X399 Fatal1ty Memory reserved issue
Printed From: ASRock.com
Category: Technical Support
Forum Name: AMD Motherboards
Forum Description: Question about ASRock AMD motherboards
URL: https://forum.asrock.com/forum_posts.asp?TID=6082
Printed Date: 27 Dec 2024 at 3:57pm Software Version: Web Wiz Forums 12.04 - http://www.webwizforums.com
Topic: X399 Fatal1ty Memory reserved issue
Posted By: geek2_21
Subject: X399 Fatal1ty Memory reserved issue
Date Posted: 15 Sep 2017 at 2:27am
" rel="nofollow - Hey, I've just installed Window 10 PRO 64Bit on my new setup and I noticed under system I only have 23.9GB usable. I believe my memory to be good so I'm think its a memory remapping issue but I can't find the setting in the bios. Any ideas??? Thanks guys
Specs - Threadripper 1900x ASRock X399 Fatal1ty Professional Gaming TR4 ATX Motherboard Corsair 32GB (4x8GB) CMR32GX4M4A2666C16 DDR4 2666MHz Vengeance RGB 2 x Geforce 1080 Samsung 960 EVO 1TB M.2 SSD
|
Replies:
Posted By: datonyb
Date Posted: 15 Sep 2017 at 2:35am
can you try 1 stick at a time
it seems weird its exactly the total of 3 sticks
either that or one stick isnt correctly pushed down enough in slot (common problem on the new am4 ram slots)
------------- [url=https://valid.x86.fr/jpg250][/url]
3800X, powercolor reddevil vega64, gskill tridentz3866, taichix370, evga750watt gold
|
Posted By: MisterJ
Date Posted: 15 Sep 2017 at 8:34am
" rel="nofollow - geek2_21 and datonyb, Please look at the memory tab on CPU-Z. I suspect it will show 3 channels. What slots are you using? Should be A2, B2, C2 and D2. CPU-Z should also tell you where (slots) the memory is being detected. I suspect (a lot of things, really) that a stick or a slot is bad or a stick is not fully seated as datonyb asked. Good luck and enjoy, John.
------------- Fat1 X399 Pro Gaming, TR 1950X, RAID0 3xSamsung SSD 960 EVO, G.SKILL FlareX F4-3200C14Q-32GFX, Win 10 x64 Pro, Enermx Platimax 850, Enermx Liqtech TR4 CPU Cooler, Radeon RX580, BIOS 2.00, 2xHDDs WD
|
Posted By: Kyuu
Date Posted: 28 Oct 2017 at 4:47pm
" rel="nofollow - I have a similar issue on the X399 Taichi. 4x16 GB installed (Kingston ECC, KVR21E15D8K2/32 kits) but both Linux and Windows only report 48 GB available. If I install only one kit, 2x16 GB, OS reports 16 GB available and the rest as reserved. BIOS shows all memory and for example dmidecode lists all 4 memory modules just fine. BIOS is 1.70.
I have another X399 Taichi with 8x16 GB similar modules installed with no problems, OS gets the full 128 GB.
Could add-on cards factor into this? Something similar happened years ago with some Intel setups. The 128 GB one doesn't have any SATA devices for example, only one M.2 drive. Both systems have a rather legacy Quadro display controller (just need basic console) and an Intel X520 10GB adapter.
|
Posted By: MisterJ
Date Posted: 28 Oct 2017 at 10:47pm
" rel="nofollow - Kyuu, more detail on your specifications always helps. I also suggest you post them in your signature as I have. I think add-on cards and SATA should make no real difference but you did not tell us what you power supply is. If it is skimpy, perhaps 12 volt buss is dropping. What memory slots are you using? Please try A2, B2, C2 and D2 and A1, B1, C1 and D1. You can also try one module in each slot and each module alone. Please let us know what you learn. Lots of people here have been trying to get 128 GB to work - please post the make/model of your memory that does work. Enjoy, John.
------------- Fat1 X399 Pro Gaming, TR 1950X, RAID0 3xSamsung SSD 960 EVO, G.SKILL FlareX F4-3200C14Q-32GFX, Win 10 x64 Pro, Enermx Platimax 850, Enermx Liqtech TR4 CPU Cooler, Radeon RX580, BIOS 2.00, 2xHDDs WD
|
Posted By: Kyuu
Date Posted: 29 Oct 2017 at 1:36am
Apologies @MisterJ and thank you for your tips. I will provide further details on both cases. I don't believe using the signature works for me as I have about 20 systems, five of them Threadrippers. I use these as small scale servers for testing environments.
I've used solely Super Flower power supplies recently, as they've proven to have an excellent track record, and it's one of the few (three or so) brands where we know exactly where they're built as well. Specifically, I have Leadex II Gold 650W fully modular models in use. I've measured the power supplies before installation and I don't believe that's the issue, but of course I can try another one. I have a few Seasonic G-series ones and a couple Corsair AX series models, but I really prefer Super Flower nowadays.
I will try changing to different memory slots on Monday once I get it out of the rack. I'm using the recommended slots according to the manual as of now.
As for the 128 GB system, here is a rundown: - X399 Taichi, BIOS 1.70 - TR 1920X - Memory: 4x Kingston 32 GB ECC kit, type KVR21E15D8K2/32 - Super Flower Leadex II Gold 650W - 1x Samsung 960 EVO 250 GB - Nvidia Quadro 600 - Intel X520 10G ethernet adapter (my storage is via iSCSI) - HEC 4U rack case with several case coolers, the CPU cooler is a Noctua NH-U9 TR4-SP3 - OS: Windows Server 2016 with Hyper-V - I also ran a test session with Debian Stretch w/latest backports kernel and no problems there either
The only major BIOS settings change is that IOMMU is disabled, because no recent Windows version will boot with that on.
I'm going to build one Threadripper setup here at our "tweaking table" early next week so I'll have a better chance to try different combinations quickly. I'll keep you posted.
I've yet to build a single workstation setup, though that's something I'll probably get to later this year. I've found the Threadripper to be much more useful in small scale server work. Earlier, I used high end X99 motherboards with E5-2600v4 Xeon CPU's but Intel killed the option of using Xeon's and ECC memory in non-server boards so I needed a replacement. I need a hardware solution that works as both a server and workstation, because that way all the workstations double as warm hardware spares for lower tier servers. I do hope AMD gets this right, because I've lost a lot of faith in Intel due to far too many stabs in the back. If these five test setups work, there will be ten more before long. Should all go well, I'm also going to test TR's with graphical design workstations later on.
|
Posted By: MisterJ
Date Posted: 29 Oct 2017 at 2:17am
Thanks, Kyuu. My first R7 1800X board had two bad slots on the same channel. The slot suggestions is to try to determine if you have a bad slot or stick. You could also try a new power supply. I assume you are running at SPD specifications. If not open your BIOS and hit F9 (Load UEFI Defaults) then F10 (Save and Exit), turn off IOMMU, if needed . W10 will not boot without IOMMC disable if below version 1703 - don't know about Server versions. Thanks for the 128 GB information. Enjoy, John.
------------- Fat1 X399 Pro Gaming, TR 1950X, RAID0 3xSamsung SSD 960 EVO, G.SKILL FlareX F4-3200C14Q-32GFX, Win 10 x64 Pro, Enermx Platimax 850, Enermx Liqtech TR4 CPU Cooler, Radeon RX580, BIOS 2.00, 2xHDDs WD
|
Posted By: parsec
Date Posted: 29 Oct 2017 at 3:49am
" rel="nofollow -
Kyuu wrote:
I have a similar issue on the X399 Taichi. 4x16 GB installed (Kingston ECC, KVR21E15D8K2/32 kits) but both Linux and Windows only report 48 GB available. If I install only one kit, 2x16 GB, OS reports 16 GB available and the rest as reserved. BIOS shows all memory and for example dmidecode lists all 4 memory modules just fine. BIOS is 1.70.
I have another X399 Taichi with 8x16 GB similar modules installed with no problems, OS gets the full 128 GB.
Could add-on cards factor into this? Something similar happened years ago with some Intel setups. The 128 GB one doesn't have any SATA devices for example, only one M.2 drive. Both systems have a rather legacy Quadro display controller (just need basic console) and an Intel X520 10GB adapter. |
So the memory is not exactly the same model in both boards? That seems to be a factor. I know it would be a pain to swap the memory models between boards, but that would be a good test.
If all the memory is seen in the UEFI/BIOS and via software as you said, but not in Windows, that reminds me of a (not AICard) memory problem seen in Intel HEDT systems. Usually that was fixed by tweaking memory related voltages, but could sometimes be difficult to fix.
This also happens when mixing two kits of memory. If you can identify which DIMMs came from each kit, don't mix the DIMMs from each kit in a single channel. Keep all DIMMs from one kit in one channel only.
Try using the memory at stock speed if not at that speed now. Clear the UEFI/BIOS if you haven't done that. Increase the DRAM voltage by 0.5V to start. The SOC voltage should get a boost too, it's 0.900V by default on Ryzen systems, not sure what it is in X399. Try increasing SOC voltage by 0.1 - 0.2V.
------------- http://valid.x86.fr/48rujh" rel="nofollow">
|
Posted By: Kyuu
Date Posted: 30 Oct 2017 at 10:35pm
I can do some swapping and PSU tests. We've scheduled to build a test setup later Tuesday (GMT+2). I haven't made any changes to memory settings, so they're running defaults. ECC isn't the fastest solution, but it does help me sleep at night. Thus, memory speed doesn't really fit onto my priority list.
I will keep you updated.
------------- Evaluating X399 + Threadripper systems for light server use.
|
Posted By: Kyuu
Date Posted: 01 Nov 2017 at 1:16am
So we put a test setup here. Similar build to yesterday's list.
The first build was without BIOS update, no storage and no additional network card, but with memories in the slots the manual recommends, A2, B2, C2 and D2. BIOS fully on defaults except IOMMU disabled. Booted Windows Server 2016 from a single SATA SSD, 64 GB okay, no reserved memory problems. Then we installed the 10GB network card (Intel X520), still no problem with reserved memory. BIOS update to 1.70, also no problems.
At this point our setup was 1:1 identical with our problematic one without trouble. So we pulled the problematic one from the rack and swapped memory modules. Still all okay, so not a fault with the memory either.
We moved to re-test the problematic system, tried it with several sets of memory, still 16 GB reserved, no matter if we have 32GB or 64GB installed. We started swapping memory modules to different slots. Eventually we ended up trying with a single memory module. Error code C0 comes up when the single module is installed in C1 or C2 and the system won't boot. The single module works fine in any other slot.
So I figured we have a faulty CPU (memory controller) or motherboard. So, we changed the CPU. The newer, originally working system still worked with the CPU from the faulty setup. And to make it more fun - after this operation, the faulty system stopped being faulty as well.
So, very long story short: reseating the CPU fixed the problem.
------------- Evaluating X399 + Threadripper systems for light server use.
|
Posted By: MisterJ
Date Posted: 01 Nov 2017 at 3:42am
Thanks, Kyuu for keeping us informed. Enjoy, John.
------------- Fat1 X399 Pro Gaming, TR 1950X, RAID0 3xSamsung SSD 960 EVO, G.SKILL FlareX F4-3200C14Q-32GFX, Win 10 x64 Pro, Enermx Platimax 850, Enermx Liqtech TR4 CPU Cooler, Radeon RX580, BIOS 2.00, 2xHDDs WD
|
|