跳到主要内容Skip to main content

← 返回硬件算力← Back to Compute

加速卡Accelerators未公布单价No public price

AMD Instinct MI440XAMD Instinct MI440X

未公布单价No public price未公布单价No public price · 官方页未给公开单价;这类产品通常按整机、云实例或询价成交,不拿经销商价冒充官方价。No public unit price on the official page; these usually sell as systems, cloud instances or on request, so dealer quotes are not presented as official.

回答「MI400 家族里给标准 8 卡服务器的是哪一档」。MI440X 是 AMD 的 8 卡服务器 drop-in 型号,配 EPYC Venice;显存与带宽沿用 MI400 系列的 432GB HBM4 / 19.6TB/s,FP4/FP8 分别约 40 / 20 PFLOPS。注意:AMD 尚未单独发布 MI440X 规格页,以上算力与显存数字来自 MI400 系列同代资料与第三方汇总,卡片按「厂商口径」标记,不当作已复核的独立规格。Answers which MI400 family member targets standard 8-GPU servers. MI440X is AMD’s drop-in 8-GPU server part paired with EPYC Venice; it carries the MI400 platform’s 432GB of HBM4 and 19.6TB/s of bandwidth, with roughly 40 / 20 PFLOPS at FP4 / FP8. Caveat: AMD has not published a standalone MI440X spec sheet, so those compute and memory figures come from same-generation MI400 material and third-party summaries; this card marks them as vendor claims, not independently verified specs.

关键参数Key specs

architecture
CDNA 5
form
standard 8-GPU OAM server, drop-in with EPYC Venice
memory
432 GB HBM4 (MI400 platform)
memory bandwidth
19.6 TB/s (MI400 platform)
FP4
~40 PFLOPS (platform figure)
FP8
~20 PFLOPS (platform figure)
scale-up
Infinity Fabric; UALink scale-out

官方原文怎么说What the source says

8 GPU Drop-In Ready “Venice” + MI440X
1 https://www.amd.com/content/dam/amd/en/documents/corporate/events/amd-ces-2026-distribution-deck.pdf
It carries the same 432 GB of HBM4 and 19.6 TB/s memory bandwidth as the MI455X
2 https://www.aiwiki.ai/wiki/amd_mi400

资料与许可Source and licence

资料时间Source checked
2026-11-19
许可 / 类型Licence / kind
vendor-claim
许可说明Licence evidence
https://www.amd.com/content/dam/amd/en/documents/corporate/events/amd-ces-2026-distribution-deck.pdf
类目Category
加速卡Accelerators
状态Status
已发布 / 未出货Announced
价格口径Price basis
未公布单价No public price · not-published

同类目还有More in this category

下一步Next step

去官网看原页Open the official page →

← 返回硬件算力← Back to Compute