Vcore and vdimm with A64

[Tony]

Guys this is most important

If you run a high vdimm and a low vcore with E die etc you are running the risk of killing the cpu's. So if you have VX etc and you are not raising the vcore on your cpu you stand a chance of pumping a lot of current into the memory controller and causing permenant damage.

Im seeing people running $1k cpu's at default vcore and BH5 at 3.4V, this is a sure way to kill the cpu and is NOTHING to do with an issue on the board. If you are going to run high vdimm run a higher vcore also, and keep your CPU's ok.

[kLLa142]

Uhh how high is high? 1.4 V enough? Will just testing it in memtest86 cause a problem?

[Tony]

if your cpu defaults to 1.35, run say 1.45V, you don't need to bash the cpu but you do need to strengthen the diodes that keep the CPU ok.

[4rory]

When was this discovered? Why is it only with e dies? because they are lover volts?

[Kenetixx]

Very Interesting , Thanks Tony

[Tony]

When was this discovered? Why is it only with e dies? because they are lover volts?

I have been shouting this for months but no one listens, so i decided to get a sticky here.

E die is the one with issues mainly due to lower vcore and smaller die

[ozzimark]

i think i understand what's going on, but would you mind explaining it from an engineering standpoint tony?

[rookiekiller]

Guys this is most important

If you run a high vdimm and a low vcore with E die

what is E die?

[tatts]

very interesting and glad im running tccd

[VoodooProphetII]

Okay...one question now. I run my BH6 memory at 3V Vdimm stock, with the Vcore at stock 1.36V... It's a 4000+ Sandy, would that be okay, or do I need to up my core as well.....?

[ozzimark]

what is E die?

revE, aka venice, san diego, manchester, toledo, newark, lancaster, denmark, venus, and a few other opteron cores i don't feel like remembering right now

[dexx]

I imagine that 3.5V may be the Vdimm you are supplying. But i cant see it being the voltage used for the bus between cpu and memory. What voltage do E6 chips expect for the memory bus? What voltage would overvolted BH5 present?

[ozzimark]

i'm not 100% certain, but i think the voltage the memory controller sees is VTT, which is normally 1/2 vdimm.

i need tony to clear up the specifics though, which is why i asked for an explaination from an engineer standpoint

[eva2000]

First i heard of this :o

haven't had any probs with 3700, 4000, fx-57 sandies and 5x opterons so far heh

but rarely at stock vcore heh

[[ r2 ]]

So what would be a safe vdimm to run while you running stock vcore?

[Monge]

Tony:Thanks for the info

I wonder if there's any relation between this and memtest #7 errors at high vdimm eg. 3.4-3.5v and vcore set to low eg. 1.4v?

So the recommended vcore offset is +0.1v when running high vdimm voltages?

[i found nemo]

but doesn't tccd draw higher current at most given voltages than utt??? hence the fact that they run hotter

[Tony]

the newer CPUs run a much lower vcore, IE 1.4 or 1.35V. what you are looking at is the difference between the vcore and the VDD supplied to the memory controller from vdimm in bios. Now if the gap between the vcore and the VDD is huge you stand a chance of breaking down the diodes that protect the core from the high VDD.

All you have to do is strengthen the cpu's defences, you do this with a bump of vcore...thats all.

Now i know some of you will say I have had no issues so i think this is BS, what I will tell you is if AMD say this to me i listen...and you should listen also.

It does not take a lot of vcore, just a bump from say 1.35 to 1.425 or so to help. The older 13nm cpu's don't look as much at risk although many did die quickly with high vdimm if you all think back.....this will have been the reason they died.

[aicjofs]

Seems completely logical. Do you have a link with any more in depth discussion?

[Bladesinger7x]

I have one question: What is considered to be high vdimm voltage?

(i.e., is 3.2v vdimm considered high or is as high considered anything more than 2.9v)

[da-key]

Has anyone cooked there E chip from overvolting it? I mean these chips can take some abuse within the limits of your cooling. And if what your saying is correct then the nf4 Expert version should contain no flaws since the memory voltage jumper issue is moot.
Good news, or is it, for those blaming the boards when all along its been AMD not ponying up. Whos liable ? What a mess if you ask me.

[don_vercetti]

it's hardly AMD's fault. I'm sure they don't recommend using more than the JEDEC max to the RAM, so that's what, 2.85v?

It does seem that a lot of the deaths have been people running high volts through the RAM (3.4v+) while keeping their fx-57s at stock voltage. At least we know what's causing it now, and people should stop having these troubles.

[Kameleon]

Interesting. I'll have to spread the word about this. Could this also be a cause of instability on all A64s when pushing the vDIMM up to ~4v (what I tend to do with my BH-5/booster), do you need to stick the vCore up above 1.6v or so?

Something I'll have to look out for if I decide to pull this BH-5 and booster out for fun with the Opteron...it only needs 1.35v to max out on my board, so I could've easily fried it with 4v through the RAM.

[RyderOCZ]

Has anyone cooked there E chip from overvolting it? I mean these chips can take some abuse within the limits of your cooling. And if what your saying is correct then the nf4 Expert version should contain no flaws since the memory voltage jumper issue is moot.

Tony is not talking about cooking the chip by applying to much Vcore...he is talking about the memory controller and too much Vdimm with too Low Vcore to compensate for the increase in Vdimm.

[lowdog]

Can one thing be clarified here:

If I'm running an x2 4800 stock with Vcore 1.35V and have cool n' quite enabled which drops the Vcore to 1.1V, have my OCZ EL Platinum running with 2.65V is this a problem. Meaning when the cpu is running with 1.1V due to cool n' quite and the mem with 2.65V....is that a problem.

WHat the hell IS safe.