Core i7 Vdimm/VTT Graveyard :D

[saaya]

I killed my XE C0

how? vcore? vtt? what voltages?

MSI Eclipse SLI...wouldn't have been my first choice, but I needed to do some homework for the THG Competition in Paris on Dec 13th & 14th...Team IORNMODS is representing as Team USA. Unfortunately, looks like my time to do homework has come to an end. I lost a second memory channel today. And the setup is quite unstable in single channel mode, even with stock clocks....I hope I learned enough to represent in Paris.

I think when this is over, I'll have to get the P6T or a GB

did you always run higher than default vtt with 1.8v vdimm?
are you sure you increased vtt to 1.45v?
how long did you run it this way?
what stepping was your cpu? B or C?
ES or retail?

sorry to hear about the cpu dieing... i hope it wasnt one you bought :S

just to be clear VTT = QPI/DRAM voltage.
1.65v has been my sweetspot, i need it to run 2133mhz RAM; the UCLK is the killer - can't run faster than 4300. thenagain I'm still on stock hs

qpi/dram voltage is VTT... the real and propper term is VTT imo, thats what its called in all intel papers. giving a voltage a marketing name or name after the effects it has when beeing increased only results in confusion :P

Saaya, so wich are concidered to be "safe" voltages for D9JNL chips with nehalem?
I´m using 1.8vDimm and 1.35V QPI with GA-EX58-UD5 right now
This is my best bandwith shot:
http://www.aijaa.com/v.php?i=3202172.jpg

why should it be different than on x48? if the drive strength is the same it should be the same voltage the chips can take before you test your luck and see if one of the chips on your sticks is sensitive to high vdimm and will blow or not.
i never saw any d9jnl die at a voltage it could run memtest 86 stable at...

That sound very "strange" to me, the catastrophics diode protections are stopping any heat crazy stuff, and I never saw in our lab a test of transistor dieing so fast ... you are doing something very wrong. Did you modify your board or something like this? With enough cooling, I never saw the life so short.

i didnt kill any core i7 with high vdimm so far, but ive only played with C0 chips...
our engineers did with a renessance and B2/B3 (?) sample and that was months ago
they didnt do anything crazy at all...
when our engineers told our intel contacts they werent surprised at all and said "yeah, dont run more than 1.8v with B, itll degrade or kill the imc within hours to days on some chips. if it degrades you need more and more vdimm to run the same mem speeds stable"
i can tell you who at intel we worked with if you want... fugger knows them too... dont wanna mention their names here tho, just pm me and ill let you know.

Do you still have the processor that you claimed are dead? I am interested in getting them for analysis, since we never saw that kind of extreme death.
Francois

i dont think they still have it, that was before idf and the chip has been returned to intel a long time ago afaik...

I have ran my C0 revision (I think it is) 965 with patriot memory at 1.9v and stock Vtt through prime95 for 24hrs and through 10 days general use before the actual memory started to fail and refused to post.

Now I am running Mushkin 2x6Gb kits at 1.64v at 1333mhz because the latest bios update refuses to allow speeds of 1600mhz

was that a retail or ES chip?
is the memory fine now on the new cpu?
what board? you know that the asus boards set vtt vcore and vdimm automatically even when it shows default, if you set high mem speeds, right?

sorry to hear your cpu died
how exactly did it happen?
no boot, or instability?

I have been running 2v 24/7 on phase

how long now? must bee 2 months at least right?
and your running what vtt? still around 1.6?

I've been running the following for 3 weeks now

CPU Voltage- 1.44v
VTT Voltage- 1.34v
Ram Voltage-1.65v

Everything seems ok. It even runs with HT Enabled. I'm under water idling at 40C and loading at 68C. When i enable HT i Load at around 78C. It's crazy how much of a difference it makes on the temps of the chip.

As for my CPU Voltage, have a run out of room? Am i shortining the life of my CPU at that CPU Voltage?? or can i try to hit 4.4 at 1.48 cpu voltage?

those voltages are fine, its the temps that are a tad high for 24/7 imo... you should try to get into the 60s... then again im not 100% sure how the core i7 temp ranges are, they seemed to have moved the temp probes within the cpu, so they might be at a hotter spot now than they were with core2duo/quad, so the temps we read are higher but still fine...

we will find out i guess... so far a couple of people are running close to 80 loaded and seem to be fine... i wouldnt recommend it though...

Erm.. you know you're not really running a 21 mult right? It's just the turbo mode kicking in on single threaded apps..

dont know about other boards, but on BloodRage you can
you can force turbo to always on and run 21x all the time... well not during boot, but thats it.

just wanted to pitch in a datapoint.
benched air for a few days at ~1.4vcore, 1.6~1.65 QPI/DRAM (VTT), and 1.6-1.7 vDRAM.
DI bench session on friday at ~1.6-68vcore, 1.75 QPI/DRAM (VTT), and 1.65-1.75 vDRAM.

Back on air now and behaving as well as ever. Interestingly enough my Kingston hyperx 2000mhz kit XMP sets QPI/DRAM to 1.65v automatically; which will get uclk up to 4300 on air, 1.75 is good up to 4500 and maybe higher.

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

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.

cool, thanks for the headsup! i didnt know that... but 1.75 vtt for stock speed... thats a LOT for 24/7...

Honestly i'm a bit disappointed with the Kingston ram. Insofar as I've tried I haven't gotten above 2133 for 32M stable on CL9, and the highest i've achieved thus far for CL8 is 1800. These are all on air, however, so QPI/DRAM stayed at 1.65v. Did not scale with volts. In my DI session I did a SPi32M run at 4560mhz uclk, but was not pushing the RAM (i think it was at 2000 maybe), so next sesh (tomorrow?) I'll be seeing how far this kit can go given sufficient uclk capacity. but I'm not anticipating results akin to fugger's corsair kit. I've seen a few kits of new ram getting better clocks in 2T, so I might try that next.
965. It has runaway amp at 1.7 under DI, but is pretty tolerant.

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..

nice Thread saaya will be same dead GTR Thread for used

i hope not, the gtr thread went way offtopic

thanks for the feedback everybody!
ill update the first post now and add your infos