Memory Error with Dual Quad Kits, None with Single

Message Topic Search Topic Options Post Reply Create New Topic Printable Version Translate Topic

   Relevant System Details:
--------------------------------
Processor: Intel i7 5820K
Motherboard: ASRock Fatal1ty X99X Killer
Ram:  Two kits of G.Skill Ripjaws-4 F4-2400C15Q-32GRK
Power Supply: Corsair RM850x
Graphics Card: Sapphire Radeon R9 280X 3GB Vapor-X

PCPartPicker part list: 
--------------------------------
http://pcpartpicker.com/p/MTzg8d

General Background
--------------------------------
Please bear with my ignorance wherever evident, I have very little experience in building computer systems. 

I am building a workstation (not for gaming), every component is brand new. I haven't installed an OS yet. Running Memtest86 from the UEFI bootable USB disk provided from their site.

The Processor, Motherboard and Ram Kits are compatible with each other according to their respective QVLs.

There are 8 DIMM Slots in the Motherboard. There are two sets of 4 DIMMS, colour coded Red and Black. The Black ones take one Quad set, and the Red ones another Quad set. The Red ones can't be populated if the Black ones aren't filled first.

I have 2 Quad-Channel kits of 32gb each (4x8gb). When I install either of the two kits  into the Black Slots and run memtest86, I get no errors. No errors with XMP, stock JEDEC and Motherboard Auto settings. No errors with Single, Parallel or RoundRobin processor cycling in Memtest 86.

The Problem
--------------------------------
However, whenever I install both the Kits (in both permutations) Memtest86 fails with Stock JEDEC and Motherboard Default settings. 

With the XMP profile selected, and both Kits installed, the UEFI mostly fails to boot, with the POST error code "bd" or "bF". If it sometimes boots, it crashes, freezes or starts to boot-loop within 10 seconds.

My Observations
--------------------------------
Even though all of the Memtest86 processor modes eventually produce errors within the first pass, the errors are particularly forthcoming in the "Parallel" processor mode, producing thousands within minutes.

From all of the tests results I have seen, I can't discern any persistent Memory Addresses that are related to the errors. 

I have done the tests before and after updating to the latest UEFI Firmware (2.10), there were no changes. I have reset the CMOS (using button, jumper and battery) and there were no changes.

Request for Help
--------------------------------
For those interested, all the MemTest86 error reports and logs (in html, pdf, zip formats) can be found at:  https://drive.google.com/folderview?id=0B96Y5aG-7pwUZ0ZiZTFPNlVNeG8&usp=sharing#list

Given all of the above, and that any MemTest86 errors are not acceptable for my purpose, I would like to know from those with more experience in such matters:

Which is most likely to be at fault, the RAM, Motherboard or the CPU?

What else could be the cause?

What should I do next?

Thank You
--------------------------------
I would be extremely grateful for any help or feedback, and thank you in advance for your time and effort.

Author	Message Topic Search Topic Options Post Reply Create New Topic Printable Version Translate Topic
on&off Members Profile Send Private Message Find Members Posts Add to Buddy List Newbie Joined: 02 Feb 2016 Status: Offline Points: 9	Post Options Post Reply Quote on&off Report Post Thanks(0) Quote Reply Topic: Memory Error with Dual Quad Kits, None with Single Posted: 19 Feb 2016 at 6:10am
	Relevant System Details: -------------------------------- Processor: Intel i7 5820K Motherboard: ASRock Fatal1ty X99X Killer Ram: Two kits of G.Skill Ripjaws-4 F4-2400C15Q-32GRK Power Supply: Corsair RM850x Graphics Card: Sapphire Radeon R9 280X 3GB Vapor-X PCPartPicker part list: -------------------------------- http://pcpartpicker.com/p/MTzg8d General Background -------------------------------- Please bear with my ignorance wherever evident, I have very little experience in building computer systems. I am building a workstation (not for gaming), every component is brand new. I haven't installed an OS yet. Running Memtest86 from the UEFI bootable USB disk provided from their site. The Processor, Motherboard and Ram Kits are compatible with each other according to their respective QVLs. There are 8 DIMM Slots in the Motherboard. There are two sets of 4 DIMMS, colour coded Red and Black. The Black ones take one Quad set, and the Red ones another Quad set. The Red ones can't be populated if the Black ones aren't filled first. I have 2 Quad-Channel kits of 32gb each (4x8gb). When I install either of the two kits into the Black Slots and run memtest86, I get no errors. No errors with XMP, stock JEDEC and Motherboard Auto settings. No errors with Single, Parallel or RoundRobin processor cycling in Memtest 86. The Problem -------------------------------- However, whenever I install both the Kits (in both permutations) Memtest86 fails with Stock JEDEC and Motherboard Default settings. With the XMP profile selected, and both Kits installed, the UEFI mostly fails to boot, with the POST error code "bd" or "bF". If it sometimes boots, it crashes, freezes or starts to boot-loop within 10 seconds. My Observations -------------------------------- Even though all of the Memtest86 processor modes eventually produce errors within the first pass, the errors are particularly forthcoming in the "Parallel" processor mode, producing thousands within minutes. From all of the tests results I have seen, I can't discern any persistent Memory Addresses that are related to the errors. I have done the tests before and after updating to the latest UEFI Firmware (2.10), there were no changes. I have reset the CMOS (using button, jumper and battery) and there were no changes. Request for Help -------------------------------- For those interested, all the MemTest86 error reports and logs (in html, pdf, zip formats) can be found at: https://drive.google.com/folderview?id=0B96Y5aG-7pwUZ0ZiZTFPNlVNeG8&usp=sharing#list Given all of the above, and that any MemTest86 errors are not acceptable for my purpose, I would like to know from those with more experience in such matters: Which is most likely to be at fault, the RAM, Motherboard or the CPU? What else could be the cause? What should I do next? Thank You -------------------------------- I would be extremely grateful for any help or feedback, and thank you in advance for your time and effort.

parsec Members Profile Send Private Message Find Members Posts Add to Buddy List Moderator Group Joined: 04 May 2015 Location: USA Status: Offline Points: 4996	Post Options Post Reply Quote parsec Report Post Thanks(1) Quote Reply Posted: 19 Feb 2016 at 3:00pm
	Some information and a correction and then comments and more information. Your board and CPU supports Quad channel memory. The four memory channels are labeled A, B, C, and D. Each memory channel has two slots, and can accept two memory DIMMs per channel. The memory slots are labeled A1 A2, B1 B2, C1 C2, and D1 D2. Note that EACH memory channel has one red slot and one black slot. For example, slot A1 is black, slot A2 is red. You purchased two, 4 x 8GB kits of memory. Each kit is considered a matched set. The two kits are not matched sets. You installed one matched kit in the black slots, and the other matched set in the red slots. The result is two unmatched DIMMs in each channel. Since both kits tested fine individually, but errors occur when the kits are mixed, that would tend to indicate a mismatch problem. Hopefully we can deal with that. I hope you can tell which DIMMs belong to each set. G.SKILL usually has sequential serial numbers on memory kits, so you should be able to sort them out. Your main re-configuration is to use DIMMs from the same kit in a channel. Whether or not it makes a difference in the sequence of DIMMs kits in each channel (ie Kit1 Kit1 Kit2 Kit2 Kit2 Kit2 Kit1 Kit1) is unknown. Or if you can use a mix of two kits regardless of their location. Personally, I've had good luck using mixed kits of DDR4 memory, from a different manufacture. That is only one instance of that working for me, so cannot be taken as an overall trend IMO. You may get lucky, or not. You seem to not be lucky so far. Intel guarantees your CPU to work with DDR4 memory at speeds up to 2133. So your 2400 memory is considered a memory OC. While DDR4 memory tends to OC well in general, the more channels being used and larger capacities of memory cause the OC capability to diminish quickly. Suggestions for the memory error issue: Reconfigure DIMMs as above. Set memory speed to 2133. Check for any difference between the memory voltage set in the UEFI, and the actual value shown in the HW Monitor screen. Usually the actual voltage is higher than the option setting. You'll want to increase the DRAM voltage when using eight DIMMs. If spec is 1.2V, use 1.3V+. You can always reduce the voltage once the errors disappear, hunting for the optimal level. Besides DRAM voltage, you'll need to increase VCCSA (System Agent) voltage, and VCCIO voltage. Again you can over-shoot with the values and reduce them if the errors disappear. Keep the memory timings and other settings at Auto until you are happy with the memory performance. Then you can attempt to set the timings, etc, to better values, if possible. Only try to increase memory speed, or use the XMP profile once your memory is working well. This is an important change in the memory settings, and should be done and kept that way while doing all of the above: Find a memory option called MRC Fast Boot, normally set to Enabled. Change it to Disabled, which will cause more "memory training" to occur. This can be a big help. You may need to keep that option Disabled. It slows down POST time, but is a minor trade off IMO. If you can't find that option, look for anything that has "training" in the name, and set it to enable training. This is all a fair amount of work and will take time. You may find you are unable to mix the two kits of memory. G.SKILL is usually good about accommodating their customers, you may be able to get a fully matched set for a little more than you spent now. Let us know what happens.


on&off Members Profile Send Private Message Find Members Posts Add to Buddy List Newbie Joined: 02 Feb 2016 Status: Offline Points: 9	Post Options Post Reply Quote on&off Report Post Thanks(0) Quote Reply Posted: 20 Feb 2016 at 6:44am
	Thank you so much for your amazingly elaborate and comprehensive reply. The reason I went with the 2x32gb kit rather than the 1x64gb kit is because of the QVL documentation as presented at both the ASRock and G.Skill websites for the relevant components, which made me conclude (in my inexperience) that the 4x8gb kits were compatible with the motherboard, whereas the 8x8gb kit was not. I find it necessary to point out how I came to that conclusion, in the hope that these WebPages will be updated in the future with more accurate information, to make things clearer for everyone . I am attaching some PDF captures of the mentioned WebPages as they appear now, in case the information is updated in the future. In the ASRock Memory QVL for the X99X Killer MotherBoard the G.Skill 8x8gb kits (e.g. F4-2133C15Q2-64GRR) are not listed, whereas the 4x8gb kits are (e.g. F4-2400C15Q-32GRR). Secondly, according to "Note1: Quad Channel: V = supports 4 modules and 8 modules" the F4-2400C15Q-32GRR supports 8 modules. Since F4-2400C15Q-32GRR is a 4x8gb kit, I logically assumed that it meant 2 kits of 4, especially since the 8x8gb kits were not listed. Site / PDF When I referred to the G.Skill Ram Configurator for this motherboard there were no 8x8gb kits available in the "DDR4 2400MHz" category, but it listed the 4x8gb kits as compatible. Site / PDF When I performed a product search on the G.Skill website for "DDR4", "64gb (8gbx8)", "Quad Channel Kit" the only 2400MHz option was the F4-2400C15Q2-64GRK. Site / PDF The QVL List of F4-2400C15Q2-64GRK, from the search above, doesn't show the ASRock X99X Killer MotherBoard as compatible. Site / PDF Whereas, the QVL List of F4-2400C15Q2-32GRK, which you get as a search result of "DDR4", "32gb (8gbx4)", "Quad Channel Kit", clearly shows the ASRock X99X Killer MotherBoard as compatible. Site / PDF Which led me to conclude that the 8x8gb kits were not compatible with my motherboard, whereas the 4x8gb were. I don't know if the QVL Specification of F4-2400C15Q2-64GRK is an error, oversight or an omission, but it totally skewed my interpretation. I would have definitely gone for the 8x8gb kit had I not misinterpreted the QVLs (or been misled by the absence of certain information) and realised earlier (pardon my ignorance) that the 8 matched kits are always likely to be more compatible. I have already spent quite a bit extra, in trying to get these matched kits rather than regular ones, to avoid the pitfalls of Voltage Adjustments. I don't have the knowledge or experience to get into the subtleties of OverClocking Memory or the time it would take to get it absolutely right. Neither can I afford to have it even slightly unstable, since this is a system meant for serious work. I was hoping to get the requisite help from the XMP profiles which I paid all the more for. I'll try to contact G.Skill support and see if they can accommodate.

on&off Members Profile Send Private Message Find Members Posts Add to Buddy List Newbie Joined: 02 Feb 2016 Status: Offline Points: 9	Post Options Post Reply Quote on&off Report Post Thanks(0) Quote Reply Posted: 25 Feb 2016 at 11:53pm
	I was trying out your suggestions, while I am waiting for a response, but can't seem to find the VCCIO setting. What other names could it be called, under which category? Also, VCCSA directly asks me to input the value, without any indication of the current or default value. Can you please tell me what the default / standard values would be, and what minimum increment I should try at first for both VCCSA and VCCIO (since the input for VCCIO is likely to have the same problem, wherever it's located). Thank you so much for your time and effort.

on&off Members Profile Send Private Message Find Members Posts Add to Buddy List Newbie Joined: 02 Feb 2016 Status: Offline Points: 9	Post Options Post Reply Quote on&off Report Post Thanks(0) Quote Reply Posted: 26 Feb 2016 at 7:27am
	Sorry, figured it out. VCCIO is found at OC Tweaker > Voltage Configuration > CPU I/O Voltage in UEFI Version - X99X Killer P2.10. Default value is 1.050V. Would like to know some sensible initial increment value. VCCSA is found at OC Tweaker > FIVR Configuration > System Agent Voltage Offset. Default value is 'Auto', so cant figure out the actual voltage. Description says Min Offset is -0.300V and Max is 0.600V. Would like to know the actual voltage and some sensible initial offset value.

njakov Members Profile Send Private Message Find Members Posts Add to Buddy List Newbie Joined: 24 Apr 2016 Status: Offline Points: 1	Post Options Post Reply Quote njakov Report Post Thanks(0) Quote Reply Posted: 24 Apr 2016 at 2:36pm
	Had same problem, no luck with pre-defined Asrock OC profiles, got same errors and endless boot loops, chrashes, blue screens etc, etc.. Board: Fatal1ty X99X Killer, CPU 5820K, Ram - 2x 16GB Corsair Venegance 3000Mhz DDR4, Module Name Corsair CMK16GX4M1B3000C15 Corsair is not on memmory support list but it works :) Bios version 1.90 does not even recognize these mem modules, so i had to upgrade to 2.10 borrowed 1 Kingston 8GB module from friend for that Only solution that works for me is: 1. Plug in single memmory module. 2. Activate XMP 2.1 DDR4 Memmory profile (this pushes BCLK to 125Mhz) 3. Set CPU multiplier to 32 *that is 4000Mhz 4. Set Override CPU core voltage to 1.2 Volts 5. Reboot, than test sistem 6. If works plug in extra modules to your board Here are some screenshots, from my system