1.65V i5 3570K

[Binge]

Originally Posted by cadaveca

Sure do have that. Many /SNBIVB system has more than that, as I stated above. I've seen 0.860V- 0.950 V for IMC, and 1.5V is default...immeidately, any stock CPU goes outside of that ratio.


What changed? The sockets? The CPUs? That only applied to P55/1366; we've had 1155 for some time now(nearly 2 years), and no, it does not matter, since it works as I stated above.

With X58 the memory controller voltage and mem voltage ratios were within the .5V. That much is obvious and it is also very obvious the difference between the IMC and mem voltage with 1155 in some cases is well outside of the .5V. For the most part the .5V rule was only applicable for users who were going over 1.5V on the x58 IMC which was known as the QPI. I don't think I'll get an answer to my question here. You have not stated how it works; you've just reposted some stock voltages and said nothing about overclocking tolerances.

[cadaveca]

Originally Posted by Binge

For the most part the .5V rule was only applicable for users who were going over 1.5V on the x58 IMC which was known as the QPI.

Ah, well ,you know, QPI stands for Quick Path Interconnect...what the CPU's bus was called, not the IMC. QPI replaced the Front Side Bus.

http://www.intel.com/content/www/us/...ure-paper.html

Now we have BCLK.

The lack of that ratio to remain is because how those two things work together is differently. Nehalem had L3 cache on one side, and memory control on the other(physically), while with SNB/IVB L3 is connected to the memory contorller this time via the system agent, and is right next to the L3. Also, with SNB and IVB, the L3 runs at the same speed as the core, while on Nehalem the L3 ran at a seperate speed.

As to the exact reason why, no I cannot tell you, as Intel has not really disclosed that info at this point, other than that the whoel northbridge sub-sytem is now very much different. Personally, I think it's becuase the system agent decouples VTT from IMC.

I'm just an enthusiast, not an engineer, so I can only relate info that Intel gives, and I still do not have direct contact with them for this sort of information, unfortunately. Just another goal to attain.

Might have something to do with them adding the PU, even I couldn't tell ya.

And with that said, you'll probably have to ask Intel directly on that one, to get something other than "the electrical layout is different now".

[Binge]

Found a bit of info from XS about user tested voltages vs intel standard. Looking for more information on how the IMC communicated with which parts of the processor. Physical placement of the components in the silicon are 1/2 of the equation. Definitely interested in finding out some more facts about voltages and the new roll of the IMC as Cad has made the observation that it is a drastically different design.

[cadaveca]

SIn posts here as well under the UID StevenB. That chart is here as well, too.

He's the only user really posting such stuff.


Part of it is that with the tuning warranty, and how these chips clock, there's really nothing to worry about. You're actually pretty safe from killing these chips....except that the raw CPU speed scales with votlage, even under high temps. This may cause some users to push too far, and shorten the lifespan of the chip drastically.

But otherwise, as long as you keep temps i ncheck, and know a bit about how voltages can fluctuate, there's not a lot you can do that is really going to hurt things.

Overclocking today is more about buying the right parts to work together, mostly memory and motherboard i nthat, too, and then getting some luck on how good your IMC is, or so it's been related by many othre reviewers...I still do not think I have found the max of my 3770K's memory controller. But then, maybe i have..still working on that one.

[Binge]

Yeah if I had known that would have been one of the first things I posted. Great to get anecdotal bits from users who have had the option of testing a number of chips. Most of the x58 knowledge was trial & error which busted a number of chips.

[cadaveca]

Originally Posted by Binge

Yeah if I had known that would have been one of the first things I posted. Great to get anecdotal bits from users who have had the option of testing a number of chips. Most of the x58 knowledge was trial & error which busted a number of chips.

That's always the case, but now there is actually so little left to user intervention, there's no a lot of info to be found. Intel has made it so easy, you jsut buy the ram you want...buy a decfent board, adn then set ram, and set turbo multis...and maybe some CPU voltage. Most users won't even have to play with IMC voltage much.


I run 2666 MHz daily with my 3770K, 10-12-12-35, 0.95 V on IMC. 10.05 V on VTT(stock votlage for both), and 1.65 V on ram. I can guarantee that any such concerns from the past and having that specific ratio between ram and IMC/VTT are long since gone.


In fact, you'll find many reports of VTT being what's needed to be adjusted when clocknig ram..not IMC voltage...

[Sasqui]

Originally Posted by cadaveca

I run 2666 MHz daily with my 3770K, 10-12-12-35, 0.95 V on IMC. 10.05 V on VTT(stock votlage for both), and 1.65 V on ram. I can guarantee that any such concerns from the past and having that specific ratio between ram and IMC/VTT are long since gone.

You're running 32GB of 2133 CL9 G.Skill at that speed? Nice...

[cadaveca]

Originally Posted by Sasqui

You're running 32GB of 2133 CL9 G.Skill at that speed? Nice...

ha, no, I was running, mhaven't updated system specs in a while.

I've never really ran less than 2133 MHz though, unless benchmarking.

[Sasqui]

Originally Posted by cadaveca

ha, no, I was running, mhaven't updated system specs in a while.

I've never really ran less than 2133 MHz though, unless benchmarking.

I'm curious what I could get The G.Skill Sniper 2133 CL9 to run at using CL11

[cadaveca]

Originally Posted by Sasqui

I'm curious what I could get The G.Skill Sniper 2133 CL9 to run at using CL11

Hard to say just by model number, depends on the ICs how much potential they might have.

[EarthDog]

Originally Posted by Sasqui

I'm curious what I could get The G.Skill Sniper 2133 CL9 to run at using CL11

On your current rig, I wouldnt imagine a terrible amount more. On an IB, as I think those are Hynix IC's (not sure), those could easily see 2400Mhz+. The memory controller on SOME of those IB CPUs are mad good.

[Sasqui]

Originally Posted by EarthDog

On your current rig, I wouldnt imagine a terrible amount more.

My current rig only supports DDR2, the backstory... I bought 32 GB of the stuff as I will soon buy a CPU and MB but haven't made up my mind yet.

[Hood]

Originally Posted by cadaveca

1.65 V ram is fine. Intel says max 1.85 V.

Somehow I missed where Intel said 1.85v is max for Ivy Bridge memory, could you point me to the article or tech document? I was under the impression that the safe max is 1.65. So if I can already overclock my basic 1.65v 1600 RAM to 1866, what happens if I bump the voltage to 1.7? (not gonna do it, just asking the question). Will I be able to hit 2000 or maybe tighten the timings a bit by adding the extra juice? And what am I gonna fry, the CPU, the RAM, or both? Anyone out there tried this?

[cadaveca]

Originally Posted by Hood

could you point me to the article or tech document?

There are only three or four on intel's site for each CPU??? It's not hard to find...

lots of guys benching with 1.8 or so, Maximus V Formula and Maximus V Gene even have 1.85 V TridentX profile....


Personally, I wouldn't recommend running it though. Intel says max VID for SNB is 1.525 or something, too, and I know that long-term that votlage will shortne the life of the chip.


With vDIMM, since it's kinda decoupled, I could seee hwo it's possible, since PLL is 1.8 V too...but I dunno 1000%. I do know that 1.85 V is listed in whitepapers though, if it hasn't been editted.