GeForce RTX 5090 vs AMD Instinct MI355X

Agent access

1. ai-native userPoint an agent at llms.txt or agent-oriented docs

   weight 2 · round to AMD Instinct MI355X
   GeForce RTX 5090~partialprobed5/10

   NVIDIA's developer portal serves a valid llms.txt (HTTP 200, descriptive content) that an agent could use to discover CUDA/GPU-related docs, but this is at the general developer.nvidia.com domain rather than an RTX-5090-specific surface, and adjacent agent-friendly affordances (markdown doc exports, OpenAPI spec) are absent (404s). missing for 10: RTX-5090-specific llms.txt or agent doc entry point, working markdown/API exports, independent confirmation agents actually use this pathway successfully.

   • [probe] “PROBE llms.txt: HTTP 200 at https://developer.nvidia.com/llms.txt # NVIDIA Developer > Comprehensive developer portal for NVIDIA accelerate…”
   • [probe] “PROBE docs-md: HTTP 404 at https://developer.nvidia.com/cuda-toolkit.md”
   • [probe] “PROBE openapi: all candidate paths 404 (https://developer.nvidia.com/openapi.json, https://developer.nvidia.com/swagger.json, https://develo…”

   proof ↗⚑ flag

   AMD Instinct MI355X✓fullprobed7/10

   A live llms.txt file was confirmed at rocm.docs.amd.com/llms.txt with a 200 response, providing an agent-readable entry point into ROCm documentation, and the runtime probe confirms the docs surface is crawlable without a browser. Missing for 10: no evidence of per-page markdown/.md variants (404 on that probe), no OpenAPI/agent tool spec, and no confirmation that agent-oriented docs cover the MI355X product pages themselves rather than just ROCm software.

   • [probe] “PROBE llms.txt: HTTP 200 at https://rocm.docs.amd.com/llms.txt # ROCm documentation > Note: ROCm documentation is split across multiple pro…”
   • [probe] “PROBE runtime (recorded 2026-09-15): the ROCm documentation at rocm.docs.amd.com answered a keyless curl naming ROCm — the MI355X's entire d…”
   • [probe] “PROBE docs-md: HTTP 404 at https://rocm.docs.amd.com/en/latest/.md”
   • [probe] “PROBE openapi: all candidate paths 404 (https://rocm.docs.amd.com/openapi.json, https://rocm.docs.amd.com/swagger.json, https://rocm.docs.am…”

   proof ↗⚑ flag

2. ai-native userRun the product headlessly / in CI for automation

   weight 2 · round to AMD Instinct MI355X
   GeForce RTX 5090—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

   AMD Instinct MI355X~partialclaimed6/10

   ROCm/Instinct docs show clear headless-automation building blocks — AMD Container Toolkit for Docker, GPU Operator for Kubernetes, Spur job scheduler (Slurm-compatible), and Cluster/ROCm Validation Suites for automated testing — all of which support running GPU workloads without a UI in CI/cluster pipelines. Missing for 10: no first-party CI pipeline example (e.g., GitHub Actions/GitLab runner config) or independent hands-on report confirming headless CI use in practice.

   • [claimed-docs] “GPU Operator Deploy and manage Instinct GPUs in Kubernetes clusters.”
   • [claimed-docs] “Spur AI-native job scheduler, drop-in compatible with Slurm, with GPU-first scheduling and Raft-based state.”
   • [claimed-docs] “Cluster Validation Suite Test scripts that validate AMD AI clusters end to end.”
   • [claimed-docs] “AMD Container Toolkit Integrate Instinct GPUs with Docker and container runtimes.”
   • [claimed-docs] “ROCm Validation Suite System validation and hardware diagnostics.”

   proof ↗⚑ flag

3. ai-native userUse an official CLI

   weight 2 · round to AMD Instinct MI355X
   GeForce RTX 5090—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

   AMD Instinct MI355X~partialclaimed5/10

   AMD ships official CLI tooling in its ecosystem — AMD SMI for GPU management and Spur, explicitly described as an 'AI-native job scheduler' with Slurm-compatible CLI — but these are infrastructure/ops CLIs rather than a CLI aimed at AI-native development or agentic workflows. Missing for 10: evidence of a CLI specifically designed for AI-agent/dev workflows (e.g., code generation, model interaction, agent orchestration) and independent hands-on confirmation of these CLIs' AI-native usability.

   • [claimed-docs] “AMD SMI Unified user-space tool to manage and monitor GPUs and drivers.”
   • [claimed-docs] “Spur AI-native job scheduler, drop-in compatible with Slurm, with GPU-first scheduling and Raft-based state.”
   • [claimed-docs] “AMD Container Toolkit Integrate Instinct GPUs with Docker and container runtimes.”

   proof ↗⚑ flag

4. ai-native userDrive the product through a documented public API

   weight 3 · round to AMD Instinct MI355X
   GeForce RTX 5090~partialprobed3/10

   CUDA Toolkit documentation describes a programmatic API (CUDA, cuTile Python/C++) for building GPU-accelerated applications, which is the closest analog to a 'documented public API' for this hardware product, but this is a low-level compute-kernel API, not an agent-drivable control interface, and direct probes found no OpenAPI/Swagger spec or machine-readable docs (404s). Missing for 10: a structured/machine-readable API spec (OpenAPI/REST), any agentic control surface, and independent corroboration of AI-native programmatic access beyond raw CUDA kernel programming.

   • [claimed-docs] “The NVIDIA® CUDA® Toolkit provides a development environment for creating high-performance, GPU-accelerated applications.”
   • [claimed-docs] “cuTile Python is an expression of the CUDA Tile programming model in Python.”
   • [probe] “PROBE docs-md: HTTP 404 at https://developer.nvidia.com/cuda-toolkit.md”
   • [probe] “PROBE openapi: all candidate paths 404 (https://developer.nvidia.com/openapi.json, https://developer.nvidia.com/swagger.json, https://develo…”

   proof ↗⚑ flag

   AMD Instinct MI355X~partialprobed4/10

   AMD documents extensive low-level APIs for driving the GPU (HIP runtime, ROCm libraries, AMD SMI, metrics exporter) and even exposes an llms.txt for machine consumption, but there is no unified, documented public API (e.g., OpenAPI/REST spec) that an AI-native agent could call to drive the product — probes for openapi.json/swagger.json all 404. missing for 10: a formal public API spec/SDK reference for agentic automation, evidence of programmatic/agent-driven control beyond developer-level HIP/ROCm libraries, independent confirmation of agent usage.

   • [claimed-docs] “HIP C++ Learn the HIP programming model.”
   • [claimed-docs] “AMD SMI Unified user-space tool to manage and monitor GPUs and drivers.”
   • [claimed-docs] “Device Metrics Exporter Prometheus-format GPU metrics for HPC and AI environments.”
   • [probe] “PROBE llms.txt: HTTP 200 at https://rocm.docs.amd.com/llms.txt # ROCm documentation > Note: ROCm documentation is split across multiple pro…”
   • [probe] “PROBE openapi: all candidate paths 404 (https://rocm.docs.amd.com/openapi.json, https://rocm.docs.amd.com/swagger.json, https://rocm.docs.am…”

   proof ↗⚑ flag

5. ai-native userBuild against official SDKs

   weight 2 · round to AMD Instinct MI355X
   GeForce RTX 5090✓fullclaimed7/10

   NVIDIA provides official SDKs (CUDA Toolkit, CUDA Tile C++/cuTile Python, RTX Neural Shaders SDK, RTX Mega Geometry, Nsight tools) that developers can build AI/graphics applications against, all documented on NVIDIA's developer portal. Missing for 10: independent/hands-on developer corroboration of building against these SDKs, and deeper API reference/sample documentation beyond marketing-style descriptions.

   • [claimed-docs] “The NVIDIA® CUDA® Toolkit provides a development environment for creating high-performance, GPU-accelerated applications.”
   • [claimed-docs] “CUDA Tile C++ is an expression of the CUDA Tile programming model in C++. It's built on top of the CUDA Tile IR specification and allows you…”
   • [claimed-docs] “cuTile Python is an expression of the CUDA Tile programming model in Python.”
   • [claimed-docs] “The RTX Neural Shaders SDK lets developers train shader data on an RTX PRO workstation and accelerate neural representations with NVIDIA Ten…”
   • [claimed-docs] “RTX Mega Geometry dramatically increases the geometric detail possible in ray-traced scenes, accelerating BVH building to enable up to 100x …”
   • [claimed-docs] “NVIDIA Nsight Compute and Nsight System suite of tools designed to help developers optimize and increase performance of their applications.”

   proof ↗⚑ flag

   AMD Instinct MI355X✓fullprobed8/10

   AMD ships extensive official ROCm SDK documentation covering HIP runtime, OpenMP, math/communication libraries, container toolkit, framework integrations (vLLM, SGLang), and cluster/scheduling tools, with docs confirmed live and crawlable. This is a strong official SDK ecosystem AI-native developers can build against, though it lacks an OpenAPI/programmatic API spec and independent third-party validation of SDK quality beyond skepticism about software support. Missing for 10: machine-readable API spec (openapi probe 404s), independent hands-on developer corroboration of SDK usability.

   • [claimed-docs] “The foundational libraries, runtimes, and tools for GPU computing on AMD hardware — math and compute libraries, communication primitives, HI…”
   • [claimed-docs] “Full-stack documentation and recipes to deploy AI workloads on AMD GPUs using popular ROCm-enabled frameworks.”
   • [claimed-docs] “Inference * [vLLM](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/vllm.html) * [SGLang](https://rocm.docs.amd.com/…”
   • [claimed-docs] “HIP C++ Learn the HIP programming model.”
   • [claimed-docs] “OpenMP Explore the OpenMP programming model.”
   • [claimed-docs] “AMD Container Toolkit Integrate Instinct GPUs with Docker and container runtimes.”
   • [probe] “PROBE llms.txt: HTTP 200 at https://rocm.docs.amd.com/llms.txt # ROCm documentation > Note: ROCm documentation is split across multiple pro…”
   • [probe] “PROBE runtime (recorded 2026-09-15): the ROCm documentation at rocm.docs.amd.com answered a keyless curl naming ROCm — the MI355X's entire d…”

   proof ↗⚑ flag

Agentic features

1. ai-native userGet AI-generated insights and suggestions from my data inside the product

   weight 2 · round drawn
   GeForce RTX 5090—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

   AMD Instinct MI355X—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

2. ai-native userDelegate tasks to a built-in AI assistant inside the product

   weight 3 · round drawn
   GeForce RTX 5090—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

   AMD Instinct MI355X—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

3. ai-native userOperate the product with natural-language commands

   weight 2 · round drawn
   GeForce RTX 5090—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

   AMD Instinct MI355X—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

Api quality

1. ai-native userDownload a machine-readable API spec (OpenAPI or equivalent)

   weight 2 · round drawn
   GeForce RTX 5090—none0/10

   This is a consumer GPU product; while NVIDIA's developer portal exists, an explicit probe found no OpenAPI/machine-readable API spec (all candidate paths 404), so there's no evidence of a downloadable machine-readable API spec.

   • [probe] “PROBE openapi: all candidate paths 404 (https://developer.nvidia.com/openapi.json, https://developer.nvidia.com/swagger.json, https://develo…”

   proof ↗⚑ flag

   AMD Instinct MI355X—none0/10

   The evidence pack includes a direct probe showing all standard OpenAPI/Swagger endpoints return 404 on AMD's ROCm docs site, and no other citation mentions a machine-readable API spec for MI355X's software stack or management tools.

   • [probe] “PROBE openapi: all candidate paths 404 (https://rocm.docs.amd.com/openapi.json, https://rocm.docs.amd.com/swagger.json, https://rocm.docs.am…”
   • [probe] “PROBE docs-md: HTTP 404 at https://rocm.docs.amd.com/en/latest/.md”

   proof ↗⚑ flag

Inference stack

1. ai-native userThis GPU has a documented LLM inference story — low-precision formats (FP8/FP4) and supported serving stacks (TensorRT-LLM, vLLM, ROCm, llama.cpp) for this part

   weight 3 · round to AMD Instinct MI355X
   GeForce RTX 5090~partialcommunity5/10

   NVIDIA's RTX documentation confirms fifth-gen Tensor Cores with FP4/FP8/FP6 low-precision support for deep learning workloads, and community discussion references RTX 5090 use for local LLM inference (though cost concerns are raised). However, there is no evidence of explicit support/documentation for named serving stacks like TensorRT-LLM, vLLM, or llama.cpp on this specific part. Missing for 10: documented compatibility with TensorRT-LLM, vLLM, or llama.cpp serving frameworks, benchmark data showing actual inference throughput on this GPU.

   • [claimed-docs] “Fifth-generation Tensor Cores deliver up to 3x higher throughput for deep learning, with new FP4 support for massive performance gains along…”
   • [community] “"If the prices for the RTX 5090 remain at 3500€, they will likely remain insignificant for the DIY crowd" for local LLM use compared to used…”
   • [claimed-docs] “The GeForce RTX 5090 is powered by the NVIDIA Blackwell architecture and equipped with 32 GB of super-fast GDDR7 memory, so you can do it al…”

   proof ↗⚑ flag

   AMD Instinct MI355X~partialclaimed6/10

   Docs confirm ROCm-based inference stack support for vLLM and SGLang, plus explicit MXFP6/MXFP4 low-precision datatype support and large HBM3E memory for LLM inference. However, the story also asks about TensorRT-LLM and llama.cpp support, and explicit FP8 support, none of which appear in the evidence pack. missing for 10: TensorRT-LLM support evidence, llama.cpp support evidence, explicit FP8 precision documentation, independent benchmarks validating inference throughput on these stacks.

   • [claimed-docs] “Inference * [vLLM](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/vllm.html) * [SGLang](https://rocm.docs.amd.com/…”
   • [claimed-docs] “AMD Instinct™ MI355X GPUs deliver leadership AI and HPC performance enabling high density infrastructures with 288GB HBM3E memory, 8TB/s ban…”
   • [claimed-docs] “The foundational libraries, runtimes, and tools for GPU computing on AMD hardware — math and compute libraries, communication primitives, HI…”
   • [claimed-docs] “Full-stack documentation and recipes to deploy AI workloads on AMD GPUs using popular ROCm-enabled frameworks.”

   proof ↗⚑ flag

Tensor specs

1. ml engineerSize training and inference from published tensor throughput — TFLOPS or TOPS with precision and sparsity stated, not a bare marketing number

   weight 3 · round to GeForce RTX 5090
   GeForce RTX 5090~partialprobed3/10

   Docs mention fifth-gen Tensor Cores with specific precision support (FP4, TF32, BF16, FP16, FP8, FP6) and a relative claim of 'up to 3x higher throughput,' but no evidence gives an absolute TFLOPS/TOPS figure paired with sparsity state — the story explicitly wants a non-bare number with precision AND sparsity documented. Missing for 10: dense vs sparse TOPS/TFLOPS tables per precision, official spec-sheet numeric throughput figures, and independent benchmark corroboration of those numbers for ML workload sizing.

   • [claimed-docs] “Fifth-generation Tensor Cores deliver up to 3x higher throughput for deep learning, with new FP4 support for massive performance gains along…”
   • [probe] “PROBE runtime (recorded 2026-09-15): NVIDIA's RTX 5090 product/spec page answered a keyless curl and names the part — the vendor spec surfac…”

   proof ↗⚑ flag

   AMD Instinct MI355X—none0/10

   The evidence includes only bare marketing multipliers (e.g., 'Up to 2.2X AI performance') and mentions of supported datatypes (MXFP6/MXFP4) without any published absolute TFLOPS/TOPS figures broken out by precision (FP8/FP6/FP4/BF16) or with/without sparsity — exactly the kind of unqualified claim the story asks to avoid. Community sources discuss architectural comparisons qualitatively but never cite AMD's actual per-precision throughput spec table.

   • [claimed-docs] “Up to 2.2X the AI performance vs. competitive accelerators1”
   • [claimed-docs] “AMD Instinct™ MI355X GPUs deliver leadership AI and HPC performance enabling high density infrastructures with 288GB HBM3E memory, 8TB/s ban…”
   • [community] “Compared to Nvidia's B200 SMs, CDNA 4 CUs have half the per-clock throughput across many 16-bit/8-bit data types; AMD still relies on a bigg…”

   proof ↗⚑ flag

1. ai-native userPerform bulk operations across many items at once

   weight 2 · round to AMD Instinct MI355X
   GeForce RTX 5090—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

   AMD Instinct MI355X~partialclaimed4/10

   The ROCm/Instinct docs point to inference frameworks (vLLM, SGLang) that natively support batched/continuous-batch inference, implying MI355X can process many requests or items in bulk, and GPU partitioning/cluster tooling suggest large-scale parallel job handling. However, there is no direct documentation of a bulk-operation API, batch job submission interface, or AI-native bulk-processing workflow specific to MI355X itself — it's inferred through third-party software rather than demonstrated first-party capability. Missing for 10: explicit bulk/batch API documentation, benchmarked throughput for batch workloads, and independent verification of batch processing at scale.

   • [claimed-docs] “Inference * [vLLM](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/vllm.html) * [SGLang](https://rocm.docs.amd.com/…”
   • [claimed-docs] “Full-stack documentation and recipes to deploy AI workloads on AMD GPUs using popular ROCm-enabled frameworks.”
   • [claimed-docs] “GPU Partitioning Split compute units and memory to partition a single GPU.”

   proof ↗⚑ flag

2. ai-native userSchedule recurring jobs or workflows

   weight 2 · round to AMD Instinct MI355X
   GeForce RTX 5090—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

   AMD Instinct MI355X~partialclaimed4/10

   AMD ships 'Spur', described as an AI-native, Slurm-compatible job scheduler with GPU-first scheduling for Instinct clusters, which implies workflow/job scheduling capability, but the evidence never explicitly confirms recurring/cron-style job scheduling or automation-depth features. Missing for 10: documentation on recurring/cron scheduling semantics, workflow orchestration examples, and independent/hands-on validation of Spur's scheduling capabilities.

   • [claimed-docs] “Spur AI-native job scheduler, drop-in compatible with Slurm, with GPU-first scheduling and Raft-based state.”

   proof ↗⚑ flag

Media engines

1. creatorHardware media engines and creator-app acceleration are documented — AV1/HEVC encoders, and professional or ISV-certified driver support where the vendor claims it

   weight 2 · round drawn
   GeForce RTX 5090—none0/10

   The evidence pack covers CUDA, DLSS, RT Cores, and gaming features but contains no documentation of NVENC/AV1/HEVC hardware encoders, Studio Driver certification, or ISV-certified professional app support for the RTX 5090. Missing for 10: any mention of hardware encoder specs, Studio Driver program, or ISV certification for creator applications.

   ⚑ flag

   AMD Instinct MI355X—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

Scale out

1. ml engineerTrain and serve at datacenter scale on this part — documented high-bandwidth interconnect (NVLink, Infinity Fabric), multi-GPU systems, and rack-scale deployment

   weight 3 · round to AMD Instinct MI355X
   GeForce RTX 5090—none0/10

   The evidence pack contains no mention of NVLink, Infinity Fabric, multi-GPU interconnect, or rack-scale deployment for the RTX 5090; all specs describe a single consumer GPU (GDDR7, PCIe 5.0) and community commentary discusses it only as a DIY/local-LLM card, not datacenter-scale training/serving infrastructure.

   • [claimed-docs] “The GeForce RTX 5090 is powered by the NVIDIA Blackwell architecture and equipped with 32 GB of super-fast GDDR7 memory, so you can do it al…”
   • [community] “Pros: Fastest GPU around (usually), 32GB GDDR7 on 512-bit bus, PCIe 5.0, potent AI performance. Cons: Driver issues in some games/apps, extr…”
   • [community] “"If the prices for the RTX 5090 remain at 3500€, they will likely remain insignificant for the DIY crowd" for local LLM use compared to used…”

   proof ↗⚑ flag

   AMD Instinct MI355X~partialcommunity5/10

   Evidence documents rack-scale deployment tooling (Kubernetes GPU Operator, Container Toolkit, Cluster Validation Suite, Spur scheduler, SR-IOV, AMD SMI) and high memory bandwidth (8TB/s HBM3E), supporting datacenter-scale operations, but no citation explicitly documents the GPU-to-GPU interconnect fabric (e.g., Infinity Fabric link topology or scale-up fabric analogous to NVLink) that the story specifically calls out. Community sources focus on compute/memory comparisons, not interconnect topology, so this axis is only partially evidenced. Missing for 10: explicit Infinity Fabric/interconnect bandwidth specs and multi-GPU topology documentation, independent multi-node training benchmarks confirming interconnect scaling.

   • [claimed-docs] “GPU Operator Deploy and manage Instinct GPUs in Kubernetes clusters.”
   • [claimed-docs] “Spur AI-native job scheduler, drop-in compatible with Slurm, with GPU-first scheduling and Raft-based state.”
   • [claimed-docs] “Cluster Validation Suite Test scripts that validate AMD AI clusters end to end.”
   • [claimed-docs] “AMD Container Toolkit Integrate Instinct GPUs with Docker and container runtimes.”
   • [claimed-docs] “SR-IOV Yes”
   • [claimed-docs] “AMD Instinct™ MI355X GPUs deliver leadership AI and HPC performance enabling high density infrastructures with 288GB HBM3E memory, 8TB/s ban…”
   • [community] “MI355X's HBM3E subsystem gives it 288GB capacity at 8 TB/s vs Nvidia B200's 180GB at 7.7 TB/s, maintaining AMD's memory capacity/bandwidth l…”

   proof ↗⚑ flag

Linux support

1. developerLinux is a first-class citizen for this GPU — documented Linux driver releases and independent Linux testing of this part

   weight 2 · round to AMD Instinct MI355X
   GeForce RTX 5090~partialcommunity5/10

   Independent Phoronix benchmarking confirms the RTX 5090 is tested and functional on Linux across 60+ compute workloads, showing real-world Linux usability, but the evidence pack contains no first-party documentation of dedicated Linux driver releases, changelogs, or Linux-specific support pages from NVIDIA. Missing for 10: documented NVIDIA Linux driver release notes/changelog, official Linux support/compatibility docs, broader independent Linux gaming/compute test corroboration beyond one Phoronix reference.

   • [community] “Phoronix Linux benchmarks found the GeForce RTX 5090 delivering 1.42x the performance of the RTX 4090 across 60+ compute benchmarks, but on …”

   proof ↗⚑ flag

   AMD Instinct MI355X~partialcommunity6/10

   ROCm's entire documented stack (HIP, container toolkit, Kubernetes GPU operator, AMD SMI, cluster validation) is Linux-native and independent hands-on benchmarking (chipsandcheese architecture deep-dive, HN wafer.ai comparisons) treats MI355X as a real, testable part. However, no evidence cites explicit Linux kernel driver release notes/versioning or independent Linux-specific driver validation reports. Missing for 10: explicit Linux driver release notes/changelog, independent third-party Linux driver-level testing (not just architecture/benchmark commentary), confirmation that cited benchmarks ran on Linux.

   • [claimed-docs] “The foundational libraries, runtimes, and tools for GPU computing on AMD hardware — math and compute libraries, communication primitives, HI…”
   • [claimed-docs] “HIP C++ Learn the HIP programming model.”
   • [claimed-docs] “AMD Container Toolkit Integrate Instinct GPUs with Docker and container runtimes.”
   • [claimed-docs] “GPU Operator Deploy and manage Instinct GPUs in Kubernetes clusters.”
   • [community] “Compared to Nvidia's B200 SMs, CDNA 4 CUs have half the per-clock throughput across many 16-bit/8-bit data types; AMD still relies on a bigg…”
   • [community] “On every row the B300 beat the MI355X in a wafer.ai comparison; critics say the price comparison (MI355X at $2.5/hr) is unrealistic since no…”

   proof ↗⚑ flag

Open drivers

1. developerRun this GPU on an open driver — open-source kernel modules or upstream Linux support documented by the vendor

   weight 2 · round drawn
   GeForce RTX 5090—none0/10

   The evidence pack contains no mention of open-source kernel modules (e.g., NVIDIA's open GPU kernel module project) or upstream Linux driver support for the RTX 5090; all driver-related community mentions are about closed-driver bugs/issues, not open-source availability. This is a fair axis for a GPU (vendors like NVIDIA do ship open kernel modules), but nothing in the evidence documents it for this card.

   ⚑ flag

   AMD Instinct MI355X—none0/10

   The evidence pack extensively documents ROCm's open-source user-space stack (libraries, runtimes, tools, monitoring, Kubernetes operator) but never mentions the amdgpu kernel driver, its open-source licensing, or upstream Linux kernel inclusion — the specific claim this story asks about is absent. Missing for 10: any vendor documentation of open-source kernel modules, upstream kernel driver support, or distro-inclusion status for the MI355X.

   ⚑ flag

4k gaming

1. gamerThis card drives high-refresh 4K gaming — vendor performance claims corroborated by independent game benchmarks

   weight 3 · round to GeForce RTX 5090
   GeForce RTX 5090~partialcommunity6/10

   Independent Tom's Hardware and Phoronix benchmarks corroborate the RTX 5090 as the fastest consumer GPU (1.42x over 4090), supporting vendor claims of top-tier gaming performance, and NVIDIA's DLSS/Reflex/path-tracing feature docs target high-refresh 4K use cases. However, community evidence raises real caveats: reviewers flag Founders Edition thermal issues, similar-or-worse perf-per-watt vs 4080/4090, and specifically critique Multi Frame Generation as 'marketing' since input sampling doesn't scale with the reported FPS boost, undercutting the headline frame-rate claims. missing for 10: dedicated 4K high-refresh benchmark suites (e.g. specific FPS-at-4K numbers across many AAA titles), resolution of the MFG input-latency criticism, and confirmation thermal/efficiency issues don't limit sustained high-refresh performance.

   • [claimed-docs] “Powered by GeForce RTX 50 Series and fifth-generation Tensor Cores, new DLSS Multi Frame Generation boosts FPS by using AI to generate up to…”
   • [claimed-docs] “Reflex technologies optimize the graphics pipeline for ultimate responsiveness, providing faster target acquisition, quicker reaction times,…”
   • [claimed-docs] “The NVIDIA Blackwell architecture unlocks the game-changing realism of path tracing. Experience cinematic quality visuals at unprecedented s…”
   • [community] “Pros: Fastest GPU around (usually), 32GB GDDR7 on 512-bit bus, PCIe 5.0, potent AI performance. Cons: Driver issues in some games/apps, extr…”
   • [community] “We've now docked half a star, due to concerns specifically with the Founders Edition running hot.”
   • [community] “MFG as an example running at 240 FPS would mean user input only gets sampled at 60 FPS. That's not the same as a game running at 240 FPS nat…”
   • [community] “Phoronix Linux benchmarks found the GeForce RTX 5090 delivering 1.42x the performance of the RTX 4090 across 60+ compute benchmarks, but on …”

   proof ↗⚑ flag

   AMD Instinct MI355X—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

Upscaling

1. gamerAI upscaling and frame generation are supported on this card — the DLSS or FSR generation is documented for this part, with broad game support

   weight 2 · round to GeForce RTX 5090
   GeForce RTX 5090✓fullcommunity9/10

   NVIDIA's official docs detail DLSS 4 with Multi Frame Generation, Super Resolution, Ray Reconstruction, and DLAA on RTX 5090/50-series, plus broad game rollout via the NVIDIA app supporting hundreds of titles. Independent review (Tom's Hardware) confirms these features work in practice, though it flags marketing caveats around Multi Frame Generation's input latency implications. Missing for 10: independent per-game compatibility list/count and more third-party benchmarking corroboration beyond one review.

   • [claimed-docs] “Powered by GeForce RTX 50 Series and fifth-generation Tensor Cores, new DLSS Multi Frame Generation boosts FPS by using AI to generate up to…”
   • [claimed-docs] “Dynamically adjust your multiplier to maximize smoothness across different games and scenes on GeForce RTX 50 Series GPUs.”
   • [claimed-docs] “Enhances image quality by using AI to generate additional pixels for intensive ray-traced scenes.”
   • [claimed-docs] “Boosts performance by using AI to output higher-resolution frames from a lower-resolution input.”
   • [claimed-docs] “Provides higher image quality with an AI-based anti-aliasing technique. DLAA uses the same Super Resolution technology developed for DLSS, c…”
   • [claimed-docs] “With the NVIDIA app you can update hundreds of games to use the latest DLSS features including Multi Frame Generation, and the newest AI mod…”
   • [community] “Pros: Fastest GPU around (usually), 32GB GDDR7 on 512-bit bus, PCIe 5.0, potent AI performance. Cons: Driver issues in some games/apps, extr…”
   • [community] “MFG as an example running at 240 FPS would mean user input only gets sampled at 60 FPS. That's not the same as a game running at 240 FPS nat…”

   proof ↗⚑ flag

   AMD Instinct MI355X—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

Llm memory

1. ai-native userRun a 70B-class quantized LLM on this GPU — published VRAM capacity and memory bandwidth that make local or single-node inference practical

   weight 3 · round to AMD Instinct MI355X
   GeForce RTX 5090~partialcommunity6/10

   NVIDIA publishes the RTX 5090's 32GB GDDR7 VRAM (docs-18) and strong Tensor Core throughput (docs-22), which is enough for some 70B-class models at aggressive quantization, but the evidence never states a specific memory-bandwidth figure and community discussion (comm-5) notes the GPU's price makes it 'insignificant for the DIY crowd' compared to used 3090s or Mac unified memory for local LLM work, undercutting the 'practical' framing. Missing for 10: published memory-bandwidth spec (GB/s), explicit vendor guidance on running 70B-class quantized models, and independent benchmarks confirming inference throughput at that scale.

   • [claimed-docs] “The GeForce RTX 5090 is powered by the NVIDIA Blackwell architecture and equipped with 32 GB of super-fast GDDR7 memory, so you can do it al…”
   • [claimed-docs] “Fifth-generation Tensor Cores deliver up to 3x higher throughput for deep learning, with new FP4 support for massive performance gains along…”
   • [community] “"If the prices for the RTX 5090 remain at 3500€, they will likely remain insignificant for the DIY crowd" for local LLM use compared to used…”

   proof ↗⚑ flag

   AMD Instinct MI355X✓fullcommunity8/10

   AMD publishes 288GB HBM3E capacity and 8TB/s bandwidth for MI355X, far exceeding what's needed for 70B-class quantized inference, and documents vLLM/SGLang inference stacks for deployment; independent analysis corroborates the memory capacity/bandwidth lead over competing accelerators. Missing for 10: no direct hands-on benchmark or published throughput numbers specifically for a 70B model at a given quantization on this GPU.

   • [claimed-docs] “AMD Instinct™ MI355X GPUs deliver leadership AI and HPC performance enabling high density infrastructures with 288GB HBM3E memory, 8TB/s ban…”
   • [claimed-docs] “Inference * [vLLM](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/vllm.html) * [SGLang](https://rocm.docs.amd.com/…”
   • [community] “MI355X's HBM3E subsystem gives it 288GB capacity at 8 TB/s vs Nvidia B200's 180GB at 7.7 TB/s, maintaining AMD's memory capacity/bandwidth l…”

   proof ↗⚑ flag

Memory spec

1. ml engineerMemory specs are published in full for this exact part — capacity, memory type, bus width, and bandwidth

   weight 2 · round to GeForce RTX 5090
   GeForce RTX 5090~partialprobed7/10

   NVIDIA's own product page confirms capacity (32GB) and memory type (GDDR7), and community review corroborates a 512-bit bus, but no evidence pack item states the exact memory bandwidth figure (GB/s) for this part. Missing for 10: an explicit published bandwidth number (GB/s) from a first-party spec sheet.

   • [claimed-docs] “The GeForce RTX 5090 is powered by the NVIDIA Blackwell architecture and equipped with 32 GB of super-fast GDDR7 memory, so you can do it al…”
   • [community] “Pros: Fastest GPU around (usually), 32GB GDDR7 on 512-bit bus, PCIe 5.0, potent AI performance. Cons: Driver issues in some games/apps, extr…”
   • [probe] “PROBE runtime (recorded 2026-09-15): NVIDIA's RTX 5090 product/spec page answered a keyless curl and names the part — the vendor spec surfac…”

   proof ↗⚑ flag

   AMD Instinct MI355X~partialcommunity6/10

   AMD's official product page confirms capacity (288GB), memory type (HBM3E), and bandwidth (8TB/s) for the MI355X, corroborated independently by chipsandcheese's comparison data. However, no evidence anywhere states the memory bus width. missing for 10: published bus width (bits) for the HBM3E memory subsystem.

   • [claimed-docs] “AMD Instinct™ MI355X GPUs deliver leadership AI and HPC performance enabling high density infrastructures with 288GB HBM3E memory, 8TB/s ban…”
   • [community] “MI355X's HBM3E subsystem gives it 288GB capacity at 8 TB/s vs Nvidia B200's 180GB at 7.7 TB/s, maintaining AMD's memory capacity/bandwidth l…”

   proof ↗⚑ flag

1. ai-native userRead the product's source under an open license

   weight 2 · round drawn
   GeForce RTX 5090—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

   AMD Instinct MI355X—none0/10

   The evidence pack describes ROCm documentation, libraries, and tools but never states an open-source license or provides a link to source code repositories; no license terms are mentioned anywhere in the pack.

   ⚑ flag

2. ai-native userSelf-host the core product

   weight 3 · round to AMD Instinct MI355X
   GeForce RTX 5090—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

   AMD Instinct MI355X✓fullcommunity9/10

   The MI355X is sold as on-prem hardware, and AMD provides a full deployment stack for self-hosting: ROCm runtime/libraries, GPU Operator for Kubernetes, AMD Container Toolkit for Docker, Cluster Validation Suite, Device Metrics Exporter, and the Spur scheduler, all documented for running the GPU in a customer-owned datacenter. Community commentary corroborates real-world self-hosted comparisons (e.g., wafer.ai benchmarks) confirming the hardware is deployed and operated independently by third parties. Missing for 10: independent hands-on report specifically walking through a full self-host bring-up (rather than benchmark-only) confirming ease of deployment.

   • [claimed-docs] “GPU Operator Deploy and manage Instinct GPUs in Kubernetes clusters.”
   • [claimed-docs] “AMD Container Toolkit Integrate Instinct GPUs with Docker and container runtimes.”
   • [claimed-docs] “Cluster Validation Suite Test scripts that validate AMD AI clusters end to end.”
   • [claimed-docs] “Device Metrics Exporter Prometheus-format GPU metrics for HPC and AI environments.”
   • [claimed-docs] “Spur AI-native job scheduler, drop-in compatible with Slurm, with GPU-first scheduling and Raft-based state.”
   • [community] “On every row the B300 beat the MI355X in a wafer.ai comparison; critics say the price comparison (MI355X at $2.5/hr) is unrealistic since no…”

   proof ↗⚑ flag

Efficiency

1. ml engineerSustained workloads are power-efficient on this part — documented power envelopes with independent performance-per-watt testing

   weight 2 · round to GeForce RTX 5090
   GeForce RTX 5090!disputedcontradicted5/10

   NVIDIA's spec page exposes board power/TOPS figures (probe-rt-1) but no explicit efficiency/perf-per-watt marketing claim is cited; independent testing (Phoronix via HN, comm-6/7/8) directly contradicts any efficiency narrative, finding RTX 5090 perf-per-watt is similar to or worse than the 4080/4090 despite higher absolute performance, and Tom's Hardware also flags thermal/power concerns (comm-2). Missing for 10: a documented power-envelope/TDP spec sheet from NVIDIA explicitly framed for ML workloads, and independent perf-per-watt benchmarks that confirm rather than undercut efficiency gains.

   • [probe] “PROBE runtime (recorded 2026-09-15): NVIDIA's RTX 5090 product/spec page answered a keyless curl and names the part — the vendor spec surfac…”
   • [community] “TL;DR; performance isn't bad, but perf per Watt isn't better than 4080 or 4090 and can even be significantly lower than 4090 in certain cont…”
   • [community] “"the 4080 Super doing well compared to the 5080 and 5090... seems to have better performance per watt ratio than them while also having some…”
   • [community] “Phoronix Linux benchmarks found the GeForce RTX 5090 delivering 1.42x the performance of the RTX 4090 across 60+ compute benchmarks, but on …”
   • [community] “We've now docked half a star, due to concerns specifically with the Founders Edition running hot.”

   proof ↗⚑ flag

   AMD Instinct MI355X—none0/10

   The evidence pack lacks any documented TDP/power envelope specs or independent performance-per-watt benchmarking for the MI355X; community sources focus on raw throughput and memory bandwidth comparisons (chipsandcheese, wafer.ai) but never normalize for power draw. Missing for 10: official power envelope/TDP documentation, independent perf-per-watt benchmarks, sustained-workload thermal/power test data.

   • [community] “Compared to Nvidia's B200 SMs, CDNA 4 CUs have half the per-clock throughput across many 16-bit/8-bit data types; AMD still relies on a bigg…”
   • [community] “MI355X's HBM3E subsystem gives it 288GB capacity at 8 TB/s vs Nvidia B200's 180GB at 7.7 TB/s, maintaining AMD's memory capacity/bandwidth l…”
   • [community] “On every row the B300 beat the MI355X in a wafer.ai comparison; critics say the price comparison (MI355X at $2.5/hr) is unrealistic since no…”

   proof ↗⚑ flag

Psu planning

1. gamerSpec a build around published board power — TDP/TGP, connector requirements, and cooling guidance for this exact card

   weight 2 · round to GeForce RTX 5090
   GeForce RTX 5090!disputedcontradicted4/10

   Evidence confirms NVIDIA's official 5090 page exposes board-power and spec data (probe-rt-1) and the community references the 12VHPWR/12V-2x6 connector spec, but no evidence pack item actually cites the published TDP/TGP wattage, PSU wattage recommendation, or explicit cooling/case guidance for this exact card. More importantly, community evidence (comm-4) documents real-world 12VHPWR connector melting incidents tied to the card's power-connector design, and comm-2 reports the Founders Edition running hot — concretely undermining confidence that a build spec'd purely around the vendor's connector/cooling guidance will be reliable. missing for 10: explicit first-party TDP/TGP wattage figure, official PSU/connector spec sheet, first-party cooling/case airflow guidance, resolution of the connector-melting safety concern.

   • [probe] “PROBE runtime (recorded 2026-09-15): NVIDIA's RTX 5090 product/spec page answered a keyless curl and names the part — the vendor spec surfac…”
   • [community] “Discussion centers on a reported RTX 5090 FE 12VHPWR connector melting; users debate that the FE's single 12V bus-bar design relies purely o…”
   • [community] “We've now docked half a star, due to concerns specifically with the Founders Edition running hot.”
   • [claimed-docs] “The GeForce RTX 5090 is powered by the NVIDIA Blackwell architecture and equipped with 32 GB of super-fast GDDR7 memory, so you can do it al…”

   proof ↗⚑ flag

   AMD Instinct MI355X—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

1. ai-native userPrevent my data from being used to train AI models

   weight 3 · round drawn
   GeForce RTX 5090—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

   AMD Instinct MI355X—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

Compute stack

1. developerShip GPU-compute workloads on the vendor's toolchain — CUDA or ROCm/HIP documentation lists this part as a supported target

   weight 3 · round to AMD Instinct MI355X
   GeForce RTX 5090✓fullcommunity8/10

   NVIDIA's CUDA Toolkit docs explicitly target GeForce RTX GPUs including Blackwell-architecture cards like the 5090, and independent Phoronix benchmarks (cited via comm-8) confirm real compute workloads running on the 5090 with 1.42x uplift over the 4090 across 60+ compute benchmarks, corroborating that it functions as a genuine CUDA compute target. Missing for 10: no explicit CUDA compute-capability/SM version listing naming '5090' directly in vendor docs, and no independent report specifically validating professional GPU-compute (non-gaming) toolchains beyond Phoronix's Linux compute suite.

   • [claimed-docs] “The NVIDIA® CUDA® Toolkit provides a development environment for creating high-performance, GPU-accelerated applications.”
   • [claimed-docs] “CUDA Tile C++ is an expression of the CUDA Tile programming model in C++. It's built on top of the CUDA Tile IR specification and allows you…”
   • [claimed-docs] “cuTile Python is an expression of the CUDA Tile programming model in Python.”
   • [claimed-docs] “Fifth-generation Tensor Cores deliver up to 3x higher throughput for deep learning, with new FP4 support for massive performance gains along…”
   • [community] “Phoronix Linux benchmarks found the GeForce RTX 5090 delivering 1.42x the performance of the RTX 4090 across 60+ compute benchmarks, but on …”

   proof ↗⚑ flag

   AMD Instinct MI355X✓fullprobed9/10

   ROCm documentation explicitly covers HIP runtime, libraries, frameworks, and inference stacks (vLLM, SGLang) targeting Instinct GPUs, and Instinct-specific docs list HIP C++, OpenMP, and other toolchain components for the MI355X line, confirming it as a supported ROCm/HIP target. missing for 10: no explicit MI355X-named code sample or compatibility matrix entry pinpointing this exact SKU rather than the Instinct family generally.

   • [claimed-docs] “The foundational libraries, runtimes, and tools for GPU computing on AMD hardware — math and compute libraries, communication primitives, HI…”
   • [claimed-docs] “Inference * [vLLM](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/vllm.html) * [SGLang](https://rocm.docs.amd.com/…”
   • [claimed-docs] “HIP C++ Learn the HIP programming model.”
   • [claimed-docs] “OpenMP Explore the OpenMP programming model.”
   • [probe] “PROBE runtime (recorded 2026-09-15): the ROCm documentation at rocm.docs.amd.com answered a keyless curl naming ROCm — the MI355X's entire d…”

   proof ↗⚑ flag

Frameworks

1. ml engineerPyTorch and mainstream ML frameworks run on this GPU through officially documented builds and support matrices

   weight 2 · round to AMD Instinct MI355X
   GeForce RTX 5090~partialcommunity3/10

   Evidence confirms NVIDIA ships a general CUDA Toolkit and Tensor Core architecture details relevant to deep learning, but there is no explicit official PyTorch/TensorFlow support matrix, compute-capability listing, or documented install path referencing RTX 5090/Blackwell specifically; a community post even flags 5090 as impractical for local LLM work due to price, not toolchain issues. missing for 10: explicit PyTorch build/support-matrix docs naming RTX 5090 or Blackwell (sm_120) compute capability, cuDNN/cuBLAS version compatibility notes, and independent confirmation of PyTorch working out-of-the-box on the card.

   • [claimed-docs] “The NVIDIA® CUDA® Toolkit provides a development environment for creating high-performance, GPU-accelerated applications.”
   • [claimed-docs] “Fifth-generation Tensor Cores deliver up to 3x higher throughput for deep learning, with new FP4 support for massive performance gains along…”
   • [community] “"If the prices for the RTX 5090 remain at 3500€, they will likely remain insignificant for the DIY crowd" for local LLM use compared to used…”

   proof ↗⚑ flag

   AMD Instinct MI355X✓fullprobed8/10

   AMD documents ROCm-enabled framework support with specific recipes for PyTorch-adjacent ML stacks and inference frameworks (vLLM, SGLang) tied to Instinct GPUs, backed by a live, crawlable documentation surface. Missing for 10: an explicit official PyTorch build/support-matrix page or version-compatibility table directly citing PyTorch, and independent hands-on confirmation of PyTorch working smoothly on MI355X specifically (only general software-support skepticism exists in community commentary).

   • [claimed-docs] “Full-stack documentation and recipes to deploy AI workloads on AMD GPUs using popular ROCm-enabled frameworks.”
   • [claimed-docs] “Inference * [vLLM](https://rocm.docs.amd.com/projects/ai-ecosystem/en/latest/inference/vllm.html) * [SGLang](https://rocm.docs.amd.com/…”
   • [claimed-docs] “The foundational libraries, runtimes, and tools for GPU computing on AMD hardware — math and compute libraries, communication primitives, HI…”
   • [probe] “PROBE runtime (recorded 2026-09-15): the ROCm documentation at rocm.docs.amd.com answered a keyless curl naming ROCm — the MI355X's entire d…”
   • [community] “On the MI355X (288GB HBM3E) and MI325X announcement, a commenter noted AMD's pricing looks good vs Nvidia H100/B100 at around $15k, but expr…”

   proof ↗⚑ flag