Vcore and vdimm with A64

[Kameleon]

Do the new opterons suffer from this problem?

I'm running Ocz VX atm at 3.4v

Well they're E cores, so you work it out

[uOpt]

Do the new opterons suffer from this problem?

I'm running Ocz VX atm at 3.4v

They are actually most affected because some of them overclock so well on low volts.

[uOpt]

Does damaging the diodes mean that the CPU dies instalntly or does it still run, but very unstable?

You try and tell us

I wonder whether those who report degrading memory controllers had encountered this problem, but it is hard to proove.

[Fr3ak]

You try and tell us

I wonder whether those who report degrading memory controllers had encountered this problem, but it is hard to proove.

I was asking because I run high vdimm with stock score for a while and somehow my opteron fails prime after 1 and a half hours with the same settings it run 32hours 20mins before....
Havent figured out whats causing this, thats why I am askin
My psu might went nuts, gotta test it later tonight

[Vassili]

First of all, the CPU doesn't have to do directly with the VDD, but with Vtt/Vref, wich should vary from 1.25 to 2V when you adjust the RAM from 2.5 to 4V. Maximum JEDEC Vdd voltage is 2.85V, and lowest voltage for A64 CPU's (CnQ activated) is 1.1V, so we know for sure that a difference of 0.325V between Vcore and Vtt is safe. This means that for a 1.35V Vcore voltage, we can have a safe Vdd of at least 3.35V, and for 1.5V Vcore -> 3.65V Vdd.
Probably the real danger is when those diodes go near the opening (conduction) voltage, wich should be ~0.6V for Si-based junction; if this is true, 1.35V Vcore and above 3.9V Vdd should kill the CPU instantly ... anyone wanna try ?

Someone went very close:


XtremeSystems Forums > Xtreme > Xtreme Bandwidth > @ My Kingston @ 280 2-2-2-5 8M stable @ 3,68v

Strange, Oh and don't get me wrong here, I do believe you Tony

[uOpt]

Someone went very close:


XtremeSystems Forums > Xtreme > Xtreme Bandwidth > @ My Kingston @ 280 2-2-2-5 8M stable @ 3,68v

Strange, Oh and don't get me wrong here, I do believe you Tony

That person is obviously running Vcore 1.4. The 1.36 reported is what DFI boards report for what is actually 1.4.

[Micutzu]

That person is obviously running Vcore 1.4. The 1.36 reported is what DFI boards report for what is actually 1.4.

1.4V in BIOS is more like 1.45V in real life ... just put a multimeter on the mobo. On the other hand, 3.68V is almost 3.8V, DFI's voltage monitoring shows less than what you are actually feeding to those poor parts

[uOpt]

1.4V in BIOS is more like 1.45V in real life ... just put a multimeter on the mobo. On the other hand, 3.68V is almost 3.8V, DFI's voltage monitoring shows less than what you are actually feeding to those poor parts

First part - not for me. On my SLI-DR I actually get what I tell it in the BIOS (measured with a quality multimeter) with much less than a percent error. I am actually impressed how well my board volts.

Second part - yes, the sensors display less than what you actually have.

[Call me Ugly]

Hello all!

I registered here just because of this thread. I have an Opteron 148 "CABNE 0528" that I have been running at stock volts untill this thread (thanks Tony!). Now I'm running 1.45v cpu for 3.3-3.4 ram (2x512 Redlines).

Is there any other component this issue may apply to? HTT and LDT are a little new to me, so I'm having trouble trying to logic out if having every other compnent overvolted may hurt the LDT bus (stock at 1.2v). Sorry if thats a newbish question, but this whole issue is surpriseing and unexpected to me, I thought I was doing my chip a favor by leaving it at stock volts. For referrance:

CPU= Opteron 148 @ 2816MHz 1.45v (now)
RAM= 2x512 Mushkin Redline XP4000 @ 256MHz 1.5-2-2-5 1T 3.4V
CHIPSET= 1.6v
LDT BUS= stock 1.2v Should it be?

[Micutzu]

First part - not for me. On my SLI-DR I actually get what I tell it in the BIOS (measured with a quality multimeter) with much less than a percent error. I am actually impressed how well my board volts.

Second part - yes, the sensors display less than what you actually have.

Right, at default i have 1.36V in SG, 1.4V in BIOS, 1.422 on the multimeter. But at 1.55V set in BIOS, it's more like 1.52/1.55/1.63. Another interesting thing is that the voltage raises slightly in load (~0.02V).

[hardnrg]

that's what i get

[Fr3ak]

Stupid question, but where can you measure vcore using a multimeter? I never did something like this before

[brschmid]

OK, i have a good (bad for me) situation that applies to this thread.

I Have killed 3 cpus now, all by overclocking and not using more than stock vcore. I didn't want to raise vcore because i was afraid of killing the cpu, well they died anyway.

I have killed a 3500+ Winnie, 3200+ Venice, and now at 3800X2. each lasted no more than 2months, then died on me.

So if this is the case that high vdimm or at least a big difference between vdimm and vcore is killing cpus. i have found my problem

[ozzimark]

what vdimm were you running at? i remember you posting at [H] and saying that you had TCCD for one of those cpu's

[brschmid]

your right i was mostly using vdimm under 3.0v but i was still using as low of vcore as possible

[alpha0ne]

I was using 1.42Vcore with 3.5Vdimm benching @ 3GHz then remembered this thread and upped Vcore to 1.44v, is this safe ??, this as per my sig using my new 146 that is simply an amazing chip

If it died I would cry for a week

[Hound]

im not an engineer of any sort so take this with a grain of salt, but here goes my definition to the problem (i could be totally wrong btw)

that the Vtt is half the Vdimm, so if you set it really high, (like 4.0v) then your Vtt is 2.0v and even if you run your cpu @ 1.6v (maybe with high drive strength) then the current/pressure would force its way through the diodes, either instantly, or over a long period killing them, because if theres a diode there, the current probably isnt meant to flow that way...

this is assuming that Vtt is the voltage that goes into the memory controller correct? if so, DDRII would be interesting, as (a not extrodinarily extreme) 2.2v/2 = 1.1v and that to me means, that the mem controller is getting less volts...could be good for the next generation of amd cpu's to run high Vdimm, or we could get exactly the oppisite problem.. of course im not underestimating amd's engineers, and their knowledge, me being 15 (with very limited electronics knowledge )probably doesnt help either

[mcnbns]

I have some questions based on this scenario: I've been running my 3200+ @ 1.4V for about 4 months and 3.2-3.4V on the RAM nearly 24/7.

1. Should it be dead by now?
2. If so, why isn't it?

3.Does making the difference between VDD and VCORE help STABILITY as well?

[Hound]

well, 3.4v/2 = 1.7v and thats a difference of less than 0.6v...

however if amd has recommended that the max volage for mem should be 3.0v...and RevE procs run on 1.35/1.4v, the Vdd = 1.5 and then the diff is..0.1 volt?!?!

[widman]

I'm really doubt this theory, cause in reality I just out of limit.
I'm running CPU AMD64 4000+ @ 0.875V and VDimm at 2.6V (default). The different are 0.425V. It has been run for 6 months without any problem.

yeah, it's 130nm die, but where is the exception? At least it should be die slower than 90nm. Or at least it would show the instability. But still it's rock stable.

[uOpt]

I have some questions based on this scenario: I've been running my 3200+ @ 1.4V for about 4 months and 3.2-3.4V on the RAM nearly 24/7.

1. Should it be dead by now?
2. If so, why isn't it?

Why do you say that?

Even the most "careful" table says 1.4 V set in BIOS is good for 3.45 Vdimm or so.

Also, is your 1.4 set in the BIOS or what the sensors report? The sensors are nonsense, trust your settings.

[uOpt]

I'm really doubt this theory, cause in reality I just out of limit.
I'm running CPU AMD64 4000+ @ 0.875V and VDimm at 2.6V (default). The different are 0.425V. It has been run for 6 months without any problem.

Why and how do you run 0.875V on that CPU?

Even cool'n'quiet doesn't go below 1.1 V.

Got a screenshot?

[widman]

Why and how do you run 0.875V on that CPU?

Even cool'n'quiet doesn't go below 1.1 V.

Got a screenshot?

I run Clawhammer on 0.875V. It runs much cooler than 1.1V on qnc. And I prefer run manual qnc.

For newer chip, it refuse to set voltage lower than 1.1V. I tested with venice core and manchester core.

here screenshot


here my manual qnc

[uOpt]

I run Clawhammer on 0.875V. It runs much cooler than 1.1V on qnc. And I prefer run manual qnc.

OK, I see.

But that is still inside what we computed would be the "probably absolute limits" table.

0.85 V Vcore should be good for 2.9 V Vdimm in that table.

[Hound]

however, that is definitly not a RevE proc, and that is what is affected, i think.. lol