Samsung DDR3 G-die

[ObscureParadox]

37 minutes ago, ground said:

Tried again a couple days later. Got 1992MHz at least 6 times across multiple boots, yet not a single time managed any higher. Something is walling hard again. Vdimm doesn't matter 2.3-2.6V, VTT doesn't matter 1.45-1.55V, dram vtt 0.95-1.2V. CPU temp -60 to -80 no difference either, so suspecting either board or mem to blame here.

Time to try good MFR on X58?

[ground]

32 minutes ago, ObscureParadox said:

Time to try good MFR on X58?

the loosest timings you can run on x58 are 11-15-15-31 with trfc 255, MFR is out for that mainly due to trfc.

[TerraRaptor]

maxmem, spd edits didn't help either?

[ground]

24 minutes ago, TerraRaptor said:

maxmem, spd edits didn't help either?

Was already running 64MB maxmem, didn't try lower. SPD edit in which sense you mean? Max bootable on this config is around 1840MHz.

[TerraRaptor]

Like forcing 512mbit instead of 1gbit per ic. It was either you or Gigelz to do it for ddr2 if i'm not mistaken:)

[ground]

39 minutes ago, TerraRaptor said:

Like forcing 512mbit instead of 1gbit per ic. It was either you or Gigelz to do it for ddr2 if i'm not mistaken:)

Yeah, I did that a while ago. I would have thought with maxmem active that would at best allow posting higher settings, but I can try next time I have LN2 on hand. 

[somerandomtechyboi]

On 9/2/2024 at 6:30 PM, ground said:

the loosest timings you can run on x58 are 11-15-15-31 with trfc 255, MFR is out for that mainly due to trfc.

i think the asus p6x58de i have can run over 255 if i leave it at auto subtimings so im curious if that might work on giga boards

my x58a ud3r shows >15 trp trcd and >31 tras if i set them at auto for high freq but in windows checking cpuz only shows the usual 11-15-15-31, might still be worth trying though

ill have to go check the bios settings for that asus board again cause when i tried to set it at trfc 240 or someting its not stable at 1900 9-12-12 2v, ics are micron d9qbj aka hot garbage should have bought 2gbit d die or cfr instead cause quadrank should fare better than whatever this garbage is

thatd really suck if being able to set loose timings is only on ami bios considering the artificial uncore limit, but if there were a bios mod thatd allow freely setting uncore independantly from the ram multi then maybe those rampage boards arent good for nothing after all

 

i am considering per ic binning myself but i have no clue how to do it on ddr3 and where to source the pcbs and i only have a heatgun rather than a rework station

maybe per ic binning will help me get 3000 tri channel stable but i think ill focus on imc binning cause the first channel is complete garbage on my w3680 compared to the 2nd and 3rd where i was able to get 3008c11 running for 2 hours p95 before crashing with two sticks (suspecting both stick and boardlimit cause that outer channel will not detect on boot if its over 3008), since you are already that close to 4000 i think ill just focus on 45nm frequency and 32nm tri channel stability cause theres no way my crappy sticks and mildly binned cpu is gonna compete with that even under ln2

 

ive also got some ecc gdie sticks now

dualrank ones at that but for whatever reason both my w3670 and w3680 refuse to detect them albiet my w3503s can, and i cannot get them clocking past 2000 for some reason which i am suspecting pcb limit cause these are low profile dimms, probably just gonna salvage the ics when i get around to per ic binning otherwise these are trash sticks

 

i wonder if maxmem will help with stability over 2.3v cause early on when i was toying around with my i7 930 before i even managed 3200 on it due to being an idiot and running 2 sticks i did manage to get 3136c9 to post but windows refused to boot, trying to push past 2.3v on any of my stable ocs caused them to crash

this isnt really that useful if im only going up to 2.6v but i do intend on voltmodding my board to go well over 3v vdimm just to toy around with extreme voltage and id like to run an actually good ic at those volts see how well it does instead of just torturing my poor samsung f die stick just because it shows minor scaling at those volts

Edited September 6 by somerandomtechyboi

[somerandomtechyboi]

On 8/28/2024 at 2:24 PM, ground said:

3700 isn’t platform limit of x58, seems like it was just the limit of my memory sticks at the time. 

Last week, Tagg Shaggy and I were building some custom Hypers and in the process also decided to take apart the best 3 sticks of DDR3L ECC gdie I had and slap the best ICs in order on new better PCBs. 

First try with best stick of the lot (2948 trp 12 binned air) - 3780. Rather disappointing improvement. Took 2nd best stick (~2900 trp 12 air) and a different IMC on LN2 and relatively easily got to ~3820 Added capacitors to the back (2x 330uF) and got to 3880. 

After that, took the best stick, added capacitors and threw it on my good IMC on LN2, and got far to close to 4000 to not try again next couple days. 

https://hwbot.org/submission/5631354

also cap modding rams? ive heard of gpu cap modding but not cap modding rams, and it seems to be a decent improvement aswell compared to gpu cap mods

where the hell do you even solder the caps on anyways?

edit: did not mean to add another post cause i thought i was editing my post and adding this quote to it, whoops

Edited September 6 by somerandomtechyboi

[ground]

1 hour ago, somerandomtechyboi said:

also cap modding rams? ive heard of gpu cap modding but not cap modding rams, and it seems to be a decent improvement aswell compared to gpu cap mods

where the hell do you even solder the caps on anyways?

edit: did not mean to add another post cause i thought i was editing my post and adding this quote to it, whoops

very carefully dremel the backside of a single sided dimm to expose the gnd plane. Just as carefully dig with the dremel one layer deep to expose the vdimm plane for the second pad. Cover exposed gnd with soldermask accordingly.

https://imgur.com/a/trying-to-achieve-ddr3-4000-on-x58-rHGRQn0

Edited September 6 by ground

[somerandomtechyboi]

9 hours ago, ground said:

very carefully dremel the backside of a single sided dimm to expose the gnd plane. Just as carefully dig with the dremel one layer deep to expose the vdimm plane for the second pad. Cover exposed gnd with soldermask accordingly.

https://imgur.com/a/trying-to-achieve-ddr3-4000-on-x58-rHGRQn0

i am really illiterate when it comes to electronics so i have no clue where the vdimm plane is but i assume the ground plane is basically everywhere, im guessing you can check the vdimm plane with a multimeter?

i do have a dremel but its one of those cheapo china dremels, i assume you want to use a sandpaper tip rather than a diamond tip for removing solder mask and cutting a layer

pics wise those look to be tantalum capacitors, not sure what those metal silver looking things are on the green stick (ceramic caps?) but since the best stick is using all tantalum caps ill just assume theyre the best ones to use, theyre dirt cheap aswell so thats nice, wonder how well a normal solid cap or an electrolytic cap would do for this but oh well thats something to test for myself not that theyd be any better but who knows

looks pretty doable with some basic soldering tools which i do have, at worst i can always just use some low temp solder paste and my heatgun if i cant solder with an iron

aside from cap modding the ram itself would cap modding the vrms that supply vdimm and vtt be worth it or no? i guess theyd be pretty hard to mod due to being on the power plane that absorbs all of the heat from a soldering iron

Edited September 7 by somerandomtechyboi

[ground]

if you have no clue what you are doing avoid this modification. 

[somerandomtechyboi]

5 hours ago, ground said:

if you have no clue what you are doing avoid this modification. 

still gonna do it anyways

i dont really care if i kill a stick or two, besides i have a trash gdie stick i dont mind killing and these x58a ud3rs are a dime a dozen if i somehow manage to kill one not to mention i have three of them available albiet one needs its main bios resoldered

just have to learn abit about electronics but ill have to remind myself to cause i know my dumbass is gonna waste time watching youtube nonsense

voltmods capacitor mods vrm mods all seem pretty simple, just need to read up on this stuff and actually execute it, cant exactly execute with no knowledge and theory can only get you so far

maybe i should start with simple things like epowering my shitty f die stick to run 3v+ (though i could just hook it up directly to the 3.3v rail) then move on to more complex stuff

i was definitely oc illiterate 5 years ago with my first build but once i got a better board even though its not optimal (p5q) i just kept getting info online did trial by fire testing to get even more knowledge and it just kept on after that, basic ebb swap bios mod with mmtool for higher fsb, screwing around with crossflashing bioses, etc. just lots and lots of exploring without any real limit cause i knew if i managed to somehow kill something the hardware is dirt cheap anyways

as long as i stick to the same principals i use for oc (lots of gathering info, lots of trial by fire, no arbitrary limits) im pretty sure ill get good at this in no time, basically if i dont have a clue then go get a clue rather than avoiding it entirely

currently in the process of moving though so itll be awhile till i can actually put any of this into practice, the only real limiter i have is my laziness cause i still havent soldered the m bios on that broken ud3r despite buying a soldering kit months ago (still basically untouched) nor have i done much learning on electronics or sold my piles upon piles of boards both working and dead

[somerandomtechyboi]

oh and i just realized that i never actually posted any of my stable freq on my w3670/w3680 so ill just put these here

and yes im putting here as i reference them alot =p

timinngs are as above with vtt 1.52v and vdimm 2.3v with vioh at 1.3v pll at 1.4v and both clock drives at 700mv though this will be annoying to restabilize for the w3680 due to inconsistent mounting on x58 and my dumbass forgot to set cl10 on the w3670 but oh well looks like gdie doesnt mind overvolting aside from going past 2.3v where it just completely destabilizes which sucks

and now ive got my hands on a g1 assassin, bios seems to be abit too new to run 1:1 uncore memclk on 45nm (similar date to ff bios on x58a ud3r ver2) so ill probably have to mod the bios and combine those two, even found this helpful repository of bios mod tools with the awardbios tools but ill do this later when i have some spare time, i hope itd be able to clock higher but the extra distance between the slots and the cpu is kinda concerning

ive even managed to get my hands on an i7 980x since someone was selling it with a working sabertooth x58 for 30$ albiet dead slots (presumably the inner slots from the main ones), if i manage to revive them like i have with this board ill definitely just resell to some idiot thatd still pay a premium for an asus board otherwise ill probably turn it into a build, i have yet to test the 980x other than knowing it works alongside the board via looking at the postcode but moving is almost done and ill have access to my ud3r again to do some testing

now i just have to source some more hynix cfr sticks to test 5 and 6 stick configs as 5 stick ive managed to get booted at 2600 albiet not a full boot as it was horrendously unstable due to mixing 2gbit cfr and 2gbit d die which do detect at this speed when combined but very unstable while running both cfrs in the innermost channel allowed me to boot into windows at 2600c11