Sandy Bridge Processor Absolute Minimum and Maximum Ratings

[dejanh]

Quick question...

I've noticed that the datasheet (http://download.intel.com/design/processor/datashts/324641.pdf) for the new SB chips does not have absolute minimum and maximum ratings, only the VID rating as shown below...

http://img.techpowerup.org/110107/i7-2600K_voltage_current_specs.png

The datasheet also states that each processor is calibrated with an individual maximum VID which may be different processor to processor though based on how I understand it not higher than the MAX VID of 1.52V.

So, anyone know what are actually the absolute minimum and maximum ratings?

[dejanh]

Really, 176 views and no reply? :confused:

[A_Noob]

On another forum it is stated by someone who has worked with i7 2600K ES samples for a while that you can go up to 1.6v and still be safe. But AFAIK, I'm unsure if it will degrade over time if you run it that high of a voltage. Sandy Bridge is not friendly to extreme cooling.

[zalbard]

I'll repost this here, I guess.


I could only find this:

http://i301.photobucket.com/albums/nn79/zalbard/sandy_bridge_voltages.png
@zalbard:

That 1.52V is not an actual safe voltage, just the maximum voltage that can be represented in the new SVID encoding, with 5mv steps instead of the 6.25mv in older CPUs. You can confim that by checking the SVID table in section 7.4, that coincides exactly with the 0.25V - 1.52V range in the table you posted.
I'd recommend holding back on vCore, maybe 1.4-1.45V max 24/7. But these are the only numbers we have...

[OC Nub]

@Zalbard: memory needs to be no more than 1.575v, or am I reading this wrong? I have a lot of memory, would hate to have to buy more.

[Atragon]

@Zalbard: memory needs to be no more than 1.575v, or am I reading this wrong? I have a lot of memory, would hate to have to buy more.

Eek, I apparantly need to play with my RAM OC a bit, the XMS3 I have is rated at 1.65, which is what I've been running it at. (I should note that the UD7 marks voltages 1.66 and above in a different colour in BIOS, so I was assuming that 1.65 was safe...)

[zalbard]

I think 1.65V should be OK, really, it's the same story as with Bloomfield. A lot of SB compatible RAM coming out right now is rated at 1.65V (at least in XMP).

[OC Nub]

Thanks, I hope so. Cant find any high clocking 1.5v ram anywhere.

On another note, the 1.52v absolute max vcore is sexy. I guess there will be a lot of 4.9-5.0ghz 24/7 rigs in the near future.

[zalbard]

That's not actually a max safe voltage. As you can read above, this is just a maximal possible VID. At least from my understanding...

[Falkentyne]

Thanks, I hope so. Cant find any high clocking 1.5v ram anywhere.

On another note, the 1.52v absolute max vcore is sexy. I guess there will be a lot of 4.9-5.0ghz 24/7 rigs in the near future.

No it's NOT that sexy. The core i7 (45nm) had an "absolute max" of 1.55v. But you don't see ANYONE trying to put 1.55 on air for 24/7 on an i7....they only go that high on air for quick benches .....

I said this before and I'll say it again.
On the Core 2's, Intel had a "Vid range" which I believe, was 0.85v-1.3625v, which I believe most people considered acceptably safe, and an Absolute minimum and maximum rating, which ranged from something like NEGATIVE vcore (Now someone explain to me how you can get negative vcore, please?), up to 1.45v. Notice the vid range and the absolute max were not the same. This was already "changed" from the Pentium D's when Intel listed the safe operational range, AND absolutes.

Then on the i7, the absolute max got bumped up (no one explained why, though). Then got bumped down to 1.4v for the 980X 32nm, and now got bumped back up to 1.52v for another 32nm chip? I don't get it...And now they removed "Absolute maximum..."

Still would love to know what negative vcore is....

[OC Nub]

I had figured since 32nm GT and Clarkdale had a max vcore of 1.4v, SB would also. But it looks like SB will be higher, which will allow for higher overclocking.

Noone pushed vcore on Bloomfield i7's to 1.55v on air 24/7 because they couldnt (due to high temps), not because the desire wasnt there.

I was going to try a 980x, or 990x when they come out but decided not to. If I got a poor clocking chip, because of the price (1000.00) I would be less willing to push the vcore. However a 300.00 chip will be much more fun for me, because I wont be so worried about killing the chip. I could buy 3 2600k's for the price of 1 GT 980x and still have a few dollars left over.