Asus Sabertooth X79, RipjawsZ F3-12800CL9Q2-32GBZL (4Gx8) BIOS settings?

[powerhouse]

I guess in around 24 hours you could be here

Hot weather is guaranteed.

Later I will check with the repair shop to see whats new and take the PC.

[Tradesman]

Keep us updated

[powerhouse]

BIOS update to 1203 seems to help. The repair shop ran Aida64 for 48 hours with no errors so far.

They had to increase some voltages though. VCCSA is at 1.25V and Vcore (I believe) at 1.65V. The standard XMP settings aren't stable.

I decided to leave the PC there to have them run more comprehensive tests. Once I get it back I will save the BIOS settings and try to reduce them a bit.

I'm not sure that running the system at maximum possible voltages is such a good idea. Any comments on that?

[Tradesman]

Hope the 1.65 is the DRAM voltage or a typo, 1.65 will kill the CPU

[powerhouse]

Hope the 1.65 is the DRAM voltage or a typo, 1.65 will kill the CPU

Sorry, of course it's the DRAM voltage. With so many settings I'm already totally confused.

By the way, are there any definite settings for the MB / CPU / memory combination, or is it often a trial and error kind of approach with 8 DIMMs / 32GB of memory?

[Tradesman]

If the mobo maker has the BIOS right, it should basically be plug and play (install, enable XMP and go), unfortunately most of the mobo makers didn't anticipate that people would jump into using 32 and more of DRAM.....and apparently decided there was no rush to have the BIOS ready

[powerhouse]

OK, here the news from today.

My report from Friday was incorrect. The technician updated the BIOS to 1203 but used the increased voltage settings with XMP, to see if the BIOS upgrade makes a difference (remember, with the 1104 BIOS there were errors after around 17 hours of testing).

After the system ran smooth over the weekend with the increased voltages, the technician reset the CMOS and put everything on default, with XMP enabled. Since yesterday Prime95 runs without errors at default settings and XMP. Today he is going to run Aida64 for a day to see if there are any errors showing up. If not, I will take the PC back and leave the settings at XMP plus default settings.

So far it seems that the new BIOS 1203 has cured the problems, but it's a bit too early to come to final conclusions.

I need to give credit to the technician who's been working on this for 3 weeks now. I paid around $25 for a system check and I think I got my moneys worth back.

I had considered replacing the 8x4GB for 4x8GB G.Skill memory which seems to be less demanding on the MOBO, and got excellent reviews on Newegg. But if the new BIOS solves the problems and doesn't demand extreme voltage settings, I'm happy with what I've got.

If it's indeed a BIOS issue, I wonder how people made the 8x4GB G.Skill DDR3-1600 work in this MOBO/CPU combination? Or was the 1104 BIOS faulty? I never tried another BIOS revision except now using the latest 1203 revision.

It should be noted that using Kingston 16GB did work with the 1104 BIOS, but the technician didn't try running 16GB out of the 32GB G.Skill memory. Any experience on BIOS/CPU/memory mismatches?

[Tradesman]

Keep us updated

[powerhouse]

Here the news: Aida64 passed. However, using the RST Pro board, it showed some errors after a few hours. Prime95 passed, but it also shut down a core after some hours. The technician believes that the core shutdown in Prime95 is a sign of memory problems. I don't know what to make of it.

I need a stable system. I bought it to run Linux with a Xen kernel running a domU (virtual machine) with Windows 7 Ultimate 64 and VGA passthrough. I will also install some guest OSes as domU with remote desktops. The Windows domU will be used for photo processing/editing and it must be solid as a rock. For normal use (Internet, mail, etc.) I want to use the Linux dom0.

This is a setup very few people will have or require, but in my case its the only real solution. I'm fed up with dual boot (Linux and Windows), since the only reason I run Windows is for the photo processing and monitor calibration (I have a high end monitor with internal LUT working in 14 bit color depth). My graphics card also supports 14 (or 16?) bits color depth.

It will be challenging enough to make all that work. I don't need any hardware problems.

What do you think? Will the memory/system be stable? Or should I be concerned about the Prime95 core shutdown and RST errors?

From my experience Linux is more stable or sometimes more forgiving with memory issues than Windows. In windows you get the blue screen. In Linux you get all sorts of errors such as segmentation faults or file system errors that often let you continue to work, but the system is nevertheless screwed up. I also expect some challenges in installing the Linux Xen kernel with the Nvidia driver that I want to use. This means that the hardware must be reliable to eliminate yet further complications.

What do you suggest?

Do you think that the DDR3-1600 4x8GB memory kit will be more stable or easier to setup?

[Tradesman]

Just my opinion, but I think you'll be fine with the rig

[powerhouse]

Ok, I finally returned the PC.

Here is the current setup:

Everything on default, except:

XMP enabled
VT-d enabled
Surge protection disabled

I ran memtest86+ v4.2 today and get memory errors.

Here are my settings in detail:





All other memory settings are "Auto".



Here the voltages (I changed nothing from default):



All the other power settings are on Auto.





CPU performance settings:




What do you suggest? The memory is obviously not stable right now. Which settings need adjustment?

Thanks!

[Tradesman]

I'd try raising the DRAM Ref Cycle Time (tRFC), for 4 8GB sticks, would lean towards 355-278 or so,

[powerhouse]

Hello tradesman, I have 8x4GB sticks, that is the RipjawsZ F3-12800CL9Q2-32GBBZL memory kit. It's a DDR3-1600MHz quad channel, 9-9-9-24 kit.

What do you suggest here?

In the meantime I have tried the following adjustments and still get memory errors:

VCCSA to 1.1, 1.15 and 1.2V with LLC set to Regular
as well as all the memory related settings published here http://www.gskill.us/forum/showthread.php?t=10512.

This includes all the recommendations shown on the print screens, including increasing DRAM AB/CD Current capability to 120% and Optimized power phase control.

IMPORTANT:

No matter what I do, I invariably receive the same error with memtest86+ v4.2, pointing to memory address 00344008e40 - 13376.0 MB. It's always a single bit error, which shows up at test #4-6 (moving patterns or inversions).

Is this a sign of faulty memory?

[powerhouse]

I've given it a try without success. It seems quite obvious that one of the DIMMs is faulty (I always get an error at the same exact address, no matter which BIOS setting or test).

I removed the memory from the PC and am going to return it tomorrow for replacement.

Tradesman: have you got any connection to G.Skill management / worldwide sales? The reason I'm asking is that I have had really bad experiences with the retail chain here that also happens to be the G.Skill distributor. I wanted to replace the memory for 8GB DIMMs (also G.Skill) and pay the extra money but they sent me around in circles.

Also, can you confirm that this looks like a DIMM defect?

[Tradesman]

I'd guess so, if the problem continually shows at the same address, did you try each stick by itself to narrow down to a specific stick or sticks......also check your PMs for one from me, will fire it off shortly