A64 3000+ Venice Overclocking Help? (DFI Ultra-D)

Discussion in 'Overclocking & Cooling' started by Lukano, May 12, 2005.

  1. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    I have NO idea how long it's gonna take me, but I'll also eventually have some SuperPi times up to compare to.....I'm taking the long/thourough way of finding my max 1:1 OC and then tweaking the hell out of the ram, and then doing the same with the 9/10 divider, cus from what I've tried....I can do quite a bit better with the 9/10...and it's not THAT much lost...

    Oh well, tough to say for now, but I wanna get my max clock/fully tweaked out times for my max 1:1 and then the same for 9/10 on the ram, regardless of the lost bandwidth, if the second pair of times with the 9/10 divider proves to be faster, then so be it...I'll use what's best.
     
  2. Lukano

    Lukano Limp Gawd

    Messages:
    344
    Joined:
    Mar 11, 2003
    Depending on how limited you are at 9:10 or 1:1 compared to say 5:6 - you may want to give the latter a try too. You might find that the sheer extra clock makes up for a bit of ram bw (at worst 30-50mhz difference between 9:10 and 5:6).
     
  3. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    I know, depending on how long it takes me to get through my 1:1 and 9/10 testing I'll probly give 5/6 a shot, just for some comparisons.
     
  4. Lukano

    Lukano Limp Gawd

    Messages:
    344
    Joined:
    Mar 11, 2003
    Any updates?
     
  5. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    so far 265x9 is perfectly fine at stock voltage.....

    270x9 is locking up within the first couple passes instead of just giving errors like it had up until now so I think I'm starting to hit a voltage limitation as opposed to a BIOS related one.

    I'm still gonna go through 'em all and make SURE though.....just to know that I tried everything I could before raising the CPU voltage....and raising the chipset or VDT does nothing at this point...so it's either the BIOS, or cpu voltage keeping me from 270x9 right now...but 265x9 is at 1.36v, so it's not doing TOO bad.....better then I had ever hit before even for 1M or 8M SuperPi stable cus I never did as much experiementation with all the different BIOS and memory slot combinations before.

    Gonna get back on it right away here, I'll take a quick screenshot of the table I've got made up in Word so you guys can see what I'm talking about.....pretty fun. ;)
     
  6. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    Here it is....just typing in the settings that I NEED for each HTT step....writing the rest down on a peice of paper.....

    I've hit a wall though and have to start on voltage work, cus I've tried about 8 consecutive BIOS' and at least one "version" from each dated release and not gotten anywhere.....up from 1.4v to 1.425v in the BIOS (running at only 1.376v) and 270x9 is now 32M stable....movin up slowly...not bad, cus last time I tried this I didnt do any of this BIOS work and could barely get 260x9 32M stable.

    [​IMG]
     
  7. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    Not that I want to start thead hijacking.....but it seems my extra work has paid off.....

    Even stressing the memory controller by keeping it at 1:1, I just finished SuperPi 32M at 290x9, DDR580 and a managable 1.568v on the CPU so far.

    I'm excited to see what this thing'll do even with just the 9/10 dram divider in place, but I just thought I'd want to keep you (Lukano) updated with what's going on, as I've been talkin' with ya on MSN.
     
  8. peterhoang2002

    peterhoang2002 Gawd

    Messages:
    705
    Joined:
    Jan 15, 2004
    just a quick question, did you see better performance at 290*9 then at 335 * 8 (as it says in your sig) or is it just pretty much the same?
     
  9. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    Not sure yet, never ran any SuperPi at 335x8 because at the time, I didn't have enough stability to do any real testing. 295x9 1:1 @ 1.648v is SuperPi 32M stable so far....gonna go for 300x9 in a bit...but I'm quite happy with this thing (now) actually.

    I'm going to collect my results/best times and scores from my 1:1 best OC, do the same with the 9/10 divider on the memory, and then redo my 8x cpu multiplier OC to see which puts out the better numbers...'till then I'm leaving my sig as is...cus I'm gonna be playing with this thing for a little while still. ;)
     
  10. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    1:1 is done.....296x9 for 2.66GHz with running at 1.66v....stock cooler, SuperPi 32M stable. :D :cool:

    At this rate I'm really excited to see what I can do with the ram at 9/10.....wish me luck. :)
     
  11. Herulach

    Herulach Gawd

    Messages:
    997
    Joined:
    Dec 24, 2004
    Some pretty sweet results, i wish you guys would hurry up and finish with this thread, every post makes me want a venice more.
     
  12. Lukano

    Lukano Limp Gawd

    Messages:
    344
    Joined:
    Mar 11, 2003
    I'm still sitting at 5hrs prime stable (haven't run 32M Pi yet, but I will today!) but no higher.

    Each time I step up the juice a bit on the vcore, it gets a little more stable for a little longer. I'll keep going until I hit 1.8 (and can anyone tell me here and now if 1.8's going to result in a little puddle of molten metal or if it's fine for a venice with adequate cooling?).

    The machine is "desktop" stable still though. Had it running for 48hrs without a lockup or crash, playing World of Warcraft and City of Heroes interspersed in there.
     
  13. Lukano

    Lukano Limp Gawd

    Messages:
    344
    Joined:
    Mar 11, 2003
  14. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    Go check out my SuperPi Challenge thread in OC and Cooling and download the 1.4 anticheat version of SuperPi.....it's the generally accepted version almost anywhere you look.

    I tried going up to 1.8v (in the BIOS, not running in windows) on stock cooling...didn't result in any dmg at all, but didnt help me either.....generally I dont tend to see any gains with this chips above 1.7 or 1.75v in BIOS. You can try 1.8v in BIOS if you want, but I doubt it'll gain you much over running 1.7 or 1.75v.
     
  15. Lukano

    Lukano Limp Gawd

    Messages:
    344
    Joined:
    Mar 11, 2003
    I managed 4hrs32m Prime on 1.70, I'll try stepping up a notch each time and record my Prime stability as I go.
     
  16. Lukano

    Lukano Limp Gawd

    Messages:
    344
    Joined:
    Mar 11, 2003
    1M with mod1.4 is 32.844s - running 32M now.
     
  17. joker_927

    joker_927 n00b

    Messages:
    40
    Joined:
    May 30, 2005
    Ok, i have been reading this post from page 1to9 and it has been VERY helpful.

    I hvave a new venice 3000+ and a dfi ultra-d (you can see why this post helped me out.)

    I started overclocking on my own 10 mhz at a time and couldnt past 230 mhz. And this was with keeping the fsb/ram ratio so that my ram is always under 400mhz.

    I went to one of the guides @ dfi-street and set all my voltages and what not.

    260 FSB
    x3 ldt multi
    x9 chip multi
    1.425 +110% (1.54)
    1.5 LTD
    1.8 Chipset
    2.9 ram

    I guess this is supposed to be a very conservative o/c according to the other forums because voltages are pretty high and the o/c is not very high. I guess you SHOULD be able to do the above o/c on stock ait cooling.

    Well here I am with an xp-120 with a 85CFM panaflo & AS5, keeping my ram always under 200mhz even though it's OCZ 3200 el platinum rev 2, with a Tagan 480w PSU and I can't even get this overclock.

    I can boot to windows fine, heck im typing this message right now, but I left Prime running last night and when i got back windows had crashed. Crashed so hard i coudnt see how many iteratons it had gone through.

    So what could be the problem if windows boots prime runs for a while, but windows crashes sometimes. Oh, and it sometimes crashes right when windows starts but its just a crash where u can click Dont Send.

    So that is basically where I am at, I can't even hit 260 stable.
     
  18. Lukano

    Lukano Limp Gawd

    Messages:
    344
    Joined:
    Mar 11, 2003
    I would say the first step is to juice the processor up a bit. It's a matter of opinion, but I'm comfortable (and have found it's required for 260+ on the 3000+ Venice's) with 1.6-1.7v (using VID special control to hit over 1.55) but bear in mind YMMV and it's not my fault if you've just got a bad chip that doesn't like too much power.

    Next step is volt up the ram so it can keep up with the rest of the system. It shouldn't really be required if you're keeping it within spec, but I noticed I actually had better luck with an overclocked pair of sticks (3.2-3.5v) of my OCZ PC3200 VX EL Gold than I did with it at a 2:3 divider 192mhz 2.6-3.2v. I can't really explain why.

    All in all, give it a bit more power and see if stability increases. If it does, then start overclocking it bit by bit again until you hit another threshold. If you're comfortable giving more power then, do so and rinse/repeat until you hit your max or you start getting concerned about too much juice or heat.
     
  19. Lukano

    Lukano Limp Gawd

    Messages:
    344
    Joined:
    Mar 11, 2003
    Also, anyone else read teh DFI 5v jumper / sudden death thread? Anyone as concerned about that as I am (threads in the same forum as this one).

    edit - here's the link to the thread in question

    http://hardforum.com/showthread.php?t=908803
     
  20. joker_927

    joker_927 n00b

    Messages:
    40
    Joined:
    May 30, 2005
    the 5v thing, if it is a true problem, doesnt effect me since i would never set my ram over 3.2v.

    Okay, right after I made my first post, I had prime running for about 5 min and then I started up WOW and my computer rebooted. :(

    I increased that vcore to +13% and CPU-Z states it is @ 1.568V

    Well, as I write this, Prime has been running for 20 min, and I have Wow minimized and have been playing for about 20 also.

    I hope that all it needed was a little more juice. My goal when Ibught the chip was 311 FSB x 9. My goal is now 300. I have been deprived of wow cause of the past 3 days of o/c ing so i'm gonna see if the comp is stable by playing for the next hour or so.

    I will be back :)
     
  21. Lukano

    Lukano Limp Gawd

    Messages:
    344
    Joined:
    Mar 11, 2003
    I know how you feel. I've neglected WoW and City of Heroes for two weeks now simply because I was so caught up in tweaking my machine. Still not more than 4-5hrs Prime Stable at the settings in my sig, but it's enough for me to leave the machine desktop stable 24/7 and play WoW for the 2-3 hrs I manage once in a blue moon. I'll just tweak further for prime stability as I have time (usually try something new, leave prime running overnight).
     
  22. joker_927

    joker_927 n00b

    Messages:
    40
    Joined:
    May 30, 2005
    260 x 9
    1.584v

    27 hours Prime95 stable, sweet!

    Raised it to 270 and running prime right now.

    Do you think Prime95 is the best tester? I think for long burn-in it would definately be, but are there any program that you don't have to run for hours to find an error? Just wondering...
     
  23. Lukano

    Lukano Limp Gawd

    Messages:
    344
    Joined:
    Mar 11, 2003
    There's no program that will show errors quicker. Basically every other tool will keep on chugging while Prime will error even on small stuff. It's the best yardstick for an absolute, "as stable as stock" overclock.
     
  24. Herulach

    Herulach Gawd

    Messages:
    997
    Joined:
    Dec 24, 2004
    Although as you and cornelius have illustrated, SuperPi is a good quick test, assuming you verify the result.
    [​IMG]
     
  25. joker_927

    joker_927 n00b

    Messages:
    40
    Joined:
    May 30, 2005
    Ok, so i'll stick with Prime95 then.

    Ok, so a temp question. What is too much? A few websites have said that 65C is the max and @ 1.616 V I'm getting 34C Idle and 47C Load with my xp-120.

    47 is still fine in my eyes, I just wondered at what point I should go OMG and turn off the computer. Also, is that a relatively good temp because i paid a lot more money for this cpu cooling setup than stock and it I don't think it's performace is "blowing me away".

    Ambient is pretty hot in this apartment though. I dont think the DFI gives ambient temp but my thermostat on my wall says 78F.

    The thing I am MOST concerned with is the chipset temp. Mine idels at 46C and gets up to 54C sometimes. The fan blade that is on there was touching my video card so i pressed it a little bit and thought it may have come unseated, but I reached my hand in there and press on it really hard. Should I put some AS5 on it? It would be rather difficult to put it on while the mobo is in the case, but I would do it if it would help. Most people are reporting mid 40's for thier chipset.

    Thanks for all the quick and helpful replies.
     
  26. peterhoang2002

    peterhoang2002 Gawd

    Messages:
    705
    Joined:
    Jan 15, 2004
    i think 65 is the max id go. some people dont like pushing it above 50 but at 65, it should still last a very long time. youll probably replace it long before it dies.
     
  27. joker_927

    joker_927 n00b

    Messages:
    40
    Joined:
    May 30, 2005
    well here is the update.

    I got 260 as stated before.and tried to go to 275, didn't work. Kept giving it voltage the way to 1.69V. Wouldn't work.

    I saw another person on this forum get 335 x 8, and i found the 8 multiplier interesting, so i switched my fsb to 315 x 8, and still nothing. So I raised voltage steadily until 1.7 and then gave up. I tried 310 and even 300 with the 8 multiplier. No go. So I went back 9 and tried 290,280,275,270 and nothing. 265 worked though @ 1.59V. By no go I mean sometimes it got to the desktop and rebooted, sometimes I get all the way into a Wow map and then rebooted, it was all reboots.

    Well, then I remembered. 260FSB x 2 x 4 = 2080! Omg! I had forgotten about the LDR multiplier. I thought 4 was gonna be enough. Well I put the multiplier to 3x.

    I am now in windows typing this message and running prime on forth iteration @.............. 330x8 = 2640, highest ever so far. Mind you its running at a whopping 1.696, but im sure that will go down a little.Next step.... 9 multiplier!
     
  28. Herulach

    Herulach Gawd

    Messages:
    997
    Joined:
    Dec 24, 2004
    DOH!

    I wouldnt worry about temps, keep it below 50 ish, and thank the gods you dont have a prescott to cool.

    As as chipset goes i cant really comment, but 50degs ish doesnt sound too bad.

    [​IMG]
     
  29. joker_927

    joker_927 n00b

    Messages:
    40
    Joined:
    May 30, 2005
    I am now discovering that the memory is what is making my overclocks fussy. If I go too far under 200 mhz, XP crashes, i was able to go all the way just past 250 (DDR500) before xp started crashing again. I only had the voltage at 2.9 so i took it to 3.0v and windows works fine. Raised the FSB to 340 and windows crashed. Lowered the miltiplier on my memory and now im typing this message in windows.

    340 x 8 = 2720

    I got my goal!
     
  30. Lukano

    Lukano Limp Gawd

    Messages:
    344
    Joined:
    Mar 11, 2003
    I think I might give this a try too.. I'm getting frustrated with the x9 and memory controller wonkiness... it'll be my project for tomorrow morning and I'll post back then.
     
  31. peterhoang2002

    peterhoang2002 Gawd

    Messages:
    705
    Joined:
    Jan 15, 2004
    i got my xp-90 this weekend but the guy i bought it from forgot to ship the black heatsink mount :( so now ill have to wait maybe another week to get it.

    well i ran some memtest86+ for another stability check and tweaking and now it gets errors at 260mhz, 3.6v, 2-2-2-8. i had to tun it down to 240mhz, 3.5, 2-2-2-8 to get rid of the errors... but this is quite hard as this looks like the sweet spot and this sweet spot doesnt have much range, up or down. to get memory at 240mhz id have to run at 294HTT or around 270HTT. 270 yields lower overall clock and i cant reach 294 at my current voltage so im really dying to slap the xp-90 w/ a vantec tornado to up my voltage a bit.

    i might try with 8x multiplier but probably wont do any testing with it till this weekend.

    well lukano, did you ever get a chance to finish that 32M run? 1M test didnt look like there was much variation so id like to see how much difference there would be in the 32M test.
     
  32. Lukano

    Lukano Limp Gawd

    Messages:
    344
    Joined:
    Mar 11, 2003
    Y'know, I keep postponing it for one thing or another. I think I'm taking the day off work sick, but it should give me some time to toy with it today. I'll get back to you (provided I remember or am not bedridden) in a while with the #'s for 32M.
     
  33. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    I'd also like to see how you're doing.....I'm starting to work more dillegently on my memory timings.....cus for now I'm temp limited at 302x9, 2.718GHz 1:1....not bad for now...but I REALLY wanna drop a Pelt on this thing.

    Awhile ago before I had done all my BIOS experiementing, I could barely get 260x9 32M stable....now that I've got 302x9 under my belt, I wanna try it again. Even with those handicapped clocks/settings, the 20C load temp in the freezer took me all the way to 2.86GHz for a CPU-Z screeny.....I REALLY wanna drop a nice, fat 226W Pelt on this chip and see if I can pull 2.9GHz out of my hat for daily use. :D
     
  34. Lukano

    Lukano Limp Gawd

    Messages:
    344
    Joined:
    Mar 11, 2003
    Okay, here's my Pi 32M time. It's on a slightly different config than the 1M time, but should not make a huge difference. Currently 293x9 1.728v 2-2-2-8 3.5v. (same as my sig a the time of posting this, once done pi I'm going to go save these settings in cmos reloaded and reboot for a prime run).

    SuperPi mod1.4 32M = 30m 01s
     
  35. Lukano

    Lukano Limp Gawd

    Messages:
    344
    Joined:
    Mar 11, 2003
    So far so good at the speeds/voltages/timings in my sig.

    4hrs Prime without a hitch, I shut down in order to rewire a bit for heat (highest I saw was 51c) and check my arctic silver application. Everything looks fantastic, so it's about as good as it's going to get on heat with the current gear.

    Now to simply leave prime running for the next 24hrs (while doing other stuff at that, like playing WoW) and then screenshot and cpu-z verify for what I hope is finally prime stable.

    edit - are you supposed to be able to do other stuff (play games, etc) while running prime? I notice if I launch a game for more than a few seconds, prime will fail. But if I leave prime running on desktop it's fine.
     
  36. joker_927

    joker_927 n00b

    Messages:
    40
    Joined:
    May 30, 2005
    I, on the other hand played Wow for almost 5 hours with prime95 running in the background fine, then I quit wow and prime failed with a rounding error about 30 seconds after it closed.

    I have gone back to the 9 multiplier. 300 x 9 which i couldnt get beforebut now I am typing this message in. Will get back later. 1.7V

    I asked some people and they said 1.7V was insane and that I shouldn't be going that high. Are they crazy, or am I?
     
  37. peterhoang2002

    peterhoang2002 Gawd

    Messages:
    705
    Joined:
    Jan 15, 2004
    i honestly dont think 1.7 is gonna kill it as long as temps are good.

    well i got a question for you people... i got 2 settings which are stable: (memtest86+ and prime 95) what will be better?

    option #1:
    2160mhz (240 * 9)
    ddr480 (240mhz) 2-2-2-8 @ 1T
    1:1

    option #2
    2502mhz (278 * 9)
    ddr500 (250mhz) 2-2-2-8 @ 2T
    9/10 divider

    i think those are my options for now... until i get something that will allow me to raise vcore a little more to achieve at least 293 and do

    options #3
    2637mhz (293 * 9)
    ddr480 (240mhz) 2-2-2-8 @ 1T
    5/6 divider
     
  38. legcramp

    legcramp [H]ardForum Junkie

    Messages:
    10,821
    Joined:
    Aug 16, 2004
    option 1 is better...by far...2t drops fps quite a bit..and the tiny extra 100mhz isnt gonna replace the 2t timings..I know..i tried it in a64 tweaker..forced 2t and my fps dropped quite a bit..
     
  39. peterhoang2002

    peterhoang2002 Gawd

    Messages:
    705
    Joined:
    Jan 15, 2004
    edit: ok option 1 had to be edited because it didnt work.
    ram at 240mhz 2-2-2-8 1T was passing memtest and so i assumed using a divider and keeping it still at 240mhz would work but apparently not. it spit our errors in every pass with test 8.

    so... 240mhz 2-2-2-8 1T only works with ram and HTT at 1:1. might work with 5:6 but havent tried as my cpu cant go that far yet. so, with option 1 slightly changed, is it still better than 2?
     
  40. Lukano

    Lukano Limp Gawd

    Messages:
    344
    Joined:
    Mar 11, 2003
    Just a quick update that I'm toying with my timings and clock as I have time, but haven't really had more than 10 minutes here and there. I've been pretty frustrated so I'm giving myself a cooldown time before I try again with newfound patience :)

    On a side (and not drama inducing, I promise) note I'm disappointed to notice corny has been banned.