I7 920 Refuses to OC with 1.25v ram

Dawginole

Honorable
May 9, 2012
8
0
10,510
Hey guys,

I've got a strange issue with my OC'ing. I usually roam around some other forums, but the issue has been so problematic a friend of mine recommended me to Tom's saying you guys can solve it.

I'm not new to OC'ing, but I'm probably still a novice. But so far I've baffled some vets.

Parts list up front:

i7 920 D0
eVGA X58 FTW
G.Skill Sniper 12Gb 1600Mhz 1.25v 9-9-9-24
HD5850, Crucial M4 and Sound Blaster (but these are probably irrelevant to this issue)

When I upgraded my mobo earlier this month I OC'd to 4Ghz fine but kept crashing to blue screen. After a few times it dawned on me I had forgotten that the new ram I purchased was no longer 1.5v but rather 1.25v ram. I cleared my OC and just wanted to get my ram timings, frequency, etc down stable before I proceeded. Set my frequency, timings, uncore, voltage and kicked the CPU VTT up 100mV to 1.2V total to get stable. I ran memtest86 and some other burn in tests and left it be. Not a single crash since then, and the puter has been stable for 3 weeks.

Yesterday I had some free time so I started attempting to OC again. And no matter what I try I simply cannot get an OC. What's more I can't even boot the machine. It won't get past POST to even get a basic OC to start fine tuning. I simply have a black screen and the machine seems to hang at start up. For whatever reason, it seems like this ram is extremely finicky about my OC.

I've exhausted my knowledge to this point and would really like to get back to 4Ghz as stock clocks seem to cause my multi-tasking, especially while gaming to take a huge performance hit.

Wondering if anyone has any experience with OC'ing with 1.25v ram and could help me at least get a basic OC going to where I could start fine tuning it.

Appreciate your feedback!
 

al360ex

Distinguished
Oct 15, 2010
300
0
18,810
Hello!

I think we have the same RAM.

I currently have this RAM installed in my PC and I've set my voltage at 1.20 in the BIOS and have no problem with obtaining a 4.2GHz overclock with my i7 920 D0 :

54207491.png


I can run IBT for 6 hours on Maximum with no crash (I've not tested longer).

I don't know if you can pull something from my specs, but it probably can't hurt !

Also, for overclocking a 1366 rig, I've always followed this guide. It's never failed me before.

EDIT : Is your ram physically set up in the slots 0, 2 and 4 ?

I hope this helped,
al360ex
 

Dawginole

Honorable
May 9, 2012
8
0
10,510
Alex,
This guide really provides a good example of my computers struggle to OC.

For instance in following this guide, I'm already at 1.4v VTT at only a 185 CPU frequency which is roughly puts me at a 2.7Ghz clock. This is the kinda voltage I would expect moving toward a 4.2-4.5Ghz OC. I would expect somewhere around 1.35v VTT at a 4Ghz OC, and I'm no where near this.

I moved from a eVGA X58 3x SLI to the eVGA X58 FTW3 to get SATAIII and USB3.0 put I'm wondering if this board is just a piece of shiznat or a lemon. :/
 

al360ex

Distinguished
Oct 15, 2010
300
0
18,810


And what is your memory frequency ? Have you manually lowered its multiplier ?

If it is set to a low frequency, like 800-1000MHz and you still have a high VTT voltage, then my guess is that you are very unlucky : You've got a very bad CPU for overclocking. It's rare, especially for a 920 D0; they usually overclock very well past 4GHz. You'll simply have to content yourself with a more moderate overclock.

Also, the motherboard is probably not the problem here, because the VTT voltage is the one for the IMC, which is located on the CPU itself, not on the motherboard.

Good luck,
 

Dawginole

Honorable
May 9, 2012
8
0
10,510



See that makes no sense though. I've had this CPU for 2.5 years. And prior to when I moved to the eVGA x58 FTW3 mobo a few weeks back my machine was OC'd at 4Ghz on the eVGA x58 SLI for all but the last 3 weeks of that 2.5 year period.

Memory frequency is at 800Mhz right now, and while this guide (just as a default reference point) doesn't want you messing with mem voltage untill you get a stable bclock I was forced to go ahead and move the mem voltage to 1.25V from Auto (1.5V) for it to remain stable.
 

al360ex

Distinguished
Oct 15, 2010
300
0
18,810
Ok, then have you tried another set of RAM 12Gb (3x4Gb) that you might have lying around ? Also, can you run the G.Skill memory it at its advertised speed ?
Have you flashed your BIOS to its latest version ?
Have you tried your new memory on another system to see if it works ?
Could your new motherboard be defective ? Is it compatible with 4GB density memory modules ?

If you tried putting everything back into the eVGA x58 SLI motherboard, can you overclock your CPU and memory ?

At this point, you have no choice to test your components independently to see which one is defective.
Also, I've never really liked EVGA's motherboards (except for the SR-2/SR-3). I prefer to trust Asus or MSI. They've never failed me before. But that just a personal preference.

If I were you, I'd try to put another set of memory in your motherboard if you have any lying around and try overclocking. If it doesn't work, you know your problem lies either with your CPU or your mobo. Since we can rule out the former, because you know you can reach 4GHz, you know it's your mobo that's the problem.

 

Dawginole

Honorable
May 9, 2012
8
0
10,510
Ok, let's roll back a step. I had something very confused in that guide based on the differences in language the board manufacturers use.

So here's where I've gotten after discovering that...

CPU ratio x15
CPU uncore x20
Merory x8
IOH 1.15

Last stable IntelBurn occurred at:
CPU freq x173
CPU VTT +50mV


After this test I cannot pass anything. I have tried to account for bclock holes skipping around up to x193 and the need of more VTT up to +200mV and cannot even get into Windows past x175 cpu freq and +100mV (which gets into Windows but fails the test).

The failure is always the same, a 0x00000124 BSOD code, which is a hardware failure. My understanding is that hardware failure points to too low of a CPU VCore. Following this guide, it is currently still on Auto which has it set at 1.0375V. Seems awfully low, but testing up to 1.36V has not stopped the 0x00000124 error yet.


Even though I've hit a road block I feel like I'm back on track somewhat compared to where I was when I first started this thread. If we could determine the cause of the BSOD code, which I believe to be voltage related then I think I would be back on track testing overclocks.

:bounce:
 

al360ex

Distinguished
Oct 15, 2010
300
0
18,810
Ok, first, have you disabled all the power-saving options on your motherboard ? That could hinder overclocking. I think in the guide I showed you, it says which ones to disable : all C-states options, down-throttling, SpeedStep (“EIST”) and etc.

When you say "CPU VTT +50mV", is it +50mV from the recommended 1.2V in the guide, so in fact 1.25V ? Or is it +50mV from the default 1.15V, so in fact 1.2V ?

The guide also says to put your memory multiplier at x6 to isolate the bclk from the memory. Yours is at x8. That could be the source of your problem. Eventually, you'll put it back at x8, but then you'll already have determined your max stable bclk, so you'll only have to play with your RAM voltage to test how much your memory modules need to run at 1600MHz, which should be at around 1.25v. Mine need only 1.2v

Your Uncore Ration should be set at x16, since your goal is to reach a bclk of 200 (200MHzx16=3600MHz which is double your rated memory frequency of 1600MHz). With Bloomfield, the Uncore ratio must always be set double the memory frequency, so double your desired SPD (which is 8). That puts your Uncore ratio at x16.

From the information you gave me, it looks like you tried to rush your overclocking a bit. That's a very common mistake to make. Everyone does it sooner or later. Overclocking takes time and LOTS of patience. Take your time and follow each step carefully.

EDIT : Also, I've just thought about something...sometimes, you cannot achieve the same bclk with a different motherboard, because of higher quality components/design or whatnot.
Maybe with your new one a bclk of 173 is the maximum you can reach. With a multiplier of 21, that would give you a max overclock of 3.6GHz.
 

Dawginole

Honorable
May 9, 2012
8
0
10,510
I have speedstep off an vdrop off. Not sure what the others you mention are.

Last passed test was at x177 freq @ +100mV (1.25V=1.15+.1).

I do not have the option of a x6 memory multiplier just 2:8.

Uncore is set to x20 (option #2 in the guide) due to working toward a 4Ghz overclock. x16 was the option to use if shooting for a 3.6Ghz overclock.

This board is one of eVGA's better boards. Lots of 4.2-4.6Ghz OC's on their forums.

In fact, when I installed this board, I didn't even mess with the voltages. Just went in and set it up for 21x191 and ran it at 4.2 for a week fine. But was getting memory failure on occasion. Like I was saying in the OP, that was when I put my clocks to stock, set my proper settings for the memory and never had another failure. Haven't been able to repeat that. It's been super nit picky since I reset that first OC on this board.
 

al360ex

Distinguished
Oct 15, 2010
300
0
18,810


The Uncore ratio has nothing to do with the CPU overclocking. In the guide, the Sample #2 is for a 4.2GHz CPU overclock with a 2000MHz memory overclock. Your memory is rated at 1600MHz and there is absolutely no room for overclocking with it. I know because I've tried (we have the same RAM). The best you can do is tighten the timings once everything else is done.



Because your Uncore (The System Memory Multiplier or SPD) is set at x20 you don't have the option to use a 2:6 ratio for the IMC (Integrated Memory Controller). Put it at x16 and you'll see the option appear for a 2:6 ratio.

You'll have to trust me on that : the Uncore multiplier is for the memory frequency.

I'll repeat what I've said in my previous post and what the guide says :

To reach your maximal Base Clock you need to isolate everything from it, in other words down-clock everything so that it does not become a limiting factor.
That's why you decrease your CPU multiplier at x15. That's why you have to decrease the IMC ratio to x6. Once you've reached your highest possible Base Clock, you'll put your CPU multiplier back up to 21 and the IMC at x8 (because you want to run your memory at 1600MHz, not 2000MHz like in the sample #2 in the guide).

Memory Frequency = Base Clock x Integrated Memory Controller (IMC) ratio
1600MHz = 200MHz x 8

The Uncore Frequency must always be double that of the Memory Frequency for Bloomfield.
Uncore Frequency = Base Clock x System Memory Multiplier (SPD)
3200MHZ = 200MHz x 16

That also explains your 2:8 ratio : 200MHz*8*2=3200MHz

So these are the values you should have :

Base clock : 133MHz
CPU Multiplier : x15
Uncore Clock Ratio : x16
Memory ratio (IMC) : x6
QPI/VTT voltage : 1.20v
IOH voltage : 1.15v

Save these settings and restart your PC.

In the guide, you are now at the step : "First overclock!" on page 3.

Remember : For the CPU settings (Base Clock and CPU multiplier) you must follow the Sample #2 and for the Memory settings (Memory and Uncore ratios) you must follow the Sample #1.
That's because your goal for the CPU is 4.2GHZ (Sample #2 200x21) and not 3.6GHz (Sample #1 180*20).
Your goal for the Memory is 1600MHz and not 2000MHz (Sample #2) or 1440MHz (Sample #1). But you'll still use the settings from the Sample #1 because your Base Clock is 20MHz higher, so 20Mhz x 8 = 160 MHz. 1600MHz - 160MHz = 1440MHz (Sample #1).

When you overclock, it's very important to know exactly what each setting does, so that you don't follow blindly a tutorial, but rather can adapt it to your needs. Currently, the frequency you want to reach with your Memory (1600MHz) is not *specifically* in the guide.
That's why I've been telling you to use the settings from the Sample #1 for the Uncore and Memory ratios.

Do you understand what I've been trying to tell you ? If not, tell me which part you don't understand and I'll try to explain it better.

 

Dawginole

Honorable
May 9, 2012
8
0
10,510
I misremembered and was away from my machine when I posted. My memory multiplier is 2:6.

Now I see what you're saying about the samples and using sample #1 for memory.

Base clock : 133MHz
CPU Multiplier : x15
Uncore Clock Ratio : x16
Memory ratio (IMC) : x6
QPI/VTT voltage : 1.20v
IOH voltage : 1.15v

Will start testing again from here.
 

Dawginole

Honorable
May 9, 2012
8
0
10,510
Stable at 4.2Ghz. Thanks for the help!

Strangely enough my final settings were almost exactly what I had manually put in the first time based on the testing from my other x58 board. But I think I had been coming up short about .025V on the Vcore which is why I was having the boot problems.

But no matter, everything is good to go now and the H100 never even let temps crack over 60C while stress testing and I'm sitting on about 31C idle. Beautiful!

Thanks again!
 

TRENDING THREADS