We both posted at the same time, I'm having that exact same issue with Patriot PX7312G2000ELK triple channel memory. I did try the Intel XMP profile in BIOS, but it did not go well. So I'm back to running at the 1Mhz mark, as you described...
Try going into the BIOS and picking the auto XMP profile to see if you have better luck than I.
Thanks for the various answers and suggestions. Kingston is telling me, that these RAMs are not supported on the DX58SO2 ! So it looks like I am out of luck, and spent a lot of money for nothing. It is not clear on their web site that it does not work on that board, but it is taht way.
To keep you posted on what I did... I tried the various profile in the BIOS, and the system hang during the POST, have to force power off, and force BIOS setup to get out. The only configuration that works fine is the 1067MHz (ValuRam at half or a third of that price will work just fine!).
I do not want to go the overclocking route, or fiddle around with all these parameters. I am relying on this system, and need it to be very stable.
Conclusion: do not use these RAMs on this board!!
Thanks to all of you,
Actually I must backtrack - turns out I'm not stable at 1600... after a couple random crashes, I discovered I'm getting rounding errors in Prime95.
The memory I have is supposed to be 1600 7-8-7-20 2T @ 1.65V. These are probably the same chips as in your 2000... considering they sell CL7, CL8 and CL9 at 1600MHz.
It works fine if I set it to auto, which results in 1333 9-9-9-27 @ 1.5V.
I've contacted Kingston for support. Note that I did confirm with them that this memory would work correctly with this board, in 12GB configurations. Perhaps they were wrong.
Hoping there's a compromise setting to get 1600 working.... like 1600 8-8-8-24 or something. I'll try some stuff out later tonight if I don't hear from them.
yes, Kingston send me that same page on their site. But all the 1600 modules are 2 ways, not 3 ways like we need on that board. So if you go by the book, your modules do not seem to be supported.
The issue is the 'wanna go faster' at the bottom of that page that lead you to a buch of expensive RAMs that are not supported !! that really misleading! that's how I got to buy those module that are not supported, and I have 12GB total! I am quite upset, I must say.
So far I have tested 12 hours stable at the following settings (1333MHz):
Uncore Multiplier 20 Memory Multiplier 10 tCL 9 tRCD 9 tRP 9 tRASmin 24 tRFC 74 tRRD 4 tWR 10 tWTR 5 tRTP 5 tRC 32 tFAW 20 Memory Voltage 1.5125 QPI Voltage 1.225 Command Rate Auto
And I have tested 4 hours stable with these settings (1600MHz):
Uncore Multiplier 18 Memory Multiplier 12 tCL 9 tRCD 9 tRP 9 tRASmin 24 tRFC 80 tRRD 5 tWR 12 tWTR 6 tRTP 6 tRC 36 tFAW 24 Memory Voltage 1.6 QPI Voltage 1.225 Command Rate Auto
These are just based on the SPD/XMP profiles and my tweaking. Haven't heard back from Kingston yet (wish they had a user forum).
Thanks for these settings, I will try to have them 'verified' by Kingston, if I can get them to answer my mail, and will let you know.
I am puzle, that the uncore multiplier is higher at 1333 than at 1600? or is it a typing mistake? I thought thse multiplier (uncore and memory) where going the same way.
I'll keep you posted,
That uncore setting got me curious as well - those are accurate, and were the automatic settings for 1333 (2667 uncore) and 1600 (2400 uncore).
I did a bunch of reading on that, and apparantly with the 970 (? forget exactly) and higher it's only necessary to keep uncore at 1.5 times memory speed (lower i7 chips have to be kept at 2 times memory speed). The BIOS has a note on this. A few forum postings seem to recommend 980X owners use 18X, 20X or 22X for 1600MHz memory. I may try 20X after everything else is stable, but the advantage to running lower uncore speed is the board can use lower voltages, and raising it apparantly doesn't add any much if any performance advantage.
So far I've been trying numerous settings combinations. What's odd is this is unstable after 2 minutes of testing:
Uncore Multiplier 18 Memory Multiplier 12 tCL 8 tRCD 8 tRP 8 tRASmin 24 tRFC 88 tRRD 5 tWR 12 tWTR 6 tRTP 6 tRC 36 tFAW 24 Memory Voltage 1.65 QPI Voltage 1.225 Command Rate Auto
But this has been stable for the past two hours of Prime95 stress testing:
Uncore Multiplier 18 Memory Multiplier 12 tCL 9 tRCD 9 tRP 9 tRASmin 24 tRFC 88 tRRD 5 tWR 12 tWTR 6 tRTP 6 tRC 36 tFAW 24 Memory Voltage 1.5125 QPI Voltage 1.225 Command Rate Auto
Note the extra low voltage. I can't quite figure out why 9-9-9 @ 1.5125V is perfectly stable, but the exact same settings with 8-8-8 @ 1.65V is unstable. I'm hoping Kingston can figure out some tweaking of the other settings (tRFC all the way down to tFAW) that will allow CL8 to work.
Well, the best I can figure out so far is that these modules really don't want to run at 1600Mhz faster than CL9 when there's six of them in this board. I've had some initial success running CL7 with just 3 modules (didn't test extensively however it didn't bomb out in Prime95 within the normal amount of time that 12GB does).
Suspiciously, Kingston doesn't even offer faster than CL9 in their stock 6 x 2GB kits. Which is why I emailed them first to see if this would work. I'm beginning to think it won't - at least not any better than the CL9 I have them working at now.
I mean the real-world practical difference between CL7/CL8/CL9 is minimal, but the annoyance here is I was told by Kingston that it would work, I have spent _countless_ hours trying to figure out why it isn't working and what might work, and Kingston's tech support has been very unhelpful.
The initial response was to try the kits 1 at a time and use Memtest86+ to check if the modules were defective (they seem to work fine 3 at a time), the second response was that if they weren't obviously defective to contact Intel for help. I haven't even had a response to my third message where I requested (for the second time) suggested timings to try at CL8 and CL9. Kingston does not publish *complete* timings for their modules, and all I can extract from the modules themselves is what they use for 1333MHz CL9 and 1600MHz CL7. What I wanted was correct tRFC, tRRD, tWR, tWTR, tRTP, tRC & tFAW values for use with these at CL8 and CL9, or for them to at least confirm what what I was using was correct. Google searches show that these can vary wildly depending on the chips used.
Hopefully my 4th message where I detail everything I've tested lately warrants a response.
If I wanted CL9 I could have spent $90 less...