XIO

xWire

Posts from every feed QUEST has found across the web — blogs, news, podcasts, channels and more.

NVIDIA Developer Forums - Latest posts

Latest posts

JP7.2 RT kernel: BPMP clock request never completes after wk2xxx sets 4800, SPI runtime PM and thermal block for ~600s

Hardware NVIDIA Jetson AGX Orin Developer Kit Super BPMP firmware: 684d9490f24f1c33cad8-bfff9e30f72 Software Kernel: 6.8.12-1021-rt-tegra #1 SMP PREEMPT_RT Custom SPI-UART: wk2xxx V2.4 (2022.07.24) on spi1.0 / c260000.spi (spi-tegra114), ports ttysWK0–ttysWK3 SPI slave in the same failure: 3230000.spi (spi-tegra124-slave), used by a userspace spidev process (spi_rx) We do not have a serial-console capture of either incident. We have not yet reproduced this on a non-RT kernel. Summary A wk2xxx…

Is it worth buying a second DGX Spark now?

WIth the political movements in the US this week, local AI may become very interesting: The likes of OpenAI / Anthropic / SpaceX will likely not be able to deploy their cloud mega-DCs as quickly as they would like - with a focus to give back more to communities - further restricting supply of cloud AI and pushing up prices. The memory market on the other hand may have more supply for consumer goods in the short-term, although what happens in the longer term is still anyone’s guess. Prepare for…

Jetson Thor: flashing / USB recovery (RCM) from a non-x86 embedded host (ARM or RISC-V)

fbua: Is there a supported way to flash or recover Jetson Thor from an ARM64 or RISC-V host? Can NVIDIA provide ARM64 builds of the flashing tools (tegrarcm_v2 and related), or a host-side flashing library suitable for an embedded system? Is the RCM protocol specification available, under NDA if necessary? We don’t have flash tool for ARM64 or RISC-V host, and no RCM protocol specification to provide. Thanks

AGX Orin (L4T 36.3) — initrd flash kernel fails to boot, USB device mode dead

Hi, gautam_shet: Is there any way to reach the eMMC on this unit given that the initrd flash kernel won’t boot? I’m considering booting an arm64 live USB via the F11 boot manager and mounting mmcblk0p1 from there. Is there a better approach, or a way to make the initrd path work? Please flash with the USB as a rootfs and you should mount the emmc and backup your data.’ gautam_shet: Does the initrd failure itself suggest anything? RCM boot completes cleanly and the target then never enumerates.…

Hook cudaLaunchKernel failed

I want to hook cudaKernelLaunch and get kernel name. Here is my code: #define _GNU_SOURCE #include <stdio.h> #include <dlfcn.h> #include <cuda_runtime.h> typedef cudaError_t (*cudaLaunchKernel_t)(const void *func, dim3 gridDim, dim3 blockDim, void **args, size_t sharedMem, cudaStream_t stream); typedef cudaError_t (*cudaFuncGetName_t)(const char **name, const void *func); cudaLaunchKernel_t real_cudaLaunchKernel = NULL; cudaFuncGetName_t real_cudaFuncGetName = NULL; static void init_real_func()…

Request to enable Public API Endpoints – 404 Function Not Found for account

Hi NVIDIA Support Team, I am encountering a persistent 404 "Function Not Found for account" error when trying to call the NIM API (e.g., POST /v1/chat/completions), even though the models endpoint (GET /v1/models) works perfectly. Based on community discussions, I understand this error indicates that my personal account is missing the "Public API Endpoints" permission. Here are the details of my account and the error: - Account Hash: NVX-leN7i4EBjcm7beyi96ez_PUyd8OuMgzHSixMV5I - Error Message:…

Cannot receive SMS OTP verification code for phone number +84 - Account Verification Issue

Hello NVIDIA Support Team, I am having trouble verifying my phone number to access NVIDIA services (NIM / ://nvidia.com). I am located in Vietnam and using the country code +84, but I am not receiving any SMS OTP verification codes. I have tried multiple times but the system now says I have exceeded the allowed verification limits. Could you please help me verify my account manually or reset the verification status? Here are my account details: - NVIDIA Account Email: [sak***n3@gmail.com] -…

Recommended way to show product / module name on the first-boot OOBE page (Jetson Thor, R39.2)

Following seems like a reasonable approach. For the factory-check use case (packer powers on the unit and must see T4000 vs T5000 before boxing), here is a way to show the module on screen: a small OEM-owned .deb with one oneshot systemd unit, modelled on the nvfb-* units in nvidia-l4t-firstboot. oem-som-banner.service is WantedBy=nv-oobe.target and ordered Before=nv-oobe.service display-manager.service, so it runs on every boot while OOBE is still pending (factory power-on), and never again…

Blackwell Pro 6000 MHz degrading

The PCB of pro 6000 WS and 5090 FE are not exactly the same. The 6000 has VRam on both sides of the PCB, the 5090 FE does not. I can almost guarantee that the shunt resistor on your card is fckd, like with all the others here. Especially if you get zero compute errors or ECC errors, then you know the chips are fine which is good news. If you want me to repair the card, PM me (it will void the warranty though). All of the cards with this issue were Pro 6000 Blackwell WS cards, only very few…

Request to Enable Public API Endpoints Permission for NVIDIA NIM

Account access issues are not handled here through the forums. Please check the pinned posts: Account Access & Verification Update Access/Accounts Dear NVIDIA Developer Community, It is no secret that the verification process on build.nvidia.com has been challenging for both the community and NVIDIA. It’s honestly been pretty frustrating from both sides, but please understand that we hear your feedback and are working through it. This is a complex problem, and one in which we cannot be open…

615 release feedback & discussion

List of issues that still affect the 615.78.08 driver as of Nov 1: Random black screens when playing intensive games, sometimes resulting in a partial GPU driver crash. It is sometimes possible to switch to a TTY, but attempting to switch back to a graphical session will crash the driver for good. Desktop memory usage is excessively high(over 2GB sometimes). A large amount of GPU Memory is unaccounted for in nvidia-smi / NVML. Applications crash when VRAM is full. Shared RAM under Linux seems…

Jetson Thor failed to enter SC7 mode on the custom board

Hi @tank.xue Based on the logs, it seems that the system is reaching the SC7 suspend/resume sequence, as indicated by the Boot-mode : SC7 Exit message. Could you confirm whether the board wakes up immediately after running sudo systemctl suspend? I also noticed a PCIe timeout during suspend: tegra264_pcie_rp_deinit PCIE C2 L2 timeout This might be worth investigating, although it is not necessarily the root cause. I would suggest checking /sys/kernel/debug/wakeup_sources and…

Request to Enable Public API Endpoints Permission for NVIDIA NIM

Hello NVIDIA Support, I am experiencing a 403 Authorization failed error when using the NVIDIA NIM API. GET /v1/models works successfully (HTTP 200). POST /v1/chat/completions returns HTTP 403 Authorization failed. I am using an active NVIDIA API key generated from build.nvidia.com. My organization is Personal. Could you please check whether Public API Endpoints permission is enabled for my organization and enable it if necessary? I would like to use NVIDIA-hosted inference endpoints for…

Isaaclab camera render robot in other environemnt

Hi @thuaj, @WangCheng23, @nathankau – sorry this sat unanswered for so long. There is now a native answer for this in current releases, so here is how it works on Isaac Sim 6.2.0. Answer: RTX scene partitioning lets you restrict which prims a given camera renders. It is described in the Omniverse Cameras documentation, “Scene Partitions”, and it was introduced in the Kit 110 renderer (see the 110.0 release highlights), which is what Isaac Sim 6.x is built on. The doc describes this exact use…

NVIDIA NIM API — GPT-OSS-20B Timeout and HTTP 410 — TechScales Production Integration

Hello NVIDIA Developer Support Team, We are TechScales, a software and AI automation company integrating NVIDIA-hosted AI inference into our website, https://www.techscales.co. We are encountering persistent API inference failures with our existing NVIDIA configuration. Technical environment: Hosting: Hostinger Business, Node.js 24 API endpoint: https://integrate.api.nvidia.com/v1 Primary model: openai/gpt-oss-20b Integration: OpenAI-compatible API Issues observed: GPT-OSS-20B: Chat completion…