 7 = 3.03GHz (auto volt) - RAM: 433MHz 4-4-4-12 @ 2.1v (1 error in Prime95 after 5-10 minutes)
    NO BOOT FSB 450 x 7 = 3.16GHz (auto volt) - RAM: 450MHz 4-4-4-12 @ 2.1v
    Tested the RAM all the way to 2.38v, CPU voltage up to 1.6v. Also tried SB 1.2v.
    FSB 1.28v and NB: 1.35v where used throughout the other voltages.
    NO POST FSB 366 x 8 = 2.93GHz (auto volt) - RAM: 366MHz 3-3-3-9 @ 2.3v
    NO POST FSB 350 x 8 = 2.8GHz (auto volt) - RAM: 350MHz 3-3-3-9 @ 2.3v
    NO BOOT FSB 444 x 7 = 3.1GHz (auto volt) - RAM: 444MHz 4-4-4-12 @ 2.1v
    BOOT FSB 444 x 7 = 3.1GHz (auto volt) - RAM: 444MHz 4-4-4-12 @ 2.2v

    So I just need to increase the voltage a tickle to get 433MHz stable. The 3-3-3-9 timings seem pretty impossible for a significant setting above 333MHz. I might look into tweaking the settings more specifically soon. Time to get that CPU going !
     
  21. antaholics

    antaholics [H]Lite

    Messages:
    106
    Joined:
    Jan 9, 2009
    isn't corsair dominators 5-5-5-15?

    I have a feeling you're tweaking too many settings at once... and you should always control the CPU voltage in the BIOS, or at the very least monitor it with CPU-Z (so you know how much to bump it up manually).

    try tweaking one thing at once (first get FSB/mobo stable with low CPU multiplier... then get CPU stable with higher multiplier and tweak volts... and get the Ram stable last)
     
  22. Phuncz

    Phuncz 2[H]4U

    Messages:
    2,630
    Joined:
    Apr 12, 2009
    Yep, it's PC8500 (533MHz) 5-5-5-15 at 2.1v. But as I found out, 433MHz also handles 4-4-4-12 at 2.1v very well.

    That's what I'm doing, getting to know what the limits are of every component. I now know that my max FSB is not too far away from 500MHz. I know my CPU's max clock is somewhere between 3.5GHz and 3.8GHz. My memory can handle 3-3-3-9 not far from 333MHz and is stable with 433MHz 4-4-4-12.

    I'm now trying to find the middle ground: getting the FSB up, getting the memory latencies down and hope to improve the CPU clock speed altogether.

    New test:
    BOOT FSB 433 x 8 = 3.46GHz (1.6v) - RAM: 433MHz 4-4-4-12 @ 2.14v
    1 error in Prime95 after 5-10 minutes, the 10 passes complete on the other 3 cores. So I need to find out if I need a little more voltage on the memory or CPU. But with this cooler I'm getting 94-97°C which is dangerously close to my own tolerance of acceptance. I'll be running this setting for a while to see if it is at least 99.9% stable, since Prime95 is an extreme scenario I almost never run into.
     
  23. antaholics

    antaholics [H]Lite

    Messages:
    106
    Joined:
    Jan 9, 2009
    intel specs for the QX6850 is 0.85V - 1.5V, up to 64.5C

    you're about 30C over the thermal limit....

    you should probably get a better cooler before baking your chip! =O
     
  24. Phuncz

    Phuncz 2[H]4U

    Messages:
    2,630
    Joined:
    Apr 12, 2009
    Yeah it's on order as we speak :) It idles at 70°C most of the time. I'll be running it on lower multiplier and voltage until it arrives.

    I tried to get the 433MHz stable on the memory in combination with the lower timings and I couldn't get it 100% stable. I tested it stable at these settings:
    STABLE FSB 420 x 8 = 3.36GHz (1.525v) - RAM: 420MHz 4-4-4-12 @ 2.1v
    FSB voltage: AUTO - NB voltage: AUTO

    However, also this setting is 100% stable:
    STABLE FSB 480 x 7 = 3.36GHz (1.525v) - RAM: 480MHz 5-5-5-15 @ 2.1v
    FSB voltage: 1.28v - NB voltage: 1.35v

    The first reaches to the limit of the RAM, the second reaches the limit of the FSB. Which one would be faster or would it not matter much ?
     
  25. antaholics

    antaholics [H]Lite

    Messages:
    106
    Joined:
    Jan 9, 2009
    I am not too sure about how much performance difference you will see with the ram speed being 120MHz lower with tighter timings... but...

    70C on idle is EXTREMELY high, and it means that your CPU is at least 5-6C over the Intel thermal specs ALL THE TIME. That's really bad for the CPU. It might be a good idea to run it at stock speeds before you get a better cooler.

    Note: Tjmax is NOT the thermal limit before damage... you will have to look in the intel specs, and for the Q6850, it's 64.5C.
     
  26. NichoTL

    NichoTL Gawd

    Messages:
    772
    Joined:
    Apr 15, 2008
    The problem is that the NT06-E is not going to improve your temps, or at least not significantly.
     
  27. dosu23

    dosu23 [H]Lite

    Messages:
    70
    Joined:
    Jan 3, 2009
    for a q6600 tjmax always should be 90? i dont know why realtemps tjmax always moves around and is like 30-40, im not sure what to trust

    my cpu v in bios is 1.35 and in cpuz its at 400x8 3200mhz and goes between 1.29-1.31, not bad?
     
  28. antaholics

    antaholics [H]Lite

    Messages:
    106
    Joined:
    Jan 9, 2009
    @dosu:

    don't worry about tjmax... just worry about core temperatures and tcase temperatures... vs the intel thermal specs

    if you did the pencil mod, that looks about right.
     
  29. dosu23

    dosu23 [H]Lite

    Messages:
    70
    Joined:
    Jan 3, 2009
    well doesnt tjmax affect the core temps? im using hwmonitor, real temp and the other i forget and they seem different till i put the tjmax to 90c then they seem the same
     
  30. antaholics

    antaholics [H]Lite

    Messages:
    106
    Joined:
    Jan 9, 2009
    i usually use hwmonitor, core temp, and everest... they all give the same values for tcase and tcores
    speedfan has core temperatures exactly 5C off

    i'm not entirely sure what effects tjmax has on core temps... but i know it isn't the "safe limit" on core temperatures. I usually ignore it and go with the intel thermal specs

    link here: http://processorfinder.intel.com/Default.aspx
     
  31. dosu23

    dosu23 [H]Lite

    Messages:
    70
    Joined:
    Jan 3, 2009
    like real temp has tjmax on 30c so shows my load cores around 71c, then if i change tjmax to 90c it shows my cores around 53c
     
  32. antaholics

    antaholics [H]Lite

    Messages:
    106
    Joined:
    Jan 9, 2009
    use core temp or hwmonitor... and check with another app to make sure you're getting the right temperatures :)
     
  33. dosu23

    dosu23 [H]Lite

    Messages:
    70
    Joined:
    Jan 3, 2009
    i use both of those and they give me the same temp because bot tjmax's are at 90c
     
  34. antaholics

    antaholics [H]Lite

    Messages:
    106
    Joined:
    Jan 9, 2009
    i would personally trust those apps more than realtemp...

    on the other hand, if you're really not sure, you can try putting your finger close to the base of the HSF and see how it feels... if it's only 53C then it should feel slightly warm, and if it's 70+C it should feel quite warm or hot
     
  35. Phuncz

    Phuncz 2[H]4U

    Messages:
    2,630
    Joined:
    Apr 12, 2009
    Well I'm not looking for the best cooler, just one that fits and improves the outrageous temps. I'll look into something better when I have more time. I'm currently running 500MHz FSB x 6 so I can use the AUTO voltage which doesn't/shouldn't exceed 1.5v until the NT06-E arrives.
     
  36. NichoTL

    NichoTL Gawd

    Messages:
    772
    Joined:
    Apr 15, 2008
    Understood, but I think this is the wrong chip for this type of enclosure... just my 2 cents.
     
  37. Ruiner

    Ruiner 2[H]4U

    Messages:
    3,044
    Joined:
    Aug 22, 2000
    Have any of the micro G4* boards caught up to the p5e-vm in terms of OCing?
     
  38. NichoTL

    NichoTL Gawd

    Messages:
    772
    Joined:
    Apr 15, 2008
    Not that we are aware of. The EG45M-UD2H hasn't been tested around here but the BIOS isn't promising...
     
  39. antaholics

    antaholics [H]Lite

    Messages:
    106
    Joined:
    Jan 9, 2009
    I am having a slightly annoying problem... it's been the second time in about two weeks or so that I had a BSOD in Win7, with the following message:

    "A clock interrupt was not received on a secondary processor within the allocated time interval."

    Both times this happened, the computer was idle, while i was AFK. Could it be my overclock failing?

    I have:
    - P5E-VM
    - Intel E5200 2.5ghz @ 3.7Ghz
    - 2x2GB G.Skill DDR2-1066 RAM
    - EVGA 9800gtx+ superclocked
    - OCZ ModXStream 500W psu

    I tested the OC at least twice over 24 hours back in feb/march.. so hopefully it's not the OC failing now.

    I did some quick searching and some people have this issue in vista x64 with non-OC'ed processors, and apparently for some people it was a BIOS issue. Any idea?
     
  40. dworley

    dworley Gawd

    Messages:
    625
    Joined:
    Sep 30, 2004