Another B650 chipset pcie card spotted, Leikuo Southbridge Max

starshipeleven · August 17, 2026, 11:56pm

While doing some pcie-related shopping on Aliexpress, I found this card, which does look like a B650 chipset card

They call it Leikuo Southbridge Max and mentions B650 as the “chip”. Also the layout and ports available are similar to the other cards we have seen from other vendors (now out of stock), like this one WisdPi PROM21 All in Expansion Card although it’s not just a straight copy-paste.
It claims to have three 10gbit usb 3.0 ports (two usb A connectors and one usb-c) and a 20gbit usb-c port, four M.2 slots and 4 sata ports in a possibly inconvenient place near one of the m.2 slots. It also has an auxiliary 6pin power connector.

In the description they mention low availability but it may or may not be FOMO marketing

Notice / Limited Stock:

Limited to 10 units per day, 1 unit per customer.

Orders exceeding this limit may be canceled.

Anyways I just bought one and I’ll report back eventually when it shows up.

Here is the link to the current item (if it does not work for you, try searching manually for the card name)
https://it.aliexpress.com/item/1005012704088998.html

they even took some effort to make a pdf “manual” listing specifications
Sb02e0ecca5614b6ab222258dcedc5a1aO.pdf (4.5 MB)

TheArkratos · August 25, 2026, 6:31pm

I’m also curious, I kind of want to frankenstein one of these onto my framework desktop.
This would allow me to use the 4x slot (or one of the M.2s via and M.2 → PCIe riser) to break out into everything I want for my VMs. USB I can use for hot swap, dedicated storage for the VM, etc. You can even use another M.2 → PCIe riser to do some low bandwidth PCIe devices.

So I’m really curious how the IOMMU groups will look.

Not looking for specific models but I was shown this one this morning: WisdPi PROM21 All in Expansion Card

For clarity: I’m already using a Intel Arc B50 pro (via a powered riser) with my framework desktop for SR-IOV for accelerating VMs

starshipeleven · August 27, 2026, 10:34pm

So, the thing arrived and after some quick testing to verify it was not going to blow up I installed it in my “dodgy stuff testing rig” aka a cheap chinese x79 board with a Xeon CPU.
The aux power connector seems to feed the 12v into the same power lines that the card also takes from the pcie slot, there are a couple diodes to prevent backfeeding into the pcie slot. So it’s actually powering the whole card, not just the usb ports as stated in the description. This is good I think.

IOMMU groups look good on the x79 board. Everything is split up, the onboard Sata controller running the 4 ports, the USB controller, and the m.2 slots. In this test there are 4 nvme drives installed on the card, the “Sandisk Corp WD PC SN810” you see in this output. Each one has its “pcie switch downstream port” in its iommu group.
Used this script to collect and paginate the information in a more readable format

gist.github.com

https://gist.github.com/flungo/428c374c040de1d0a30fd4a593d39040

lsiommu

#!/bin/bash
for d in $(find /sys/kernel/iommu_groups/ -type l | sort -n -k5 -t/); do 
    n=${d#*/iommu_groups/*}; n=${n%%/*}
    printf 'IOMMU Group %s ' "$n"
    lspci -nns "${d##*/}"
done;

#!/bin/bash
for d in $(find /sys/kernel/iommu_groups/ -type l | sort -n -k5 -t/); do 
    n=${d#*/iommu_groups/*}; n=${n%%/*}
    printf 'IOMMU Group %s ' "$n"
    lspci -nns "${d##*/}"
done;

Here is the relevant output

IOMMU Group 25 03:00.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Upstream Port [1022:43f4] (rev 01)
IOMMU Group 26 04:00.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 26 05:00.0 Non-Volatile memory controller [0108]: Sandisk Corp WD PC SN810 / Black SN850 NVMe SSD [15b7:5011] (rev 01)
IOMMU Group 27 04:02.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 27 06:00.0 Non-Volatile memory controller [0108]: Sandisk Corp WD PC SN810 / Black SN850 NVMe SSD [15b7:5011] (rev 01)
IOMMU Group 28 04:04.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 29 04:05.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 30 04:06.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 31 04:07.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 32 04:08.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 32 0b:00.0 Non-Volatile memory controller [0108]: Sandisk Corp WD PC SN810 / Black SN850 NVMe SSD [15b7:5011] (rev 01)
IOMMU Group 33 04:0a.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 33 0c:00.0 Non-Volatile memory controller [0108]: Sandisk Corp WD PC SN810 / Black SN850 NVMe SSD [15b7:5011] (rev 01)
IOMMU Group 34 04:0c.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 34 0d:00.0 USB controller [0c03]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset USB 3.2 Controller [1022:43f7] (rev 01)
IOMMU Group 35 04:0d.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 35 0e:00.0 SATA controller [0106]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset SATA Controller [1022:43f6] (rev 01)

The usb controller also appears with lsusb. The usb 3.whatever gen whatever aka the 20gbit controller is detected as “20000M/x2” so it’s correct.
In this output I have connected a 10gbit usb hub (the "ID 2109:0822 VIA Labs, Inc. ") and it’s negotiated at the right speed, the “10000M”, and there is a usb 3.0 card reader plugged into it just to show it’s detected.

It does not seem to be “plug and play” for some reason. To get it to see what I connect I have to send a
lsusb -vv, which is odd behaviour for USB.

/:  Bus 003.Port 001: Dev 001, Class=root_hub, Driver=xhci_hcd/12p, 480M
    ID 1d6b:0002 Linux Foundation 2.0 root hub
    /sys/bus/usb/devices/usb3  /dev/bus/usb/003/001
    |__ Port 004: Dev 002, If 0, Class=Hub, Driver=hub/5p, 480M
        ID 2109:2822 VIA Labs, Inc. 
        /sys/bus/usb/devices/3-4  /dev/bus/usb/003/002
        |__ Port 005: Dev 003, If 0, Class=Billboard, Driver=[none], 480M
            ID 2109:8822 VIA Labs, Inc. 
            /sys/bus/usb/devices/3-4.5  /dev/bus/usb/003/003
/:  Bus 004.Port 001: Dev 001, Class=root_hub, Driver=xhci_hcd/5p, 20000M/x2
    ID 1d6b:0003 Linux Foundation 3.0 root hub
    /sys/bus/usb/devices/usb4  /dev/bus/usb/004/001
    |__ Port 004: Dev 002, If 0, Class=Hub, Driver=hub/4p, 10000M
        ID 2109:0822 VIA Labs, Inc. 
        /sys/bus/usb/devices/4-4  /dev/bus/usb/004/002
        |__ Port 004: Dev 003, If 0, Class=Mass Storage, Driver=usb-storage, 5000M
            ID 0bda:0316 Realtek Semiconductor Corp. Card Reader
            /sys/bus/usb/devices/4-4.4  /dev/bus/usb/004/003

I don’t have a 20gbit usb device to test.

The generous size heatsink does get hot, a small fan or some airflow is better.

I don’t know how to read sensor information from this thing to check temperature, but “sensors” command does pick up the 4 ssd thermal sensors (note the name, it’s the pcie address of the ssds above)

nvme-pci-0500
Adapter: PCI adapter
Composite:    +43.9°C  (low  =  -5.2°C, high = +83.8°C)
                       (crit = +87.8°C)

nvme-pci-0c00
Adapter: PCI adapter
Composite:    +55.9°C  (low  =  -5.2°C, high = +83.8°C)
                       (crit = +87.8°C)

nvme-pci-0b00
Adapter: PCI adapter
Composite:    +42.9°C  (low  =  -5.2°C, high = +83.8°C)
                       (crit = +87.8°C)

nvme-pci-0600
Adapter: PCI adapter
Composite:    +49.9°C  (low  =  -5.2°C, high = +83.8°C)
                       (crit = +87.8°C)

Testing this again on a Ryzen 5 PRO 5650G shows a different situation though. All the stuff on the card is now in the same iommu group because screw you I guess. It’s the same OS drive as on the “test rig” above so the only difference is the hardware and possibly BIOS settings (different motherboard/CPU obviously). What is this horseshit AMD? Are you softlocking IOMMU groups on the chipsets on purpose now?
Tried changing various BIOS settings and even reset defaults (and then enable IOMMU again), no changes, always everything in the same IOMMU group.

IOMMU Group 0 00:01.0 Host bridge [0600]: Advanced Micro Devices, Inc. [AMD] Renoir PCIe Dummy Host Bridge [1022:1632]
IOMMU Group 0 00:01.1 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] Renoir PCIe GPP Bridge [1022:1633]
IOMMU Group 0 01:00.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Upstream Port [1022:43f4] (rev 01)
IOMMU Group 0 02:00.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 0 02:02.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 0 02:04.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 0 02:05.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 0 02:06.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 0 02:07.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 0 02:08.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 0 02:0a.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 0 02:0c.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 0 02:0d.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 0 03:00.0 Non-Volatile memory controller [0108]: Sandisk Corp WD PC SN810 / Black SN850 NVMe SSD [15b7:5011] (rev 01)
IOMMU Group 0 04:00.0 Non-Volatile memory controller [0108]: Sandisk Corp WD PC SN810 / Black SN850 NVMe SSD [15b7:5011] (rev 01)
IOMMU Group 0 09:00.0 Non-Volatile memory controller [0108]: Sandisk Corp WD PC SN810 / Black SN850 NVMe SSD [15b7:5011] (rev 01)
IOMMU Group 0 0a:00.0 Non-Volatile memory controller [0108]: Sandisk Corp WD PC SN810 / Black SN850 NVMe SSD [15b7:5011] (rev 01)
IOMMU Group 0 0b:00.0 USB controller [0c03]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset USB 3.2 Controller [1022:43f7] (rev 01)
IOMMU Group 0 0c:00.0 SATA controller [0106]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset SATA Controller [1022:43f6] (rev 01)

Testing this again, on my gaming rig with an asus motherboard and a non-APU CPU, a Ryzen 7 3700X, shows the IOMMU groups split correctly, like on the x79 motherboard at the start. So it seems this is yet another way the APUs on AM4 socket are gimped, pcie 3.0, no 4x4x4x4 bifurcation and also this bullshit limitations on IOMMU groups. Tomorrow I’ll try swapping the CPU on the board that has the Ryzen 5 PRO 5650G to confirm if this is a CPU thing or a motherboard thing. I also have a Gigabyte AM4 motherboard in my “main” PC that I can try out.
Thankfully I personally don’t need IOMMU for passing through stuff so this is not a deal breaker to use with an APU for my usecase (I just needed more m.2 slots for activities and the USB ports)

IOMMU Group 14 01:00.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Upstream Port [1022:43f4] (rev 01)
IOMMU Group 15 02:00.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 15 03:00.0 Non-Volatile memory controller [0108]: Sandisk Corp WD PC SN810 / Black SN850 NVMe SSD [15b7:5011] (rev 01)
IOMMU Group 16 02:02.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 16 04:00.0 Non-Volatile memory controller [0108]: Sandisk Corp WD PC SN810 / Black SN850 NVMe SSD [15b7:5011] (rev 01)
IOMMU Group 17 02:04.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 18 02:05.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 19 02:06.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 20 02:07.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 21 02:08.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 21 09:00.0 Non-Volatile memory controller [0108]: Sandisk Corp WD PC SN810 / Black SN850 NVMe SSD [15b7:5011] (rev 01)
IOMMU Group 22 02:0a.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 22 0a:00.0 Non-Volatile memory controller [0108]: Sandisk Corp WD PC SN810 / Black SN850 NVMe SSD [15b7:5011] (rev 01)
IOMMU Group 23 02:0c.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 23 0b:00.0 USB controller [0c03]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset USB 3.2 Controller [1022:43f7] (rev 01)
IOMMU Group 24 02:0d.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset PCIe Switch Downstream Port [1022:43f5] (rev 01)
IOMMU Group 24 0c:00.0 SATA controller [0106]: Advanced Micro Devices, Inc. [AMD] 600 Series Chipset SATA Controller [1022:43f6] (rev 01)

the usb behaviour as described at the start is unchanged on either other systems, the 10gbit usb hub and other USB devices are detected and negotiates at the right speed but only after I launch a lsusb -vv. Very weird, but even if I can’t find a workaround I can live with that, I had no luck finding other usb controllers that negotiate at 10gbit at all on Linux besides onboard Intel/AMD, there is a bug open in kernel but nobody seems to care about looking into it.

I’ll do some more tests in a few days.

TheArkratos · August 31, 2026, 2:42am

Whelp, I also ordered the Lekuo one based on your positive experience and the fact that I need 2 more USB-c ports.

My plan is a bit crazy, but I plan to use my framework desktop with an Intel B70 pro in the slot (currently running a b50 there) using a powered riser and to use this off the top M.2, which currently has a usb-c card running off an M.2 to PCIe riser. The current setup works great, but with this I get to keep the 2 usb-c ports I use while getting access to sata and not just the M.2 I lost but also more if I ever need to. Good find, thanks for the share!

if you are wondering, the plan is AI workloads but also VM workloads, being able to pass thru M.2s and a usb controller to the VM(s) should be much nicer with this.

starshipeleven · August 31, 2026, 7:53am

Some more testing, swapped the 3700X CPU on the motherboard that had the 5650G, set IOMMU to Enabled in BIOS settings.
Also enabled “Above 4G decoding” and “Resizable bar support” (they are needed too)
The IOMMU groups are now split up properly. I don’t post the terminal output again here because it’s the same as above.

Going back to the 5650G on that board, same settings, and IOMMU groups are not split.

So that’s “confirmed”, the APUs IOMMU is gimped somehow while with a normal AM4 CPU without integrated graphics or on an old xeon cpu on a x79 board this works fine. Not a card issue.

Also now it seems the USB ports work properly and detect the 10gbit hub and other devices on hotplug, as expected for USB. No idea why that is working fine now and not during the previous testing above.

Sudo_Apt · August 31, 2026, 6:32pm

That sounds like a sweet Frankenstein use. I also just ordered one, mostly to put all my little Optanes into a storage server for various accelerating workloads.

taggart · September 2, 2026, 5:47pm

Phoronix just did a review of this card

https://www.phoronix.com/review/lekuo-southbridge-max

Sudo_Apt · September 7, 2026, 3:31pm

I put 4 Optane M10 Drives in one, just to see if it sullied the latency.

It did not

Sudo_Apt · September 7, 2026, 8:29pm

Jeffrey1910 · September 8, 2026, 3:18am

I also ordered one of them, as it seem a interesting option to extend the IO or add bifurcation support on unsupported platforms.
I’m actually surprised we have not seen more attempts as this way sooner.
Probably the costs and availability of the chipsets played a role.

I remember Asrock offered a B650 chipset card like this when AM5 launched.
All be it as a proprietary version for the Livemixer motherboard series. (https://www.youtube.com/watch?v=NrKZeEmu6UA&t=1s)
I remember Wendell did try it in other boards for fun and it just worked.
If I recall correctly the proprietary cable was to add hotplug support and video out for pcie tunneling on usb4.
Without the cable you missed video support and needed to reboot the pc to make it detect new pcie hardware over usb4, the regular usb4 stuff did work without issue.

So we’ll probably won’t get a full featured usb4 40gbe variant of this card type.
Although I don’t mind to be proven wrong.

So now the important question, how many cards are able to be daisy chained (with adapters) until no new cards are detected anymore or the system just refuses to post.
Would also be intrestring to test if any latency will eventually be measurable when multiple are chained.