I recently got a Lynnfield kit from someone who failed to get it stable. The kit consists of the following:
- Intel Core i7 860 processor (codename Lynnfield), socket 1156
- Asus P7P55D Deluxe motherboard
- Corsair Dominator 8 GB (4×2 GB) DDR3-1600 kit (CMD8GX3M4A1600C8)
Well, on the surface this all looks nice. Reasonably high-end products from wellknown brands. So, it is up to me to see why it is not working.
The problem itself is reasonably hard to reproduce. The system will generally boot okay, and run stable for hours at a time. But sooner or later, a BSOD will occur. The memory quickly became the prime suspect. Tweaking the memory settings in the BIOS seems to directly affect the stability.
I know from experience that using 4 DIMMs can be rather troublesome, and sometimes you will not be able to run 4 DIMMs at the rated specs. These DIMMs are extra troublesome since they are early overclocking DIMMs for Core i7. They require 1.65v to operate at their rated speed of 1600 MHz. This voltage is on the upper limit of the Lynnfield specification. 1600 MHz is actually outside Lynnfield specification, since Intel gives 1333 as the maximum speed (although the Asus motherboard claims to be able to go up to 2200 MHz, so the IMC of Lynnfield is probably capable of much more than the Intel specs).
The motherboard manual also warns us:
“Due to Intel spec definition, X.M.P. DIMMs and DDR3-1600 are supported for one DIMM per channel only.”
So strictly taken, this combination is not going to work (even though the memory is sold with a big “Intel Core i7 compatible” sticker on it, specifically aimed at Lynnfield systems). It also appears that we’re not the only ones having problems running this memory. On the Corsair forum there are a number of people who can’t get it running right.
Then again, specs have never stopped us before, have they? Running a system a bit out-of-spec is generally not a problem, so since the system is *almost* stable, can we figure out what to do to get it 100% stable?
I started by just running the memory at a stock setting with the JEDEC 1333 MHz profile. This didn’t require the CPU’s memory controller to run out of spec at all. 1333 MHZ and 1.5v are perfectly fine settings for the CPU. But even with the stock settings of 1333, the system would not be 100% stable. In fact, it didn’t seem to make much of a difference with the 1600 settings. So my guess is that it is a problem with the voltage somewhere. It just doesn’t want to drive 4 DIMMs happily. So I figured I might as well go back to the 1600 settings, and try to get that to work. Now there are basically two options here: the IMC voltage (Vtt) and the DDR3 voltage (Vddq).
I didn’t think the DDR3 voltage would be all that suspect. After all, Corsair already tells you to run 1.65v instead of the standard 1.5v for DDR3. They will probably use a reasonable margin when binning their chips for 1600 MHz at 1.65v, rather than cutting it too close.
The IMC voltage however, that’s another story. The CPU, board and memory are all from the very early Lynnfield era. There are a couple of interesting facts here:
- Although Lynnfield is a close cousin of the first Core i7 series known as Bloomfield, the IMC voltage specifications were much lower. Bloomfield was spec’ed up to 1.4v, where Lynnfield was limited to 1.21v.
- The Intel specifications for Lynnfield have received an update at a later time, now listing the maximum Vtt as 1.4v, as with Bloomfield.
As a result, the IMC in Lynnfield has some rather anemic settings by default. Even if I enable the XMP profile for the memory, my board will only use 1.15v on the IMC, and it considers anything over 1.2v to be in the ‘danger zone’ (the voltage values are displayed in red). At first I wasn’t aware of the spec change to 1.4v, so I thought my motherboard was correct… And you have to be careful with overvolting modern CPUs. They run on very low voltages anyway, and large currents can easily damage them.
Initially I tried pushing the IMC to 1.2v, and pushing the DDR3 to 1.7v. This way both were still reasonably safe. But it did not seem to solve the problem. After some experimenting, I found that the IMC voltage affects stability more than the DDR3 voltage. I found that the DDR3 will run fine at the 1.65v setting, as long as the IMC voltage is high enough. I am currently running the IMC at 1.22v, and the stability seems to be fine. It is slightly outside the original spec, but still well below the updated spec of 1.4v. So I guess that stevevnicks was on the right track on the Corsair forum, when he suggested to set the IMC voltage to 1.25v.
So to conclude: If you are having issues with your Lynnfield, be it with getting it to run stable period, or with getting good overclocking results, the key seems to be in ignoring what your motherboard says about the IMC voltage, and just push it a bit further. I think it is quite safe to go with Intel’s 1.4v figure. Firstly, it is in their official spec sheet. And secondly, Bloomfield is a very similar CPU, also made on the same 45 nm manufacturing process, and it is also rated at 1.4v. In fact, when I browsed some Lynnfield reviews and overclocking sites, I noticed that many of them ran the IMC at voltages around 1.4v or higher anyway. Perhaps they weren’t even aware of the lower spec of Lynnfield, and just applied the same values as they had on Bloomfield. I would not suggest going over 1.4v though, there still is a realistic chance of damaging the CPU. I would suggest to keep it as low as possible.
AMD FAN! No just joking! 🙂
I try to remain neutral 🙂
Intel systems aren’t always without their flaws either. Apparently the Lynnfield memory controller is not very good with 4 DIMMs.
Speaming of AMD. *facepalm at Bulldozer*
In most cases still comparable or slower than a Thuban.
Even a 8150 can’t reach a 2500K performance.
Pingback: AMD Bulldozer, can it get even worse? | Scali's blog
Pingback: Another FreeBSD server upgrade… | Scali's OpenBlog™
Pingback: [Problema] Patriot Viper 3: 3,95GB utlizables de 16GB