GPUs & AI Accelerators arenaGPUs & AI Accelerators
Current GPUs and AI accelerators — GeForce RTX 50-series, Radeon RDNA 4, RTX PRO workstations, and H200/B200/Instinct MI355X in the datacenter — judged on stated tensor throughput, VRAM, toolchain openness (CUDA/ROCm), gaming performance, and datacenter scale.
43 user stories · 344 judged cells · updated 2026-09-15 · Evidence as of 2026-09-15
Leaderboard — every product ranked by evidenceLeaderboard
Rank by
| 1 |  AMD Instinct MI355Xoem-channel vs Radeon RX 9070 XT ↗ | n/a | untested | n/a | 54/100 | 24/100 | 10/15 verified | 71/100 integrity | |||
| 2 |  Radeon RX 9070 XThardware-purchase vs AMD Instinct MI355X ↗ | n/a | untested | n/a | 54/100 | untested | 6/14 verified | 0/100 integrity | |||
| 3 |  NVIDIA B200oem-channel vs AMD Instinct MI355X ↗ | n/a | untested | n/a | 48/100 | untested | 7/9 verified | 100/100 integrity | |||
| 4 |  RTX PRO 6000 Blackwelloem-channel vs AMD Instinct MI355X ↗ | n/a | untested | n/a | 42/100 | untested | 6/10 verified · 2 disputed | 75/100 integrity | |||
| 5 |  GeForce RTX 5070 Tihardware-purchase vs AMD Instinct MI355X ↗ | n/a | 20/100 | n/a | untested | untested | 5/11 verified | 40/100 integrity | |||
| 6 |  NVIDIA H200 (SXM)oem-channel vs AMD Instinct MI355X ↗ | n/a | untested | n/a | 42/100 | untested | 11/11 verified | 100/100 integrity | |||
| 7 |  GeForce RTX 5080hardware-purchase vs AMD Instinct MI355X ↗ | n/a | 27/100 | n/a | untested | untested | 5/12 verified | 20/100 integrity | |||
| 8 |  GeForce RTX 5090hardware-purchase vs AMD Instinct MI355X ↗ | n/a | untested | n/a | untested | untested | 11/14 verified · 2 disputed | 100/100 integrity |
Best by user type — persona-weighted winnersBest by user type
Per persona, the product with the highest persona-weighted coverage over just that persona's stories — not the same ranking as the overall PA Score leaderboard above.
Best for developer
Radeon RX 9070 XT
55/100
Runner-up: AMD Instinct MI355X (49/100)
3 developer stories scored
Best for creator
GeForce RTX 5070 Ti
36/100
Runner-up: RTX PRO 6000 Blackwell (36/100)
1 creator story scored
Best for ml-engineer
NVIDIA H200 (SXM)
30/100
Runner-up: AMD Instinct MI355X (27/100)
5 ml-engineer stories scored
Best for ai-native
AMD Instinct MI355X
29/100
Runner-up: Radeon RX 9070 XT (22/100)
31 ai-native stories scored
Story matrix — every product × every judged storyStory matrix
43/43 stories shown · legend
Agenticness — how well agents can access and operate the productAgenticness
Agent access
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Agenticness — how well agents can access and operate the productPoint an agent at llms.txt or agent-oriented docs | ai-native | partialT 5/10 | partialT 6/10 | fullT 7/10 | fullT 8/10 | partialT 6/10 | partialT 5/10 | fullT 8/10 | fullT 7/10 |
| Agenticness — how well agents can access and operate the productRun the product headlessly / in CI for automation | ai-native | none 0/10 | none 0/10 | none 0/10 | partialC 4/10 | none 0/10 | none 0/10 | partialC 6/10 | partialC 6/10 |
| Agenticness — how well agents can access and operate the productPlug MCP servers into this product so it can use their tools | ai-native | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
| Agenticness — how well agents can access and operate the productConnect an agent via an official MCP server | ai-native | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
| Agenticness — how well agents can access and operate the productUse an official CLI | ai-native | none 0/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | partialC 5/10 |
| Agenticness — how well agents can access and operate the productDrive the product through a documented public API | ai-native | partialT 3/10 | partialT 4/10 | none 0/10 | partialT 4/10 | none 0/10 | none 0/10 | partialT 4/10 | partialT 4/10 |
| Agenticness — how well agents can access and operate the productIssue scoped/least-privilege API credentials for an agent | ai-native | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
| Agenticness — how well agents can access and operate the productBuild against official SDKs | ai-native | fullC 7/10 | fullT 7/10 | fullT 7/10 | fullT 7/10 | partialX 6/10 | partialT 5/10 | none 0/10 | fullT 8/10 |
| Agenticness — how well agents can access and operate the productSubscribe to events via webhooks | ai-native | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
Agentic features
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Agenticness — how well agents can access and operate the productGet AI-generated insights and suggestions from my data inside the product | ai-native | none 0/10 | partialC 4/10 | partialC 4/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 |
| Agenticness — how well agents can access and operate the productSet up automations that run autonomously in the background | ai-native | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
| Agenticness — how well agents can access and operate the productDelegate tasks to a built-in AI assistant inside the product | ai-native | none 0/10 | partialC 5/10 | partialC 5/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 |
| Agenticness — how well agents can access and operate the productOperate the product with natural-language commands | ai-native | none 0/10 | partialC 4/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 |
Api quality
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Agenticness — how well agents can access and operate the productExplore an interactive API reference with runnable examples | ai-native | n/a | none 0/10 | n/a | n/a | n/a | n/a | n/a | none 0/10 |
| Agenticness — how well agents can access and operate the productDownload a machine-readable API spec (OpenAPI or equivalent) | ai-native | none 0/10 | n/a | n/a | n/a | n/a | none 0/10 | none 0/10 | none 0/10 |
| Agenticness — how well agents can access and operate the productTest against a sandbox environment without touching production data | ai-native | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
| Agenticness — how well agents can access and operate the productRely on versioned APIs with a documented deprecation policy | ai-native | n/a | n/a | n/a | n/a | n/a | n/a | n/a | none 0/10 |
Ai compute — stories about ai compute in this arenaAi compute
Inference stack
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Ai compute — stories about ai compute in this arenaThis 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 | ai-native | partialX 5/10 | partialC 4/10 | partialC 3/10 | partialC 5/10 | disputedD 4/10 | partialT 4/10 | partialX 5/10 | partialC 6/10 |
Tensor specs
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Ai compute — stories about ai compute in this arenaSize training and inference from published tensor throughput — TFLOPS or TOPS with precision and sparsity stated, not a bare marketing number | ml-engineer | partialT 3/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | partialT 4/10 | none 0/10 | none 0/10 |
Automation depth — how much of the product can run unattendedAutomation depth
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Automation depth — how much of the product can run unattendedPerform bulk operations across many items at once | ai-native | none 0/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | partialC 4/10 |
| Automation depth — how much of the product can run unattendedDefine rules that trigger actions automatically on events | ai-native | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
| Automation depth — how much of the product can run unattendedSchedule recurring jobs or workflows | ai-native | none 0/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | partialC 4/10 |
| Automation depth — how much of the product can run unattendedVersion, review, and roll back my automations | ai-native | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
Creator media — stories about creator media in this arenaCreator media
Media engines
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Creator media — stories about creator media in this arenaHardware media engines and creator-app acceleration are documented — AV1/HEVC encoders, and professional or ISV-certified driver support where the vendor claims it | creator | none 0/10 | partialC 3/10 | partialC 6/10 | partialC 4/10 | partialC 6/10 | none 0/10 | none 0/10 | none 0/10 |
Datacenter scale — stories about datacenter scale in this arenaDatacenter scale
Scale out
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Datacenter scale — stories about datacenter scale in this arenaTrain and serve at datacenter scale on this part — documented high-bandwidth interconnect (NVLink, Infinity Fabric), multi-GPU systems, and rack-scale deployment | ml-engineer | none 0/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | partialX 6/10 | fullX 8/10 | partialX 5/10 |
Driver openness — stories about driver openness in this arenaDriver openness
Linux support
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Driver openness — stories about driver openness in this arenaLinux is a first-class citizen for this GPU — documented Linux driver releases and independent Linux testing of this part | developer | partialX 5/10 | none 0/10 | none 0/10 | partialC 6/10 | none 0/10 | none 0/10 | partialX 6/10 | partialX 6/10 |
Open drivers
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Driver openness — stories about driver openness in this arenaRun this GPU on an open driver — open-source kernel modules or upstream Linux support documented by the vendor | developer | none 0/10 | none 0/10 | none 0/10 | partialC 6/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 |
Gaming performance — stories about gaming performance in this arenaGaming performance
4k gaming
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Gaming performance — stories about gaming performance in this arenaThis card drives high-refresh 4K gaming — vendor performance claims corroborated by independent game benchmarks | gamer | partialX 6/10 | partialX 6/10 | partialX 4/10 | partialX 3/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 |
Upscaling
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Gaming performance — stories about gaming performance in this arenaAI upscaling and frame generation are supported on this card — the DLSS or FSR generation is documented for this part, with broad game support | gamer | fullX 9/10 | fullX 9/10 | fullX 9/10 | partialX 5/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 |
Memory vram — stories about memory vram in this arenaMemory vram
Llm memory
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Memory vram — stories about memory vram in this arenaRun a 70B-class quantized LLM on this GPU — published VRAM capacity and memory bandwidth that make local or single-node inference practical | ai-native | partialX 6/10 | none 0/10 | none 0/10 | none 0/10 | fullX 8/10 | fullT 9/10 | none 0/10 | fullX 8/10 |
Memory spec
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Memory vram — stories about memory vram in this arenaMemory specs are published in full for this exact part — capacity, memory type, bus width, and bandwidth | ml-engineer | partialT 7/10 | none 0/10 | none 0/10 | none 0/10 | partialX 4/10 | partialT 6/10 | none 0/10 | partialX 6/10 |
Openness — open source, data portability, and self-hosting storiesOpenness
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Openness — open source, data portability, and self-hosting storiesDo everything through the API that I can do in the UI | ai-native | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
| Openness — open source, data portability, and self-hosting storiesExport all of my data in open formats and leave | ai-native | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
| Openness — open source, data portability, and self-hosting storiesRead the product's source under an open license | ai-native | none 0/10 | none 0/10 | none 0/10 | partialT 5/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 |
| Openness — open source, data portability, and self-hosting storiesSelf-host the core product | ai-native | none 0/10 | none 0/10 | none 0/10 | fullC 7/10 | fullX 7/10 | fullX 7/10 | fullC 8/10 | fullX 9/10 |
Power cooling — stories about power cooling in this arenaPower cooling
Efficiency
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Power cooling — stories about power cooling in this arenaSustained workloads are power-efficient on this part — documented power envelopes with independent performance-per-watt testing | ml-engineer | disputedD 5/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | partialX 3/10 | none 0/10 | none 0/10 |
Psu planning
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Power cooling — stories about power cooling in this arenaSpec a build around published board power — TDP/TGP, connector requirements, and cooling guidance for this exact card | gamer | disputedD 4/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 | none 0/10 |
Privacy posture — data-handling and privacy storiesPrivacy posture
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Privacy posture — data-handling and privacy storiesChoose where my data is stored (region/residency) | ai-native | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
| Privacy posture — data-handling and privacy storiesPrevent my data from being used to train AI models | ai-native | none 0/10 | none 0/10 | partialC 4/10 | none 0/10 | partialC 3/10 | none 0/10 | none 0/10 | none 0/10 |
| Privacy posture — data-handling and privacy storiesControl data retention and deletion | ai-native | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
| Privacy posture — data-handling and privacy storiesOpt out of telemetry and usage tracking | ai-native | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a |
Software toolchain — stories about software toolchain in this arenaSoftware toolchain
Compute stack
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Software toolchain — stories about software toolchain in this arenaShip GPU-compute workloads on the vendor's toolchain — CUDA or ROCm/HIP documentation lists this part as a supported target | developer | fullX 8/10 | partialC 6/10 | partialT 6/10 | fullC 8/10 | fullT 8/10 | fullT 8/10 | partialX 6/10 | fullT 9/10 |
Frameworks
| Story | Persona | GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | Radeon RX 9070 XT | RTX PRO 6000 Blackwell | NVIDIA H200 (SXM) | NVIDIA B200 | AMD Instinct MI355X |
|---|---|---|---|---|---|---|---|---|---|
| Software toolchain — stories about software toolchain in this arenaPyTorch and mainstream ML frameworks run on this GPU through officially documented builds and support matrices | ml-engineer | partialX 3/10 | partialC 4/10 | partialC 5/10 | fullC 8/10 | disputedD 4/10 | partialX 6/10 | partialX 5/10 | fullT 8/10 |
Verdict✓ fullclear evidence~ partialwith caveats! disputedevidence conflicts— noneno evidence foundn/aquestion doesn't apply to this kind of product
ProofT probedtested by usX communityusers back itC claimedvendor claim onlyD contradictedevidence disagrees⚿ auth-gatedprobe hit a live sign-in wall — verified reachable, untestable keylessly
quality 0–10 · PA Score /100 · A–D = evidence confidence · full guideAdjacent arenas — categories often shopped togetherAdjacent arenas
Shopping this category often means shopping these too.