Upper voltage limits on Winchester?

[Insomniak]

Been working with my new OC, and noticed the following:

If I go for 2.4GHz at 283x8.5, RAM is downclocked slightly, but the system is 12 hour prime stable at 1.5125 Vcore. Changing things around to 300x8, RAM is at full frequency, but the system needs 1.575 Vcore for 12 hour stability.

This normal? Would a higher HTT want more voltage to remain stable, even though overall chip speed is the same? Also, what's the upper limit on Vcore that Winchesters can safely take? I've been hearing mostly 1.6V - that the consensus?

Just wondering, as I know the memory and chip are not requesting extra volts here for stability.

 

[Zebo]

It's not the chip it's the mem controller needing more volts to stay stable.

 

[Insomniak]

Well yeah, but the mem controller is on die. I figured that though, because Prime would crap out in 1/2 minutes. Either way, since they're packaged together, that means Vcore goes up.

 

[Budman]

there isnt really a set limit but once you start upping the voltages on a A64 the temps really start to go up.

As long as your temps are good you can keep upping that voltage.

 

[Leper Messiah]

I run mine at ~1.6vcore-ish. (Epox flucuates from about 1.56 to 1.62 depending on load) and it still loads some where in the mid 40Cs, with the stock heatsink (which I'm very impressed with BTW). I would say that where I run it is the limit though, don't want to fry the controller.

 

[Insomniak]

That's my concern - frying something. I'd think Winchester would suffer an electrical disaster far before it hit its heat limit. My temps haven't gone up at all under load from the OCing - it was 44C when I ran it at stock with 1.4v, it's 44C under load now @ 2.4Ghz 1.575v. Idle tempe are up from 28C at 1.4v to 32C, but still - that's so low it hardly matters.

 

[Zebo]

Originally posted by: Insomniak
Well yeah, but the mem controller is on die. I figured that though, because Prime would crap out in 1/2 minutes. Either way, since they're packaged together, that means Vcore goes up.

No what I mean is sometimes the chip can handle the speed but the controller can't therefore more Vcore is required. This is obviously the case in your situation since at same core speed it passes with a lower HTT but fails at higher HTT. Have a look here.
http://www.xtremesystems.org/forums/showthread.php?t=53443&highlight=controller

 

[milkrocks]

Mine runs fine at 1.62V

 

[Insomniak]

Originally posted by: Zebo

Originally posted by: Insomniak
Well yeah, but the mem controller is on die. I figured that though, because Prime would crap out in 1/2 minutes. Either way, since they're packaged together, that means Vcore goes up.

No what I mean is sometimes the chip can handle the speed but the controller can't therefore more Vcore is required. This is obviously the case in your situation since at same core speed it passes with a lower HTT but fails at higher HTT. Have a look here.
http://www.xtremesystems.org/forums/showthread.php?t=53443&highlight=controller

Yeah I know - that was the point I was making - it's the mem controller, but Vcore still has to rise because they pull their power from the same place. I just don't want to fry the core in an attempt to feed the mem controller, but it looks like I'm still within spec, so all is well.