Gigabyte MJ11-EC1 EPYC 3151 Mystery

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

orbis

New Member
Oct 7, 2023
2
1
1
Hello.
I'm trying to build NAS.
Can you clarify me in some things please.

The board has
4 sata connectors and
SlimSAS SFF-8643 and
u2_2 SFF-8654 8i ?


Can I connect 4 disks using sata connectors and 4 more in SlimSAS SFF-8643 (SFF-8643 to 4 SATA)?
Can use the u2_2 has an a pci express with the refered adapter SlimSAS 8i (SFF-8654) to PCIe x8 Slot Adapter to connect a 10Gb NIC?
 
Last edited:

jnrnbt.

New Member
Sep 21, 2023
16
9
3
Germany
Hey, yeah, that should be possible. Maybe you want to use my combination of pcie adapter and cable, x4 should be enough for 10gbe and you just need a sata power cable.


Meanwhile I tried to use my dual x4 cable with another pcie board (bought another one because i was curious), but that won’t work.
 

mstoebich

New Member
Mar 26, 2017
20
6
3
28
There are also 10gbe NICs on m.2 boards (Amazon). I found one thats ~120€, maybe there are even cheaper models, these seem to all use the same AQC-107 chipset. A quick google search also suggests it's probably supported out of the box by esxi, but i didn't have the time to check yet.
 
Last edited:

iridial

Member
Oct 14, 2023
39
27
18
I got one of these mb as well, the BMC interface shows that comes from a G431-MM0-OT.
BIOS version is F09, released on 09/14/2021, did anyone find a newer version? in GIGABYTE site there is only an older version (F02)..

Anyway, I did some tests with a power meter using the following setup:
PicoPSU, a sata ssd, changed the included fan with a less noisy one, single 16gb ddr4 rdimm (M393A2K43CB2-CTD), arch linux using 6.5.7-zen1-1-zen kernel.

The cpu driver defaults to acpi-cpufreq, since this is a zen 1 cpu amd-pstate is not supported, so there are 3 available freqs (cpupower frequency-info): P0= 2700MHz, P1= 1900MHz, P2= 1200MHz
plus the boost up to 2.9 GHz. The available idle states are: POLL, C1, C2 (cpupower idle-info)

here are the results:

power OFF – BMC ON

w/ mngmt eth + 1 intel i210 connected: 5.7W
w/o any eth connected: 4.9W

power ON
cpu governor: schedutil – no powertop
  • idle: 20.5W
  • stress -c2: 30.1W
  • stress -c4: 36.7W

cpu governor: powersave – powertop --autotune
  • idle: 20.2W
  • stress -c2: 24.3W
  • stress -c4: 25.4W

In BIOS setting advanced -> amd cbs -> NBIO -> cTDP to Manual and cTDP value from 5 to 20 did not reduce power consuption in my tests. I didn't see other BIOS settings that could help decrease power usage, other than turning off cpu cores (didn't try).
I also tried to mess with the power states manually using amdctl, but couldn't figure out the right parameters to use in order to lower power consuption.
output of amdctl -g -m:

Core 0 | P-State Limits (non-turbo): Highest: 0 ; Lowest 2 | Current P-State: 0
Pstate Status CpuFid CpuDid CpuVid CpuMult CpuFreq CpuVolt IddVal IddDiv CpuCurr CpuPower
0 1 108 8 96 27.00x 2700.00MHz 950mV 27 10 37.00A 35.15W
1 1 95 10 104 19.00x 1900.00MHz 900mV 19 10 29.00A 26.10W
2 1 96 16 108 12.00x 1200.00MHz 875mV 12 10 22.00A 19.25W


Using an atx psu increases power consumption by 4-5W. Another 2-3W increase while using the BMC html5 KVM.

If anyone knows some tools/settings to make it sip less power, let me know :)
 

TAK

New Member
Oct 14, 2023
1
1
3
I have had my board for a month and it works fine. I use SK Hynix HMA84GR7CJR4N-VK which clocks at 2133 instead of 2400 because I use four of them and only one or two dimm can run at 2400 as the manual stated. The 128GB ECC RAM works fine but my CPU configuration is strange. As you can see in the following output it shows 4 cores with 1 thread per core which result in 4 logical processes instead of the stated 8.
Bash:
root@homeserver:~# lscpu
Architecture:            x86_64
  CPU op-mode(s):        32-bit, 64-bit
  Address sizes:         43 bits physical, 48 bits virtual
  Byte Order:            Little Endian
CPU(s):                  4
  On-line CPU(s) list:   0-3
Vendor ID:               AuthenticAMD
  Model name:            AMD EPYC 3151 4-Core Processor
    CPU family:          23
    Model:               1
    Thread(s) per core:  1
    Core(s) per socket:  4
    Socket(s):           1
    Stepping:            2
    Frequency boost:     enabled
    CPU max MHz:         2700.0000
    CPU min MHz:         1200.0000
    BogoMIPS:            5400.32
    Flags:               fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush
                          mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc
                          rep_good nopl nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq moni
                         tor ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm
                          cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ski
                         nit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb hw_pst
                         ate ssbd ibpb vmmcall fsgsbase bmi1 avx2 smep bmi2 rdseed adx smap clflushopt s
                         ha_ni xsaveopt xsavec xgetbv1 clzero irperf xsaveerptr arat npt lbrv svm_lock n
                         rip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold
                          avic v_vmsave_vmload vgif overflow_recov succor smca sme sev sev_es
Virtualization features:
  Virtualization:        AMD-V
Caches (sum of all):
  L1d:                   128 KiB (4 instances)
  L1i:                   256 KiB (4 instances)
  L2:                    2 MiB (4 instances)
  L3:                    16 MiB (2 instances)
NUMA:
  NUMA node(s):          1
  NUMA node0 CPU(s):     0-3
Vulnerabilities:
  Gather data sampling:  Not affected
  Itlb multihit:         Not affected
  L1tf:                  Not affected
  Mds:                   Not affected
  Meltdown:              Not affected
  Mmio stale data:       Not affected
  Retbleed:              Mitigation; untrained return thunk; SMT disabled
  Spec rstack overflow:  Mitigation; SMT disabled
  Spec store bypass:     Mitigation; Speculative Store Bypass disabled via prctl and seccomp
  Spectre v1:            Mitigation; usercopy/swapgs barriers and __user pointer sanitization
  Spectre v2:            Mitigation; Retpolines, IBPB conditional, STIBP disabled, RSB filling, PBRSB-eI
                         BRS Not affected
  Srbds:                 Not affected
  Tsx async abort:       Not affected
I plan to use the u2 socket with two 4x PCIe adapters but it depends on the BIOS capability. It must support PCIe bifurcation 4x4x4x4 or 8x4x4 to get the second adapter running and I haven't checked my BIOS yet. If the BIOS don't allow to change the PCIe lane mode then I will be using a single 8x adapter with a PCIe splitter.
 
  • Like
Reactions: iridial

iridial

Member
Oct 14, 2023
39
27
18
same for mine, only 4 cores / 4 threads, also shown in the bios cpu info tab. While in the BMC cpu inventory tab shows 4c/8t instead .-.
 

Attachments

jnrnbt.

New Member
Sep 21, 2023
16
9
3
Germany
Hey Folks,
i spent some time with the board now but needed to put it into service for it´s real purpose. So i ordered another one for playing around.

The Hyper Threading seems to be disabled in the G431 boxes per default, you need to enable it in the BIOS (i don´t know where it was, but it was buried a bit deeper). Works fine.

What i found is that some of my PCIe cards just don´t work in my configuration of SFF-8654-SFF-8611-PCIe x4 chain. Thats the case with some 10gbe dual sfp cards and some tested RAID cards, where other cards work.
For example a LSI 92XX-4i4e with 4 sata ports + external sas work where the same controller with internal sas and external sas won´t. So you need to test it for yourself.
I also designed and 3d printed an ATX I/O shield, i just need to make some minor improvements and then i´ll post it here.

All in all this thing works as intended like a charm, i love it and if one PCIe x4 is enough for you you can have an epic EPYC server including adapter & cables, plus case, psu, memory and storage at around 400€
 

iridial

Member
Oct 14, 2023
39
27
18
Can confirm, the additional 4 threads were disabled in BIOS -> Advanced Tab -> AMD CBS -> Zen Common Options -> Core/thread Enablement -> Agree -> SMTEN -> AUTO.
Power consumption with this change is higher at full throttle, hitting the 45W as per TDP.

Thanks!
 

Attachments

xen0x

New Member
Jun 25, 2021
3
1
3
Germany
Hey everyone,
came across these boards by accident. As they seem to offer some great bang for the buck ordered some from ram-koeing.

Before ordering additional equipment for the PCIe-Breakout, wanted to briefly confirm some details.

- M.2 Slot is a PCIe 3.0 x4.
- SlimSAS, right next to the SATA-Ports is either standard SATA or PCIe 3.0 x4.

Most interesting to me is the U2_2 (SFF-8654) Port. It's PCIe 3.0, but do we have 4 or 8 PCIe-Lanes?
If we happen to have 8 Lanes we might be able to use something from here:

With this:
SlimSAS PCIe gen4 Device Adapter 8i to x4x4 - 2W for 60€ plus a cable, one would have 2x PCIe 3.0 x4, which should be enough for a HBA and a 2-Port 10G Ethernet Card.

@jnrnbt.
You mentioned two PCIe adapters are not working for you. Did you try one on both SFF-8612 of your SFF-8654 breakout-cable?
 
  • Like
Reactions: jnrnbt.

jnrnbt.

New Member
Sep 21, 2023
16
9
3
Germany
That PCIe Adapter seems fine, but i don´t know if that works with 2 cards.

The u2_2 should be 8 Lanes, as @SadoKitten stated Gigabyte uses an IC on their PCIe distribution board for the 10 x1 Ports. I don´t think they use x4 for upstreaming to 10x x1. Given that i´m highly confident that the board is mostly identical to the MJ11-EC0, where is one PCIe x16 Port instead of the u2_1 (which isn´t soldered and does not show up in the BIOS) and u2_2 ports i think that both of these 8654-8i ports should electrically be x8.

m.2 should indeed be PCIe 3.0 x4, i´m running a NVMe card with that specs.

It´s an excellent question for the other SlimSAS Port. I never thought about it as i already saw it´s SATA Ports in BIOS and i didn´t know that you can use it for PCIe. But if you have a look at the MJ11-EC0(!) Version from Gigabyte it states that there is PCIe on this port too.

Yes, i tried both sff-8612 ends, to no avail.
 

prime420

New Member
Oct 17, 2023
9
10
3
For every one how want to update his BMC to actuell version. Here is the download link BMC firmware update
Simple unzip the file and navigate to the folder "fw", than upload the file "rom.ima_enc" in IPMI under the point "Maintenance" -> "Firmware update".
 
  • Like
Reactions: jnrnbt. and mrpasc

xen0x

New Member
Jun 25, 2021
3
1
3
Germany
It´s an excellent question for the other SlimSAS Port. I never thought about it as i already saw it´s SATA Ports in BIOS and i didn´t know that you can use it for PCIe. But if you have a look at the MJ11-EC0(!) Version from Gigabyte it states that there is PCIe on this port too.
That's were I got that info from.

The u2_2 should be 8 Lanes, as @SadoKitten stated Gigabyte uses an IC on their PCIe distribution board for the 10 x1 Ports. I don´t think they use x4 for upstreaming to 10x x1. Given that i´m highly confident that the board is mostly identical to the MJ11-EC0, where is one PCIe x16 Port instead of the u2_1 (which isn´t soldered and does not show up in the BIOS) and u2_2 ports i think that both of these 8654-8i ports should electrically be x8.

[....]

Yes, i tried both sff-8612 ends, to no avail.
If your PCIe works on both SFF-8612, but not two in parallel, possibly no PCIe bifurcation. The MJ11-EC0, according to page 40 of the manual, supports this.
 

prime420

New Member
Oct 17, 2023
9
10
3
Has anyone of you used a pci4.0 nvme on this mainboard?
For me, a pci4.0 nvme don't work as expected. The board don't recognize this device in UEFI.