Skip to content

feat(ubenchx): add Huawei 950PR bandwidth / 新增 Huawei 950PR 带宽结果 - #1506

Merged
Oseltamivir merged 4 commits into
masterfrom
feat/ubenchx-huawei-bandwidth
Oct 26, 2026
Merged

Oseltamivir merged 4 commits into
masterfrom
feat/ubenchx-huawei-bandwidth

Conversation

@Oseltamivir

@Oseltamivir Oseltamivir commented Oct 26, 2026 •

Copy link
Copy Markdown
Contributor

Adds measured Huawei 950PR copy-bandwidth results, retaining the exact Ascend950PR_9579 SKU in metadata. The 950PR series uses FP16 rotating-buffer graph timing and a qualified 1.6 TB/s reference; its amortized small-copy times are distinguished from the existing Triton GPU protocol.

Hover guides follow the rendered point for MBU, bandwidth and latency, including after zoom. Cypress verifies these interactions, the Huawei label, the public API, and English/Chinese rendering. Static ubenchX captures do not use inference-run overlays.

Local broad dev-mode smoke hits a Next.js Performance.measure error on /zh/this-route-does-not-exist; targeted ubenchX regressions and full Chrome/Firefox CI pass.

AI model disclosure

Primary implementation, data audit and tests: the exact runtime model/version identifier was not exposed and could not be verified. claude-opus-5-5 provided an advisory Chinese-copy review; its findings were checked against the measurement source.

PR checklist
  • I have completed the AI model disclosure and kept it current
  • Added, changed, or removed filters, views, or pages include the matching API and inferencex-skills changes
  • New or changed indexable pages ship their /zh sibling
  • Changed Chinese copy follows the PR review checklist in docs/chinese-copy.md
中文

新增 Huawei 950PR 的实测拷贝带宽,元数据保留准确的 Ascend950PR_9579 SKU。数据采用 FP16 轮转缓冲区图执行计时,并说明 1.6 TB/s 参考值的限制;小消息摊销时间与现有 Triton GPU 协议明确区分。

MBU、带宽和延迟视图的悬停参考线与实际数据点对齐,缩放后仍保持一致。Cypress 已验证这些交互、Huawei 名称、公开 API 和中英文显示。静态 ubenchX 数据不使用推理运行叠加功能。

本地开发模式的广泛 smoke 测试在 /zh/this-route-does-not-exist 遇到 Next.js Performance.measure 错误;针对 ubenchX 的回归测试和完整 Chrome/Firefox CI 均通过。

AI 披露:主要实现、数据核对和测试所用模型的具体运行时版本无法核实;claude-opus-5-5 曾提供辅助中文文案审核,后续修改由主要实现模型对照英文核查。

保留 Ascend 950PR_9579 和 910B3 的完整轮转拷贝测量点、误差范围及来源哈希。
展示 Ascend 950PR_9579 与 910B3 轮转拷贝数据,明确摊销计时、带宽参考值和缺失大小,并同步 API 与技能文档。
@Oseltamivir
Oseltamivir requested a review from adibarra as a code owner October 26, 2026 12:43
@cursor

cursor Bot commented Oct 26, 2026

Copy link
Copy Markdown

Bugbot is paused — your team has an unpaid invoice

Bugbot is paused because your team has an unpaid invoice.

Pay it from the billing page in the Cursor dashboard and Bugbot resumes automatically.

(requestId: f0a57d3f-68c7-48c1-908f-ab02ed736c80)

@vercel

vercel Bot commented Oct 26, 2026 •

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated
inferencemax-app Ready Ready Preview Oct 26, 2026 1:54pm UTC

Request Review

按要求移除 910B3 带宽系列,并将展示名称统一为 Huawei 950PR。
统一 Huawei 950PR 展示名称,移除 910B3 界面及 API 数据,并使悬停参考线与当前指标和缩放后的数据点对齐。
@cursor

cursor Bot commented Oct 26, 2026

Copy link
Copy Markdown

Bugbot is paused — your team has an unpaid invoice

Bugbot is paused because your team has an unpaid invoice.

Pay it from the billing page in the Cursor dashboard and Bugbot resumes automatically.

(requestId: aec8eb5a-4a6b-413c-bc8b-54ccbfdd5151)

@Oseltamivir Oseltamivir changed the title feat(ubenchx): add Huawei copy bandwidth / 新增 Huawei 拷贝带宽结果 feat(ubenchx): add Huawei 950PR bandwidth / 新增 Huawei 950PR 带宽结果 Oct 26, 2026
@Oseltamivir
Oseltamivir merged commit aa03d97 into master Oct 26, 2026
37 checks passed
@Oseltamivir
Oseltamivir deleted the feat/ubenchx-huawei-bandwidth branch October 26, 2026 14:21

This branch was successfully deployed

1 active deployment
Preview — 83012b41 Deployed Oct 26, 2026 by vercel[bot]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant