Plan the full training path.
Record model size, precision, dataset throughput, checkpoint storage and the interconnect requirement. Keep scaling assumptions visible in the RFQ.
AI Training SystemsTurn model size, numerical precision and training goals into a system requirement with memory, interconnect and data delivery considered together.
Counts are starting points. Double-width cards, single-width cards and integrated baseboards use different platform designs.
Record model size, precision, dataset throughput, checkpoint storage and the interconnect requirement. Keep scaling assumptions visible in the RFQ.
AI Training SystemsSpecify model memory, context length, concurrency, response-time targets and deployment locations before choosing accelerator density.
AI Inference PlatformsOpen the catalog record for specifications and sources. Manufacturer names identify equipment; they do not imply an authorized relationship.
8U · Eight H100 GPUs · 640 GB GPU memory
Complete DGX system: two Xeon 8480C processors, 2 TB host memory and a 4+2 PSU design. Firmware, fabric cables and actual-unit configuration require review.
8U · Eight H200 GPUs · 1128 GB GPU memory
Keep the H200 identity distinct from H100. The reviewed system uses two Xeon 8480C processors and 2 TB host memory; verify actual hardware, software and facility requirements.
8U · Intel Xeon 5th Gen · H100/H200 platform
Machine type 7DHE: eight SXM5 GPUs, up to sixteen front NVMe drives and XCC2. The B200 system has a separate guide and is not assumed interchangeable. OEM guide marked withdrawn.
4U · Intel Xeon 6 · 32 DDR5 slots
Platform layout: 8 double-width or 16 single-width PCIe positions. The individual GPU table has a model-name inconsistency and needs review.
4U · AMD EPYC 9005 · 24 DDR5 slots
Up to 8 double-width 600W or 16 single-width 75W PCIe accelerators; specific GPU, cooling and PSU combinations require review.
6U · Dual Intel Xeon Scalable · 32 DDR5 slots
Eight-accelerator baseboard platforms include HGX H100/H200, MI300X and Gaudi3. Their baseboards, interconnects and storage rules are distinct.
2U · Intel Xeon Scalable · 32 DDR5 slots
Observed front riser layout: 4 double-width PCIe x16 or 8 single-width PCIe x8 GPU positions. A total PCIe-slot count is not a GPU count.
2U · AMD EPYC 9004 · 24 DDR5 slots
Model-level CPU and memory facts reviewed. Exact GPU population, riser and cooling support remain under review.
2U · AMD EPYC 9005 · 24 RDIMM slots
Up to 2 double-width 450W GPUs or 6 single-width GPUs. Confirm the accelerator, risers and site power together.
2U · Dual Intel Xeon Scalable · iLO 6
4 double-width or 8 single-width chassis; selected H100 options allow 2 or 4. PSU quantity and cable kits depend on chassis and GPU.
2U · AMD EPYC 9004 / 9005 · iLO 6
GPU CTO chassis supports up to 4 double-width or 8 single-width GPUs. Standard rear PCIe slots are not a substitute for the GPU CTO chassis.
2U · Intel Xeon 6 · iLO 7
H200 NVL: maximum 2 with power-cap and cooling conditions. L40S / RTX PRO 6000 options: maximum 3. This is not DL380a Gen12.
3U · AMD EPYC 9004 / 9005 · 24 DIMM slots
The specified 8-DW PCIe chassis supports up to 8 double-width GPUs. PCIe and SXM platform variants have different layouts and configuration rules.
8U air-cooled · Intel Xeon 6700P · 32 DDR5 slots
Eight NVIDIA GPUs on an HGX B300 baseboard. This integrated baseboard platform is distinct from eight discrete PCIe cards.
2U · Intel Xeon 6700P / 6500P · Multiple cooling variants
Up to 10 single-width GPUs or 2 double-width GPUs, subject to the specified chassis and riser rules. Ten single-width positions do not imply ten double-width accelerators.
4U · Dual Intel Xeon 4th / 5th Gen · 32 DIMM slots
Up to 10 double-width PCIe GPUs. Verify the selected card list, power input, cabling and platform population rules before preparing a BOM.
Exact system suffix requires source review
The located official source identifies SYS-422GA-NBRT-LCC. That suffix is not treated as an alias or used to verify the requested unsuffixed model.
8U rear-I/O · Dual Intel Xeon 4th / 5th Gen · 32 DIMM slots
Eight GPUs on an NVIDIA HGX H100 or H200 baseboard, with NVLink and NVSwitch. Review the exact baseboard version and cooling specification.
4U · Dual Intel Xeon 6900P · 24 DIMM slots
Up to 8 double-width PCIe GPUs. PSU output depends on input conditions; accelerator selection, power and cooling require a complete configuration review.
Catalog identity · OEM field review pending
Open the existing exact-model catalog record. Accelerator population, processor generation, power and cooling fields require source review before configuration planning.
Catalog identity · OEM field review pending
Preserve the complete model suffix and revision when requesting a system. No accelerator count or configuration compatibility is assigned here without field-level review.
4U · Dual Intel Xeon 6500 / 6700 · 32 DDR5 slots
Up to 8 dual-slot GPUs, each up to 600W. Preserve the E12P suffix; this is a separate identity from ESC8000-E12. Exact GPU qualification, risers, power input and the complete option BOM require review.
5U · Dual AMD EPYC 9005 / 9004 · 24 DDR5 slots
The AAX1 Rev. 3.x source specifies an HGX H100 platform with 8 SXM5 GPUs. Preserve the complete suffix and revision; H200 and liquid-cooled siblings are separate systems. Exact options, input power and current firmware support require review.
4U · Dual Intel Xeon 4th / 5th Gen · 32 DDR5 slots
Up to 10 double-width GPU cards. The published storage layouts are alternatives; maximum drive and GPU counts are not assumed to operate together. GPU qualification, drive topology, power input and redundancy combinations require review.
4U · Dual AMD EPYC 7002 / 7003 · 32 DDR4 slots
The observed A800-version datasheet specifies one HGX A800 baseboard with 8 GPUs. Older A100 revisions remain separate. The later document fetch returned 404; current document availability, exact board revision, power input and configuration rules require review.
2U · Intel Xeon 6700P / 6500P · Based on SR650a V4
The guide lists separate populations of 8 single-width L4, 4 double-width L40S or 2 H200 NVL GPUs. These are alternative configurations. GPU feature codes, cooling, power, vSAN qualification and software entitlements require review.
5U · Dual Intel Xeon 5th Gen · H200 configuration scope
The observed portfolio entry lists 8 H200 GPUs and air- or liquid-cooled configurations. Current QuickSpecs, memory, storage, power and fabric details remain unverified; H100 configurations are not inferred from this H200 entry.
Catalog entries are research and configuration references. Availability, condition, unit photos, included items and commercial terms are confirmed in the quotation.
Search entry points for exact-model research. A family name is not a verified system configuration.

Open the referenced model records to review documented specifications, outstanding checks and quotation requirements.
Open the exact catalog record, review its evidence and add it to your comparison.
Open comparisonChoose an available planning profile, then confirm the complete bill of materials.
Open server plannerKeep quantities, condition, unit-photo requests and delivery requirements in a local draft.
Open quote basketPlanning profiles cover specific systems. A model link or matching GPU count does not certify compatibility; review the exact chassis, GPU, risers, cables, cooling and power configuration.
Distinguish PCIe cards from integrated accelerator baseboards.
This guide adds the decisions, documents and next steps that support this page.
GPU platforms
Start with model size, precision, parallelism and dataset movement.
GPU platformsDefine request concurrency, latency target and model-memory footprint.
AI systemsCheck slot width, risers, auxiliary power and supported cooling.
GPU checklistDistinguish PCIe cards from integrated accelerator baseboards.
Plan host networking, GPU interconnect and storage access separately.
Request platform power requirements and facility cooling review.
A planning sequence, not a record of completed work.
Counts show records in the site catalog. They do not represent stock, available quantities or verified compatibility.
Snapshot: 2026-09-15
| Decision | Record in your brief | Evidence to retain |
|---|---|---|
| Accelerator architecture | Distinguish PCIe cards from integrated accelerator baseboards. | Complete identifier and source revision |
| Fabric & data path | Plan host networking, GPU interconnect and storage access separately. | Platform document or written clarification |
| Rack power & cooling | Request platform power requirements and facility cooling review. | Offered-unit details and agreed scope |
Tick items as you prepare them. This is your personal checklist, not a hardware validation.
Saved in this browser when local storage is available.
Reference-based generated scenes; actual-unit photos are available only through an explicit unit-specific request. Related models are separate catalog records, not substitutes for the item on this page.
Continue the relevant directory or service flow.
Open resourceUse the connected tool to develop the requirement.
Open resourceRetain the supporting reference with your project brief.
Open resourceWorkload and precision defined · Accelerator model and quantity named · Platform support documents requested · Power and cooling review assigned.