Interesting chat with intel on the 1.65v ram issue

horrorwood · 24 Apr 2012 at 09:55

Ok so it seems they are saying 1.65v ram is fine on ivybridge. I thought it would be with the amount of "ivybridge memory" that is 1.65v

In fact they even recommend 1.5v ram for use with an i7 920, even though I am pretty sure 1.5v ram was not around when it was released. I think we'd only just gone down from 1.8v ram to 1.65v!

See below for the chat I had with them.

Jenny: Hello. Thank you for using the Intel Customer Support chat service. We are glad to be of service. How can I help you today?
: Hi, I just wondered if the i5 3570k supports 1.65v DDR3 voltage?
Jenny: The recommended memory voltage for Intel Core Processors is 1.5V.
Jenny: We do not recommend going higher than the recommended voltage as it may shorten the life span of the processor.
: Ok, so for example if I was using a Core i7 920, would 1.65v DDR3 voltage be ok?
Jenny: Intel® recommends using memory that adheres to the Jedec memory specification for DDR3 memory which is 1.5 volts, plus or minus 5%.
Jenny: Anything more than this voltage can damage the processor or significantly reduce the processor life span.
Jenny: That is 1.5V +-5% or 1.575V.
: when the i7 920 was released the lowest available voltage DDR3 was 1.65v though?
Jenny: We normally get reports from customers using the recommended memory voltage which is 1.5V.
Jenny: Anyway, memory modules rated at 1.65V normally operate on just 1.5V.
: true, but some of the higher end stuff will not run at 1.5v
Jenny: You can double-check this in BIOS to ensure that the processor will not get damaged.
: ok so if I run 1.65v ram on an i7 920 will my warranty be void?
Jenny: No, if 1.65V of memory is used, we can still provide warranty replacement for the processor.
Jenny: That is still acceptable.
: ok, is that the same for the i5 2500k and the new i5 3570k?
Jenny: However, if there will be any visible burnt mark due to high memory voltage, that's the time warranty will be void.
Jenny: Yes.
: ok thank you
Jenny: You're welcome.
Jenny: Nice chatting with you.
: thats all thanks
Jenny: Thank you for visiting the Intel Customer Support chat service. I look forward to chatting with you again. Have a nice day!

Plec · 24 Apr 2012 at 11:01

I think it's always been a recomendation rather than an iron clad must be voltage - mainly to avoid them getting lots of returns before the warranty runs out. As you rightly pointed out most memory on release was 1.65 spec at higher frequencies but 1333Mhz could be had at 1.5V - but as mentioned most 12800 will run at sub 1.58V.

The interesting line was:

Jenny: However, if there will be any visible burnt mark due to high memory voltage, that's the time warranty will be void.

Still ambiguious and open for debate as i would still try and get a return under warranty if i had only been pushing 1.65V (anything over then fair enough) - but fortunately all my memory has run sub 1.5V.

But on the whole, i think the reality is, that you can run at 1.65V if necessary but the unpredictable ruduction in lifespan becomes a concern once you're out of warranty (especially OEM purchases.). Fortunately most sticks run at SPD voltages at higher frequencies - mainly because the timings are much higher.

horrorwood · 24 Apr 2012 at 12:37

I do think they are just quoting the JEDEC specification all of the time to be honest.

I wonder what they'd say if I asked them what voltage DDR3 ram I can use with a core 2 series CPU, 1.5v ram was not out then!!

8 Pack · 24 Apr 2012 at 16:34

The 1.5v thing is utter nonsense if I am honest loads of RAM that is certified for Intel chipsets has XMP of 1.65v. Its all Intel blurb.

Surveyor · 24 Apr 2012 at 18:45

Not exactly a ringing endorsement to Sandy Bridge and Ivy Bridge users.

"You can use 1.65V but if we can tell it damaged your processor don't come crying to us".

It's almost like they expect it to.

tribz · 24 Apr 2012 at 20:11

Just out of interest, what is the issue with running +1.5v and how does it shorten the life of the CPU? Thanks.

horrorwood · 24 Apr 2012 at 20:20

Surveyor said:
Not exactly a ringing endorsement to Sandy Bridge and Ivy Bridge users.

"You can use 1.65V but if we can tell it damaged your processor don't come crying to us".

It's almost like they expect it to.

Their quote was "visible burn mark", which I have never seen myself on any recent CPU, have you?

A quick google search of i5 2500k burn mark brings up nothing!

tribz said:
Just out of interest, what is the issue with running +1.5v and how does it shorten the life of the CPU? Thanks.

I and many others do not think there is a problem with running 1.65v

deusexlumina · 24 Apr 2012 at 20:27

"that's all" ...

she's not your maid

Intel are just covering thier backs, like most companies do, 1.65V is fine and will never cause burn marks anyway. Its an extra .15V for goodness sake.

Surveyor · 24 Apr 2012 at 21:32

horrorwood said:
Their quote was "visible burn mark", which I have never seen myself on any recent CPU, have you?

A quick google search of i5 2500k burn mark brings up nothing!

I can't say that I've gone out of my way to look.

There's a thread here where it looks like the CPU socket pins have been burnt to the processor.

horrorwood · 24 Apr 2012 at 22:56

Surveyor said:
There's a thread here where it looks like the CPU socket pins have been burnt to the processor.

Pretty sure its impossible for a CPU socket pin to melt to a CPU.

The pins are made of a copper alloy with gold plating and a nickel underplate, not likely to melt afaik!

I am not saying it could not happen but I highly doubt it, if your pc is getting that hot to melt the socket then I think you'll have more to worry about... like a fire.. lol

Surveyor · 24 Apr 2012 at 23:02

horrorwood said:
Pretty sure its impossible for a CPU socket pin to melt to a CPU.

The pins are made of a copper alloy with gold plating and a nickel underplate, not likely to melt afaik!

I'm not saying it melted but they appear to have become fused together.

horrorwood · 25 Apr 2012 at 00:05

Surveyor said:
I'm not saying it melted but they appear to have become fused together.

Which would mean that one of the metals would have to melt slightly surely?

It would also be almost impossible to do as for it to get that hot and cause no damage to the pins around it would be unbelievable.

I'd guess it was more like the original owner damaged a pin so it was upright and then tried to install a CPU in the socket.

Anyway this isn't really the point, point is I've not heard/seen anybody with a burnt CPU. Even the guy on here who decided to put 1.8v core voltage through his sandybridge which killed it

Surveyor · 25 Apr 2012 at 00:22

horrorwood said:
Anyway this isn't really the point, point is I've not heard/seen anybody with a burnt CPU. Even the guy on here who decided to put 1.8v core voltage through his sandybridge which killed it

Krono5 · 25 Apr 2012 at 00:35

Surely you could just try blagging it with intel anyway and say it was the chip being faulty?

Do you think it's possible for some poor guy to get a faulty chip running at stock and it to cause damage?

horrorwood · 25 Apr 2012 at 08:45

Surveyor said:
Pics
http://imageshack.us/photo/my-images/31/98033448.jpg/

Both of those are from when foxconn made faulty sockets. Your pic is from here: http://www.anandtech.com/show/2859

Surveyor · 25 Apr 2012 at 18:36

horrorwood said:
Both of those are from when foxconn made faulty sockets. Your pic is from here: http://www.anandtech.com/show/2859

Some people are never happy.

You wanted pictures of burnt CPUs and that's what I gave you.

Justintime · 25 Apr 2012 at 19:13

To be fair it had nothing to do with RAM being the culprit though, it was a well publicised thing with the Foxconn sockets. Not seen any dead cpus due to 1.6v+ memory in any generation of cpu thus far.

Surveyor · 25 Apr 2012 at 19:27

Justintime said:
To be fair it had nothing to do with RAM being the culprit though, it was a well publicised thing with the Foxconn sockets. Not seen any dead cpus due to 1.6v+ memory in any generation of cpu thus far.

To be fair I never said it did.

horrorwood said he'd not heard/seen anybody with a burnt CPU so I gave him a couple of pictures.

OcUK say greater than 1.60V could limit lifespan of the CPU.

I say that for the vast majority there's no need to exceed that voltage so why chance it?

There's plenty of RAM available which runs at less than 1.60V and even the RAM tested at 1.65V will usually run with less.

biuro74 · 25 Apr 2012 at 20:45

Blah, blah... technical chatting with Customer Service is like having a date with a married woman: in both cases they are not proper person

In "Core" series CPU (i3/i5/i7) most safe is difference between CPU Vcore and CPU VCCIO (1366/1156) or System Agent (1155) less or equal 0.5 V - to prevent long term damage of memory controller which is part of CPU. By default VCCIO is 1.1V which gives 1.65 V voltage for memory (it includes +/- 5% margin), and SA in Sandy Bridge is ~0.93 V which gives 1.5 V for maximum memory voltage.

However, CPU controller in 1366/1156/1155 is very good reworked which means its default voltage, heat emission and overclocking potential is just big. Best example is default memory support: 1366: 1066, 1156/SB: 1333 (plus o/c upto 2133) and Ivy Bridge: 1600 (plus o/c upto 2.8 GHz).. looks like guys in Intel tech noticed very efficient point that might be easily "drained up" by 3rd party vendors (memory suppliers).
So as soon as difference between memory controller and memory itself is less than 0.5V, CPU is safe regarding to common rules. It is why SB might use memory with 1.65V, probably all mobo BIOSes raise a bit SA voltage automatically once higher memory voltage is detected, or they decrease memory freq regarding to SPD/JEDEC values to avoid damage - regarding to my chat to some MSI guys and my experience - I used 1.8 V memory with SB which downgraded it to 1333 MHz/1.50V safe values. 1366/1156 user must set their VCCIO manually, if they want not to smoke their CPUs, and if SA is not increased by BIOS automatically in SB, their users should do it manually, too.

Justintime · 25 Apr 2012 at 21:02

biuro74 said:
Blah, blah... technical chatting with Customer Service is like having a date with a married woman: in both cases they are not proper person

In "Core" series CPU (i3/i5/i7) most safe is difference between CPU Vcore and CPU VCCIO (1366/1156) or System Agent (1155) less or equal 0.5 V - to prevent long term damage of memory controller which is part of CPU. By default VCCIO is 1.1V which gives 1.65 V voltage for memory (it includes +/- 5% margin), and SA in Sandy Bridge is ~0.93 V which gives 1.5 V for maximum memory voltage.

However, CPU controller in 1366/1156/1155 is very good reworked which means its default voltage, heat emission and overclocking potential is just big. Best example is default memory support: 1366: 1066, 1156/SB: 1333 (plus o/c upto 2133) and Ivy Bridge: 1600 (plus o/c upto 2.8 GHz).. looks like guys in Intel tech noticed very efficient point that might be easily "drained up" by 3rd party vendors (memory suppliers).
So as soon as difference between memory controller and memory itself is less than 0.5V, CPU is safe regarding to common rules. It is why SB might use memory with 1.65V, probably all mobo BIOSes raise a bit SA voltage automatically once higher memory voltage is detected, or they decrease memory freq regarding to SPD/JEDEC values to avoid damage - regarding to my chat to some MSI guys and my experience - I used 1.8 V memory with SB which downgraded it to 1333 MHz/1.50V safe values. 1366/1156 user must set their VCCIO manually, if they want not to smoke their CPUs, and if SA is not increased by BIOS automatically in SB, their users should do it manually, too.

Never heard any relation to system agent voltage and memory voltage and from the beginning it was generally regarded as not having any effect on clocks/stability and generally to be left alone at stock, the VTT/VCCIO on SB is about 1.05v default which a .5 differential would be 1.55v and intel states 1.575v being the max, if theres a differential of .5 between the mem controller voltage and the RAM voltage then using say 1.15v VTT should increase that limit to about 1.65v etc..