3400+ dead and too much vdimm is the cause

[Calcu]

After modding yesterday, the mods worked perfectly. To start I had 1.85v when set to 1.7v and 3.29v when set to 2.9v which was perfect for what I wanted. Only problem was that even with 3.2v, I still couldn't get above 230FSB stable with 1 or 2x 512MB Mushkin Lvl II PC3500 memory in any slot. I even tried my 512MB CH5 HyperX PC3500 and the samething, all this at 2x LDT. So I decided to raise the vdimm a little more to 3.5 to test, same problem, so I turned the pot back down to 3.3(2.9 in BIOS). So now I was starting to think that these A64s don't like double sided memory and I don't have any single sided memory here to test with. Anyways, I turned off the PC to insert both memory modules back and tried to power up and beep, beep. So reinstalled the memory modules and even tried the HyperX and samething. So I started to think that the problem could be and then immediately thought, DEAD CPU. Lucky I had another CPU here, popped it in and it fired right up. So from my testing, my CPU died from too much vdimm voltage and not from vdd, LDT voltage(mine were set to 1.7v vdd & 1.3v LDT) or CPU voltage(1.65v while testing) and my setup is watercooled, so temps weren't a problem. Well back to testing(below 3.5v now)

Calcu

[CodeRed]

damn shame about that CPU.

good to know what the definite cause of all these deaths is though.

now I am worried, my kingston needs 3.6V to really come good (267 2-2-2-5-10-16) ... hmm, should I risk it again.

[texuspete00]

What kind of cooling you got? Might still be able to get great performance and max your cpu even if you cant really ultilize lower available multipliers.

Hmmm.... well everyone will eventually be forced to DDR II. This will probably dissapear. One good thing at least. I heard amd moves to 1000MHz (250*4) Hypertransport and boards will be utilizing either type of DDR 500 Dual Channel Ram but dependant on board you get.

[Evil_Spork]

thats a bummer to hear! alot of us want to run BH5 with high vdimm on these.. ohwell.. 939 + DDRII here i come

[xgman]

You know it is quite possible that what killed it was the turning of the pot too abruptly or whatever (assuming you turned it with the computer powered on). Maybe even a slight voltage spike may have sent shivers to the cpu mem controller. The odd thing is that it died after you turned it back on.

[Calcu]

Actually I adjust the pots with the PC off, I've had that habit for a while now, so I know that's not the problem. As for the PSU, I have a brand new 550W Enermax and voltages are rock solid, 100% load or not, so I doubt that's the problem. But from now on, I will try your method Mickey, leave it on Auto and adjust the pot but let's see what happens now,

Calcu

[luihed]

Ive lost 2 a64 now and both are from high vdimm I believe...... one at 3.54v cpu aircooled and one at 4.06v with my cpu waterchilled...... im sure my psu has rock solid voltage considering its a Turbocool 510 dlx....

[saaya]

memory at 4.06v? what memory was that?

turbocool... is that pc power&cooling? i heard their cpus have very very high volttage fluctuations compared to other psus!

anandtech had a review about it i think...

[Ragnarok]

well running 4v vdimm was kinda asking for it

the extra power delivered is most likely dissipated in built in mem controller (due to internal resistance)..

[IamAnoobieCheez]

By looking at calcu's first post, I see what *could* easily be the problem.

The good'ol 32bit AMD platforms such as nforce2 and every other AMD & Intel platforms can have their memory controller/ AGP controller get "whooped".. meaning "Automatic Deaths" when messing around with the memory sticks by changing them during the system overclocked state. Doing that very easily to "destroy" the motherboards. Now, with the AMD 64 systems, they blow up the CPUs (integrated memory controller in the chip) when they mess around with the memory sticks during the overclocked settings. Makes sense to me.

Of course, i'm not saying all the AMD64 deaths is caused by this. There are of course many factors that are involved as you all know. but this is something to think about as well.

[IamAnoobieCheez]

and it don't matter they are supercooled or nitro cooled.. as long as you change your memory sticks during the overclocked settings, it will destroy your memory controller... not the memory chips itself, also the possibility of damaging the circuit where it handles between the memory slots on your board to your chip... once again, related with "memory controller". That's my opinion.

[Calcu]

Actually the system was not overclocked, I only overclock from Windows, through Clockgen. IamAnoobieCheez, I take it you haven't played with the A64, ah? Overclocking is not like it was it the nForce2, every time I power up, the system resets the multiplier and FSB to default(200 x 11) but with the Y BIOS, at least evertime I restart the system, the multiplier stays at what I set it at with Clockgen. So I doubt that was the problem,

Calcu

[IamAnoobieCheez]

Originally posted by Calcu
Actually the system was not overclocked, I only overclock from Windows, through Clockgen. IamAnoobieCheez, I take it you haven't played with the A64, ah? Overclocking is not like it was it the nForce2, every time I power up, the system resets the multiplier and FSB to default(200 x 11) but with the Y BIOS, at least evertime I restart the system, the multiplier stays at what I set it at with Clockgen. So I doubt that was the problem,

Calcu

Absolutely... I didn't even lay my finger or even smelled any one of the A64 systems.

so you always reset your CMOS before you change around your memory sticks during the overclocked states?.. i doubt it..

Adding one stick from the exsisting memory stick on your board is even more dangerous.. Clear the CMOS "every time" before you change your memory sticks.

who doesn't... overclocking is not the same with the nforce2 and A64 systems.

no... no... no..... 200FSB isn't overclocked setting for the CPU but not the memory department... Even though memory can easily do 200FSB (250FSB for that matter) *can* still completely destroy the "memory controller" when you mess around with the memory sticks by changing them.. which means... Taking out the memory stick(s) or/ switching around, even more importantly, Adding "a" memory stick when you already have one stick of memory in the slot..

[IamAnoobieCheez]

when I say "overclocked settings/states", I'm refering to:

- Overclocked FSB for the memroy department (memory slots on the board, circuit on the PCB, and all the way to the integrated memory controller inside your chip.

- Yes.. Vmem voltages too. Any high volts.


It can "still" destroy the memory controller even without the high volts of Vmem...

[luihed]

Originally posted by saaya
memory at 4.06v? what memory was that?

turbocool... is that pc power&cooling? i heard their cpus have very very high volttage fluctuations compared to other psus!

anandtech had a review about it i think...

Yeah its a PC power and cooling..... I never read teh Anand review but from my own testing, the rails are rock stable..... Ive tested it running idle and then running prime95 and the voltages doesnt even twitch ......

Before, I thought that cooling makes a difference with how high you can crank the vdimm ....I killed my first a64 on stock cooling and I believe MrIcee did the same on his....... but I guess I was wrong.....

[Zeus]

Calcu, from where did you read Vdimm, from the mosfet with a multimeter, or in bios?

[Calcu]

With a MM at the mosfet and used the BIOS rarely but it's pretty accurate, so I will probably use it now,

Calcu

[Astennu]

Originally posted by MickeyMouse
thats whats killing the chips its on a reboot, lol

ya cant set 3.3v at say 2.6v reboot change it to 2.9v and expect to see 3.5-3.6v..

do this right now

boot to bios with no mod hooked up set 2.9v reboot enter bios make sure its on 2.9v shut down hook up mod enter bios set it via 2k 15 turn vr or 2x 1k's in series to 3.5-3.6v at 2.9v via VR shut down remove the vr and measure the resistince

no add in a fixed resistor with that value or higher and use a 1k vr

I say 760ohm fixed and a 1k vr in series

this means no more voltage splikes thta can kill I would go wiht 3.4v as a cap though 3.4v can get ya 265mhz 1:1 at 2-2-2-6 on any nf3/via a64 mobo..

this boards vdimm has no tolerance for psu/user or voltage error

Maybe it’s a idea to make a database for the AN50R like 760K fixed would give ya …. V-Dimm so less people will destroy there Cpu. I don’t know Mikey if you know any values ?

I hope to have my new Shuttle friday but i don't like to blow the cpu up with a mem mod

[`schr0et]

Hah well, I tried the VDimm, mod was a success perfect soldering job by me.

I had 2x 1K VR's combined were set to 1.984K ohms, turned on comp read 3,17 vdimm

few seconds later I saw a spike to 3,78

I was like the F UCK?

I have a PC Power & Cooling 425 W, and I already checked my lines with a multimeter thru the mosfets, rock solid lines.

Seems these A64's are fragile as heck!

Oh well I still haven't exceeded the warranty time for the CPU can still RMA it

Hopefully this time I'll get a more persistent CPU

[Zeus]

Did you turn one of the VR's with power on?
I thought that was causing spikes and killing the CPU?

[CodeRed]

its odd how you dont hear about too many FX-51's dying the same way ...

is it just the A64 chips?

[Zeus]

Originally posted by MickeyMouse
NOPE its people wiht the shuttle nf3 not vdimm modding the board right or rushing not understanding the way the boards vregs work and so on and so on..

What's the secret then, Jason?
I'm planning to do the Vdimm mod as soon as i get my AN50R.

Is it the senseability of the VR that causes spikes?

[`schr0et]

Sry my statement wasn't clear as to what caused my lines to fluctuate.

It was something stupid I did.

1st: My CPU + Mobo worked fine when my VDDR was set to "auto", the line was very stable @ 3,17

Only when I set my VDDR to 2.6 VDimm did my lines go crazy upwards 3,78 VDimm and then it wouldn't reboot

So pretty much so long as your always using Auto and only change your volts thru VR's you should be just fine.

[xgman]

PR that may be worth a try, but it seems the best thing to do is find a sweet spot, "set it, and forget it" on these 64's., although there seem to be way more reports of this on NF3 boards with few if any on Via boards.

[Astennu]

so if i would use a 2k vr and leave the VDDR @ Auto nothing bad will happen ? I think i'm gonna try that if i have my shuttle in tomorrow