SECTIONS
EDITIONS
NETWORK
AIXX Pedia The encyclopedia of compute hardware

AIXX Pedia › Model Entry › MI350X / MI355X

Official image of AMD Instinct-series OAM integrated system scenarios (vendor image)
Scene Gallery · AIXX Pedia (AMD official site; MI350 series product images to be added)

MI350X / MI355X

AMD CDNA4 · 288GB HBM3E · Model Entry #015

Overview OVERVIEW

MI350X / MI355X It's the new flagship of the AMD Instinct family with The CDNA4 Architecture Flagship (Released June 12, 2025): 288GB HBM3E per card, 8TB/s bandwidth, 185 billion transistors on 3nm—the two models share the same die and differ in cooling and power: MI350X 1000W Air-cooled (Passive OAM) 、 MI355X 1400W with Liquid-cooled support , with MXFP4 throughput of 9.2/10.1 PFLOPS respectively (per AMD's official two-page spec sheet [1]).

It is the The direct successor to the MI325X : Memory stays in the 288GB class but bandwidth rises to 8TB/s (MI325X was 6TB/s), the architecture moves from CDNA3 to CDNA4, and FP4/FP6 precision gets native support — the delivery on the promise of 'one generation per year' upgrades to both memory and architecture (per AMD's official generational framing [2]). AMD's official B200 comparison cites the current 180GB/7.7TB/s tier — the 288GB capacity gap is the flagship selling point of the AMD camp (official comparison baseline, same as the B200 entry in this library).

In the 2026 spot market, the MI350 series circulates primarily as 8-GPU OAM complete systems, still ramping in volume: the maturity of the ROCm 6.x+ support matrix for CDNA4, and the price spread against MI300X/MI325X inventory, are the core of the three key procurement questions. This entry carries no market band—actual price subject to the seller's quote.

Use Cases FIT

AIXX Index (0–100): Estimated in this Pedia via weighted public specs — memory 40% + interconnect 30% + ecosystem 30%.

Extra-large model inference 91AIXX Index
288GB per card + 8TB/s: single-card hosting of hundred-billion-parameter models plus ultra-long context — AMD's answer to the memory ceiling (per official snapshot [1]).
FP4/FP6 inference pipelines 88AIXX Index
CDNA4 with native MXFP4/FP6—AMD's flagship option for new-precision inference pipelines (a generational capability the MI325X lacks).
Liquid-cooled data center training 84AIXX Index
MI355X 1400W liquid-cooled form factor—an 8-GPU training solution for liquid-cooled data centers; air-cooled facilities should choose the MI350X.
CUDA legacy production lines 45AIXX Index
Migration within the MI-series frameworks—CUDA-deeply-bound pipelines should be evaluated on total cost (per the ROCm entry).

History HISTORY

On June 12, 2025 , MI350X/MI355X officially launched (per the official Launch Date framing [1]): CDNA4 architecture, 3nm, 185 billion transistors, 288GB HBM3E/8TB/s — memory and bandwidth take another step up from MI325X (256GB/6TB/s), with MXFP4/FP6 new precision natively added.

2025-2026 , the two models ship in divergent form factors: the Liquid-cooled MI355X (1400W/2400MHz) targets liquid-cooled data centers and leading cloud providers, while the Air-cooled MI350X OAM (1000W/2200MHz) covers air-cooled AI compute centers — a dual-track approach covering different facility conditions (per the official two-page spec framing [1]).

2026 , spot ramp-up: 8-GPU OAM complete systems flowing in growing volume, with the ROCm 6.x+ support matrix for CDNA4 continuing to mature—the MI350 series becomes the AMD camp's current flagship (against the NVIDIA B300 generation). Cross-border circulation policy is still evolving; procurement should follow competent-authority policy and professional advice (per public reporting).

Technical Specifications SPECIFICATIONS

ItemMI350XMI355X
Architecture / ProcessCDNA4 · TSMC 3nm · 185 billion transistors
Memory288GB HBM3E · 8TB/s (both models)
Peak Clock2200 MHz2400 MHz
MXFP4 Throughput9.2 PFLOPS10.1 PFLOPS
Power (TBP)1000W1400W
CoolingPassive OAM (air-cooled)Passive & Active (supports liquid cooling)
Shared SpecsInfinity Fabric 8 links · PCIe Gen5 · OAM module · SR-IOV

Specifications follow AMD's official two-page spec sheet (official snapshot as of 2026-10 [1], cross-verified by both C5/C6 generation scouts); the Launch Date of 6/12/2025 is per official labeling.

MI350X · 1000W · Air-cooledMXFP4 9.2 PFLOPS · 2200MHz MI355X · 1400W · Liquid-cooledMXFP4 10.1 PFLOPS · 2400MHz Two-model comparison: same 288GB HBM3E / 8TB/s (per AMD official specs)

Three AMD Flagship Generations Side by Side THREE GEN

ItemMI300XMI325XMI350X/MI355X
ArchitectureCDNA3CDNA3 (refinement)CDNA4
Memory192GB HBM3256GB HBM3E288GB HBM3E
Bandwidth5.3TB/s6TB/s8TB/s
New PrecisionFP8FP8MXFP4 / FP6
Power750W1000W1000W / 1400W
One-line buying adviceValue on existing stockTransitional TierCurrent Flagship

The three-generation comparison uses each model's official snapshot (see individual entries for MI300X/MI325X); CDNA4's MXFP4/FP6 is the new-precision generational capability.

Family Members FAMILY

The Instinct family tiers by generation; the MI350 series is the CDNA4 current flagship:

MI350X (1000W air-cooled OAM · this entry)MI355X (1400W with liquid-cooling support · same-die high spec)MI325X (256GB · previous-gen flagship · see entry)MI300X / MI300A (192GB generation · see entry)MI400 series (next generation, per AMD's public roadmap)

Cooling & Deployment COOLING

The MI350 series deployment divide is cooling form factor : the MI350X is a Passive OAM air-cooled design (1000W) that can land in air-cooled AI centers; the MI355X supports Liquid cooling (1400W), the density pick for liquid-cooled facilities. The 8-card OAM system (UBB baseboard) is the standard delivery form — rack-level cooling and power (54V power bus) for 1000W/1400W×8 = 8–11kW-class loads must be planned holistically (per the official two-page spec framing [1]).

Deployment differences vs. the MI325X (also OAM): the higher power density of CDNA4 makes liquid cooling go from optional to mandatory for the MI355X —air-cooled facilities choose the MI350X, liquid-cooled facilities choose the MI355X by density—the "fix facility conditions first, then pick the model" selection order holds (per our library's compilation).

Model Compatibility MODEL FIT

Model / ScenarioMemory RequirementMI350 Series Deployment NotesAIXX Index
DeepSeek R1 full (671B)FP8 ~690GB8-GPU 2.3TB fits with ample headroom—MXFP4 inference path depends on ROCm framework support88
Ultra-long-context inferenceKV cache scales linearly with length288GB per card offers the largest KV cache space in the MI series91
LLM training / fine-tuning8-GPU IF networkingThe CDNA4 training stack matures with ROCm—factor tuning labor into the cost82
FP64 HPCOfficial snapshot is authoritativeCDNA4 FP64 figures per the official spec sheet—anything not listed here = unverified—

Memory is estimated as FP16 weights ≈ parameter count × 2 bytes (estimated framing); FP4/FP6 paths are subject to actual ROCm framework support (official framing [1]).

Official AMD Instinct-series image
AMD Instinct OAM complete-system scene (official image · AMD website; MI350 series product images to be added)

Software & Ecosystem ECOSYSTEM

ItemRequirements / Status
Software stackROCm 6.x+ (CDNA4 support matrix per official docs, same as the ROCm entry)
FrameworksPyTorch ROCm build / vLLM / SGLang—MI350 series support matures with versions
VirtualizationSR-IOV support (per the official spec snapshot [1])
Observabilityrocm-smi / amd-smi
ComparisonSee the ROCm entry migration framework for the CUDA ecosystem comparison

Ecosystem info compiled from public sources; versions evolve over time — defer to AMD's official releases.

Purchasing & Inspection PROCUREMENT

Three inspection points when buying MI350 series:

1
Model vs. cooling form factor cross-check
MI350X (1000W air-cooled) and MI355X (1400W liquid-cooling support) are two different deliveries—verify model and power tier with rocm-smi, and specify the cooling form factor in the contract.
2
Memory and compute testing
Measured 288GB memory, measured MXFP4/FP6 benchmarks (≥80% of rated = Pass, estimated framing)—CDNA4's new precision is this generation's core selling point; skipping the test means buying blind.
3
ROCm version and support matrix
The MI350 series is new: verifying the three-way matrix of ROCm version × card model × framework is a procurement prerequisite (per the ROCm entry)—check compatibility first for "new card, old stack" combinations.

Acceptance Quick Reference QUICK CHECK

Four-step quick check for the MI350 series:

1
Identity
rocm-smi: MI350X/MI355X model, 288GB memory, serial number—cross-check against shipping documents.
2
Memory Bandwidth
Measured 8TB/s-class bandwidth benchmark (≥85% = Pass, estimated framing)—HBM3E health is the first gate for used units.
3
IF interconnect
8-GPU rccl-tests bus bandwidth—IF link connectivity is the core acceptance item for OAM systems.
4
Full-load Burn-in Test
Full Load at the model's power tier (1000W/1400W): stable temperatures, zero ECC growth (per the precedent framework)—the MI355X liquid-cooled form follows the liquid-cooling acceptance procedure.

Shipping & Delivery LOGISTICS

1
Delivery and compliance
The MI350 series circulates as 8-GPU OAM complete systems: shock-proofing and anti-static measures for whole-system shipping; for cross-border compliance status rely on professional advice—our descriptions reflect public reporting and do not constitute legal advice.

Glossary Cards GLOSSARY

CDNA4 and MXFP4/FP6

AMD's fourth-generation data center architecture: adds MXFP4/FP6 microscaling floating-point precisions—further lowering the memory and bandwidth cost of large-model inference (against NVIDIA's FP4 route), the generational capability gap CDNA4 holds over CDNA3.

MI350X vs MI355X

Two models on the same silicon: MI350X at 1000W Air-cooled OAM (2200MHz/MXFP4 9.2PF), MI355X with 1400W Liquid-cooled support (2400MHz/MXFP4 10.1PF) — both share 288GB/8TB/s of memory; choose based on facility cooling conditions and density requirements (per the official two-page framing [1]).

288GB HBM3E / 8TB/s

The MI350 series memory spec: 288GB per card, 8TB/s system-wide bandwidth—the current peak of the MI series' "memory arms race" (against the MI325X at 256GB/6TB/s and the MI300X at 192GB/5.3TB/s).

OAM and UBB

The OAM module + UBB baseboard 8-GPU system is the MI350 series' standard delivery form—the carrier for the 54V power bus and IF interconnect; "loose modules" off the baseboard cannot be used standalone.

Limitations & Controversies LIMITS

  • The ROCm migration ledger : same as other MI-series frameworks—cost CUDA-installed pipelines in person-months (per the ROCm entry); the MI350 series is new, so framework maturity for CDNA4 support must be verified item by item.
  • The MI355X liquid-cooling threshold : 1400W power + liquid-cooled form factor—liquid-cooling retrofit budget and O&M capability are prerequisites; air-cooled facilities should choose the MI350X (per our library's compilation).
  • No established used-market yardstick : the MI350 series only began shipping in mid-2025, so used-inspection practice is limited—memory health and IF link condition are the inspection focus (per our library's inspection experience).
  • The B200 comparison baseline framing : AMD's official comparison uses the B200's current 180GB/7.7TB/s framing—note each side's chosen framing in cross-camp comparisons (per our library's compilation).
  • Cross-border trade policy variables : AMD high-end cards also fall under export-control contexts (per public reporting) — for cross-border procurement compliance, defer to professional advice; this Pedia does not constitute legal advice.

Market Notes MARKET NOTE

MI350X / MI355X · Market band not yet opened

The MI350 series circulates mainly as 8-GPU OAM complete systems, and the dual-model + cooling-form-factor differences make price dispersion wide—no verifiable unified market snapshot yet; AIXX Pedia honesty principle: no market band; prices are subject to the seller's quote 。

Spot search: AIXX Exchange Hall › (Zero-commission Matching, hence neutral data).

FAQ FAQ

MI350X or MI355X—how to choose?

Let facility cooling conditions decide: air-cooled AI centers pick the MI350X (1000W air-cooled OAM); liquid-cooled facilities or high-density needs pick the MI355X (1400W liquid-cooled, higher clocks and MXFP4 throughput)—memory is the same 288GB/8TB/s.

MI350 series or MI325X—how to choose?

With sufficient budget for new purchases, pick the MI350 series: the architecture gap (CDNA4's MXFP4/FP6) plus memory bandwidth (8TB/s vs. 6TB/s); MI325X inventory offers a friendlier price spread but no new precision—training pipelines should go straight to CDNA4.

Can the MI350 series run the full DeepSeek R1?

Yes on an FP8 basis: full R1 in FP8 is about 690GB, fitting within the 8-GPU 2.3TB memory with ample headroom—with the ROCm builds of inference stacks vLLM/SGLang, run benchmarks to validate before deployment.

What to test for MI350 series acceptance?

Four steps: rocm-smi identity check (model/288GB/serial) → measured 8TB/s memory bandwidth → 8-GPU rccl-tests interconnect → Burn-in Test at Full Load per the power tier (MI355X liquid-cooled follows the liquid-cooling procedure). Verifying the ROCm support matrix is a procurement prerequisite.

MI350 series or B200—how to choose?

A dilemma of ecosystem vs. capacity: the MI350 series' 288GB memory beats the B200's 192GB-class (current framing), but the B200 is backed by the CUDA ecosystem—new open-source-stack pipelines can weigh the MI350 series, while CUDA-installed pipeline migration should be costed in total (per the ROCm entry framework).

MI325XMI300XROCmB200Liquid-cooled CDUNVMe All-Flash StorageMI400 series (next generation)UBB baseboardMXFP4/FP6 precision

Model comparison and selection COMPARE

References REFERENCES

  1. AMD website — Instinct MI350X product page (official two-page spec sheet, official snapshot as of 2026-10) https://www.amd.com/en/products/accelerators/instinct/mi350/…
  2. AMD website — Instinct MI355X product page (liquid-cooled high-spec framing) https://www.amd.com/en/products/accelerators/instinct/mi350/…
  3. MI355X teaser and MI350P PCIe variant coverage (third-party cross-reference) https://www.servethehome.com/
  4. AMD Official Documentation — ROCm Documentation Site (support matrix) https://rocm.docs.amd.com/
  5. Listing search — AIXX Exchange (Zero-commission Matching; no market snapshot for the MI350 series yet—actual price subject to the seller's quote) https://www.aixx.vip/hall.html?model=MI350
Versions and errata | This entry is maintained by the AIXX Pedia editorial team, AI-assisted drafting with human fact-checking. Updated 2026-10-10 · Version v1.0 · All changes tracked and auditable. Spotted an error? Let us know via Corrections Channel ; corrections are published after verification and recorded in version history.

Video Overview VIDEO

This Library
Editorial PolicyAbout This Library (Disclosure of Interests)Corrections Channel
Editorial Constitution Summary
Every claim sourcedDrawbacks always stated · corrections publicly loggedNo paid entries acceptedZero traffic-funneling or sales elements within entriesMarket data attached only as data modules
AIXX Family
AIXX PediaAIXX ExchangeAIXX News
Market overview: AIXX Exchange · © 2026 AIXX Pedia · An AIXX Product