Sign in with
Sign up | Sign in
Your question

Is 1.65V memory bad on Z77?

Last response: in Memory
Share
September 1, 2012 11:57:04 PM

Someone told me that...

Voltage greater than 1.55-1.6 volts are dangerous for Intel CPU's and runs the risk of frying the controller.


I have had a X55 system for over 2years with patriot ram running at 1.64V, never had a problem with the ram, mobo or CPU.


Has things changed with the Z77 board and Ivy Bridge chip (overclocked)? Where 1.5V is desired.

More about : 65v memory bad z77

a b } Memory
a b à CPUs
September 2, 2012 12:03:04 AM

1.5 is preference. I think you can technically run 1.65 but itll deteriorate your memory controller or something that I've heard. i have no first hand experience on it though
a c 129 } Memory
a c 169 à CPUs
September 2, 2012 12:05:34 AM

The Sandybridge and Ivybridge processors are standardized around the 1.5 volt JEDEC standard. 1.65 volt memory is outside of Intel's specifications for the Sandybridge and Ivybridge processors but still well under the absolute maximum which is 1.8 volts.

There were reports on early 6 series motherboards that high voltage memory was frying the CPU but I believe it turned out to be faulty electrical design on some sockets from Foxconn.

There has been at least one report from an Intel engineer saying that 1.65 volt memory will work without issue, and at least one report from an Intel service representative that using 1.65 volt memory will not void the warranty.
Related resources
a b } Memory
a b à CPUs
September 2, 2012 12:07:18 AM

do you mean x58?

1.5v memory requirement hasn’t changed since Nehalem.

Some people might get away with it but it’s not recommended. 2yrs is still early days considering how often the computer is on and stressed.

The Northbridge on the motherboard and the memory controller on the cpu will be affected if something shorts out due to overvaulting.
a c 129 } Memory
a c 169 à CPUs
September 2, 2012 1:17:55 AM

boju said:
do you mean x58?

1.5v memory requirement hasn’t changed since Nehalem.

Some people might get away with it but it’s not recommended. 2yrs is still early days considering how often the computer is on and stressed.

The Northbridge on the motherboard and the memory controller on the cpu will be affected if something shorts out due to overvaulting.


To be completely fair, if your PC is vaulting over anything you have much bigger issues at hand
September 2, 2012 2:34:55 AM

Pinhedd said:
To be completely fair, if your PC is vaulting over anything you have much bigger issues at hand


Why do you say that?
Like it said, been running v1.64 for years on i7 930 (3.8Ghz, air). I just make sure the vtt is max 0.5 lower then the ram volt, so i have to raise the default vtt a bit, but that has to be done anyway because my ram where to fast for the Uncore to be x2 the rams speed without raising vtt.
Just don´t understand why 1.65v it is a big deal with Sandy/Ivy when I have never had a problem with it before. Plus why are those rams even sold by companies that dont even focus on AMD products.


And yes... I made a typo, X58... a gigabyte UD5 to be exact...
a c 129 } Memory
a c 169 à CPUs
September 2, 2012 2:42:48 AM

irlwizard said:
Why do you say that?
Like it said, been running v1.64 for years on i7 930 (3.8Ghz, air). I just make sure the vtt is max 0.5 lower then the ram volt, so i have to raise the default vtt a bit, but that has to be done anyway because my ram where to fast for the Uncore to be x2 the rams speed without raising vtt.
Just don´t understand why 1.65v it is a big deal with Sandy/Ivy when I have never had a problem with it before. Plus why are those rams even sold by companies that dont even focus on AMD products.


And yes... I made a typo, X58... a gigabyte UD5 to be exact...


I was making a joke about his typo. He said overvaulting when he presumably meant overvolting.

Putting CPUVTT at 1.1 to 1.2 volts and the CPUVDDQ (DDR3 VCC) at 1.65 should be fine. I haven't found a single reputable source pointing to 1.65 volt memory as being the culprit of a failure and multiple sources (including Intel's datasheet) saying otherwise.
September 2, 2012 2:44:19 AM

Look at this...
http://www.corsair.com/en/memory-by-product-family/domi...
"•Guaranteed to work on all triple channel Intel platforms" default 1.65v...

Is it just double cannel that can´t handle more then 1.5v?

Edit: To be honest, I opened this thread because I am considering recommending some dual channel Kingston Hyper X (1.65V) to a friend who needs a ram upgrade for $40. The 1.5V versions cost $60 with the exact same 9-9-9 latencies. Seems like a waste of money to me.
a c 129 } Memory
a c 169 à CPUs
September 2, 2012 2:53:25 AM

irlwizard said:
Look at this...
http://www.corsair.com/en/memory-by-product-family/domi...
"•Guaranteed to work on all triple channel Intel platforms" default 1.65v...

Is it just double cannel that can´t handle more then 1.5v?


Intel only has two platforms which implement triple channel memory. The first is the first set of Nehalem processors, the i7-900 series. The second is the 2-way capable Xeon E5-2400 (socket 1356, not to be confused with socket 1366) series processors which are half way between Sandybridge and Sandybridge-E.

Since corsair doesn't sell memory for servers their triple channel sets are targeted only at the Intel i7-900 series processors. Each memory channel is addressed and controlled individually so the pinout for a quad channel architecture is identical to that of a dual channel architecture. You can use dual channel memory on a triple channel platform and triple channel on a quad channel platform. The difference is that triple and quad channel sets come in sets of 3 and 4, which is more convenient. High speed enthusiast memory may have actually been hand picked for quality and tested for stability before hand as well.

Funnily enough, the i7-900 series processors are also rated at 1.5 volts and not 1.65 volts, which is why I tend to believe that the "1.65 volts is bad for sandybridge" was simply pulled out of someones ass and repeated to the point that everyone believed it.
a b } Memory
a c 283 à CPUs
September 2, 2012 3:19:11 AM

Pinhedd said:
Funnily enough, the i7-900 series processors are also rated at 1.5 volts and not 1.65 volts, which is why I tend to believe that the "1.65 volts is bad for sandybridge" was simply pulled out of someones ass and repeated to the point that everyone believed it.


If you listen to and trust jaquith (which I do), it's not necessarily 1.65V RAM in and of itself that's the problem. It's the VCCIO/VCCSA voltage required to run 1.65V kits (and even some 1.5V kits, like the G.Skill Ares line) at their rated speed that's the killer.

The high VCCIO/VCCSA voltage is what kills the IMC.
September 2, 2012 3:22:49 AM

Pinhedd said:
Funnily enough, the i7-900 series processors are also rated at 1.5 volts and not 1.65 volts, which is why I tend to believe that the "1.65 volts is bad for sandybridge" was simply pulled out of someones ass and repeated to the point that everyone believed it.



Starting to think that as well. This is a fun discussion.

Have no idea what VICCIO/VCCSA stands for but ok... Jaquith is the ram god.
a c 129 } Memory
a c 169 à CPUs
September 2, 2012 3:36:37 AM

DJDeCiBeL said:
If you listen to and trust jaquith (which I do), it's not necessarily 1.65V RAM in and of itself that's the problem. It's the VCCIO/VCCSA voltage required to run 1.65V kits (and even some 1.5V kits, like the G.Skill Ares line) at their rated speed that's the killer.

The high VCCIO/VCCSA voltage is what kills the IMC.


Intel's specification allows for up to 1.4 volts on CPUVTT/VCCSA and 2.1 volts on the CPUPLL. The recommended maximum for VTT and VCCSA is 1.2 volts, and I'm sure Jaquith would agree with this.

There are a few motherboards that grossly overcompensate on the VTT/SA voltages when higher memory is used. My Rampage IV Extreme will push both of them to 1.4 volts if I leave them on Auto, this is far too high. This is also not necessarily the fault of the memory manufacturers, its the fault of the firmware designers. it is recommended to keep the DDR3 supply voltage and signalling voltage within a certain range but the line drivers should be able to compensate for this.

The reason I call "bullshit" on the claim is that many of the "1.65 volt memory is burning out the IMC omg!" claims popped up within months of Sandybridge being released, far too soon for the effects of the high voltages to have appeared. They then suddenly disappeared. Similarly no one made the same claims on Nehalem and Westmere despite the specification being the same for the respective absolute maximums and typical ranges.
a b } Memory
a c 283 à CPUs
September 2, 2012 3:46:01 AM

It may be true that there's absolutely no truth to it, but I'd still rather err on the safe side with what I know is "safe".

I have the Performance Tuning Protection Plan anyway (questionable as to whether I even should have bought into it, I know), so it doesn't really matter what voltages I use for anything, I suppose. I'm free to burn my chip to the ground if I wanted, but I'd still rather not, lol.
a c 129 } Memory
a c 169 à CPUs
September 2, 2012 3:47:57 AM

irlwizard said:
Starting to think that as well. This is a fun discussion.

Btw why are the 900s called Nehalem. Thought i7 900s where Westmere chips with codenames like boomfield depending on performance.


Nehalem can actually be divided into 3 "steps"

The first step for Nehalem was the LGA 1366 lineup which included only the 900 series processors. These processors had 4 cores, integrated memory controllers, and 3 memory channels. They did not have integrated graphics or integrated PCIe. PCIe was still purely the domain of the chipset and this necessitated a high-speed QPI link between the CPU and the chipset.

The second set of Nehalem processors includes the 700 and 800 series processors on the LGA 1156 platform. These processors had 4 cores, integrated memory controllers, two memory channels, and 16 integrated PCIe lanes. The integrated PCIe lanes meant that a high-speed QPI link was no longer necessary and was replaced with a simpler DMI link (which is itself based on PCIe 2.0) to the chipset which was now nothing more than a storage controller and IO hub.

Last, came Westmere. Westmere itself issued replacements on both the LGA 1366 platform and the LGA 1156 platform. The LGA 1366 platform was upgraded to 6 cores. The LGA 1156 platform got integrated graphics (only on the low end).
a c 129 } Memory
a c 169 à CPUs
September 2, 2012 3:49:27 AM

DJDeCiBeL said:
It may be true that there's absolutely no truth to it, but I'd still rather err on the safe side with what I know is "safe".

I have the Performance Tuning Protection Plan anyway (questionable as to whether I even should have bought into it, I know), so it doesn't really matter what voltages I use for anything, I suppose. I'm free to burn my chip to the ground if I wanted, but I'd still rather not, lol.


Certainly won't argue with that!
a b } Memory
a b à CPUs
September 2, 2012 5:11:54 AM

Pinhedd said:
To be completely fair, if your PC is vaulting over anything you have much bigger issues at hand


LOL it would be the strongest case ever haha.

I looked at the word but didnt twig, thanks for correcting :D 
a c 129 } Memory
a c 169 à CPUs
September 2, 2012 2:53:21 PM

boju said:
LOL it would be the strongest case ever haha.

I looked at the word but didnt twig, thanks for correcting :D 


I have some old Dell workstation cases that could withstand a nuclear blast. I also have some HP ProLiant servers which could be used as armor for a tracked vehicle in a pinch. Damn things are indestructible.
September 2, 2012 3:15:09 PM

Pinhedd said:
Nehalem can actually be divided into 3 "steps"

The first step for Nehalem was the LGA 1366 lineup which included only the 900 series processors. These processors had 4 cores, integrated memory controllers, and 3 memory channels. They did not have integrated graphics



The good old days when CPUs didn´t waste arcitecture on integrated graphics.

I really hope they make at least ONE haswell that has NO space wasted on integrated graphics, can you imagen how fast it would be? APUs suck for desktop users.
I don´t know anyone who doesn´t have a gfx card in their desktop. And if you only want a non-gaming PC you are better of getting a laptop/tablet with integrated graphics, keep the APUs out of desktops!
a c 129 } Memory
a c 169 à CPUs
September 2, 2012 6:06:15 PM

irlwizard said:
The good old days when CPUs didn´t waste arcitecture on integrated graphics.

I really hope they make at least ONE haswell that has NO space wasted on integrated graphics, can you imagen how fast it would be? APUs suck for desktop users.
I don´t know anyone who doesn´t have a gfx card in their desktop. And if you only want a non-gaming PC you are better of getting a laptop/tablet with integrated graphics, keep the APUs out of desktops!


Sandybridge-E doesn't waste space on IGPs ;) 
a b } Memory
a c 283 à CPUs
September 2, 2012 6:14:30 PM

I don't really consider the IGP to be wasted space anyway.

No, gamers don't normally use the IGP, but gamers are a very small slice of the overall market. Besides that, Quick Sync is very nice if you do a lot of video encoding, and the IGP can be very useful, if your GPU craps out.

And the 2550K (without an IGP) does exist...
a c 129 } Memory
a c 169 à CPUs
September 2, 2012 6:23:57 PM

DJDeCiBeL said:
I don't really consider the IGP to be wasted space anyway.

No, gamers don't normally use the IGP, but gamers are a very small slice of the overall market. Besides that, Quick Sync is very nice if you do a lot of video encoding, and the IGP can be very useful, if your GPU craps out.

And the 2550K (without an IGP) does exist...


The 2550k has an IGP on the die, it's just disabled for yield purposes. There are only 5 different chips serving all 150+ Sandybridge and Sandybridge-E processors across all the desktop, server/workstation, and mobile segments. Intel created that particular processor specifically because of the 2500k's popularity resulted in it constantly running out of stock and its use as a gaming CPU meant that most users didn't need or want the IGP
a b } Memory
a c 283 à CPUs
September 2, 2012 6:31:34 PM

Pinhedd said:
The 2550k has an IGP on the die, it's just disabled for yield purposes.


Yeah, I knew that, actually, but disabled is just as good as not having it there at all, to me. It's not like that space was going to be used by anything else anyway, and as long as it's disabled, there's no extra power draw.

I've definitely noticed though, that the 2550K's aren't very popular. Not sure if that's because people don't know about them or if people just don't want them.

I did know about it when I bought my 2500K (built this rig in March), but I specifically wanted the IGP, in case I ever needed it.
September 2, 2012 6:47:57 PM

DJDeCiBeL said:
I've definitely noticed though, that the 2550K's aren't very popular. Not sure if that's because people don't know about them or if people just don't want them.


Lack of advertisemen must be the case. It sounds like an awsome chip, I didn´t know it existed. Is there something like the 2550k but Ivy Bridge version... i5 or higher? That might really help with their heat issues if the integrated graphics is disabled. Less power draw = less heat.
a b } Memory
a c 283 à CPUs
September 2, 2012 6:51:29 PM

irlwizard said:
Is there something like the 2550k but Ivy Bridge version... i5 or higher?


There isn't right now, but I don't know if one might be planned or not. The 2550K wasn't released until January of this year, so that was a full year after the 2500K.
a b } Memory
a b à CPUs
September 2, 2012 6:51:30 PM

the igpu is nice when you ever decide to sell off your gpu for money(then wait for another gpu to arrive) or for troubleshooting gpu problems. Though if you had like a p67 mobo, no reason not to get the 2550k unless it magically costs more.(which it normally does, but if you can get it at the same price, might as well get that extra .1 ghz speed.)
September 2, 2012 9:38:54 PM

Oh the CPU dicussion tought me a lot. Thanks...

But as for RAM
Which one of these 3 should I buy... They are all 1600Mhz 9-9-9s but I can´t find reviews of them by good sites. Not buying from newegg because I get sales tax (California).

Crucial 1.5v - http://www.amazon.com/Crucial-Ballistix-DDR3-1600-240-P...

Kingston 1.5v - http://www.amazon.com/Kingston-Technology-1600MHz-KHX16...

Kingston 1.35v - http://www.amazon.com/Kingston-1600MHz-Non-ECC-Voltage-...

The 1.35v seem interesting.
a c 129 } Memory
a c 169 à CPUs
September 2, 2012 10:03:33 PM

irlwizard said:
Oh the CPU dicussion tought me a lot. Thanks...

But as for RAM
Which one of these 3 should I buy... They are all 1600Mhz 9-9-9s but I can´t find reviews of them by good sites. Not buying from newegg because I get sales tax (California).

Crucial 1.5v - http://www.amazon.com/Crucial-Ballistix-DDR3-1600-240-P...

Kingston 1.5v - http://www.amazon.com/Kingston-Technology-1600MHz-KHX16...

Kingston 1.35v - http://www.amazon.com/Kingston-1600MHz-Non-ECC-Voltage-...

The 1.35v seem interesting.


Kingston HyperX is always good, as is Corsair Vengeance and Mushkin Redline
September 2, 2012 10:20:44 PM

"Kingston HyperX is always good". Kingston anything is always good. xD

But which one do you think will give me better real-time performance. 1.5 or 1.35v? Both are hyper X and I can´t find reviews on these 2012 versions. Wonder how big the power saving vs performance trade is with the 1.35v. They might save money in the long run.
a c 129 } Memory
a c 169 à CPUs
September 2, 2012 10:23:32 PM

irlwizard said:
"Kingston HyperX is always good". Kingston anything is always good. xD

But which one do you think will give me better real-time performance. 1.5 or 1.35v? Both are hyper X and I can´t find reviews on these 2012 versions. Wonder how big the power saving vs performance trade is with the 1.35v. They might save money in the long run.


If they are rated the same they will perform the same. Nehalem and Sandybridge don't have native support for 1.35 volt memory, just Ivybridge
September 2, 2012 10:42:19 PM

I´m going with i5 Ivy k version on a Z77 board incase you wonder.
a c 129 } Memory
a c 169 à CPUs
September 2, 2012 11:32:38 PM

irlwizard said:
I´m going with i5 Ivy k version on a Z77 board incase you wonder.


With an Ivy you can use 1.35, 1.5, and 1.65 volt memory. The 1.65 volt memory will usually be out of specification though, often with tighter timings than those called for by JEDEC
September 3, 2012 12:32:53 AM

Not sure wat JEDEC is. Anyway I plan to OC the Ivy i5 3570k.

Do you think the 1.35v would help the system OC overall since the mobo would be colder. Plus if I give the 1.35v ram 1.5v it might run at tighter latencies then the stock 1.5v ram, right? Won´t bother with a 1.65v ram this time around.
a c 129 } Memory
a c 169 à CPUs
September 3, 2012 12:46:30 AM

irlwizard said:
Not sure wat JEDEC is. Anyway I plan to OC the Ivy i5 3570k.

Do you think the 1.35v would help the system OC overall since the mobo would be colder. Plus if I give the 1.35v ram 1.5v it might run at tighter latencies then the stock 1.5v ram, right? Won´t bother with a 1.65v ram this time around.


Memory doesn't really get hot anyway. Even the 1.65 volt DDR3-2133 modules in my PC barely get warm. All 1.35 volt memory is also rated to run at 1.5 volts at the same timings. I'm not sure if they will run tighter, the answer is usually no.
September 3, 2012 1:15:24 AM

Pinhedd said:
Memory doesn't really get hot anyway. Even the 1.65 volt DDR3-2133 modules in my PC barely get warm. All 1.35 volt memory is also rated to run at 1.5 volts at the same timings. I'm not sure if they will run tighter, the answer is usually no.


Never been worried about the ram temps, but wont it let me run lower Vtt, then the mobo will be cooler?

By the way I found a review for the LoVo that answered my own question about the power consumption, for a desktop it is pointless because you only save 5W on load. Here is the review for people that want to see. http://www.custompcreview.com/reviews/kingston-hyperx-g...

Could save 9W by underclocking it but that is not an option my rig will be used for gaming. Think the 1.5V Plug and Play ram might tolerate more tweaking. Going to try and run it with 8-8-8 t1.

Edit: Hmm the PnP is made for Sandybridge chips. Might skip on Kingston, sigh... Patriot is always good at tighter latencies.
!