Announcement

Collapse
No announcement yet.

Cannot get XMP to work using DDR4 PC4-24000 (F4-3000C15Q-32GRK)

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #16
    So I started systematically checking my ram sticks and dimm slots to see if any of them were defective. I tested 1 stick with XMP profile DDR4-3000 in the dram channel A with memtest, 8 passes of test #5 and 2 passes of #7 (anything more is just too much time). It passed. Went to channel B with 2 passes of #2, passed. Same for channel D then channel C.

    I did the same for another stick, passed in all the tested slots.

    I tried 2 #5 tests with both sticks in dual channel mode in channel A and C (the way the motherboard recommends), and it passed

    I did the same for a 3rd stick but started in dram channel B because i figure'd i would test the hardest in the channel i intended to put that stick in, and tested the other channels with 2 #5 tests. All passed.

    I put in all 3 tested sticks (#1 in A, #2 in C, #3 in B) and it passed 2 passes of #5.

    I finally get around to the fourth stick. It won't POST with XMP 3000 in any DRAM slot. So I figured I had a bad RAM stick.

    Well, maybe not. I try to go back to the good triple channel config.

    It won't POST at XMP 3000

    WTF!!!

    Now none of the sticks will post at XMP 3000 in ANY slot, even individually. They still work at 2133 mode, but whats the point? I'm using the latest BIOS for my motherboard. A beta one was released last night which is the one I was testing on. I am really getting pissed off at this scenario. Either my CPU is a dud or the RAM is a dud. I have no idea which.

    I need some inspiring words. What can I do to fix this?!

    edit: I reseated the CPU and started at 1 DIMM again and was able to POST at DDR4-3000 XMP. I tried all 4 DIMMS after than and it just goes into failed OC mode (resets to default speeds). I pulled out the extra sticks, then the single DIMM once again would not POST at 3000MHz again. Reset CMOS, booted to defaults, loaded the XMP, and it POSTED. Put in another stick, and have dual channel DDR4-3000 and loaded into OS just fine. I swear.. this thing is loading wrong timings or something. It's really finicky.

    Here are the BIOS settings after loading up XMP 3000 on 2 DIMMs. There are a lot of pictures so here are just the links.

    https://dl.dropboxusercontent.com/u/...napShot_05.png
    https://dl.dropboxusercontent.com/u/...napShot_06.png
    https://dl.dropboxusercontent.com/u/...napShot_07.png
    https://dl.dropboxusercontent.com/u/...napShot_08.png
    https://dl.dropboxusercontent.com/u/...napShot_09.png
    https://dl.dropboxusercontent.com/u/...napShot_10.png
    https://dl.dropboxusercontent.com/u/...napShot_11.png

    edit: 3rd ram stick put in, got triple channel up. Didn't work right away though. I tried the "bad stick" first, posted, but freezes when loading OS/memtest. Tried the good stick, same thing. Tried the good stick in a different ram slot, it allowed memtest to run (no errors) and OS loaded. If I can get this 'bad' stick to test good @ 3000... then bad ram channel, in which, how would I be able to fix that?

    edit: i subbed in the good stick with the 'bad' one and it booted right up. passed memtest no problem, loaded up OS fine. So is my CPU's memory controller bad? All the ram runs fine at default 2133MHz, but its so slow I spent a lot of money on the RAM and the CPU (i7-5960x). Was hoping an extreme edition cpu would at least be capable of this stuff... What can I do to get my ram working at the correct speeds at this point?
    Last edited by ssateneth; 09-29-2014, 09:08 PM.

    Comment


    • #17
      It's a shame that you have to go through this much pain spending that much money. I feel your pain. If I were you, I'd check the CPU first by having the RAM at around 2666 and see how far you can OC your CPU. If you can OC to 4.3Ghz using around 1.28v, your CPU and/or MB may be ok.

      Comment


      • #18
        You need to test the limits of the motherboard and CPU to have a better idea of what it is capable of. I would suggest manually inputting some voltages even though XMP is enabled. Some motherboard BIOS/EFI still need some work so you need to prevent fluctuation and have everything manually input for consistency and accuracy.

        Try stock CPU frequency, then see what the highest DRAM frequency value you can achieve. Then you can back down a notch and see what CPU OC you can achieve. During the process of this, you should be able to get an idea of what Voltage values you need. Take note.

        Keep us posted, we'll be here to help.

        Comment


        • #19
          Yeah I was using near-stock speeds at the time. 3.5GHz core up from 3, just kept it like that. Uncore also was bumped up to 3.5GHz up from 3 since it seemed to benefit memory write speeds a great deal according to AIDA. I was able to tweak 4 channel 2750MHz timings for a while. The only benchmarks I did were AIDA64's memory read/write/copy/latency benchmarks (best out of ~5 times), and running LinX 5 times using 16GB, 8 thread, realtime priority and saving the best time/GFLOPs.

          There are more memory timings in the BIOS than are listed below, but I just was lazy and used the same format from my X79 chipset motherboard. I wish I could set the tREF timing higher to 15.6μs but the BIOS maxes out the timing at 32767 and loops around, so 42900 = 10132, which resulted in slightly slower benchmarks

          Code:
          DDR4-2750	42900 = 15.6μs, 21450 = 7.8μs, 10725 = 3.9μs 3.5Ghz Core/Ring
          CL	tRCD	tRP	tRAS	tRFC	CR	tCCD	tRRD	tWR	tWTR	tRTP	tFAW	tWCL	tAL	tCKE	tREF	BL	Read	Write	Copy	Ltcy	GFLOPs
          15	15	15	35	390	1	4	6	20	4	10	30	14	0	6	10400	8	69196	55566	71187	64.1	386.4040
          15	15	15	35	390	1	4	6	20	4	10	30	14	0	6	32767	8	70804	55784	72278	61.6	389.1870
          15	15	15	35	285	1	4	6	20	4	10	30	14	0	6	32767	8	70915	55851	72385	61.1	389.4681
          14	15	15	35	285	1	4	6	20	4	10	30	14	0	6	32767	8	71039	55858	72569	60.4	390.0160
          14	13	15	35	285	1	4	6	20	4	10	30	14	0	6	32767	8	70698	55869	71262	60.0	389.8454
          The sucky part is I can't run a command rate of 1 anymore. I think my memory controller got fried, or broke it. I only ran "auto" voltage for system agent at the time and there is no sensor for that voltage, so no idea what voltage it actually applied.

          edit: I hate jumping on bandwagons, but I may need to resort to an ASUS motherboard. Can't get my RAM write speeds up without a high ring speed. Maybe their "OC socket" will improve the RAM capability..
          Last edited by ssateneth; 09-30-2014, 05:25 PM.

          Comment


          • #20
            Hardware has been very rushed, so companies are consistently working on updates and improvements. As far as max OC and performance, ASUS and Gigabyte have the most mature firmware for this platform at this time.

            Comment


            • #21
              Bleh. Well I've gone ahead and ordered an asus rampage V. Hopefully I get better results with that. Would like to be able to get the rated speeds out of my RAM stick. But damn, this thing runs HOT.

              The i7-5960x under custom liquid loop (140mm x 4 x 80mm thick radiator with 8 140mm fans in push pull, swiftech apogee hd waterblock, swiftech MCP35X pump, and coollaboratory liquid pro thermal 'paste') shows 80C under linpack stress with 4.4GHz 1.3vcore, and the coolant temperature usually hovers about 25-28C.

              I got a few hangs / thermal shutdowns stressing it at those speeds. I can let place the radiator OUTSIDE in the cold and get coolant temperatures down to about 5C, so CPU only gets to about 60C, but it acts funny under such low idle temperatures (Weird boot hangs, freezes, etc. There hasn't been any condensation since my room is very dry).

              Welp, will post again when I get my new motherboard and post my findings.

              Comment


              • #22
                G.SKILL TECH, after about of week of not touching settings or rebooting, I got around to rebooting to fiddling with settings again. Main target was to see if enabling "SVID communication" would fix the CPU package power sensor to be accurate when under a 125MHz BClk (Currently broken on both MSI and ASUS motherboards)

                This motherboard, or CPU, is crazy.

                Nothing I would do would allow the motherboard to boot under 125MHz BClk. Stock multipliers, reduced multipliers, stock memory, reduced memory, stock volts, higher volts, fiddling with various CPU features, C states, even default settings.. nothing. Wouldn't even let me post at 125MHz BClk.

                I recall CMOS resets would cause the motherboard to act weird too, so I gave that a shot. Held down CMOS reset button for 30 seconds, started up, posted at default settings (100 BClk). Figures what the hell, try the XMP-3000 setting again, and a couple other settings to force constant core speed (not adaptive).

                IT POSTED?!?!?!

                But it wasn't stable. Wouldn't load OS, or load memtest. So I loaded default settings again, then just the XMP-3000 profile, rebooted..

                Memtest rock stable.
                Windows rock stable.

                I just... I don't even. I need some explanation to what is going on here. Is there some sort of training period when powering on? Does it store these settings somewhere? Is it possible to see these or save them? I'm afraid of powering off the PC because I might lose the capability of running 3000MHz memory again. It's like this motherboard/CPU has a mind of it's own.

                I'm going to turn off the PC and back on and see what happens. If it continues to POST, I'll try disabling hyperthreading as I usually do and post back with any results.



                Edit: I guess it's behaving .. for now. The RAM itself is giving out before the memory controller now, as seen in memtest runs. In my experience so far, memory controller malfunctions cause outright freezes/hangs in loading/running memtest, loading OS, etc, whereas malfunctions in the RAM itself has manifested in errors detected in memtest, or BSOD's in windows.

                The RAM itself is running at the states speeds, if not FASTER! I'm tweaking timings as I post this. Have already maxxed the tREF to 32767 (largest bandwidth increaser and latency reducer I've found), then reduced tRFC about 20%, reduced command rate from 2 to 1, and am working on CAS from 15 to 14 now (memtest stable), all at 3000MHz RAM. So it does indeed look like you do get what you pay for. Just make sure you don't have bottlenecks/malfunctions from other hardware.
                Last edited by ssateneth; 10-08-2014, 12:26 PM.

                Comment


                • #23
                  Nice! This is all with the new RVE mobo?

                  Comment


                  • #24
                    No, it was with the MSI X99S XPOWER AC motherboard. At the time of posting this. I am now using an RVE (begrudgingly). Feels kind of bandwagon-ish using an ASUS motherboard, but will post with my results after I've cleaned up the house, unless you wish me to start posting in a different thread.

                    Comment


                    • #25
                      I feel for you all. I think there was numerous motherboard bios issues originally, which have been mostly solved now.
                      I am running the Asus X99E-WS with DDR 3000 no problems at XMP or manual mode. I have also been able to get 4.6GHz the I7-5960x,but dropped down to 4.5GHz for lack of trust for long term stability.

                      Comment

                      Working...
                      X