OCZ1600LV6GK + MSI x58 Pro-E + i7 920

[Vihsadas]

Hello all,
I've recently built a new i7-920 system and I'm having trouble getting it stable at stock speeds. Here are the specs:

i7-920
******* H50 Liq. Cooled
MSI x58 Pro-E (latest bios installed version 8.7)
6B Triple channel kit of 1600LV6GK Platinum (spec'd to run at 7-7-7-24 @ 1.65V)
eVGA GTX 295
Antec TP-750 Watts

Temps are not a problem yet:
CPU: 42C load
IOH: 73C load
System: 35C load

Originally I threw the RAM in the system and immediately started receiving BSODs/errors when trying to install Windows 7 (Triple Channel). I know the ram needs 1.65Volts so I increased that and was able to install windows. However, I am still receiving instabilities (can’t install Unreal Tournament 3), and many errors in prime 95 and Memtest. I then decided to play with some of the settings. While troubleshooting I’ve decided to keep the voltages I’ve listed below as constant and just change timings and ratios, because this board/RAM needs to be able to do at least stock settings at these voltages. As of yet, I haven’t been able to get it to do that.

Settings:
EIST @ Disabled
C-State @ Disabled
Bclk @ 133 (133 – default)/ratio @ 20
DRAM V @ 1.65
QPI @ 1.35
CPU Volts at +.030 (it seems to undervolt slightly at stock (1.192V) so I upped it slightly)
Spread Spectrum off

First set of timings I tried on all three sticks (SINGLE CHANNEL ONE STICK AT A TIME):
7-7-7-24-88-2T @ 800Mhz (DDR3-1600), Uncore @ 3200

Stick 1:
Memtest – errors on test 8
Prime 95 small FFTs In place – passed up to 5 hours
Prime 95 blended – one thread failed in 2-3 mins, a second in 4-5 mins, a third in 10 mins
Unreal tournament Install and unpack fails giving error 1335 (corrupt archive, memory or read error)

Stick 2:
Memtest – massive errors on test 6
Prime 95 small FFTs – passes
Prime 95 blended – all but one thread fails in about 9 mins
Unreal Tournament Install and unpack fails at Data5.cab (USB)

Stick 3:
Memtest – Passed standard tests (once)
Prime 95 small FFTs In place – passes
Prime 95 blended – one thread failed in 4 mins, another in 5 minutes
Unreal tournament Install and unpack fails (corrupt archive, memory or read error)

So, re-tried everything again at:
9-9-9-24-88-2T @ 666 (DDR3-1333), Uncore @ 2666
1 stick at a time, Single Channel!
Voltages unchanged

Stick 1:
Passes memtest (1 pass)
Prime95 small FFTs – seems to be okay
Prime95 Blend – Fails at 3 minutes
Plays unreal tournament but doesn’t seem to crash


Stick 2:
Passes memtest (1 pass)
Prime95 small FFTs – seems to be okay
Prime95 Blend – epically fails after 10 seconds on two or more threads (sigh…)
Unreal Tournament Plays for a bit and then crashes

Stick 3:
Passes memtest (1 pass)
Prime95 small FFTs – seems to be okay
Prime95 Blend – passes for 30 mins
Unreal Tournament properly installs and seems to play...

And finally, I’m running them all right now in triple channel mode at 9-9-9-24-88-2T @ 666, Uncore @2666, 1.65VDIMM 1.35QPI, hyperthreading diabled, everything else the same.
Prime 95 Blend – errors within 3 minutes
Prime 95 small FFTs – seems to be okay
Memtest fails with multiple errors on test 3

In conclusion, none of the sticks can run stock settings in single channel mode. One of them seems okay at 9-9-9-24 at DDR-1333 (below spec on both counts). One cant’ even do that in single channel mode.

I’ve tried reseating the CPU since I know the memory controller is on the CPU, no go. The lever arm that holds the CPU in, however, did need quite a bit of force to put in the CPU, but I don't really think that has anything to do with it (computer installed windows and is running right now).

It seems weird to me that an entire kit is bad, especially because OCZ claims to test everyone by hand. Any ideas? I'm starting to think I should have stuck with my previous choice of manufacturer who shall remain unnamed as per the "profanity" filter lol...

[Vihsadas]

Trying some crazy stuff...

I read somewhere that some RAM doesn't like the 20 mult on the CPU, go figure. So I upped the bclk to 160 and kept the RAM at 1600, decided to be crazy and set the timings to:
7-7-7-28-88-2T.
Triple Channel memtest revealed a recurrent memory error on test 3 at a very specific memory address. There are probably many other errors in the other tests, but this one was easy to isolate.

By switching the modules from DIMM to DIMM in a logical manner, I decided to see if it was a specific memory stick or the DIMM itself. Doing this, I found out the memory address changed only when RAM module 1 was moved. It changes from:

001a4e961f0 to
001a4e96170 to
001a4e961b0

depending on whether Module 1 is in DIMM A, B, or C.
So I'm fairly certain that at least one of the problems I'm having is a faulty stick. Now, let's see if the other two sticks can handle 1600Mhz at a CPU mult of 17. Memtest with the other two sticks in dual channel mode seems to be working @ 7-7-7-28-88-2T. Could the CPU mult really have been part of the problem? I will report back after some Prime95ing and more extensive memory testing.

[Vihsadas]

Even doing dual channel right now with stick 1 taken out of the machine @ 9-9-9-24-88-2T @ 800Mhz (DDR3-1600) I'm still not memtest stable or prime stable. This is at:
160 Bclk
17 CPU mult
5 Ram Mult
10 Uncore mult

QPI is set to auto

1.20 Vcore
1.65 VDIMM
1.35 QPI
1.80 CPU PLL

What should I do?

[SteveOCZ]

Do you have any other memory you could test in the system? This could also be related to a bad CPU/memory controller.

[Vihsadas]

I haven't discounted that fact, but RAM and CPUs are expensive so I don't have any spares. I wish I could have tested that to start with and I may not have had to deal with this. However, because the other CPU tests have been mostly flawless (prime 95 small FFTs), and the problem is modifiable by using different modules I'm still leaning towards the idea that this is partly RAM related...but weirder things have happened.

With stick number 1, I get consistent errors in memtest # 3 at the exact memory addresses above (depending on which DIMM), regardless of the position of the other two sticks, and this particular error disappears when I remove the stick. Isn't that a good sign that at least part of the problem is module related?

I just read something interesting on the internet. I'm going to try something and I'll report back.

[Vihsadas]

I went away for christmas break so I wasn't able to fiddle with this until now. I'm still not getting any stability.

I've reseated my CPU 2x times now, and I've checked for bent pins both times. I found none. I am now trying to release the pressure on the socket bit by bit (quarter screw turns on the heatsink holding the CPU in Place), and I've locked my QPI @ 4.8GT/s.

I don't have another stick of RAM to try, but I do not believe this to be a CPU problem as some of my modules seem to be worse off than others at stock speeds. as a last ditch effort I'm trying 11-11-11-28 @ 1066 with all the recommended voltages...I don't have my hopes up.

[jeffqwer]

I went away for christmas break so I wasn't able to fiddle with this until now. I'm still not getting any stability.

I've reseated my CPU 2x times now, and I've checked for bent pins both times. I found none. I am now trying to release the pressure on the socket bit by bit (quarter screw turns on the heatsink holding the CPU in Place), and I've locked my QPI @ 4.8GT/s.

I don't have another stick of RAM to try, but I do not believe this to be a CPU problem as some of my modules seem to be worse off than others at stock speeds. as a last ditch effort I'm trying 11-11-11-28 @ 1066 with all the recommended voltages...I don't have my hopes up.

try this

1600 9-9-9-28-2T tRFC 99 1.60v all DIMMvref 0.800

forget @7-7-7. It doesn't work on MSI X58 Series

My Settings:

EIST Disabled
C-State Disabled
Bclk 200 /ratio 20
CPU PPL 1.86
DRAM V 1.60
QPI V 1.30
CPU Volts +1.900 (1.248~1.272 in windows)
Spread Spectrum off

[Vihsadas]

try this

1600 9-9-9-28-2T tRFC 99 1.60v all DIMMvref 0.800

forget @7-7-7. It doesn't work on MSI X58 Series

My Settings:

EIST Disabled
C-State Disabled
Bclk 200 /ratio 20
CPU PPL 1.86
DRAM V 1.60
QPI V 1.30
CPU Volts +1.900 (1.248~1.272 in windows)
Spread Spectrum off

Hey thanks for the suggestions. I tried this (and pretty much everytying else) and still no go. Still not prime stable, or memtest stable.

[SteveOCZ]

Can you test one stick at a time for me quickly?

Thanks

[Vihsadas]

Can you test one stick at a time for me quickly?

Thanks

I did that previously.

One of the DIMMs has many, many errors in memtest at 1600 w/ suggested settings, another sticks has less errors, and the third on has a few errors.

This occurs at any speed settings I choose from 1066 to 1600, with all sorts of different timings and such. I have ordered a second kit which is verified working and compatible with the MSI so I'll report back when I get it.

[SteveOCZ]

Ok thanks

[Vihsadas]

******* 6GB set - TR3X6G1600C9 work perfectly...Prime stable after throwing them in the box without much tweaking:

9-9-9-24, 1.62 Volts
1600Mhz,
160 Bus CPU @ 3.2 Ghz

The sticks are either all faulty, or this set is NOT compatible with MSI boards. Now I'm just annoyed because I spent an extra $200 for no reason...

[SimonOCZ]

Sent PM with RMA Instructions, at least you will be sure the modules you get in exchange have been hand-tested.

Simon