epilectrik
/
spark
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
5 Commits
1 Branch
0 Tags
5.0 MiB
 
Branch: master
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from 'master'
${ noResults }
spark/context/open-questions.md

11 KiB
Raw Permalink Blame History

id title status source_sections related_topics
open-questions Open Questions active Aggregated from all context files [gb10-superchip memory-and-storage connectivity dgx-os-software ai-frameworks ai-workloads multi-unit-stacking physical-specs setup-and-config skus-and-pricing]

Open Questions

Catalog of known unknowns, research gaps, and unresolved questions about the Dell Pro Max GB10.

Hardware

GB10 Superchip

  • Q: What are the exact clock speeds for CPU and GPU dies under sustained load?
    • Status: Unknown. No official boost/base clocks published.
    • Would resolve: Performance prediction, thermal modeling
  • Q: What is the detailed per-precision TFLOPS breakdown (FP4/FP8/FP16/FP32/FP64)?
    • Status: FP4 = 1,000 TFLOPS (official). FP64 HPL = ~675 GFLOPS (benchmarked). Others inferred.
    • Would resolve: Accurate workload performance estimation

Memory

  • Q: Is the LPDDR5X soldered or socketed?
    • Status: Resolved — see Resolved Questions below.

Storage

  • Q: Which specific SSD model/brand is used in Dell vs. DGX Spark units?
    • Status: Unknown. Only form factor (M.2 2242 PCIe Gen5) confirmed.
    • Would resolve: Drive performance expectations, replacement sourcing
  • Q: What are the exact sequential and random IOPS?
    • Status: Unknown. No benchmarks published.
    • Would resolve: Storage performance expectations

Software

DGX OS

  • Q: Can stock Ubuntu 24.04 ARM be installed instead of DGX OS?
    • Status: Likely possible but unsupported. Not documented. Requires NVIDIA kernel.
    • Would resolve: OS flexibility
  • Q: Full list of pre-installed NVIDIA packages and exact versions?
    • Status: Major components known (CUDA, cuDNN, Docker, NGC, AI Enterprise, DGX Dashboard). Exact versions not published.
    • Would resolve: Development environment baseline
  • Q: Update cadence and EOL timeline details?
    • Status: 2-year guarantee mentioned (Jeff Geerling). Exact cadence unknown.
    • Would resolve: Long-term maintenance planning

AI Frameworks

  • Q: TensorFlow support status on ARM GB10?
    • Status: Unknown. Official vs. community builds unclear.
    • Would resolve: Framework selection for TF users
  • Q: Full NGC catalog availability for GB10?
    • Status: Unknown. Which containers have ARM64 builds.
    • Would resolve: Software ecosystem breadth
  • Q: vLLM or other inference server support on ARM Blackwell?
    • Status: Unknown.
    • Would resolve: Production inference deployment options
  • Q: JAX support status?
    • Status: Unknown.
    • Would resolve: Framework selection for JAX users

Networking / Multi-Unit

  • Q: Can more than 2 units be stacked?
    • Status: Only 2-unit documented. Slurm/K8s support suggests possible. Not confirmed.
    • Would resolve: Maximum scaling potential
  • Q: Performance overhead of inter-unit communication (quantified)?
    • Status: 200GbE RDMA link, but no latency/overhead benchmarks published.
    • Would resolve: Stacked vs. single performance expectations
  • Q: Actual tokens/sec for 405B models on stacked config?
    • Status: Unknown.
    • Would resolve: Real-world stacking value proposition

Physical / Environmental

  • Q: VESA mount compatibility?
    • Status: Unknown.
    • Would resolve: Mounting options
  • Q: Exact heatsink dimensions and material?
    • Status: Dual-fan + dense heatsink confirmed, but exact specs unknown.
    • Would resolve: Aftermarket cooling or case modding potential

Performance Benchmarks

  • Q: Tokens/sec for Llama 3.3 70B specifically?
    • Status: Only Llama 3.2 3B (~100 tok/s) and GPT-OSS-120B (~14.5 tok/s) benchmarked.
    • Would resolve: Most common use case performance
  • Q: Fine-tuning time estimates for common model sizes?
    • Status: Partially resolved — scripts and methods documented (Full SFT 3B, LoRA 8B, QLoRA 70B) but wall-clock times not published.
    • Would resolve: Training workflow planning
  • Q: Stable Diffusion / image generation performance?
    • Status: Partially resolved — ComfyUI confirmed working with SD 1.5. Quantitative benchmarks (images/sec) not published.
    • Would resolve: Non-LLM AI workload suitability
  • Q: Speculative decoding speedup factor?
    • Status: EAGLE-3 and Draft-Target methods documented. Quantitative speedup (tokens/sec improvement) not published.
    • Would resolve: Inference optimization ROI

Resolved Questions

Date Question Resolution Source
2026-02-14 Is the M.2 SSD user-replaceable? Yes — FRU. 4 Torx screws + 1 M2x2. Supports M.2 2230 AND 2242. PCIe Gen4. Dell Owner's Manual Rev A01
2026-02-14 Memory channel configuration? 256-bit interface, 16 channels LPDDR5X 8533 NVIDIA DGX Spark User Guide
2026-02-14 Does DGX OS include Docker? Yes — Docker + NVIDIA Container Runtime pre-installed NVIDIA DGX Spark User Guide
2026-02-14 Operating temperature range? Dell: 0-35°C; NVIDIA: 5-30°C Dell Owner's Manual + NVIDIA UG
2026-02-14 Humidity range? 10-90% non-condensing NVIDIA DGX Spark User Guide
2026-02-14 Operating altitude? Up to 3,000m (9,843 ft) NVIDIA DGX Spark User Guide
2026-02-14 Noise levels under load? Never exceeded 40 dB at 1-1.5m (non-stress) ServeTheHome
2026-02-14 Cooling solution? Dual-fan + dense heatsink, front-to-back airflow Jeff Geerling, ServeTheHome
2026-02-14 BIOS/firmware update procedure? apt upgrade + fwupdmgr, or DGX Dashboard GUI ServeTheHome
2026-02-14 Network boot (PXE)? Supported via UEFI → Advanced → Network Stack Configuration NVIDIA DGX Spark User Guide
2026-02-14 First-boot wizard steps? 10-step wizard documented (language, account, network, etc.) NVIDIA DGX Spark User Guide
2026-02-14 QSFP cables for stacking? Amphenol NJAAKK-N911/0006 or Luxshare LMTQF022-SD-R NVIDIA DGX Spark User Guide
2026-02-14 Software config for stacking? Netplan + SSH + MPI + NCCL v2.28.3 NVIDIA DGX Spark User Guide
2026-02-14 Does stacking appear as single device? No — 2-node distributed cluster, requires multi-node code NVIDIA DGX Spark User Guide
2026-02-14 Can QSFP ports be used for general networking? Yes — Ethernet config, 200GbE RDMA capable Jeff Geerling
2026-02-14 Tokens/sec for common LLMs? Llama 3.2 3B: ~100 tok/s; GPT-OSS-120B: ~14.5 tok/s Jeff Geerling, ServeTheHome
2026-02-14 Thermal throttling behavior (Dell)? Dell design prevents throttling; quieter than DGX Spark Jeff Geerling
2026-02-14 Is LPDDR5X soldered? Yes — soldered, not upgradeable Form factor / LPDDR5X standard
2026-02-14 OTA update mechanism? apt + fwupdmgr (CLI) or DGX Dashboard (GUI) ServeTheHome
2026-02-14 HDMI version on Dell? HDMI 2.1a (not 2.1b). Max 8K@30. Dell Owner's Manual Rev A01
2026-02-14 SSD interface speed? PCIe Gen4 NVMe up to 64 GT/s (NOT Gen5 as forums suggested) Dell Owner's Manual Rev A01
2026-02-14 SSD form factors supported? Both M.2 2230 and M.2 2242 Dell Owner's Manual Rev A01
2026-02-14 Max display resolution? USB-C DP: 8K@60; HDMI: 8K@30 Dell Owner's Manual Rev A01
2026-02-14 BIOS entry method? Delete key for BIOS; F7 for one-time boot Dell Owner's Manual Rev A01
2026-02-14 Full BIOS menu structure? Main/Advanced/Security/Boot/Save&Exit fully documented Dell Owner's Manual Rev A01
2026-02-14 Wireless module model? AzureWave AW-EM637, 2.4/5/6 GHz, Wi-Fi 7, BT 5.4 Dell Owner's Manual Rev A01
2026-02-14 Network controller model? Realtek RTL8127-CG (10GbE) + NVIDIA ConnectX-7 (QSFP) Dell Owner's Manual Rev A01
2026-02-14 Processor cache? 16 MB Dell Owner's Manual Rev A01
2026-02-14 Weight range? 1.22-1.34 kg depending on configuration Dell Owner's Manual Rev A01
2026-02-14 Power adapter dimensions? 23 x 78 x 162 mm, multi-voltage output (5V-48V) Dell Owner's Manual Rev A01
2026-02-14 USB-C MST support? Not supported (single display per port only) Dell Owner's Manual Rev A01
2026-02-14 Service tools required? Phillips #0, T5 or T8 Torx screwdriver Dell Owner's Manual Rev A01
2026-02-14 CUDA compute capability / SM architecture? sm_121 (compile with -DCMAKE_CUDA_ARCHITECTURES="121") build.nvidia.com/spark
2026-02-14 CUDA toolkit version? CUDA 13.0 (PyTorch wheels: cu130) build.nvidia.com/spark
2026-02-14 DGX Dashboard URL/port? http://localhost:11000 build.nvidia.com/spark
2026-02-14 TensorRT-LLM availability? Confirmed — container tensorrt-llm/release:1.2.0rc6 build.nvidia.com/spark
2026-02-14 Fine-tuning methods supported? Full SFT (3B), LoRA (8B), QLoRA 4-bit (70B), FSDP multi-node build.nvidia.com/spark
2026-02-14 Image generation support? ComfyUI confirmed (SD, SDXL, Flux) on port 8188 build.nvidia.com/spark
2026-02-14 Ollama / Open WebUI support? Yes — Docker container, port 12000 (Sync) or 8080 (direct) build.nvidia.com/spark
2026-02-14 NVIDIA Sync details? Cross-platform app, SSH key automation, VS Code/Cursor/Dashboard launch, port forwarding build.nvidia.com/spark
2026-02-14 PyTorch NGC container? nvcr.io/nvidia/pytorch:25.11-py3 (ARM64) build.nvidia.com/spark
2026-02-14 Speculative decoding methods? EAGLE-3 (built-in drafting) and Draft-Target (8B+70B) build.nvidia.com/spark