Latest BIOS for NUC6i7KYK and "Above 4G decoding"

[agrt]

Hello, 

I need to enable the "above 4g decoding option" but i'm surprised to see the skull canyon does not offer such switch trough the visual bios. I'm trying to plug a egpu requiring larger pci memory space. Consequently when doing a lspci -v, i can see the gpu recognised ; with unassigned address.

You guys have ideas / advices ? Is it feasible on the nuc ?

[Oscar_A_Intel1]

Hello agrt, Thank you for posting on the Intel Community. I understand you're trying to enable the above 4g decoding option, however, this system doesn't have that option available in the BIOS. We have a list of external GPUs validated and tested on this system by Intel. Please check it on the link below. http://compatibleproducts.intel.com/ProductDetails?EPMID=89187# Please let me check if there's any workaround about what you're trying to do and I'll be posting back with the outcome. Regards, Oscar A. Intel Customer Support Technician Under Contract to Intel Corporation

[MRoss5200]

Just curious, I have seen other eGPUs work on KY NUC. What is different about the one you are using that requires this?

[agrt]

Hello Mrosss,

The egpu is a tesla k80 - (2x k40). This gpu require larger pcie addressable memory space.

Inspecting kernel logs at boot give the following :

agruet@gpu02:~$ dmesg | grep nvidia [ 1.408582] nvidia: loading out-of-tree module taints kernel. [ 1.408587] nvidia: module license 'NVIDIA' taints kernel. [ 1.417870] nvidia: module verification failed: signature and/or required key missing - tainting kernel [ 1.423558] nvidia-nvlink: Nvlink Core is being initialized, major device number 242 [ 1.423883] nvidia: probe of 0000:0b:00.0 failed with error -1 [ 1.423961] nvidia: probe of 0000:0c:00.0 failed with error -1 [ 1.424046] nvidia-nvlink: Unregistered the Nvlink Core, major device number 242 [ 1.473897] nvidia-nvlink: Nvlink Core is being initialized, major device number 242 [ 1.474097] nvidia: probe of 0000:0b:00.0 failed with error -1 [ 1.474105] nvidia: probe of 0000:0c:00.0 failed with error -1 [ 1.474171] nvidia-nvlink: Unregistered the Nvlink Core, major device number 242 [ 1.563575] nvidia-nvlink: Nvlink Core is being initialized, major device number 242 [ 1.563791] nvidia: probe of 0000:0b:00.0 failed with error -1 [ 1.563799] nvidia: probe of 0000:0c:00.0 failed with error -1 [ 1.563871] nvidia-nvlink: Unregistered the Nvlink Core, major device number 242 [ 7.707224] nvidia-nvlink: Nvlink Core is being initialized, major device number 239 [ 7.707392] nvidia: probe of 0000:0b:00.0 failed with error -1 [ 7.707402] nvidia: probe of 0000:0c:00.0 failed with error -1 [ 7.707627] nvidia-nvlink: Unregistered the Nvlink Core, major device number 239 [ 7.714423] nvidia_modeset: Unknown symbol nvidia_register_module (err -2) [ 7.714428] nvidia_modeset: Unknown symbol nv_kthread_q_schedule_q_item (err -2) [ 7.714434] nvidia_modeset: Unknown symbol nvidia_get_rm_ops (err -2) [ 7.714441] nvidia_modeset: Unknown symbol nv_kthread_q_item_init (err -2) [ 7.714443] nvidia_modeset: Unknown symbol nv_kthread_q_stop (err -2) [ 7.714444] nvidia_modeset: Unknown symbol nvidia_unregister_module (err -2) [ 7.714450] nvidia_modeset: Unknown symbol nv_kthread_q_init (err -2) [ 7.792437] nvidia-nvlink: Nvlink Core is being initialized, major device number 239 [ 7.792615] nvidia: probe of 0000:0b:00.0 failed with error -1 [ 7.792626] nvidia: probe of 0000:0c:00.0 failed with error -1 [ 7.792722] nvidia-nvlink: Unregistered the Nvlink Core, major device number 239

From this :

agruet@gpu02:~$ lspci -tv -[0000:00]-+-00.0 Intel Corporation Sky Lake Host Bridge/DRAM Registers +-02.0 Intel Corporation Sky Lake Integrated Graphics +-08.0 Intel Corporation Sky Lake Gaussian Mixture Model +-14.0 Intel Corporation Sunrise Point-H USB 3.0 xHCI Controller +-14.2 Intel Corporation Sunrise Point-H Thermal subsystem +-16.0 Intel Corporation Sunrise Point-H CSME HECI #1 +-1c.0-[01]-- +-1c.1-[02]----00.0 O2 Micro, Inc. Device 8621 +-1c.2-[03]----00.0 Intel Corporation Wireless 8260 +-1c.4-[04-3c]----00.0-[05-3c]--+-00.0-[06]----00.0 Intel Corporation Device 1575 | +-01.0-[07-3b]----00.0-[08-3b]--+-01.0-[09-0c]----00.0-[0a-0c]--+-08.0-[0b]----00.0 NVIDIA Corporation GK210GL [Tesla K80] | | | \-10.0-[0c]----00.0 NVIDIA Corporation GK210GL [Tesla K80] | | \-04.0-[0d-3b]-- | \-02.0-[3c]-- +-1d.0-[3d]----00.0 Samsung Electronics Co Ltd Device a804 +-1f.0 Intel Corporation Sunrise Point-H LPC Controller +-1f.2 Intel Corporation Sunrise Point-H PMC +-1f.3 Intel Corporation Sunrise Point-H HD Audio +-1f.4 Intel Corporation Sunrise Point-H SMBus \-1f.6 Intel Corporation Ethernet Connection (2) I219-LM

Then :

agruet@gpu02:~$ lspci -v ... 0b:00.0 3D controller: NVIDIA Corporation GK210GL [Tesla K80] (rev a1) Subsystem: NVIDIA Corporation GK210GL [Tesla K80] Flags: bus master, fast devsel, latency 0, IRQ 18 Memory at c4000000 (32-bit, non-prefetchable) [size=16M] Memory at <unassigned> (64-bit, prefetchable) Memory at <ignored> (64-bit, prefetchable) Capabilities: <access denied> Kernel modules: nvidiafb, nouveau, nvidia_384_drm, nvidia_384   0c:00.0 3D controller: NVIDIA Corporation GK210GL [Tesla K80] (rev a1) Subsystem: NVIDIA Corporation GK210GL [Tesla K80] Flags: bus master, fast devsel, latency 0, IRQ 18 Memory at c5000000 (32-bit, non-prefetchable) [size=16M] Memory at <unassigned> (64-bit, prefetchable) Memory at <ignored> (64-bit, prefetchable) Capabilities: <access denied> Kernel modules: nvidiafb, nouveau, nvidia_384_drm, nvidia_384 ...

As You can see Memory is unassigned. This is due to the BIOS limiting access to extended memory space thus why I'm asking for enabling "above 4g decoding" switch

Any ideas ?

All the best. rgds a

[Oscar_A_Intel1]

Hello agrt, Thank you for the information provided. We're investigating the behavior you're experiencing. I'll be posting back as soon as I have the outcome. Best regards, Oscar A. Intel Customer Support Technician Under Contract to Intel Corporation

[agrt]

Hello Oscar and thanks for considering my issue. Actually there is maybe a better way to ask for Intel engineers :

Is the NUC Support 64-bit BARs or not ? Otherwise I'm maybe wrong but I don't see any workaround to this issue. All the best.

[Ronny_G_Intel · Accepted solution]

Hi agrt, I just got confirmation that this feature is not supported on Skull Canyon or on any other NUC out on the market. We are currently looking into the possibility of adding this feature to new products but this is still "to be defined". Regards, Ronny G