feat(gpu): render RTX PRO 6000 Blackwell benchmark data / 新增 RTX PRO 6000 Blackwell GPU SKU 渲染支持#618
Merged
Merged
Conversation
Add the NVIDIA RTX PRO 6000 Blackwell Server Edition (GB202, PCIe Gen5, 96 GB GDDR7) to HW_REGISTRY so MiniMax-M3 benchmarks on this SKU render across the inference, evaluation, and calculator charts. Display labels, colors, sort order, and cost/power lookups all derive from this single entry, so no other frontend edits are needed. hwToGpuKey already strips the new `-lat` artifact suffix via its first-segment logic; added a normalizer test to lock that in. All-in power and the three cost tiers are 9.99 placeholders pending official values from the SemiAnalysis AI Cloud TCO model; TDP is the datasheet 600 W max board power. $/token and per-MW metrics for this SKU are not meaningful until the real numbers land. 中文：将 NVIDIA RTX PRO 6000 Blackwell Server Edition（GB202，PCIe Gen5， 96 GB GDDR7）加入 HW_REGISTRY，使该 SKU 上的 MiniMax-M3 基准测试能在推理、 评估与计算器（calculator）图表中正常渲染——展示标签、配色、排序及成本/功耗 查询均由这一条目自动派生，无需改动其他前端代码。hwToGpuKey 的首段截取逻辑 已能剥离新的 `-lat` 产物后缀，补充一个 normalizer 测试加以固化。全能耗 （all-in power）与三档成本暂用 9.99 占位，等待 SemiAnalysis AI Cloud TCO 模型的官方数值；TDP 采用数据手册的 600 W 最大板卡功耗。在真实数值补齐前， 该 SKU 的 $/token 与 per-MW 指标无实际意义。
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
There was a problem hiding this comment.
Pull request overview
Adds support for rendering benchmark artifacts that report the new rtx6000pro-lat hardware id by introducing an rtx6000pro GPU SKU in the shared hardware registry, and locking in the expected DB ingest normalization behavior via a unit test.
Changes:
- Add
rtx6000protoHW_REGISTRYso the new SKU can be labeled/sorted and participate in derived metrics. - Add a
hwToGpuKey()test case assertingrtx6000pro-latnormalizes tortx6000pro.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
| packages/constants/src/gpu-keys.ts | Adds rtx6000pro entry to HW_REGISTRY (vendor/arch/label/sort/TDP + power/cost fields). |
| packages/db/src/etl/normalizers.test.ts | Adds unit coverage for hwToGpuKey('rtx6000pro-lat') normalization. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds the NVIDIA RTX PRO 6000 Blackwell Server Edition (GB202, PCIe Gen5, 96 GB GDDR7) as a new GPU SKU so the MiniMax-M3 benchmark from
SemiAnalysisAI/InferenceXrun 29987453043 renders on the dashboard. Artifact hardware id:rtx6000pro-lat.The only structural change is a single
HW_REGISTRYentry (the single source of truth). Display labels, chart colors, legend sort order, dropdowns, and cost/power lookups all derive from it automatically — so the point renders across the inference, evaluation, and calculator charts (and the?unofficialrun=overlay path, which resolves hardware through the same registry) with no other frontend edits.Changes
packages/constants/src/gpu-keys.ts— newrtx6000proentry: vendorNVIDIA, archBlackwell, labelRTX PRO 6000,sort: 9(sorts last, after the datacenter SXM/OAM parts),tdp: 600(datasheet max board power).packages/db/src/etl/normalizers.test.ts— test locking inhwToGpuKey('rtx6000pro-lat') → rtx6000pro. No normalizer source change was needed:hwToGpuKeyalready takes the first segment before-, so the-latsuffix is stripped automatically.Known follow-ups (intentionally deferred)
9.99placeholders (flagged with a// TODO). All-in power (kW/GPU) and the three$/GPU/hrcost tiers await official numbers from the SemiAnalysis AI Cloud TCO model. Latency/throughput/interactivity charts are unaffected;$/tokenand per-MW metrics for this SKU are not meaningful until real numbers land.GPU_SPECS) entry not added. It needs scale-out cluster-fabric fields that aren't defined for a single-node 8×PCIe box, plus agetScaleUpTopologyConfigPCIe branch and updates to the hardcodedgpu-specs.test.tsassertions.29987453043is ingested + cache invalidated.No
/zhpage changes: this adds a hardware SKU to a data registry, not an indexable page/tab/post, and the label is a hardware SKU name (stays English per the translation rules).Test plan
pnpm typecheck✅pnpm lint/pnpm fmt✅ (also via pre-commit hook)-latnormalizer case)中文说明
新增 NVIDIA RTX PRO 6000 Blackwell Server Edition（GB202，PCIe Gen5，96 GB GDDR7）作为新的 GPU SKU，使
SemiAnalysisAI/InferenceXrun 29987453043 的 MiniMax-M3 基准测试能在仪表板中渲染。产物硬件标识为rtx6000pro-lat。唯一的结构性改动是在
HW_REGISTRY（单一事实来源）中新增一个条目。展示标签、图表配色、图例排序、下拉选项以及成本/功耗查询均由其自动派生——因此该数据点可在推理、评估与计算器（calculator）图表中正常渲染（?unofficialrun=叠加路径也通过同一注册表解析硬件），无需改动其他前端代码。改动
packages/constants/src/gpu-keys.ts— 新增rtx6000pro条目：vendorNVIDIA、archBlackwell、labelRTX PRO 6000、sort: 9（排在数据中心 SXM/OAM 产品之后）、tdp: 600（数据手册最大板卡功耗）。packages/db/src/etl/normalizers.test.ts— 固化hwToGpuKey('rtx6000pro-lat') → rtx6000pro的测试。无需改动 normalizer 源码：hwToGpuKey已取-前的首段，-lat后缀会被自动剥离。已知的后续事项（有意暂缓）
9.99占位（已用// TODO标注）。全能耗（kW/GPU）与三档$/GPU/hr成本等待 SemiAnalysis AI Cloud TCO 模型的官方数值。延迟/吞吐量/交互性图表不受影响；在真实数值补齐前，该 SKU 的$/token与 per-MW 指标无实际意义。GPU_SPECS）条目。 单节点 8×PCIe 机型缺少 scale-out 集群网络（cluster-fabric）字段，且需要为getScaleUpTopologyConfig增加 PCIe 分支并更新gpu-specs.test.ts中的硬编码断言。29987453043完成 ingest 并使缓存失效后，数据点才会出现。无
/zh页面改动：本次仅向数据注册表新增硬件 SKU，并非可索引的页面/标签页/文章，且标签为硬件 SKU 名称（按翻译规范保持英文）。测试
pnpm typecheck✅pnpm lint/pnpm fmt✅（并经 pre-commit 钩子验证）-latnormalizer 用例）Note
Low Risk
Registry and test-only change; placeholder cost/power may mislead cost charts until real TCO values replace 9.99.
Overview
Adds
rtx6000protoHW_REGISTRYso benchmark hardware ids likertx6000pro-latresolve to a known NVIDIA Blackwell PCIe SKU (label RTX PRO 6000,sort: 9,tdp: 600). Labels, chart ordering, and cost/power lookups pick it up from the registry without other UI changes.Power and all three
$/GPU/hrtiers are9.99placeholders (documented with a TODO) until SemiAnalysis TCO numbers land—$/tokenand per-MW views for this SKU are not meaningful yet.Adds a unit test asserting
hwToGpuKey('rtx6000pro-lat')→rtx6000pro; no normalizer code change (existing “first segment before-” behavior already strips-lat).Reviewed by Cursor Bugbot for commit b79b003. Bugbot is set up for automated code reviews on this repo. Configure here.