rumours about high memory clock freq / volts adversely affecting sandybridge lifespan

K

kandor

Gawd
Joined
Aug 24, 2005
Messages
632
Hi,

I've been seeing/hearing rumours that because the memory controller is on the cpu die that running it faster than spec or with more volts (i.e. 1.65 vs. 1.5) that you can radically shorten the life of your processor.

My i7 2600k is oc'd to 4.9 (at 1.35 volts) memory is mushkin blackline 4 x 2gb sticks 1.65 volts at 1600. Is this anything to worry about as I would expect this proc to last for a while. Is there any truth to this scare mongering?
 
I think Intel said that they don't recommend running 1.65V RAM long-term, but I can't find the link right now.
 
The truth is that Intel wants people to standardize on 1.5v.

There is no hard evidence that 1.5v is an *absolute* necessity for Sandy Bridge, but given the extra effort Intel has gone through to encourage major OEMs to replace enthusiast 1.65v modules with 1.5v stuff, there may be some truth to it.

Intel hardly lifted a finger when memory makers sold 1.65v modules like candy for Sockets 1366 and 1156, but there's an obvious push to change things now with 1155 (See Corsair Vengeance, for example). Whether that means you should be worried, I have no idea.

If you want, you can probably just back your module down to 1333 and it should probably run at 1.5v. I seriously doubt this will affect your performance (even overclocked). See here:

http://www.xbitlabs.com/articles/memory/display/sandy-bridge-ddr3_5.html#sect0

In an overclclocked system, the difference between 1600 CL7 and 1333 CL9 is under 10% for most things.
 
Hmm, I guess I am worried a bit now. I have 16gb corsair vengence in my p67 system which is 1.5v, However my Z68 system has 8gb of mushkin PC16000 that can run at 2000 which is 1.65v and originally purchased for an old i5-750 p55 chipset. I wouldn't mind replacing it except that at the time ddr-3 was really expensive and those 8gb probably cost close to 600$ 2 years ago. It's going to look awkward sitting on my shelf gathering dust - lol.
 
kandor said:
Hmm, I guess I am worried a bit now. I have 16gb corsair vengence in my p67 system which is 1.5v, However my Z68 system has 8gb of mushkin PC16000 that can run at 2000 which is 1.65v and originally purchased for an old i5-750 p55 chipset. I wouldn't mind replacing it except that at the time ddr-3 was really expensive and those 8gb probably cost close to 600$ 2 years ago. It's going to look awkward sitting on my shelf gathering dust - lol.
Click to expand...

Most high-speed modules have lower speeds where they are specced for 1.5v. Try looking at the Mushkin site for complete module specs.
 
Yeah, if it is DDR3-2000 at 1.65V it will most likely run fine at 1600 at 1.5V, and if not, certainly at 1333. Just dial down the speed (which doesn't matter for performance anyway) and run lower volts.
 
Some 1.65 ram modules ran at their stock speed when downvolted too. Try dropping to 1.60 and stress test it. Maybe even lower, down to 1.55V.
 
First, running your RAM at 1.65V on SB shouldn't "radically" reduce the life span. It may shorten the total lifespan, but that may mean instead of running for 20 years, it only runs for 15 years. It is hard to tell just how much it may hurt the total lifespan. I doubt it would cause it to fail within 3 years, but I do not believe there is not enough evidence right now to show the long term effects.

That being said, most of the 1.65V RAM modules I have seen work in 1.5V spec as well. The OC modules like the 2000, may not work at those speeds at 1.5V. The best thing to do is just try different settings to see what the max speed you can get them to go at 1.5V or maybe a little higher like 1.55V as suggested above.
 
Wait, now they are saying faster than 1333 is damaging not just the volts?
 
skizzled said:
Wait, now they are saying faster than 1333 is damaging not just the volts?
Click to expand...

What? I think you may be confusing what people are saying. Typically with any kind of electrical device you need more volts to get faster operation. So for memory to achieve the 2000 Mhz, you need to up the volts to 1.65V. When running on Sandy Bridge, that may cause issues according to Intel. Now if you run that memory at 1.5V it may not operate at 2000 Mhz, it may only run at 1600 Mhz.

So what people are saying is trying to run at higher speeds may mean higher volts which may damage the lifespan of the CPU.
 
You can also use CPU-z to check what the JEDEC profiles are on the RAM - that'll show you what voltage it'll run at a lower speed. So the top profile would be DDR3-2000 at 1.65V, but it is quite likely that the profile for DDR3-1600 is at 1.5V. That happens fairly often, as NoOther said.
 
NoOther said:
What? I think you may be confusing what people are saying. Typically with any kind of electrical device you need more volts to get faster operation. So for memory to achieve the 2000 Mhz, you need to up the volts to 1.65V. When running on Sandy Bridge, that may cause issues according to Intel. Now if you run that memory at 1.5V it may not operate at 2000 Mhz, it may only run at 1600 Mhz.

So what people are saying is trying to run at higher speeds may mean higher volts which may damage the lifespan of the CPU.
Click to expand...
I understand that but the wording of the OP makes it seem like its either or.
running it faster than spec or with more volts
Click to expand...
 
Thanks for all the info, here is my spd page, so I see the JEDEC #3 for 1333 operation but I suppose I could just experiment as mentioned for 1600. What I don't understand is why the slower speeds have somewhat looser timings? i.e. the 9,9,9,24 for the 1333 profile.

9AIWN.jpg
 
The timings are probably looser because the voltage is lower, if you up the voltage you could also up the timings.
 
I think superior CPU cooling will help life span as well.
 
imho, i dont think this is a rumour, i think its true.
on my system i had perfectly working ram and cpu that was quite malleable. After leaveing my ram at 1.6 volts and 2133mhz speed, after a few days my cpu was not able to even be TOUCHED meaning my cpu is just an i5 2500 at stock, and i did have 2500k. if i touch the voltage, even bump it up or down my motherboard just keeps resetting itself, only to fix it is to reset bios. same with my ram, its stuck at 1333 and is at 1.5, but when i first bought it was advertosed tobe 1866 @ 1.65. Unfortunately this is no longer the case, my ram is also having issues now, if i modify ANYTHING at all, timing, speed, voltage, my motherboard keeps resetting and as i said before, only way to fix it is to reset bios. so... moral of the story? DONT PUT THE RAM VOLTAGE TOO HIGH!!
 
Just to clear the record Intel was the one who started this "rumor" back in the days when i7 was lauched. I'm too lazy to cite my references but I do recall during reading reviews where Intel was very concerned about overvolting the memory controller. Thats why we take it so seriously.
 
I have been running Mushkin 2133 Redline at 1.65V for quite awhile now, all while pushing my 2600K over 5.0GHz. I haven't had any problems yet. But, thats not to say that it couldn't affect anything down the road after prolonged use. But, I have been OCing the crap out of this CPU since I got it and still haven't had a single issue while running the RAM at 1.65V @ 2133MHz
 
I have some Mushkin Redline that has an XMP profile for 7-8-7-24 1600MHz at 1.65v. It is currently running at 10-11-10-28 2133MHz at 1.5v just fine.

Lower the voltage, play with the timings, the quality of the RAM may surprise you.

As for this:


kandor said:
What I don't understand is why the slower speeds have somewhat looser timings? i.e. the 9,9,9,24 for the 1333 profile.[/IMG]
Click to expand...

Check out the wiki for DDR3:

http://en.wikipedia.org/wiki/DDR3_SDRAM#Modules

According to your picture, Mushkin lists this as PC3-10700H. That H is the important part. The JEDEC standard for DDR3-1333H is 9-9-9.
 
Last edited:
I run my RAM at 1600 mhz CAS 9 1.5v, and it's rated for 1333 mhz 1.65v. So I would just experiment with undervolting, I'm sure you can get it to run at 1600 mhz at 1.5v.
 
Thanks for the info!

If you're running at 2133 (1.5v) 10-11-10-28 how to you compare whether that is better/faster/stronger than running some slower speed but with tighter timings? i.e. 1600 @ 7-10-8-27?
 
old skool said:
Just to clear the record Intel was the one who started this "rumor" back in the days when i7 was lauched. I'm too lazy to cite my references but I do recall during reading reviews where Intel was very concerned about overvolting the memory controller. Thats why we take it so seriously.
Click to expand...

The only time I remember anything official on this was back in the A64 days when AMD reps hit the forums to detail this issue and advise what safe limits are.

I'm not seeing, and haven't seen, Intel do this to any degree.


Ordinarily my 'gun-h'o self would say to ignore the FUD, but in this particular case: since memory overclocks with sandybridge do fuck all for performance in games; it's totally and utterly not worth the risk for the unnoticable gain.

I'm at 1.6v for my ram, overclocked to 1866mhz. Just gunna leave it at that.
:cool:
 
wsgroves said:
http://www.newegg.com/Product/Product.aspx?Item=N82E16820231504
New memory...why do the companies come out with 1.65v modules if it could be a potential problem?
For someone looking to buy ram in the next 2 days for my x79...should I stay away from that kit and go with a lower volt version instead?
Click to expand...

Simple, higher speed needs more volts. If you are buying for X79, just get DDR3-1600, as faster RAM doesn't show any performance increase anyway.
 
kandor said:
Thanks for the info!

If you're running at 2133 (1.5v) 10-11-10-28 how to you compare whether that is better/faster/stronger than running some slower speed but with tighter timings? i.e. 1600 @ 7-10-8-27?
Click to expand...

From a synthetic / numeric standpoint, you trade slightly worse latency (8.9ns vs 9.4ns) for a bit more bandwidth (20GB/s vs 27GB/s). There was no significant difference in any of the benchmarks I ran between the two speeds. You get an extra two or three frames per second (out of 140) in an x264 first pass. That's about it.

Memory CAS timings don't really impact performance on SB chips at all. But 27GB/s looks awesome :)
 
That's cool, but if I wanted to try reaching a high freq/low volt scenario with my blackline versus your redline, I wouldn't be sure where to start - maybe I'll just try your numbers - but how did you arrive at those timings?
 
kandor said:
That's cool, but if I wanted to try reaching a high freq/low volt scenario with my blackline versus your redline, I wouldn't be sure where to start - maybe I'll just try your numbers - but how did you arrive at those timings?
Click to expand...

I copied the PC3-17000 Blackline numbers off Mushkin's site, as well as a few of the others conservative timings for 1.5v:

10-11-10-28 1.5v 2133MHz
9-10-9-27 1.5v 1866MHz
9-9-9-24 1.5v 1600MHz
 
Nexus6 said:
I copied the PC3-17000 Blackline numbers off Mushkin's site, as well as a few of the others conservative timings for 1.5v:

10-11-10-28 1.5v 2133MHz
9-10-9-27 1.5v 1866MHz
9-9-9-24 1.5v 1600MHz
Click to expand...

Did you try 7-7-7-21 1.5v? Seems like they could do that speed.
 
Thanks for the advice Nexus6, and I still love watching bladerunner:) This is at 1.5v

X1rbA.jpg
 
You must log in or register to reply here.
Back
Top