Strange limit of kit Corsair D GT-2133-C9, max running @2240 Mhz with each setup.

[sandro_c]

Hello people ,
often on the forum I have seen that many are able to turn memories on the type CMT16GX3M4X2133C9 @ 2400 MHz with overclocking , but more evidence do the more I am convinced that for my PC , there is an insurmountable barrier @ 2240 Mhz on these memories and I can not understand whether the two is due to the IMC of the processor, or the memory itself.
Is there any specific test can That indicated the origin of the causes of this limit ?
In my 24/7 use use these setups described below:
CPU clock gen ........ Auto
Vdram.AB ..............read 1,575 1.56 V manual
Vdram.CD .............. read 1,575 1.59 V manual
CPU VTT V ................ read 1,175 1,181 manual
CPU VCCSA ............read 1,150 1,169 V manual
CPU PLL 1.8V......... read 1,788 manual
VTTDRR auto ................... read 0.75 V
Vcore offset ................. read 0.065 v 0.920 to 1.402
BClock ................. 102
FSBRAM ..... 1:16
Cpu clk ................ multi - 12X 46X read 1225-4648 MHz
digital power control :
cpu LLC .......... medium
Curr cap 130% cpu ....
cpu Pow Duty .... T probe
Vcore MOS volt 7v ....
VCCSA LLC ............ High
VCCSA Curr Ch ... 130 %
VCCSA fix freq ....... 350
VTT sw freq 1.3x ............
Dram abcd Curr Cap .130 %
Dram Pow Phase ....... optimiz .
PCH sw freq ............ 1.3 x
ALL OTHER param .... AUTO
Memory timings are:
primary manual ... 9-11-10-27- 1T-
secondary auto ... / read 4-128-9-4-24-4-4-8
tertiary ...... all Auto but tRWSR is manual -3

I have made many unsuccessful attempts :
Leaving part of the setup as described above ,
I edited VCCSA 1.25V , VTT 1:30 V , Vram 1,625 v, Clockgen.20MF , starting from a multi 38X up to 46x I tried to raise BClock to 101-102,103,104 , then to 105 (@1119 mhz) blocks my PC and I have to turn off to make boot. The temps are still acceptable , but it also blocks in the cold condition, while up to 104 Bclk I can still run 10 steps IntelBT .
Also changing mem. timings in AUTO, readind value of -10-12-11-28-128- 2T ... all of these settings have produced a result of more than 2 -3C ° temperature, but only this.

Nothings to do! If anyone has any better ideas please suggest me.

Hello
sandro c

[Raja]

Sandro, my recommendations are based upon the fact that you have everything running tight - its actually good for performance. If you go up higher by relaxing the timings, you won't see much benefit unless you are able to gain a lot of frequency, which is why I have advised you stick where you are (with the voltages you have right now). Up to you however 🙂

-Raja

[sandro_c]

Hi HallweenWeed, hi Raja,


The thing is getting a little bit dangerous in an attempt to overcome this " strange limit " and more and more, it becomes clear that it is not worth to overtake him at any cost. The OC that I have already achieved is not sublime but not miserable ...... and a long time ago I expressed myself talking about the futility of making critical system for a GFLOPS few more, but the "esploi" of our friend from Dubai has made me forget everything .
I thank my good friend HallweenWeed for supporting me in all attempts and the good Raja for his wise advice, but I'll stop my current results , even down a little on the setup used so far , that I will stay with the

CPU @ 4692 ,
BClock 102-46X
ram @ 2174
VCCSA - VTT 1,155 v
Vcore offset to 0.06
LLC medium
xx.current cap. 130%
idle temps around 31C°
load temps around 68 C°

.... is stable and seems quite secure from more that 8 months of use 24/7 .
With the crisis that we have in Italy , and with the fact that I am retired .... if this PC breaks down as happened to his predecessor three years ago, ..... who knows when you will see me again on this forum.

Thanks again to all of you, and excuse me but my thread stops here.

sandro c .

[HalloweenWeed]

sandro c wrote:
The thing is getting a little bit dangerous...

Sandro I would not recommend anything risky without saying it was risky. My recommendations above are well within the forum accepted ranges. VTT & VCCSA 1.325V is not risky.

[sandro_c]

Hello HalloweenWeed,
I am sure that you never would recommend to me a setup dangerous, but the previous test with the Vram at 1.65 v and VCCSA-VTT at 1.275 that did not lead to improvements, and instead gave a BSOD 0x00009c that I had ever seen, with the consequent next Boot block and an check-disk to recover some bad system files find, made ​​me reflect on the fact that .... I'd better satisfied as they are. Even if I could run @ 2400 Mhz ram, how can I use these tensions to the continuous use of the PC 24/7? I am interested only use 24/7 ... I'm not making a race of OC.
What do you think?
sandro c.

[HalloweenWeed]

sandro c wrote:
Hello HalloweenWeed,
I am sure that you never would recommend to me a setup dangerous, but the previous test with the Vram at 1.65 v and VCCSA-VTT at 1.275 that did not lead to improvements, and instead gave a BSOD 0x00009c that I had ever seen, with the consequent next Boot block and an check-disk to recover some bad system files find, made ​​me reflect on the fact that .... I'd better satisfied as they are. Even if I could run @ 2400 Mhz ram, how can I use these tensions to the continuous use of the PC 24/7? I am interested only use 24/7 ... I'm not making a race of OC.
What do you think?
sandro c.

Sandro: With the last generation CPUs, which are very similar to ours in comparison to the previous generations, it was common for users to set VTT 1.4 - 1.45V for the faster memory kits, of the same type, such as DDR3-1866. Many users found this was necessary for stability at OC; and they never reported, as a concensus, of failures over time due to this high VTT setting. I refer to the famous EVGA forums, to which I am a part of, where I have 4815 posts in just over 4 years, of which I have 38 "blue ribbon" posts, and the vast majority of my posts were helping other users with their problems such as memory/CPU/motherboard. Some people may be concerned that a different CPU generation is too different to compare, but in the case of the IMC I would argue that. The IMC (Integrated Memory Controller) is basically the same as the X58 series CPUs, with a few minor tweaks, IMHO (in my humble opinion).

That said, VCCSA is new beginning W/Sandy bridge (SB) architecture (when only considering Intel enthusiast desktop CPUs, not counting older Xeon models, of which I don't know about). One thing I found out quickly when I first got my newly released RIVE (2011) was that VCCSA needed to be increased to get a decent (respectable) OC from the CPU. One user on one of these forums led me to believe that you raise VCCSA instead of Vcore to get to higher speeds on SB-E CPUs. Later I discovered that he was partially mistaken, but the default VCCSA was still insufficient for most OC >4GHz. If you check the forums reports, there are many users out there that have ran VCCSA at ~1.3V. Not only that, your own memory recommends 1.3V in it's SPD XMP table 1. Do you think they would recommend a setting that caused degradation? I don't. Personally I have been running VCCSA at 1.225V (or a little more at times) all these (>2) years. It is apparent that certain CPUs need more VCCSA to achieve higher DDR3 speeds. I haven't heard of a competent report of degradation from this yet. There was one thread regarding degradation in 2012, but other users did not concur, and I compared it with my settings finding that mine was just about the same, and I had no degradation; so I would ignore that one.


So just to summarize, and be a bit more specific, I would consider any VTT of 1.4V or less to be "normal" and non-degrading, and would guess the same of VCCSA @ 1.35V. There are posts out there, from reputable members, that recommend raising VTT & VCCSA together, keeping them the same as each other. But if higher VTT was necessary, I would cap VCCSA at 1.325V.


What does other people think about this (max. VTT & VCCSA)? I would love to hear your opinions.


Now just so you know, I don't have a problem with you deciding not to carry this issue farther, trying more settings to get your desired memory speeds. It's just that, IMHO, you will need to be less timid (less afraid, for translation purposes) with your VTT & VCCSA settings. Otherwise, I cannot help you. The 9c BSOD 'tends to indicate' a lack of VTT, and indicates the need for higher VTT; although it could be VCCSA and/or Vcore as well (probably not). The reason I stuck to DDR3-1866 when I upgraded my memory was that people (on this forum) were reporting difficulties achieving speeds >2000 even when that speed was advertised/specified for their memory. If you don't post back, I will just assume you don't want to, and that's fine with me too. GL (good luck to you).

[sandro_c]

OK HallowweenWeed,
You say that if I do not answer Consider the topic closed because you can not help me, but far be it from me to remain silent thought.
It is clear that you are very informed about this topic, and with my experience I would not dare contradict you even for a moment, and it is also clear that in me has won the extreme caution in the experiment setup, but that does not mean i do not have confidence in you, in the strongest terms.
It just means that I do not go on like this for more fun .... but only stress for the reasons "social-economic" already explained.
However, apart from this, we assume that whit a VTT_VCCSA of 1,325 v. I finally exceed the rock of 2400 Mhz, ... you advise me to use 24/7 to take this voltage in the future? And if the CPU does not degrade, what will happen to my temperatures idle-voltage full load with this ... as I will keep at bay with my air cooling? Sorry for the delay but I've worked hard to translate well.
hello
sandro c.

[Arne_Saknussemm]

Great thread! as usual sandro c 🙂

I would be another voice to say stay with the tight timings you have at a pretty high frequency...pretty sweet timings I would say. Sure, experiment and try more voltage but I personally would shy away from "high" VCCSA or VTT. especially for 24/7 OC.

The sweet spot is always down from max performance. Performance per volt is where the sweet spot is to be found.

Of course each to their own but 1.1 for VCCSA is a nice number and anything above makes me nervous these days. I have had two CPUs degrade unfortunately...both on the same board. I am half of the mind the board was bad and half of the mind that the 1.25 VTT/VCCSA I was using at higher clocks was to blame...

Anecdotal?...for sure but there is quite a bit of anecdotal evidence pointed this way not to mention Intel's recommended 1.2v VCCSA 1.1 VTT...

[Raja]

That's a good post Arne 🙂

I can't confirm the degradation stuff mind but I do endorse using the most effective point of a CPU to clock it. That is the point before one has to make a huge jump in voltage for little or in some cases no gain (maybe even a performance loss in the case of DRAM).

[HalloweenWeed]

Maybe a 1*C rise in temperature, but that's within margins of error of measurement too; dust buildup is likely to cause more rise in temperature than what may occur from increasing VCCSA. VTT could cause a similar rise in temperature. There is always a chance that some component may fail when putting just a tiny bit more voltage or heat in it, this is the standard risk we all take when overclocking; and DDR3 >1600 is overclocking the CPU IMC. So I am not saying there is no risk. However, I am saying that the risk is so minimal as to be considered highly unlikely to cause a failure. If a failure does occur, it is much more likely that the failure would have occurred anyway, sometime not long after the failure did occur; due to defective component. I know if a failure occurs when we OC something, we always feel like we caused the failure. But in some cases the failure was going to happen soon anyway, we just have no way of knowing that. But you probably already know this, I just need to remind you for my own relief of guilt feeling in the worst-case scenario.

[HalloweenWeed]

Yeah the economy has hit me hard too, and a work loss due to serious illness in the past has been what we call a "double-whammy" here in the USA, meaning something like two hard hits in one. But our components are no longer "new" components, and if one fails a person should be able to get a suitable used replacement pretty cheap on e-bay, even in Italy I imagine.