{"id":866,"date":"2026-11-03T18:27:45","date_gmt":"2026-11-03T18:27:45","guid":{"rendered":"https:\/\/openglobe.io\/?post_type=produtos&p=866"},"modified":"2026-11-05T19:31:37","modified_gmt":"2026-11-05T19:31:37","slug":"g494-zb4-aap2","status":"publish","type":"produtos","link":"https:\/\/openglobe.io\/en\/produtos\/g494-zb4-aap2\/","title":{"rendered":"G494-ZB4-AAP2"},"content":{"rendered":"
\n\t\t\t\t
\n
\n
\n \t\t\t

Home<\/a>  \u203a <\/span> Products<\/a>  \u203a <\/span> G494-ZB4-AAP2<\/strong><\/p>\n\t\t

\n
\n \n<\/div>\n\t\t\t

GIGABYTE \u00b7 4U GPU server \u00b7 AMD EPYC<\/p>\n\t\t\n<\/div>\n\t\t\t

G494-ZB4-AAP2: 4U server with up to 8 GPUs<\/h1>\n\t\t\t\t\t

4U GPU server with dual AMD EPYC processors and up to eight dual-slot PCIe Gen5 GPUs. Validated with NVIDIA H200 NVL, NVIDIA RTX PRO 6000 Blackwell and AMD Instinct MI350P to train and serve AI models at scale.<\/p>\n\t\t

\n \t\t\t\n\tRequest a quote  \u2192\n<\/a>\n\t\t\t\t\t\n\tView datasheet\n<\/a>\n\t\t\n<\/div>\n
\n \t\t\t

Up to 8 dual-slot GPUs<\/p>\n\t\t\t\t\t

2 x AMD EPYC 9005\/9004<\/p>\n\t\t\t\t\t

48 DDR5 DIMMs<\/p>\n\t\t\t\t\t

12 NVMe Gen5 bays<\/p>\n\t\t\t\t\t

4 x 3000 W Titanium PSUs<\/p>\n\t\t\n<\/div>\n\n<\/div>\n

\n \t\t\t\"Servidor\t\t\n<\/div>\n\n<\/div>\n\n<\/header>\n
\n
\n
\n
\n \n<\/div>\n\t\t\t

Applications<\/p>\n\t\t\n<\/div>\n\t\t\t

The platform for AI at scale<\/h2>\n\t\t\t\t\t

The G494-ZB4-AAP2 brings together eight GPUs, two AMD EPYC processors and 48 memory modules in a single server. It is the foundation for training models, serving generative AI to many users or running GPU-accelerated HPC.<\/p>\n\t\t

\n
\n
\n \t\t\t\"\"\/\t\t\n<\/div>\n\t\t\t

AI training<\/h3>\n\t\t\t\t\t

Eight PCIe Gen5 GPUs, including NVIDIA H200 NVL in two sets of four with NVLink bridge.<\/p>\n\t\t\n<\/div>\n

\n
\n \t\t\t\"\"\/\t\t\n<\/div>\n\t\t\t

Inference at scale<\/h3>\n\t\t\t\t\t

Generative AI and language models serving many users at once, with up to eight GPUs per server.<\/p>\n\t\t\n<\/div>\n

\n
\n \t\t\t\"\"\/\t\t\n<\/div>\n\t\t\t

HPC and simulation<\/h3>\n\t\t\t\t\t

Two AMD EPYC processors, 48 DDR5 modules and 12 NVMe Gen5 bays for scientific computing and large-scale data analysis.<\/p>\n\t\t\n<\/div>\n

\n
\n \t\t\t\"\"\/\t\t\n<\/div>\n\t\t\t

Visual computing<\/h3>\n\t\t\t\t\t

Rendering and digital twins with up to eight NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs.<\/p>\n\t\t\n<\/div>\n\n<\/div>\n\n<\/div>\n\n<\/div>\n

\n
\n
\n
\n \n<\/div>\n\t\t\t

Engineering highlights<\/p>\n\t\t\n<\/div>\n\t\t\t

Eight GPUs and two AMD EPYC in 4U<\/h2>\n\t\t
\n
\n \t\t\t\"Servidor\t\t\n<\/div>\n
\n
\n \t\t\t

Acceleration<\/p>\n\t\t\t\t\t

Up to 8 dual-slot GPUs<\/h3>\n\t\t\t\t\t

Eight FHFL PCIe Gen5 x16 slots connected through two Broadcom PEX89144 switches. Supports GPUs up to 600 W with the optional fan kit.<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Validated GPUs<\/p>\n\t\t\t\t\t

NVIDIA and AMD<\/h3>\n\t\t\t\t\t

8 x NVIDIA H200 NVL, 8 x NVIDIA RTX PRO 6000 Blackwell Server Edition or 8 x AMD Instinct MI350P, at 25 \u00b0C ambient.<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Processors<\/p>\n\t\t\t\t\t

2 x AMD EPYC<\/h3>\n\t\t\t\t\t

9005 and 9004 series, SP5 socket, cTDP up to 400 W (500 W at 30 \u00b0C ambient).<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Memory<\/p>\n\t\t\t\t\t

48 DDR5 DIMMs<\/h3>\n\t\t\t\t\t

12 channels per CPU. Up to 5200 MT\/s with EPYC 9005 and 4800 MT\/s with EPYC 9004 (1 module per channel).<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Expansion<\/p>\n\t\t\t\t\t

6 extra PCIe Gen5 slots<\/h3>\n\t\t\t\t\t

In addition to the GPUs: 3 FHHL and 3 low-profile PCIe Gen5 x16 slots for high-speed networking and DPUs.<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Power<\/p>\n\t\t\t\t\t

4 x 3000 W Titanium<\/h3>\n\t\t\t\t\t

Four redundant power supplies with 80 PLUS Titanium certification (C19 cable).<\/p>\n\t\t\n<\/div>\n\n<\/div>\n\n<\/div>\n\n<\/div>\n\n<\/div>\n

\n
\n
\n
\n \n<\/div>\n\t\t\t

Datasheet<\/p>\n\t\t\n<\/div>\n\t\t\t

G494-ZB4-AAP2 specifications<\/h2>\n\t\t
\n
\n
\n
\n \t\t\t\"Servidor\t\t\n<\/div>\n\t\t\t

Front<\/p>\n\t\t\n<\/div>\n

\n
\n \t\t\t\"Servidor\t\t\n<\/div>\n\t\t\t

Rear<\/p>\n\t\t\n<\/div>\n\n<\/div>\n

\n
\n \t\t\t

Form factor<\/p>\n\t\t\t\t\t

4U rack \u00b7 448 x 176 x 880 mm (W x H x D) \u00b7 1000.5 mm deep with the optional fan kit<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Motherboard<\/p>\n\t\t\t\t\t

MZB3-G41<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Processor<\/p>\n\t\t\t\t\t

2 x AMD EPYC 9005 \/ 9004 series \u00b7 SP5 socket (LGA 6096) \u00b7 cTDP up to 400 W (500 W at 30 \u00b0C ambient)<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Memory<\/p>\n\t\t\t\t\t

48 DDR5 RDIMM slots \u00b7 12 channels per CPU \u00b7 up to 5200 MT\/s (EPYC 9005) and 4800 MT\/s (EPYC 9004) with 1 module per channel<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

GPU<\/p>\n\t\t\t\t\t

Up to 8 dual-slot PCIe Gen5 GPUs \u00b7 up to 600 W each with the optional fan kit<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Validated GPUs<\/p>\n\t\t\t\t\t

8 x NVIDIA H200 NVL (2 sets of 4 with NVLink bridge) \u00b7 8 x NVIDIA RTX PRO 6000 Blackwell Server Edition \u00b7 8 x AMD Instinct MI350P<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

PCIe switches<\/p>\n\t\t\t\t\t

2 x Broadcom PEX89144<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Front storage<\/p>\n\t\t\t\t\t

12 x 2.5\" Gen5 NVMe\/SATA\/SAS-4 hot-swap<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Internal storage<\/p>\n\t\t\t\t\t

2 x M.2 (2280\/22110) PCIe Gen3 x4<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

PCIe slots<\/p>\n\t\t\t\t\t

8 x FHFL PCIe Gen5 x16 for GPU \u00b7 3 x FHHL PCIe Gen5 x16 (2 rear and 1 front) \u00b7 3 x LP PCIe Gen5 x16<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Networking<\/p>\n\t\t\t\t\t

2 x 1 Gb\/s (Intel I350-AM2, front) \u00b7 1 dedicated management port<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Management<\/p>\n\t\t\t\t\t

ASPEED AST2600 BMC \u00b7 VGA output<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Security<\/p>\n\t\t\t\t\t

TPM 2.0 header (optional module)<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Power supply<\/p>\n\t\t\t\t\t

4 x 3000 W 80 PLUS Titanium redundant \u00b7 C19 cable<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Cooling<\/p>\n\t\t\t\t\t

12 x 60 x 60 x 56 mm fans and 2 x 40 x 40 x 28 mm fans<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Operating temperature<\/p>\n\t\t\t\t\t

10 \u00b0C to 35 \u00b0C<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Weight<\/p>\n\t\t\t\t\t

66.8 kg (net)<\/p>\n\t\t\n<\/div>\n

\n \t\t\t

Product code<\/p>\n\t\t\t\t\t

6NG494ZB4DR000ACP2<\/p>\n\t\t\n<\/div>\n\t\t\t

Specifications per the manufacturer. Validated GPUs require 25 \u00b0C ambient, and 600 W GPUs require the optional fan kit. The server is validated with a single GPU model at a time. GPUs sold separately. With a single processor, some memory and slots are unavailable.<\/p>\n\t\t\t\t\t

Official GIGABYTE page \u2197<\/a><\/p>\n\t\t\n<\/div>\n\n<\/div>\n\n<\/div>\n\n<\/div>\n

\n
\n
\n
\n \n<\/div>\n\t\t\t

For channels and resellers<\/p>\n\t\t\n<\/div>\n\t\t\t

Your project backed by a specialized distributor<\/h2>\n\t\t\t\t\t\"\"\/\t\t
\n
\n
\n \t\t\t\"\"\/\t\t\n<\/div>\n\t\t\t

Gigabyte certification<\/h3>\n\t\t\t\t\t

Strategic partnership with Giga Computing and access to Gigabyte server platforms.<\/p>\n\t\t\n<\/div>\n

\n
\n \t\t\t\"\"\/\t\t\n<\/div>\n\t\t\t

Consultative pre-sales<\/h3>\n\t\t\t\t\t

Technical support to choose GPU, processor, memory, networking, power and cooling based on the workload.<\/p>\n\t\t\n<\/div>\n

\n
\n \t\t\t\"\"\/\t\t\n<\/div>\n\t\t\t

Stock and invoicing<\/h3>\n\t\t\t\t\t

Stock availability and invoicing through the distributor to speed up the reseller's project closing.<\/p>\n\t\t\n<\/div>\n\n<\/div>\n\n<\/div>\n\n<\/div>\n

\n
\n
\n
\n
\n \n<\/div>\n\t\t\t

Project quote<\/p>\n\t\t\n<\/div>\n\t\t\t

Request a quote for the G494-ZB4-AAP2<\/h2>\n\t\t\t\t\t

Describe your scenario and a technical specialist will get back to you with the recommended configuration.<\/p>\n\t\t

\n
\n
\n \t\t\t\"\"\/\t\t\n<\/div>\n\t\t\t

Configuration based on workload: training, inference, HPC or visual computing<\/p>\n\t\t\n<\/div>\n

\n
\n \t\t\t\"\"\/\t\t\n<\/div>\n\t\t\t

Service for resellers, integrators and businesses<\/p>\n\t\t\n<\/div>\n

\n
\n \t\t\t\"\"\/\t\t\n<\/div>\n\t\t\t

Pre- and after-sales with a technical specialist<\/p>\n\t\t\n<\/div>\n\n<\/div>\n\n<\/div>\n

\n
\n \t\t\t