Cas 2.5-3-2-1-1T, Stable... at DDR508... at 2.2 Volts.

[Kunaak]

Probably the weirdest ram I ever came across...
this ram doesn't respond at all to voltages.
what it does do, is get more and more stable, the faster you overclock it... and the lower you let the voltages go.

for instance, 240 at cas 3-3-2-8-1T at first felt impossible, cause like anyone, I would think "ok, 2.6 volts isn't good... lets try 2.8" and so on.
well, after a time I forgot to set the voltage, when clearing a CMOS, I saw that I got into Memtest and it was running great.... at 2.5 volts.
weird I thought, and rebooted... set it to 2.6 volts... and it would get into memtest... and immediatly start to error...
out of grim curiousity I set it back to 2.5 volts... and Viola... stable.

that was with my ULtra D.
at that moment, I began thinking of trying to talk some friend of mine to mod a DFI bios to see if they could get me some lower voltages cause I had a feeling about something...
then I got a DFI Expert to use, and that has a option of going down as far as 2.3 volts... luckily it undervolts... so it really went down as low as 2.2 volts in my case... which for once, worked in my favor.

now, I am even more curious to see what would happen if I could take the voltages even lower...

the settings you see here are boot up settings, they have not been adjusted in windows at all.
I chose a tras of 1, cause tras 0 shows up as a blank spot in CPUZ... I wanted it to be clearly shown that it was tras 1.

the Memtest shot below, has been running for the last 2 days straight, with the full test, not just test #5 for a burn in session, the entire test to make sure it's solid.
the 32M test is just to make sure its Windows stable and it is.
I only have 1 videocard at the moment, and thats in my Main PC, so I have not tested this in 3DMark or Gaming yet.

I've also screwed up the OS a few days ago, cause I was trying to run 32M at cas 2-2-2-5 -1T at 215 at 1:1 .... and it didn't like it and corrupted.
since then, I can't get Sisoftsandra to work, so no bandwidth tests yet.
but I can reinstall windows, if anyones real curious to see that.

[s7e9h3n]

Maybe I missed it, but I can't find where you mentioned the brand/IC's of your sticks

[Kunaak]

Corsair 3500 LL Pro.

Infineon is what I was told they are.
I will have to confirm that later.
I can't remove the IHS however.
this ram has some LED on it, and I have no idea how thats hooked to the ram and I don't want to remove it, so I will just have to ask later.

[Vassili]

Is that also game/3D-mark stable?

Very nice result!

I've also screwed up the OS a few days ago, cause I was trying to run 32M at cas 2-2-2-5 -1T at 215 at 1:1 .... and it didn't like it and corrupted.

What is the highest 2-2-2-5 run with these sticks?

[MaSell]

vdimm like ddr2 :P

[biohead]

:|

ok how bout everyone with infenion ic's undervolt it and see what happens.
many expert users here so shouldn't be a problem.

weird stuff... there are lots of ocz eb users who cant even get 260 stable. maybe they just need to lower the volts to 2.5...

[stealth17]

how do you set 2.2v when the bios on the Ultra-D is only down to 2.5v, then it overvolts .1v even at that?

DDR Booster?

[Pyr0]

how do you set 2.2v when the bios on the Ultra-D is only down to 2.5v, then it overvolts .1v even at that?

DDR Booster?

Kunaak wrote:

then I got a DFI Expert to use, and that has a option of going down as far as 2.3 volts... luckily it undervolts... so it really went down as low as 2.2 volts in my case... which for once, worked in my favor.

[brandinb]

hey thats pretty cool kunak

i would request a mod in the volt mod section and im sure they can show you how to turn down the v-dimm even lower then 2.2v

[stealth17]

Kunaak wrote:

my bad

[metro.cl]

weird ram

[CCUABIDExORxDIE]

-5b F acted similar for ozzimark too, the lower the voltage the higher. he was thinking of a way to get less voltage, but i dont think he knew about the Expert. ill need to show him this thread.

[Vassili]

Xtreme systems has to make another sub-forum then; Xtreme Low-Volt mod's

[musk]

Corsair 3500 LL Pro.

Infineon is what I was told they are.
I will have to confirm that later.
I can't remove the IHS however.
this ram has some LED on it, and I have no idea how thats hooked to the ram and I don't want to remove it, so I will just have to ask later.

Please post when you confirm the IC's.

are these the modules?

If so;

Lurking underneath the classy but effective heat spreaders, lie Infineon -5 64M X 8 Revision B IC's that have been speed binned by Corsair for use in this product. This is BE-5, or 5 nanosecond RAM, with normal timings of 2-3-3-6. Corsair has shipped these modules with the SPD programmed at 2-3-2-6 at a 1T Command Rate as previously mentioned. In previous iterations of Corsair's 3500LL Rev 1.1 memory, the IC's used were Winbond -5 Rev C, however to due short supply Corsair moved to Infineon IC's which proved quite apt for use in their low latency memory modules.

[Redbeard]

Please post when you confirm the IC's.

are these the modules?

If so;

Lurking underneath the classy but effective heat spreaders, lie Infineon -5 64M X 8 Revision B IC's that have been speed binned by Corsair for use in this product. This is BE-5, or 5 nanosecond RAM, with normal timings of 2-3-3-6. Corsair has shipped these modules with the SPD programmed at 2-3-2-6 at a 1T Command Rate as previously mentioned. In previous iterations of Corsair's 3500LL Rev 1.1 memory, the IC's used were Winbond -5 Rev C, however to due short supply Corsair moved to Infineon IC's which proved quite apt for use in their low latency memory modules.

I work for Corsair, this is exactly right. We initially used the Infineon revCs, but when they were in short supply, we used the Infineon revB ICs.

Right now it really depends on what we have in stock, we're backlogged on the 3500LL because it's such a popular part, luckily we have been able to buy a buttload (technical term) of ICs over the past week or so and it's made a huge dent in our backlog. Keep patient, the modules are shipping out every day by the hundreds.

As for the revB vs. revC, I know revision 1.0 is supposed to be revC, but if you post your revision number in this thread I will let you know for sure which IC is on them. (To the original poster, not to everybody else, that will be a huge headache since I would have to look them up each time...heh.)

[cdelong]

it's nice to have froum support from manufacturers. Thanks!

[ozzimark]

-5b F acted similar for ozzimark too, the lower the voltage the higher. he was thinking of a way to get less voltage, but i dont think he knew about the Expert. ill need to show him this thread.

only when at 3-3-3, 1T though.. probably a limitation of the board's ability to supply power more than anything.

[celemine1Gig]

Nice to know. I bought some Infineon BE-5 based 1GB sticks for a friend. Those sticks are only able to do a stable ~205Mhz while fed with ~2.6V and timings of 2.5-3-3-10. And the worst is, even with 2.5V they get hot. I mean really hot. I think those sticks could be candidates for low voltage vmodding.
It's not that hard BTW. Normally you connect a resistor in parallel to get higher volts by slowly decreasing the resistance. To get lower volts (in contrast to this normal method to increase volts) you just need to connect a resistance in series and start to slowly increase from zero upwards.

[Kunaak]

I don't see how to find any revision number to the ram.
I am looking at it right now, and other then a serial number, theres nothing to indicate what version they are...

[lowfat]

that is very interesting and impressive

[s7e9h3n]

I don't see how to find any revision number to the ram.
I am looking at it right now, and other then a serial number, theres nothing to indicate what version they are...

They don't look like these?

[chinkgai]

so those of us with the "unlucky" mobo's that dont go 2.2...are we out of luck unless we do some soldering?

[Kunaak]

oh theres the problem then...

see the sticker these are printed on, is extremely cheap.
the ink in the "V1.2" of that picture has just rubbed off, cause thats pretty much the exact point I use to pinch the ram, when pushing it into the socket....
is it possible to get the version number from the serial code or something?
cause that area, is completly gone from both sticks.

anyone that gets this ram, will know exactly what I mean, that stick is like grease, and the ink on it, comes off with almost no effort at all.

here you can see for yourself what I mean...





the 1 and 2 on the ram, I wrote myself, so I could remember which stick was the best of the two.
for the record, the first stick does 280 at cas 3-3-2-8-1T easy.
the second stick does 274 cas 3-3-2-8-1T easy.
both stable on thier own.
combined... 262 is the max. same latency.

oh yeah, this ram absolutly must be used in the Orange Slots.
the yellow slots are pure garbage for this ram.
atleast in my case.

Heres my EXACT Bios settings.
one more note...

tras 0.... is also stable at these settings.








the Bios I use is 11-18. I believe it was called FX60 Test, when I downloaded it from techpowerup.com
I don't know if this bios is any different from any other, I was having huge troubles with the DR Expert when I got it, cause it wouldn't work with any of my 2 gig ram kits.
with a few emails to DFI and the help of RGone from DFI Street, he gave me a solution that allows me to use 2 gig ram kits in this board without much hassle (boot single channel first, then set your bios settings, shut down, then put the second stick in, if you have problems)
I tried many bios's seeing if any would help my 2 gig problems, none did.
when I finally got the solution from RGone, the 11-18 bios is what I had on there... now I don't want to change it, cause I might jinx something haha...

[Redbeard]

oh theres the problem then...

see the sticker these are printed on, is extremely cheap.
the ink in the "V1.2" of that picture has just rubbed off, cause thats pretty much the exact point I use to pinch the ram, when pushing it into the socket....
is it possible to get the version number from the serial code or something?
cause that area, is completly gone from both sticks.

Wow, good feedback. I've never seen the ink rub off our modules before, I'm going to check into that.

I'm almost 100% sure you have the Infineon revC chips there, the revBs don't really clock much over 250 usually. There's another way to know exactly, if you can get me the lot number, it usually starts with 05 or 06 on those, I can find out specifically.

There's also an outside chance it's the Aeneon revC chips, but those are really Infineon revCs anyway.

[chinkgai]

argh, i have aeneon chips that does 245 @ 2.5-3-2...im itching to try this but dont wana solder to lower volts...rofl

kunaak, what exactly did u gain going from regular voltage (2.6-2.7) to 2.2?