i5/i7 Lynnfield QPI/VTT voltage

[Yorr]

Hello!. I decided to post this in here because this place is full of really savvy people who could help me.

I don't know why, but, when it comes to overclocking the new Intel Lynnfield i5/i7 CPUs, everybody (including overclocking or specialized websites) seems to be using the QPI/VTT voltage into the 1.35v-1.4v (or more..) range, which is the maximum recommended value for the older Bloomfield i7s (1.35v says Intel Datasheet for Bloomfield). Nobody seems to know that this spec has changed for the Lynnfield, to an absolute maximum of 1.21v VTT, which, obviously, limits the overclocking potential for this new CPUs in comparison to the older ones. Is it safe to use 1.35v VTT in this new CPUs as well? Why is most people ignoring this?

This is the link to the Intel Datasheet for the Lynnfield:
http://download.intel.com/desi...or/datashts/322164.pdf
Look at the table 7-4 on page 67.

Any insight would be greatly appreciated.

 

[vshin]

You're one of the few but not the only one to notice this distinction between Bloomfield and Lynnfield. I'm sure there are a lot of testimonials that say it's ok to raise the VTT voltage above the recommended max but I think it's a mistake.

Based on that document, Lynnfield is spec'd to operate normally up to a vcore of 1.400 and a vtt of 1.155 volts (Tables 7-5, 7-6). Intel also lists an absolute max vcore of 1.55 and a vtt of 1.21 volts (Table 7-4). They stipulate that operating above the normal parameters but below the absolute max may result may reduce function and lifespan of the CPU but should function again once returning to normal specs. Operating beyond the absolute parameters may not only reduce function and lifespan, but functionality may not return even if returning to normal specs. In short, exceeding the absolute limit may result in permanent damage so leave it to xtreme overclockers who really know what they're doing and are not afraid to accept the consequences. For the casual overclocker using air or low-end water cooling, it would be best to stay under the absolute limit because if things don't work out, at least you'll still have a working CPU when you return to stock.

So to answer your question, if you are the sort that gets nervous about raising the vcore above 1.400, then you shouldn't feel comfortable raising the vtt above 1.155 volts either.

 

[RussianSensation]

I have limited my overclock to using QPI/VTT of 1.250V since going above 189 Base Clock on my board requires for me to use a whooping 1.39/1.41VTT (with the current BIOS), along with a substantial increase in Vcore from 1.30V to 1.36V to just get to 4.2ghz. My temperatures skyrocket from 70*C at my current settings to 82-84*C all for about a 7.7% increase in performance. I decided that for now 1.30Vcore and QPI of 1.25V is my sweet spot in terms of voltages and temperatures (Intel specifies temps < 72.7*C).

I suggest you overclock your processor and also attempt to find a sweet spot and then evaluate yourself what is reasonable given your cpu cooler and the time you intend to keep your processor. Remember you already void your warranty when you start overclocking.

 

[MadScientist]

I made reference to this last week in the i5-750 overclocking thread. Every oc review I have read so far of the i5-750 and i7-860, including Anandtech's review of Gigabyte's and Asrock's micro P55 MBs, have the VTT voltages set between 1.36 and 1.39V.
Like RS, I have been trying to keep the VTT voltage down.

I have found that by just bumping up the PCH voltage to 1.10 or 1.12V has increased stability while keeping VTT voltages in the 1.24 to 1.30V range from 3.8Ghz to 4.0Ghz.

I've got some more testing to do and will post results in the i5-750 overclock thread when done.

 

[Yorr]

...Well, then...would you guys consider a 1.23v VTT on a Core i7 860 safe?. I'm at 3.67 GHz, 1.30 Vcore, Prime95 and LinX stable with nice temps. The other possibility I have, which is stable too, is 3.57 GHz with a VTT 1.21v (the absolute maximum limit recommended by Intel we've been talking about), same Vcore. What do you guys think?

Another question I have is: Is VTT affected by Vdroop? Does LLC affects the VTT?.

 

[MadScientist]

Originally posted by: Yorr
...Well, then...would you guys consider a 1.23v VTT on a Core i7 860 unsafe?. I'm at 3.67 GHz, 1.30 Vcore, Prime95 and LinX stable with nice temps. The other possibility I have, which is stable too, is 3.57 GHz with a VTT 1.21v (the absolute maximum limit recommended by Intel we've been talking about), same Vcore. What do you guys think?

Another question I have is: Is VTT affected by Vdroop? Does LLC affects the VTT?.

Looks good.

 

[lopri]

http://www.anandtech.com/cpuch...owdoc.aspx?i=3634&p=19

I think the above has some bearings on different electrical characteristics of Lynnfield compared to Bloomfield. Anand says overclocking Bclock by 133MHz multiples is the way to go but I am not sure what that means. i5/i7's Bclock is 133.. so that makes the next one 266 and that can't possibly be it. (both for Bclock itself and for the resulting CPU frequency)

Can anyone comment on it? Or did he mean 33Mhz/33.3MHz, assuming that's the base clock for both PCIe and CPU? (33.3x3 for PCIe and 33.3x4 for CPU?)

 

[MrK6]

Originally posted by: lopri
http://www.anandtech.com/cpuch...owdoc.aspx?i=3634&p=19

I think the above has some bearings on different electrical characteristics of Lynnfield compared to Bloomfield. Anand says overclocking Bclock by 133MHz multiples is the way to go but I am not sure what that means. i5/i7's Bclock is 133.. so that makes the next one 266 and that can't possibly be it. (both for Bclock itself and for the resulting CPU frequency)

Can anyone comment on it? Or did he mean 33Mhz/33.3MHz, assuming that's the base clock for both PCIe and CPU? (33.3x3 for PCIe and 33.3x4 for CPU?)

I read that as well, and was also curious as to what he meant by it.

I have to start with the i5 750 is a very difficult chip to overclock stably, perhaps due to the complexity of the chip itself and all the different components of it. As far as VTT goes, it stabilizes higher BCLK clocks, but anything else as well? Right now I have mine at 1.21V per Intel's specs, running 190x20 for 3.8GHz (Vcore at 1.325V in BIOS). I tried 200MHz BCLK because I wanted to get my RAM to 2.0GHz, but it wouldn't stay stable 200x19. Oddly enough, it also didn't respond to higher VTT OR higher Vcore (within reason, I wasn't going to pump 1.4V+ through it). However, 200x17 (3.4GHz) is perfectly stable on my chip. I wonder if the limitations encountered are entirely due to the PCIe component of the CPU. I've found that Prime95 is almost completely useless in determining stability, as an overclock can be hours Prime stable but will freeze quickly in a 3D environment. I even tried running Prime95 and RTHDRIBL at the same time, but it didn't detect instability any more easily. I've actually found GTA IV to be the best test for stability . Maybe someone can chime in here, but is there any definitive sources/guides on what parts of the chip are affected by what frequency and voltage adjustments? It seems like would help clear up this conundrum.

 

[Max_Q]

Hey Yorr,

Thanks for starting this thread. P55 and Lynnfield voltages, especially as advocated by the memory manufacturers, are kind of a dirty little secret.

The way I see it, the memory manufacturers are being given tacit approval by Intel to recommend higher than published voltages, while at the same time Intel gets to cut down on their warranty exposure (and deny your warranty claim?) by publishing really low voltages. This implies that Intel's published voltages are unnecessarily conservative, because if the voltages that the memory manufacturers are recommending are actually damaging they'd be letting themselves in for a world of liability. Of course if Intel is playing it straight with us, then we're all being royally screwed by the memory manufacturers.

For their own motherboards, Intel recommends a maximum memory voltage of 1.6v, and says anything higher will damage something. Interesting considering that the maximum "normal operational limits" in their datasheet for VDDQ is 1.575v

The OCZ employees on their forums regularly recommend VDDQ of 1.65v and VTT of 1.35v to get their memory stable at advertised speeds.

Either way I just want to know if I can bump up my ram voltage to 1.65v to get the 1600 speed I paid for, without damaging my Core i7-860.

I have posted a question at the OCZ forums here:

http://www.ocztechnologyforum.com/forum/showthread.php?68140-Safe-Voltages

and at the Intel processor forums here:

http://communities.intel.com/message/81821#81821

in case you want to follow along.

 

[Munky]

Does your memory not run stable at 1600, 1.6v? I'm asking because I also have memory (not OCZ) rated at 1600 1.65v, but I find it runs stable even at just 1.6v.

 

[tweakboy]

Easy, higher you take voltage, more you can OC,and be stable. But highver voltage means higher cpu temps, have adequete cooling... gg

 

[Max_Q]

I have 8GB of OCZ3G1600LV4GK. On my GIGABYTE GA-P55A-UD4P it is stable at:

1333 @ 8,8,8,18 1T - VDDQ 1.5v, VTT 1.1v
1333 @ 7,8,7,18, 2T - VDDQ 1.56v, VTT 1.15v

Anything above that on my motherboard is above Intel's published "normal operational limits".

1600 @ 8,8,8,24 2T - VDDQ 1.6v, VTT 1.1v - Booted Windows. Error in OCCT after 29 sec.
1600 @ 8,8,8,24 2T - VDDQ 1.62v, VTT 1.15v, - Booted Windows. Blue Screen running OCCT after 34 min.

It will boot and load Windows at 1600 @ 8,8,8,24 2T - VDDQ 1.64v (which is what it's rated for), but I haven't done any extensive testing because that goes back to my original point. I don't know what it's doing to my processor and chipset, and I don't know what voltages are long term safe.

 

[bob5568]

Max...just noticed Simon responded from OCZ. IMHO, his response appears to lack the level of awareness I'd expect of someone recommending his customers exceed the oem voltage specifications.

It opens up another theoretical possibility to your guess work above, about intel and ocz's interaction. Perhaps there's been none and ocz doesn't know what their talking about?

 

[Max_Q]

If that's the case bob, we're all in a world of s**t.

No, at the engineering level I'm sure they know exactly what they're doing. And their ram is pretty and shiny and gold!

Besides, I think Simon offered me a job... so don't talk about my new employer like that. :biggrin:

 

[bob5568]

Max, response from Intel to your forum post is certainly underwhelming, huh.

 

[Max_Q]

Hi bob,

That Intel forum seems to be amazingly dead for the size and worldwide scope of the company. I'm still hopeful that they pay one of their guys to spend an afternoon answering questions now and then.

So far, OCZ is ahead on points.

Can you think of any other place I could post where I might stand a chance of getting a real answer?

 

[RussianSensation]

More discussions based around vDIMM, Vtt:

http://www.xtremesystems.org/forums/showthread.php?t=207972

Update: I've been running Vtt of 1.250V and Ram voltage of 1.568V since September 8, 2009, mostly 100&#37; cpu load 24/7 Seti@home with occasional gaming and HD movies.

 

[Max_Q]

Hi RussianSensation,

Yeah, I've read through that thread. Most (if not all) of those guys are discussing voltages as they apply to X58/920's. As Yorr noted at the beginning of this thread, Intel has actually lowered the maximums since Bloomfield. The Core i7-900 series maximum VTT was 1.35v and maximum VDDQ was 1.875v, but the Core i7-800 series on the P55's maximums (again, according to Intel) are VTT of 1.21v and VDDQ of 1.65v.

Not to be disrespectful to those guys, but that thread reminds me of the monkeys around the monolith in the beginning of "2001". There are guys posting to the thread saying "What do you think of my voltages, are they safe?". Someone will answer "Sure, I've been running those voltages for three days now, and everything's great, so don't worry". Meanwhile, one of the guys that really knows what's going on will come on the thread and say "Well, I don't know, but I've been running these voltages and I'm just praying it'll hold together". BTW, I'm paraphrasing.

The point is, nobody knows. Everybody's just feeling their way through the dark, wondering if and when their investment is going to up in smoke. Meanwhile, some very big companies are making some very big dollars while putting out conflicting specs.

Not to belabor the point, but for most of the performance ram out there, you can not reach the rated speeds unless you're willing to set your memory voltage to the ragged edge of maximum that Intel allows (remember, for the P55/860 1.65v is the maximum... 1.575 is the top end of normal). And if you do that, here's what Intel says will happen: "At conditions outside functional operation condition limits, but within absolute maximum and minimum ratings, neither functionality nor long-term reliability can be expected. If a device is returned to conditions within functional operation limits after having been subjected to conditions outside these limits (but within the absolute maximum and minimum ratings) the device may be functional, but with its lifetime degraded depending on exposure to conditions exceeding the functional operation condition limits."

In the case of OCZ, their recommendation of a VTT voltage of 1.35v blows right through Intel's published maximum.

What I'd like to see is Intel, or someone sufficiently connected to Intel, issue a statement that says, in effect, that the voltage recommendations being given out by the major ram manufacturers were arrived at in consultation with Intel and have their blessing. I mean, Intel even sold their Core logo to OCZ for a line of 1.65v ram yet, as I noted above, Intel's motherboard division says that greater than 1.6v is a no no. Figure that out.

BTW, your system looks fast, yet reasonable. Do you have turbo on?

 

[bob5568]

Max, I'm also troubled by what appears to be zero knowledgeable explanations. I'm also amazed by Simon appearing to learn about the situation because you asked questions, yet he's provided advise such as to exceed the vtt max....while Intel remains silent.

Smells like some kind of screwup at the corporate level. If there wasn't such a thing, folks with engineering knowledge would be openly responding with information that rings true.

I wonder, in response to your question up a message or so, if the motherboard forums might provoke a response. The gigabyte process of asking and getting an answer the next day actually worked for me once. Someone on these forums made it appear that the current rating on the fan connectors was easy to find if I just read my manual, yet I still couldn't find it, asked gigabyte directly both for the current rating and the reference location within the manuals, and they said 1 amp, and forget looking for a referece, there wasn't any!

I'll lay this one on them and see what they say.

Bob

 

[Max_Q]

Hey Bob,

I'm betting that the ram guys had tons of stock on hand of 1.5v DDR3 (since from what I've read, ALL DDR3 is 1.5v) that just wouldn't go very fast unless you hit it with a lot of voltage. That used to be OK, I guess. Supposedly ram voltage was not as critical with the older chipsets (and definitely the older non IMC processors). And since companies like OCZ offer a lifetime warranty on their ram, who cares. Burn it up... they'll give you more.

But then Intel hit them with Bloomfield/X58 and then Lynnfield/P55. The ram guys need to advertise their performance ram at high speeds in order to sell it, and the only way to get there is with voltage. So they picked the absolute highest voltage they could find out of Intel's spec, called it "designed specifically for the Intel® P55 Chipset and subsequent Intel® Core™ i7, i5, and i3 (Socket 1156) processors", and started binning away.

Of course, if you read Intel's spec (and you believe it) that absolute maximum voltage isn't at all "safe".

It's ironic that you buy performance ram for more umm... performance, and the result (again, according to Intel) may be degraded performance and lifetime.

Again, I can't believe that all of the ram manufacturers would be taking what would appear to be a massive chance with their customer's gear. They must have been told, back channel, by Intel that VTT of 1.35v and VDDQ of 1.65v is long term safe. Well, at least 3 years safe even for the weakest examples they put out.

Thinking about it now, if I could go back I might buy some of that nice 1.35v ram, run it at 1600 @ 1.5v with tighter timings and be happy. After all, I want a fast computer, but not at the expense of a useful lifetime. Then again, for the real world difference that ram speed makes, maybe not. I got a pretty good deal on the OCZ. :hmm:

BTW, where are you communicating with Gigabyte? I sent them a question (another subject) a few days ago, and according to their system they haven't even looked at it yet.

Let me know what you find out.

Regards,
Max

 

[SHAQ]

My rule of thumb is to not exceed .15v higher than VID on your CPU and 1.6v max on your RAM. RAM speed doesn't provide much of a performance boost anyway so I don't see why people get obsessed with it. I leave mine at 1100 cas 6 and don't worry about it.

 

[BababooeyHTJ]

Informative thread here. I'm surprised that I don't see more people mentioning this. Gives me something to think about with my 860.

 

[spinejam]

my settings:

 

[aigomorla]

MAX VTT on an LGA1156 is 1.21v?

You guys kidding me right?

That's like someone grabbing you by the B@LLs in overclocking.

Now i see why u guys need a Ton of Vcore.

Yeah, I've read through that thread. Most (if not all) of those guys are discussing voltages as they apply to X58/920's. As Yorr noted at the beginning of this thread, Intel has actually lowered the maximums since Bloomfield. The Core i7-900 series maximum VTT was 1.35v and maximum VDDQ was 1.875v, but the Core i7-800 series on the P55's maximums (again, according to Intel) are VTT of 1.21v and VDDQ of 1.65v.

OMG... thats so sad... higher bclk will demand a higher Vtt.

 

[biostud]

My run at 1.28V vCore and 1.25 Vtt, but I had to disable the higher c-states to get it stable.