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

GNU/Linux Fedora Graphics low quality Age of Empires 4 - Steam

I had the same problem on Fedora 44. My solution was also to turn Gameplay Resolution Scale to 100% (that was 50% of the problem) But also to change screen resolution with the slider. From like 1280 x 800 to something else. Some resolutions make the graphics worse, some better. Picked one that made it better. Now im a happy man who can play age of empires 4 again.

NVIDIA Nemotron API returns 403 Authorization Failed – Models endpoint works

Hello NVIDIA Community, I am currently integrating NVIDIA Nemotron 3.5 Lightning into my Python-based AI research agent project. I am experiencing an authorisation issue with the NVIDIA API. Here are my test results: GET /v1/models works successfully and returns 84 available models. POST /v1/chat/completions consistently returns HTTP 403 Forbidden. Error message: {“status”:403,“title”:“Forbidden”,“detail”:“Authorization failed”} Model: nvidia/nemotron-3.5-lightning-30b-a3b Endpoint:…

Programming the sMCU and FSI on IGX Thor systems

Hello @martin.haubro! Based on the title and content of your topic, it looks like it may receive better visibility and feedback in a different category. We took the liberty of moving it for you. If this was an incorrect assessment, please send me a direct message. Disclaimer: this moderation suggestion and message were generated with AI assistance.

Unable to receive SMS OTP for Vietnam (+84) phone verification on build.nvidia.com

Hello NVIDIA Support Team, I am trying to complete the phone number verification on build.nvidia.com to access NVIDIA NIM, but I never receive the SMS verification code on my Vietnamese phone number (+84). Could you please help me verify/activate my account manually? Account Email: duonghr2110@gmail.com Phone Number: +84 865216504 Thank you!

ConnectX-8 P6612_Ax firmware 40.50.1002 reports "IRISC not responding" on both PCI functions

The data you provided points to an issue that has been identified recently and the upcoming release will have a fix for it (40.51.xxxx). Issue description: “Resolved a firmware issue where deregistering a very large memory region (e.g., 1.7 TB) could exceed the firmware internal watchdog window, causing a false “not responding” error and an OS hang with the message “All CMD slots are stalled, aborting command.”” Your workaround seems like to right approach till then. Regards, Yaniv

SMCU reset AND gate (U184): needed without the debug MCU?

Hi NVIDIA team, A. What we see in your dev kit (to confirm with NVIDIA) On the SMCU module (P3974 page 5), U184 is an AND gate on the U2A reset. One input is the PMIC reset (RAA271084 pin 29), which you mark “push-pull”. The other input is PMIC_RST_N from the carrier, which you mark “OD” (open-drain), with a 100 kΩ pull-up. On the carrier (P4071), that input is driven by the debug MCU (U136 PA07) and by the SMCU JTAG header J111 pin 10. The debug MCU runs from the always-on rail, so this input…

NVIDIA, THIS IS UNACCEPTABLE! RTX 4080 SUPER — BLACK SCREEN ON DRIVER 621.14, 5120×2160 @ 120 Hz BROKEN, WinDbg EVIDENCE INCLUDED

I’m having the same problem. After installing driver 621.14, my RTX 5060 Ti has become unusable; the screen goes black for no apparent reason—often as soon as Windows loads, frequently right at the Windows login screen. If I disable the graphics card in Device Manager, Windows works normally; however, if I enable it, the screen goes black and Windows freezes within minutes. The only way to fix it is to boot into Safe Mode and disable the graphics card again. I’ve already performed a clean…

620 release feedback & discussion

Nvidia-powerd is still quite problematic - whether it is Balanced or Performance mode, it loves to give too much power to the CPU and overutilize it, and I feel like it underutilizes GPU under some GPU-load threshold as well. For instance at 90% load you would still gaing performance for allocating less power to the CPU and more to the GPU, but instead the CPU works its hardest and sits withing 90 degree range with fresh and healthy cooling, and the GPU would be at 60W. Throttling with cpupower…

Manual Phone Verification Request for NVIDIA NIM API (Kazakhstan)

Hello NVIDIA Team, I am trying to generate an API key for testing models on build.nvidia.com, but my country (Kazakhstan, country code +7) is not available in the phone verification dropdown menu. Could you please assist me with manual verification or enable API access for my account? Account Email: alecsandrmih2@gmail.com Purpose: Personal development and non-commercial model testing. Thank you!

Per-camera ISP override selection for custom GMSL sensors

Board NVIDIA Jetson AGX Orin Developer Kit L4T R36 (release), REVISION 3.0, GCID 36923193, 2024-07-19 Kernel 5.15.136-tegra nvidia-l4t-camera 36.3.0-20240719161631 nvidia-l4t-core 36.3.0-20240719161631 libargus 0.99.3.3 (single-process) We are integrating custom GMSL camera sensors on Jetson and need to use different ISP tuning per camera when multiple sensors are present. Based on NVIDIA docs, we configured the camera module badge using the expected pattern (module_position_part-number style),…

DGX Spark Evaluation / Sponsorship Request – Local AI Coding Agents

So I have built something I guess that is similar to the original request. I have my apps that I use in a development life cycle, where they are in a loop where they are constantly being fixed, improved or new features when I create requests. I update the stats for this every 6 hours. Partly an experiment to see what is possible with a local setup, and actual real use to improve the applications that I built for my security tooling. f4n6 – 25 Sep 26 Spark Models and Costs Running a software…

Is it worth buying a second DGX Spark now?

I only have one Spark, for my workflow, if I had to enter now, I would quite likely go for the 256GB M5 Ultra. It is cheaper than 2x Sparks (and lower idle power usage), memory is much faster and I believe in 3-5 years I can sell it for more. Yes 2x Sparks still in most cases prefill faster, but with good caching for my workflow it matters a bit less, plus the gains on decode are significant. I do prefer linux and having CUDA is a strong selling point of the Spark, plus it is more easily…

[DeepStream 9.2] PyServiceMaker local Smart Record trigger with nvmultiurisrcbin

Hardware: dGPU (x86_64) DeepStream installed locally: 9.1; asking about current 9.2 behavior PyServiceMaker: Python Pipeline API Issue type: Question I am using a single nvmultiurisrcbin with multiple RTSP sources. It is configured with matching uri-list and sensor-id-list, smart-record=2, and the Smart Record cache/output properties. The DeepStream 9.2 plugin documentation describes smart-record=2 as enabling Smart Record for cloud and local events. How can a local event in a PyServiceMaker…

NVIDIA, THIS IS UNACCEPTABLE! RTX 4080 SUPER — BLACK SCREEN ON DRIVER 621.14, 5120×2160 @ 120 Hz BROKEN, WinDbg EVIDENCE INCLUDED

One additional point I want to emphasize for NVIDIA engineering: My external display is an LG UltraGear 32GX870A-B connected over HDMI 2.1. This is a mainstream, high-end gaming monitor from a major display manufacturer, so I do not think this should be treated as an obscure or unusual compatibility edge case. A current LG UltraGear paired with modern GeForce hardware is exactly the kind of configuration I would expect to be covered by normal NVIDIA display validation. My symptoms are not…

Native vLLM for RTX Spark on Windows ARM64 (w/Qwen3.8-Flash-Next-FP4-Local)

Thanks @michael.valente for the detailed WSL2 vs native comparison and the FP8 kernel work, and thanks @dev.parkjeonghwan for folding it into the project. To summarize for anyone landing here later: Model fit: the project already supports NVIDIA’s Qwen3.8-Flash-Next-FP4-Local on 128 GB N1X systems with a 32/96 split, with NVMe offload for the n-gram embeddings. Use the dedicated GPU memory and paging-file settings from post #4 to avoid commit-related OOMs. WSL2 vs native: post #13 has the first…

NVIDIA, THIS IS UNACCEPTABLE! RTX 4080 SUPER — BLACK SCREEN ON DRIVER 621.14, 5120×2160 @ 120 Hz BROKEN, WinDbg EVIDENCE INCLUDED

I am seeing a potentially related regression on a different hardware configuration, so I wanted to add another data point for NVIDIA engineering. System: HP OMEN Max 16 NVIDIA GeForce RTX 5080 Laptop GPU LG UltraGear 32GX870A-B External monitor connected over HDMI 2.1 Windows 11 Known-good NVIDIA driver: 610.88 With 610.88, the external LG display works normally, including fullscreen gaming and the expected resolutions/refresh rates. Beginning with newer NVIDIA drivers, I have experienced…

Running AV Reasoning Models Within Onboard Compute and Latency Limits

Originally published at: Running AV Reasoning Models Within Onboard Compute and Latency Limits | NVIDIA Technical Blog Autonomous vehicle (AV) developers need reasoning models that can operate within the compute and latency constraints of an onboard system. Larger models provide a foundation for developing these capabilities, but moving them into a vehicle requires additional work to reduce model size, preserve useful reasoning, and validate performance on the target hardware. NVIDIA Alpamayo 2…

Implausible authored physics:mass on three nvidia/simready-assets packages (air hose reel, utility knife, storage shelf)

Hi @pekka3, Thanks for confirming! I’ll flag this thread with the team so they can take a closer look and follow up with you directly in the Hugging Face discussion. huggingface.co nvidia/simready-central · Implausible authored physics:mass on three... Hi, reposting here as suggested on the NVIDIA forum ( topic 388277 ).

Request to expose BPMP dram refresh temp control to allow us avoid PCIE transfer stalls

Hi, Thank you for the mr4_policy suggestion. We applied it as instructed and the module no longer boots. Our findings are below; the full UART log is attached. Platform Jetson AGX Orin, t234-A01 production silicon, production-fused, secure boot. L4T / Jetson Linux 36.4.4. MB1 1.4.0.4-t234-54845784-e89ea9bc, BL31 v2.8 l4t-r36.4.4, OP-TEE 4.2, Jetson UEFI v36.4.4. BPMP DTB: tegra234-bpmp-3701-0005-3737-0000.dtb (and the -maxn variant). Update path: A/B bootloader capsule (TEGRA_BL.Cap), installed…