GSKILL 512 LA still can't hit 300 1T on DFI Ultra D!!

Printable View

03-06-2005, 07:01 PM
ben805

GSKILL 512 LA still can't hit 300 1T on DFI Ultra D!!

I ditched the Neo2 Platinum and bought the DFI NF4 Ultra-D and a 6600 GT PCI-Ex, after messing around with this darn thing all day....I still can't get these little LA to hit 300 1T stable, SuperPI 32MB keep crapping out....what is up with this?? memory voltage cranked all the way to hell like 2.9 to even 3.2v with no luck. I had them installed on dimm 2 and 4.
Should I be using any other BIOS for this PITA mobo other than the one that came with the DFI board? Is there any special BIOS settings that I should be aware of? Can anybody please show me your Dram settings on the BIOS so I can compare? man...this is really discouraging, I'm about to RMA these 512 LA back to newegg and get me some VX and call it a day, these TCCD just plain sucks!! :stick: you think this winnie 3000 might have a bad memory controller? maybe I should just throw the whole setup out of the window and just get me a FX55!!
Please help.... :eek: :eek:
03-06-2005, 07:29 PM
Major

Don't give up on the G.Skill !!!!

set everything to Auto in Dram Bios

set 2.5 4 3 7 1t manually

Keep Vdimm to 2.8 or less

Give that a try and see what happens
03-06-2005, 07:30 PM
sauria

I bet it's the memory controller not the RAM
03-06-2005, 07:41 PM
Kanavit

hmm, maybe you should get OCZ rev2, or Corsairs. i have to really push my vdimm to 2.8v to get my Gskills stable at super low timings and high clockspeeds. Gskill will do 2-2-2-5 tho, my old kingstons couldn't. anyways, i think from users feedback OCZ seems to be most consistant performer.
03-06-2005, 07:47 PM
Onepagebook

here is the setting i can get it up to 330 when I did super pi 16m and 32m with bachus.
details here:
http://www.xtremesystems.org/forums/...ad.php?t=55093

setting:
http://img93.exs.cx/img93/8875/19lv.jpg

This version of a64 tweaker is the best ever to fit with DFI board. :toast:
Thanks Code-red.
when you are in the bios. set Drive strength to 8.
Tref choose 4708.
03-06-2005, 07:47 PM
ben805

Quote:

Originally Posted by Major Slaughter

Don't give up on the G.Skill !!!!

set everything to Auto in Dram Bios

set 2.5 4 3 7 1t manually

Keep Vdimm to 2.8 or less

Give that a try and see what happens

that's what I did, I'd also tried 2.5 4 8 4 and the rest of them auto with no luck, I'll try to mess around with the voltage more and see what happen...

This DFI mobo seems to be a bit more stable than the Neo2, still have to get familiar with all the extra settings :p:
03-06-2005, 08:06 PM
Onepagebook

Don't use auto setting.
03-06-2005, 08:07 PM
ben805

1 Attachment(s)

FINALLY, I DID IT!!! 313 @2.5-4-7-3 1T Oh YEAA BABY!!

:banana: :banana: :banana: :banana: :banana:

BIG THANKS goes to all who've been trying to help me, especially bachus_anonym and OPB, couldn't of done it without all the great advices and tips from y'all. I'll play around with the A64 tweakers more and see how high I can hit, these GSKILL 512 LA isn't too bad afterall, well...actually it's awesome!!

:toast:
03-06-2005, 08:10 PM
Major

Thanks Onepagebook !! see ! lol
03-06-2005, 08:30 PM
ben805

should I disable the DRAM Bank Interleave? and what should I set the DQS Skew Control and DQS Skew Value to? sorry about all the stupid questions but I'm trying to fine tune the settings and hopefully I can get a little bit more out of these very fine TCCD. :)
03-06-2005, 08:33 PM
Onepagebook

Quote:

Originally Posted by ben805

should I disable the DRAM Bank Interleave? and what should I set the DQS Skew Control and DQS Skew Value to? sorry about all the stupid questions but I'm trying to fine tune the settings and hopefully I can get a little bit more out of these very fine TCCD. :)

man, I got a dozen of beear again? :stick:

** Bank interleave is the big help for 3d
Usually when I do floating point calculation or OC, I disable it :)

Thank you for your effort.

** here is my suggestion again:
when you do pi, DO NOT use 200, 15.6 for tref.
it'd be unstable and really lot more slower than 100 or 133, 1.95
03-06-2005, 08:49 PM
ben805

thanks for the awesome tips OPB, owe you one again!! :toast: :toast: :toast: :toast:

:bows: :bows: :bows:
03-06-2005, 09:00 PM
Onepagebook

you are really welcome.:D
03-06-2005, 10:00 PM
FlyingHamster

Quote:

Originally Posted by Onepagebook

** here is my suggestion again:
when you do pi, DO NOT use 200, 15.6 for tref.
it'd be unstable and really lot more slower than 100 or 133, 1.95

OPB, does that tref setting have anything to do with your cpu:ram ratio? for example does using 166 1.95us change it from 1:1??
03-06-2005, 11:08 PM
TridenT

thks too...heh

read it before put my gskill to work.
works great!

http://lepclan.com.br/webfile/gskill345mhz.jpg

http://valid.x86-secret.com/show_oc.php?id=5681

gonna try if its stable...

[]´s
03-06-2005, 11:27 PM
Onepagebook

Quote:

Originally Posted by FlyingHamster

OPB, does that tref setting have anything to do with your cpu:ram ratio? for example does using 166 1.95us change it from 1:1??

No, it just the ram refresh rate
nothing to do with CPU/RAM ratio ;)
03-07-2005, 07:37 AM
saaya

thx for the tips! very helpfull!

heres another six pack for ya :D

:toast: :toast: :toast:
03-07-2005, 07:51 AM
Vapor

Quote:

Originally Posted by ben805

FINALLY, I DID IT!!! 313 @2.5-4-7-3 1T Oh YEAA BABY!!

Ummmm, if I'm not mistaken, you're only running your RAM at 296MHz, not 313MHz ('cause of half multipliers).

Try running it again with a multiplier of 8x.
03-07-2005, 08:42 AM
HARDCORECLOCKER

:D OPB - I'm very impressed again, You are showing that OCing is not only a question of money but also of knowledge !!!

Unfortunately I can't test it because still waitin' for my board but I'm sure this will help me a lot with my first DFI NF4 / G.SKILL LA setup when RMA'd board has arrived back.

:toast:
03-07-2005, 08:58 AM
clockedOut

Quote:

Originally Posted by Vapor

Ummmm, if I'm not mistaken, you're only running your RAM at 296MHz, not 313MHz ('cause of half multipliers).

Try running it again with a multiplier of 8x.

OOPS, This 1/2 multi thing needs to be posted in BIG RED LETTERS someplace.... I see people getting real excited everyday with the 1/2 multi only to get the rug pulled out from under them.
03-07-2005, 12:52 PM
ben805

Quote:

Originally Posted by clockedOut

OOPS, This 1/2 multi thing needs to be posted in BIG RED LETTERS someplace.... I see people getting real excited everyday with the 1/2 multi only to get the rug pulled out from under them.

Have already tried that but for whatever reason this 3000+ winnie does NOT like any even multiplier, so 9x, 8x and even 7x are not possible to pass SuperPI 32MB on mine, 275 x8 or x7 are totally unstable regardless of voltage being use, even IF I can get it to run 314 x8 stable....but still, that's going to give me what? 2504mhz?? :rolleyes: the slightly higher memory bus of 18mhz does not make up, and failed to compensate the overall performance lost of 165mhz over the cpu clock. In other word, running 314 x8 I can do some brief sandra and everest mem bench, but the score is way less than running at 314 x8.5!! Memory clock isn't everything, you'll have to look at the whole picture and target the overall score, NOT just the memory clock!! Yea, with the 1/2 multiplier I might be getting "ONLY" 296, but that's 296 @2.669Ghz. Sure, I would love to run 333 x8 and still being able to maintain ~2.66Ghz, that can be done if AMD fairy willing to touch my cpu with her magic stick ... :stick:
Now here's the question, why didn't the CPU-Z or BIOS or any other utilities ever report the corrected value of 296 instead? I suppose the calculation of these softwares are inferior then? :stick:
03-07-2005, 02:21 PM
clockedOut

The problem I have is with the winchester core(I think) As long as I keep my mem under 200 (I have to use a mem divider) It will Prime at x7, x8, x9 multi's it primes at 300x8 but my GSkill DDR600 is on a divider to keep it at 200. When you use the .5 multi doesnt it use a mem divider? Im gonna test my winchester theroy tonight.... I have a 3500 newcastle core to try. OH, and it has to be at 2T to Prime. It wont Prime at any setting at 1T.
03-07-2005, 03:32 PM
saaya

ben805, could be your memory, when you use a half multi the memory frequency is lower than with full multis, if you use memory diveders the memory frequency jumps around uite a bit with different multis, so i guess its your memory that cant take the high frequency... lower the memory divider some more :)

could also be a bios issue though...
03-07-2005, 04:56 PM
ben805

Quote:

Originally Posted by saaya

ben805, could be your memory, when you use a half multi the memory frequency is lower than with full multis, if you use memory diveders the memory frequency jumps around uite a bit with different multis, so i guess its your memory that cant take the high frequency... lower the memory divider some more :)

could also be a bios issue though...

hmm....I doubt it's the memory, if these sticks are unstable at 275 to begin with, then using the 1/2 multiplier on 313 or 314 with the result of 296, theoritically at this speed the SuperPI 32M and Prime95 would of failed horribly. I don't know what it is but this little winnie has it's own awkward personality that doesn't make much sense.

313 or 314 (x8.5 or x7.5) @1T, 1:1, 2.5-4-3-7= SuperPI 32M, memtest #5, and Prime95 blend test rock-stable on DFI Ultra-D.
275 (x9, x8, or x7) @1T, 1:1, 2.5-4-3-7= SuperPI crashed within 30 seconds, memtest passed, Prime95 failed within 2 seconds, constantly getting random BSOD, regardless what vcore and vmem being use, different version of BIOS made no difference. Motherboard made no difference either since this cpu carry the exact same behavior and signature from Neo2 Platinum to this DFI Ultra-D.

:stick:
03-07-2005, 05:23 PM
saaya

Quote:

Originally Posted by ben805

hmm....I doubt it's the memory, if these sticks are unstable at 275 to begin with, then using the 1/2 multiplier on 313 or 314 with the result of 296, theoritically at this speed the SuperPI 32M and Prime95 would of failed horribly. I don't know what it is but this little winnie has it's own awkward personality that doesn't make much sense.

313 or 314 (x8.5 or x7.5) @1T, 1:1, 2.5-4-3-7= SuperPI 32M, memtest #5, and Prime95 blend test rock-stable on DFI Ultra-D.
275 (x9, x8, or x7) @1T, 1:1, 2.5-4-3-7= SuperPI crashed within 30 seconds, memtest passed, Prime95 failed within 2 seconds, constantly getting random BSOD, regardless what vcore and vmem being use, different version of BIOS made no difference. Motherboard made no difference either since this cpu carry the exact same behavior and signature from Neo2 Platinum to this DFI Ultra-D.

:stick:

the memory dividers are not working totally accurate, here:
http://www.xtremesystems.org/forums/...ad.php?t=41595