Overclocking Intel to Enthusiasts: Don't Overclock RAM with lots of vDimm

Taken from the article:

Overclockers looking to bolster their new Nehalem CPUs with over-volted memory may soon be disappointed if they’re concerned about the health of their CPU. Intel is telling motherboard manufacturers not to encourage overclockers to push the voltage of their DIMMs beyond 1.65V, as anything higher could apparently damage the CPU.
...
Of course, this is still beyond the standard JEDEC DDR3 voltage spec of 1.5V, but it’s also a long way off the voltage required by overclockers’ enthusiast memory, such as Corsair’s Dominator TW3X2G2133C9DF, which needs 2V to run at its specified 2,133MHz maximum speed with timings of 9-9-9-24.


Reference URL: http://www.custompc.co.uk/news/604980/over-volted-ram-will-damage-core-i7-cpus.html

Well... hmm this should be interesting.

 

How is ram voltage linked to damaging the cpu?

 

I don't see connection between these two.

 

I didn't realise you two were masters of circuitry and Nehalem architecture.

There was a similar issue with A64 chips not that long ago, and it's all down to the memory controller being on the CPU. Put too much juice through it, and it'll inevitably suffer tbh.

 

and is it solved now?

if it's a known problem, why doesn't Intel use the same solution (if it's solved)

 

Why's the controller got to have the same voltage as the ram sticks? With DDR2 It was in the northbridge wasn't it, how much juice is going through it at the moment then 1.5 volts with sticks taking 2 volts? Or is it alot more different than simply moving it to reduce latencies?

 

Nehalem is the first Intel architecture to have an Integrated Memory Controller on the CPU itself, hence as Krikkit stated, if you put too much juice through it you are inadvertently damaging the CPU.

 

1 step forward 2 steps back by Intel?

 

Thanks, I would never have deduced that from reading your link in the first post or any other post in this thread.

 

I think it was on socket 939 chips and mostly on DFI boards that would let you put crazy volts into the RAM.

 

Don't you think that Intel would've solved it if it was reasonable to do so? With some of the best bods in the industry working on these kinds of problems for literally hundreds of thousands of man-hours, don't you think it'd be solved if it were a simple problem?

 

Krikkit clearly not.. what they wanted to do was to let trolls on tech forums slag them off for being silly.

 

Of course! How silly of me!

 

And I almost got scared for the moment...

 

Maybe for the 0.1% of people out there that feel the need to cook their RAM. The rest of the world will enjoy the performance boost of having an on-die memory controller (whether they know it or not) and will like it.

 

Didn't claim to be


Mem controller being onboard the A64 didnt stop people putting stupid amounts of voltage through BH-5 on the A64 platform either

 

here is how I see it, intel wants to give consumers the best performance possible. and this is what they have done by having an Integrated Memory Controller on the CPU. if it gave better performance to do it the other way, i'm sure they would have.

 

Will the i7 processors even notice a huge improvement from the faster memory speeds? Just having the integrated memory controllers should already give them a decrease in latency over the Penryn architecture. I suppose we will have to wait to see if those higher voltage speed settings will even have an effect on processor performance as I am sure people will overclock and test with those memory speeds. It may be that the i7 series doesn't gain much performance by higher memory speeds.

 

http://anandtech.com/memory/showdoc.aspx?i=3426

 

because IMC on Nehalem

see the diagram of the architecture which shows the RAM plugged into the CPU (not the chipset); this means the vdimm is now pegged to vcore and cannot and should not be changed independently

http://www.augustinefou.com/2008/10/core-i7-vdimm-problem-aka-is-high-vdimm.html