Core i7 Vdimm/VTT Graveyard :D

[Slovnaft]

its setting uncore voltage aka vtt to 1.75v by default? damn!


hmmm isnt that kingston 2000 mem micron based?
if yes then it needs more juice to get higher i think.
1.9-2.1 is usually the sweet spot for micron above that is where it stops scaling well, unles you have a really nice set of course..

1.65, and yeah, the XMP profile sets that, i run at 1.5v for 4ghz+ uclk benching seems stable enough without pushing too far.

Kingston said they're most likely Elpida though I wouldn't be surprised if it was Micron, I'm not going to risk ripping off the ICs again though. I'm convinced that R2E is the reason why these sticks aren't clocking well. I just put a patriot kit through the paces for a review and had the same limitations, only about 100mhz OC max on 1.8v where others have gotten 200+ at stock. I have put these sticks through 2.04v with no gain. tightened timings are a no go, they get 1800 at cl8 max. Hopefully I'll get to fool around with a gigabyte board soon.

[systemviper]

1.65, and yeah, the XMP profile sets that, i run at 1.5v for 4ghz+ uclk benching seems stable enough without pushing too far.

Kingston said they're most likely Elpida though I wouldn't be surprised if it was Micron, I'm not going to risk ripping off the ICs again though. I'm convinced that R2E is the reason why these sticks aren't clocking well. I just put a patriot kit through the paces for a review and had the same limitations, only about 100mhz OC max on 1.8v where others have gotten 200+ at stock. I have put these sticks through 2.04v with no gain. tightened timings are a no go, they get 1800 at cl8 max. Hopefully I'll get to fool around with a gigabyte board soon.

I should finally be getting my system up this sunday for it's first paces on air before hitting it with water and putting it in a MM duality modded for horizontal MB.,

I am glad that the bla kingston ram you talked about before might be the board, I can't wait to see what it will do on the Gig X58 Ex, it should be fun, but man i need to read about all the settings, not like a 775 i guess

[hiwa]

tried over 100 hours testing/bench
I7 965 retail
Vqpi: 1.60/8
Vdimm: 2.00/04

[Slovnaft]

I should finally be getting my system up this sunday for it's first paces on air before hitting it with water and putting it in a MM duality modded for horizontal MB.,

I am glad that the bla kingston ram you talked about before might be the board, I can't wait to see what it will do on the Gig X58 Ex, it should be fun, but man i need to read about all the settings, not like a 775 i guess

what kit did you get? if you have clip-on hs would you mind checking the chips for me?
It's actually very very similar to 775. not much of a learning curve. really only about 3 voltages to manipulate

[zads]

are you sure about this?
I checked with corsair and they said xmp doesnt set vtt...
and i chedked with one of our fae's, we tested several xmp profiles on the intel reference board and none of them change vtt...
hmmm

Yep I'm sure because.. I write XMP SPD profiles as part of my job

It actually depends on what revision the XMP profile is under.
The up-to-date spec, v1.2, specifies "QPI (Bus) Core Voltage Level".
This voltage is simply ignored when used in non-X58 (X48, X38) systems.

Previous XMP specs (any DIMM that has an old XMP profile on it) do not include this voltage,
and thus will not automatically set a QPI (un)core voltage.

[T_Flight]

Yes, Intel XMP SPD specification for Intel X58/Tri-channel requires that the memory module manufacturer specify in the SPD what voltage to run the DIMM as well as what voltage to run the uncore/QPI/memory controller voltage.

Yeah, and I got lied to (along with the other users there), flamed, and ran off Corsair's Forum by a rep there for trying to help somebody by explaining that to them. Needless to say that won't be happening again. I am no longer a member there and will no longer use or discuss those products ever again. No good deed goes unpunished with them I guess.

[saaya]

i can benchs with my UD5 / OCZ D9GTR at 2.2/2.22v since 4/5 hours by day (in 5 days)

It's ok , i have too on aircooling my STT 1800 at 2.36/2.4v and no pb for me

what vtt setting?

:update: Well forget about me testing XMP profile, i just tried to use it and i couldn't post

yeah, thats the thing with xmp... there isnt any certification, anybody can create an xmp profile and ship it out and it could potentially have 2.4v vdimm and 1.1v vtt programmed in it too, which would kill the cpu or can damage the mem... thats why ive never been a big fan of xmp... and most mainboard makers dont like xmp

about qpi dram voltage... well initially asus called the bclock/qpi so then it kinda made sense to call vtt qpi/dram clock cause more vtt means a higher bclock/dram clock is possible... but like i said, thats just confusing in the end.
i hope people will stick to vtt... vtt is already well known from c2d and c2q anyways, even though it was a different voltage.

Using the Corsair 1600C8 on the Asus P6T-Deluxe with everything on AUTO, turning XMP profile on sets the voltage of VTT(QPI/DRAM) to 1.35V, and VDIMM to 1.66V. All other voltage settings remain on AUTO.

yeah but are you sure the xmp sets higher vtt?
or xmp sets vdimm and the bios then automatically increases vtt to adjust it to vdimm?

i asked our contacts at corsair, ocz, adata and gskill as well as intel and they said they dont know about xmp setting vtt? 0_o supposedly theres an xmp spec v1.2 but none of our engineers received it and when they asked their intel contacts they didnt know about it either...

i guess it does exist but wasnt really promoted/sent out very efficiently... or maybe many people ignored it cause nobody really likes xmp that much

Yep I'm sure because.. I write XMP SPD profiles as part of my job

It actually depends on what revision the XMP profile is under.
The up-to-date spec, v1.2, specifies "QPI (Bus) Core Voltage Level".
This voltage is simply ignored when used in non-X58 (X48, X38) systems.

Previous XMP specs (any DIMM that has an old XMP profile on it) do not include this voltage,
and thus will not automatically set a QPI (un)core voltage.

k, thanks... i guess ill just continue to email intel guys until somebody knows about the 1.2 doc and can send it to our engineers

Yeah, and I got lied to (along with the other users there), flamed, and ran off Corsair's Forum by a rep there for trying to help somebody by explaining that to them. Needless to say that won't be happening again. I am no longer a member there and will no longer use or discuss those products ever again. No good deed goes unpunished with them I guess.

sorry to hear about that

[Ghigo]

My corsair TR3X6G1600C9 xmp also raise vtt to 1,35 if activated on ga-x58-extreme.
But i also think is motherboard raising it cos memory raised to 1,65
As a side note running np 8-8-8-20 with 12 GB of it at 1600 np

[jcool]

At what BCLK?

[Ghigo]

At what BCLK?

200x21 F3 bios, with F4 i got problems get so high , cant even get to windows at 190.
Just finished a memtest 2 pass with latest one all good at this timings 8-8-8-20 T1

[T_Flight]

what vtt setting?


yeah, thats the thing with xmp... there isnt any certification, anybody can create an xmp profile and ship it out and it could potentially have 2.4v vdimm and 1.1v vtt programmed in it too, which would kill the cpu or can damage the mem... thats why ive never been a big fan of xmp... and most mainboard makers dont like xmp

about qpi dram voltage... well initially asus called the bclock/qpi so then it kinda made sense to call vtt qpi/dram clock cause more vtt means a higher bclock/dram clock is possible... but like i said, thats just confusing in the end.
i hope people will stick to vtt... vtt is already well known from c2d and c2q anyways, even though it was a different voltage.

yeah but are you sure the xmp sets higher vtt?
or xmp sets vdimm and the bios then automatically increases vtt to adjust it to vdimm?

i asked our contacts at corsair, ocz, adata and gskill as well as intel and they said they dont know about xmp setting vtt? 0_o supposedly theres an xmp spec v1.2 but none of our engineers received it and when they asked their intel contacts they didnt know about it either...

i guess it does exist but wasnt really promoted/sent out very efficiently... or maybe many people ignored it cause nobody really likes xmp that much


k, thanks... i guess ill just continue to email intel guys until somebody knows about the 1.2 doc and can send it to our engineers

sorry to hear about that

Saaya,
I know you work with Foxconn, but I ntoced this when watching things on my Rampage II Extreme.

Conditions: Voltages Set "Auto", Set XMP DDR 1600 Profile, No OC's, Boot To WinVista64, Open TurboV, Check Voltage, Test Voltage Via ProbeIt to verify.

I know you have the experience to get to the voltage points directly on the board. If you wanna verify it, you can do it the same way i did above using Foxconn's Dekstop OC Util (assuming you guys have one), and measure the QPI/DRAM (Vtt) on the Board. I'd also keep an eye on the vDelta's, and watch for any bugs in case something decides to runaway. I haven't seen any runaway, but a user did report that in another thread here (sorry cannot find it now), and sent his board in to be looked at.

I was extremely careful, and really cautious with things related to emmory Subsystem, becasue i knew this stuff was new, and if I found anything it could potentially help alot of people, and save many greif. I also knew I had a 1000 dollar CPU, a 400 dollar motherboard, and 350 dollar RAM in there and I didn't want to screw it up, so *that* made me even more careful.

It's an easy test to set up. It might even make QVL's interesting!

[zads]

what vtt setting?
yeah, thats the thing with xmp... there isnt any certification, anybody can create an xmp profile and ship it out and it could potentially have 2.4v vdimm and 1.1v vtt programmed in it too, which would kill the cpu or can damage the mem... thats why ive never been a big fan of xmp... and most mainboard makers dont like xmp

about qpi dram voltage... well initially asus called the bclock/qpi so then it kinda made sense to call vtt qpi/dram clock cause more vtt means a higher bclock/dram clock is possible... but like i said, thats just confusing in the end.
i hope people will stick to vtt... vtt is already well known from c2d and c2q anyways, even though it was a different voltage.

yeah but are you sure the xmp sets higher vtt?
or xmp sets vdimm and the bios then automatically increases vtt to adjust it to vdimm?

i asked our contacts at corsair, ocz, adata and gskill as well as intel and they said they dont know about xmp setting vtt? 0_o supposedly theres an xmp spec v1.2 but none of our engineers received it and when they asked their intel contacts they didnt know about it either...

i guess it does exist but wasnt really promoted/sent out very efficiently... or maybe many people ignored it cause nobody really likes xmp that much


k, thanks... i guess ill just continue to email intel guys until somebody knows about the 1.2 doc and can send it to our engineers

Merry Christmas guys

Actually, there is a pretty thorough Intel certification process,
just like there is for SLI/EPP 2.0 profile memory.
Only profile 1 (Intel certified settings) is required to go through certification.

It is also possible for memory module manufacturers to just write an uncertified Intel XMP profile on their memory..
(the same goes for SLI/EPP 2.0 profile memory)
However, if they are uncertified they aren't allowed to put "Intel XMP certified" on their product,
or put any Intel Extreme certified logos on their products.
This is an example of an Intel XMP-certified module WITH logo:


Since its Christmas eve, and I'm not so busy at work today,
I've made some snippets to show you guys the Intel XMP spec and certification process:

[Xello]

Found this also in my board's manual;

Profile QPI Voltage

This value is dependent on the Extreme Memory Profile (X.M.P.) settings. When Extreme Memory Profile (X.M.P.) is set to Disabled, this item will display as 1.2V. When Extreme Memory Profile (X.M.P.) is set to Profile1, this item will display the value based on the SPD data on the XMP memory.

Weird with the 1.2v, my board's stock vtt is 1.15v.

[Achill3uS]

little bump for the topic

used this voltages for testing my retail 965XE in P6T Dlx aircooled all of them over 10 hours for SPI 32M, still nothing goes wrong, nothing died, no degrades.

vcore: 1.40-1.58~
vqpi: 1.60-1.65~
vdimm: 1.80-1.96~

[M.Beier]

little bump for the topic

used this voltages for testing my retail 965XE in P6T Dlx aircooled all of them over 10 hours for SPI 32M, still nothing goes wrong, nothing died, no degrades.

vcore: 1.40-1.58~
vqpi: 1.60-1.65~
vdimm: 1.80-1.96~

[westsider]

i have benched with
vcore 1.84
vdimm 2.25
vtt 1.78


no problems at all

but voltage above 1.65volt doesnt give much at all
i just tested my cooling system

[Achill3uS]

what chip did you tested? these are pretty high voltages on vcore/vtt for even ln2 imho... any results/gain over 1.65?
oh ya, and how long did you feed this?

[westsider]

i have watercooling with about +2 +4 celsius water


and it was 3838A

[Achill3uS]

omg so you used this volts on chilled water? quite brave ya

3838 but 920, 940 or 965? just for curious.

[westsider]

920

here are 1.7v+ for pi

[systemviper]

Sweet!~

how you getting your water that cold?

[westsider]

we have very cold winter in germany
i use anti freeze
otherwise i must warm up my radiator every morning

[Brian y.]

little bump for the topic

used this voltages for testing my retail 965XE in P6T Dlx aircooled all of them over 10 hours for SPI 32M, still nothing goes wrong, nothing died, no degrades.

vcore: 1.40-1.58~
vqpi: 1.60-1.65~
vdimm: 1.80-1.96~

That is exactly what I have kept all my testing voltages at and with that range. Most of the testing was done under lN2(CB-120) and now with all 8 cores enabled CPU fails OCCT instantly at stock settings. If HT is turned of then it is fine.

965 retail 3836A767

going to rma this week.

btw, this is the second retail CPU I know of that this has hapened to that was seeing the voltage with in the limits you have posted.

[TheKarmakazi]

going to rma this week.

Hope you get a good one, Intel usually gives out some gems in RMA

[jcool]

3835A730 i7 940, 3rd memory channel died last week. Was running only 1,7V Vdimm 24/7, around 1,45-1,5V VTT. Used 2V for like 5 minutes when I first got it, but its IMC never even could handle 1600Mhz on the rams stable. So I am guessing its IMC was flawed to begin with... on phase change.