Announcement

Collapse
No announcement yet.

MSI X58 Platnium & 12gb DDR3/1600??

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

  • MSI X58 Platnium & 12gb DDR3/1600??

    Hello all, seems that my issue is not the only one from what i have seen/read.

    Currrently have a 920 (co)
    Msi X58 Platninum (3.4 bios)
    12gb Gskill F3-12800CL8TU-6GBPI (2 kits)(corsair mem fan/cooler)
    2 4850x2's
    4 160gig sata drives
    Custom watercooling setup, proc/nb/vid cards.

    Currently runnin a 200bclk/19multi (+.180 vcore, 1.6 vram, rest auto)
    setup seems to be fine @ 3/6 mem/uncore Bandwidth is meh.... (8-8-8-21-74)
    When set to 4/8 mem/uncore (1600) (9-9-9-24-74) bandwidth is outstanding,
    cept it freezes up on the everest level 3 cache proc test.

    I have read most of the "other" posts concering this, and take it that i am going to get the "patented" "we only guarantee 6gb @ those speeds" reply, but am hoping for the best here.

    Any help with base timings/sub timings would be appreciated. I'm at a loss, i figured that if i paid for DDR3/1600, it oughta at least run at its rated speeds/timings. I hope i'm not expecting too much from my setup, i could be wrong.....

    Thank's in advance....

    ~Screwtech......

  • #2
    I actually did get this one on not more than some days ago. Never got over it.

    raising the vDIMM voltages had no impact it seemed nor REF, raising board voltages had more impact on this one, but none actually solved this same issue like the board simply wouldn't take it.

    Would love to hear answer to this one too.
    "Sex is like freeware, shareware on weekends. When do we get to open source?" -TwL

    Thanks AMD/ATI for banning legit customers who asks questions of your screw-ups:
    http://i45.tinypic.com/30j0daq.png

    Comment


    • #3
      Yeah, im getting the "proc" is the issue over @ the MSI forums, but i dont believe it. Seems that if a board/memory is rated @ a certain spec, the manufacturers oughta make SURE that it does what it is advertised to do....

      Comment


      • #4
        Well, as I said I have seen many memory sticks here not going to spec, but I have to say that I do not believe this is the case here. I do believe it is the stress caused to the board by memory, but I do not believe it is the memory itself. I honestly believe this has all tight up on boards own voltages failing to deliver enough power to sync calls from CPU > board > (in this point when they are translated to DDR understandable form) here is where things go bad and this doesn't mean memory would be the cause so this one is gonna be hard one to solve. I am here trying to figure it out myself and most probably solution will be when I get this covered that there will be either new BIOS OR I get stable voltages to the BIOS somehow all the suddenly functional.


        You didn't mention me what was your vDIMM voltages? Since I see some affect on those. might just pull it together to add there like +0.02-0.04v on top just for a test.
        Also I am wondering what does the MSI say about board itself voltages (Sorry, I do not own a single MSI board, but same chips anyway and as far it goes what I've seen on net some raises on board might actually pull this off can't suggest any since as said don't own and don't know how MSI voltage drops works).


        (When something in BIOS says value x.xxV you can be damn sure it's not. Reality in metered or, if you check those voltages on OS (much closer to reality) are totally different, if for example our 1.65v on BIOS would by MSI board be 1.58-1.60v that would explain why it doesn't work)
        Last edited by genetix; 07-08-2009, 04:24 PM.
        "Sex is like freeware, shareware on weekends. When do we get to open source?" -TwL

        Thanks AMD/ATI for banning legit customers who asks questions of your screw-ups:
        http://i45.tinypic.com/30j0daq.png

        Comment


        • #5
          Ahhh, current settings for vram is 1.6, i have went 1.65, and 1.7 highest, with no change in issues...

          Forgot to say "thanks" for the posting, i kinda figured that someone from Gskill would acknowledge this by now....
          Last edited by screwtech02; 07-08-2009, 05:58 PM.

          Comment


          • #6
            well, doesn't really matter who 'acknowledge':s this one. Basically as far it goes to G.Skill they support 1 kit being in spec not 2. This is out of the line. 2 kits will need higher latencies for sure that is why I was simply saying 'unofficial' idea what could be wrong and where it goes wrong to make some idea what could work to try.

            doesn't matter if it's 1.7 in your BIOS, if that is not an OS reading it is faulty near all cases. So as I said check the voltages on OS. That could actually help you getting this memory straight'n up.
            "Sex is like freeware, shareware on weekends. When do we get to open source?" -TwL

            Thanks AMD/ATI for banning legit customers who asks questions of your screw-ups:
            http://i45.tinypic.com/30j0daq.png

            Comment


            • #7
              Well, i cannot seem to find the vdimm reading in everest ultimate, and the msi overclocking utility does not list it either.... Any suggestions???


              Well, i downloaded HW monitor, seems my board doesnt support any type of memory votage reading, nice..........
              Last edited by screwtech02; 07-08-2009, 08:21 PM.

              Comment


              • #8
                Check the sensor page of everest it's there called 'vDIMM' and 'vDIMM-VTT'(REF 1/2 of vDIMM voltage). It's there or use CPUID HWMonitor.

                Either way, This is not gonna be easy. You are atm referring, if as 6 blocks of DDR3 at 1600Mhz would be just you check some values and go. I can enlighten you that it is not, if you populate entire controller this means it's around 1.5 times as stressful as using single channel in each controller. the controller straight simply runs out.

                Originally posted by JUST Bad Example from DDR2:
                consideration of DDR2 reference over 4 dimms to 2 dimms would look something like this: base latency for 2 dimms (separate controllers) would be at 1066Mhz around ~1.90ns meaning that when latency 5 would be recommended this would be possible to be build as 10.00ns / 1.90ns == ~5.263t = 5t(on good memory) while 4 dimms setup this would look like 10.00ns / 1.66ns == ~6.02t = 6t(as you can see there is no way in hell this will be POST/error free as 5T). So, as you can see this is exactly same what we are speaking here. However, when we bring voltage calculations in this 4 dimms set where stock 2 dimms at 5T with voltage 2.1v and raise 4 dimms controller voltages up by 2.3 this would make it possible to bring down the ~6.02T to ~5.63T which would in good memory be bootable.
                --

                Also the EPP or XMP is automated to use x1.00 strenght on all controllers. This means it's very very tight squeeze for the controller and latencies increase, is one direction to solve this to slower the DDR3 speed or the actual speed of DDR3 to get even DDR2 on full controller with exact spec would mean large voltage increase, it would mean control over the controller to exact Skew so that it slows down secondary controller slots to get room for tighter latencies. (would look something like: 1.00x on slots 1-3-5 and 0.75x strenght on slots 2-4-6).

                --

                Someone correct me on G.Skill, if I am wrong on above statements.
                Last edited by genetix; 07-10-2009, 03:34 AM.
                "Sex is like freeware, shareware on weekends. When do we get to open source?" -TwL

                Thanks AMD/ATI for banning legit customers who asks questions of your screw-ups:
                http://i45.tinypic.com/30j0daq.png

                Comment


                • #9
                  Only thing listed in the sensor area is:

                  cpu core
                  3.3v
                  +5v
                  +12v
                  +3.3v standby
                  vbat battery

                  Says onboard sensor type is Fintek F71882F (ISA A00h)

                  No mem voltages are listed, tried speedfan also, got a (.29) reading there, i would have to assume its not avail in the onboard sensor...

                  As far as skews go, only 2 avail on my board, cpu/ioh....

                  Thats what i get for trying to keep a budget....

                  Comment


                  • #10
                    Yeah, then there's still problem on sensor detection on Everest perhaps take a look to configs ways to read them. you won't get much luck on SpeedFAN as it simply doesn't understand them, atm, HWMonitor would probably be your best shot (This btw is dependable entirely by Intel SMBus driver. So, make sure you ain't using the v1.0.0.1 they accidentally added to latest driver package and that your SMBus driver is v9.x or v8.x on your device manager.). CPU/IOH aren't gonna help you those are mainly to different sort of balancing, but my point was simply to make understanding of how it is on screen. Hopefully now you got better understanding what is going wrong. Perhaps with little higher tRCD/tRP/tRFC/tWR and some voltage tweaks you can tweak the thing to 1600Mhz near the spec, but like most cases I got on table with out tCL rise it's pretty hard to get truly stable result on full controller.

                    anyway, when you do decide to test it.. Use some procedure like.. raise all to +1T, drop it to MemTest86+ run something like tests 4,6,7 for 20% (prefer 7 the worst of these for quick testing), back to BIOS and drop/raise values so long it doesn't error out. After you think you have found decent solution and back on Windows dump the box to IntelBurnInTest(v2.3 just came out-LinPack), If that still stays success then you know you got rock solid compo to work with.

                    there's my 2 cents.
                    Last edited by genetix; 07-08-2009, 09:26 PM.
                    "Sex is like freeware, shareware on weekends. When do we get to open source?" -TwL

                    Thanks AMD/ATI for banning legit customers who asks questions of your screw-ups:
                    http://i45.tinypic.com/30j0daq.png

                    Comment


                    • #11
                      Thanks man, i'll give it a go. Is this a issue with all the X58 boards/mem, or just my MSI??? Anybody got a clue as to what the secondary timmings oughta be manually set at??

                      Comment


                      • #12
                        This is an issue of the board not specifically MSI, but the way boards are build. 1 single controller has an limited bandwidth in it on 1 stick on an controller has full controller bandwidth & strenght to use while controller populates 2 sticks the streght of controller or the bandwidth won't simply be enough to give both sticks enough power to run on as low latencies. However, tweaking some latencies, and specially tweaking the smaller values will give you as fast result as you would be running on spec latencies. Your memory in other words can take advantage of higher latencies (at tRAS or tRFC) to actually be faster (give room for controller) and give even more speed out of memory than "single side"-setup. (on some occasions like directly sync with CPU/CPI/NB)

                        (o, and might mention that 775 nor 1366 setups doesn't matter. It still won't go with full controller to the spec in any case at these speeds. Just reading from other topic. Think it more of an what your CPU can passthrough at maximum when that is a hit with DDR3 that would be the speed you want to be. here's really good article perhaps worth to take a look what I mean: http://techreport.com/articles.x/15967 )
                        Last edited by genetix; 07-09-2009, 10:39 AM.
                        "Sex is like freeware, shareware on weekends. When do we get to open source?" -TwL

                        Thanks AMD/ATI for banning legit customers who asks questions of your screw-ups:
                        http://i45.tinypic.com/30j0daq.png

                        Comment


                        • #13
                          Wow, good read, and explains alot too.... Now im really thinkin i shoulda gone with a Asus board again, with the "better option" bios.... This MSI board is too limited in the voltage regulation dept it would seem.....

                          Comment


                          • #14
                            Nah, would be enough basically, if you got the low level timings with the basic tCL-tRCD-tRP-tRAS-tRRD-tRFC would help you to tweak fuither on something like tRTR, tWTW and so on. Everything together makes the bandwidth sadly it's not possible to do 'cosmetic' tweaking in all boards which in a sense kills the idea of actually speeding up entire memory.

                            on voltages you would have to go so high before you would reach an level to improve latencies that it would not be something you would want to be running 24/7 although some level increase of course helps.
                            "Sex is like freeware, shareware on weekends. When do we get to open source?" -TwL

                            Thanks AMD/ATI for banning legit customers who asks questions of your screw-ups:
                            http://i45.tinypic.com/30j0daq.png

                            Comment


                            • #15
                              Be nice if someone from Gskill would post the full tiimings needed for this... Got the 9-9-9-24 down, need the rest....

                              Comment

                              Working...
                              X