NVIDIA AI 算力、搭建部署(H100、H200、H800、RTX4090、RTX PRO6000、H20、A100、A800等）、调试、专业技术服务、 解决您 IT疑难问题解决方案（NVIDIA PRODUCT 技术服务遵守美国出口相关规定和中国出口相关规定)

NVIDIA PRODUCT 技术服务遵守美国出口相关规定和中国出口相关规定
NVIDIA AI 算力、搭建部署(H100、H200、H800、RTX4090、RTX PRO6000、H20、A100、A800等）、调试、专业技术服务、 解决您 IT疑难问题解决方案

NVIDIA H100

NVIDIA HGX H100 4-GPU
Dual 5th/4th Gen Intel Xenon Scalable Processor
5 Hot-Swap Heavy Duty Cooling Fans
32 DIMM Slot DDR5
Baseboard Area  For GPUs

Item                Description
SYS-421GU-TNXR      X13DGU-P,CSE-458GTS-R3K02P,HF,RoHS/REACH
GPU                 NVIDIA Redstoe-Next HGX H100 4-GPU with 94GB
GPU                 8562Y+ 2P  32C2. 8G300W (32/2.5,24/3/270)
Memory              64GB  DDR5-56000 2RX4 LP (16GB) ECC RDIMM
SSD                 Samsunng  PM9A3 1.9TB NVMe PCIex4 U.2 7mm  1DWPD
IB                  900-9X766-003N-SQ0 1XNDR400G IB/EW OSFP  Gen5x16 CX7

/////////////////////////////////////////////////////////////////////////////////

Item                Description
SYS-421GU-TNXR      X130DGU-P.CSE-458GTS-R3K02P,HF,RoHS/REACH
EWCSC               0%3YRS LOABOR,3YRS PARTS,1YR CRS UNDER LIMITED WARNTY
GPU-NVHGX-H100-494  NVIDIA Redstone-Next HGXH1004-GPU with 94GB
P4X-EMR8562Y-SRN6G-MCC 8562Y 2P 32C2 8G300W(32/2.5,24/3/270)60M SGX512 DS,IA.QA,DL
MEM-DR564L-SL01-ER56 64GB DDR5-56002RX1 LP(16GB)ECC RDIMM HF,RoHS
HDS-SUNO-MZQL21T9HCJRA7 Samsung PM9A3 1.9TB NVMe PCIe4x4 U.27mm 1DWPD 5YR SED Opan
ACC-CX766003N-SQ0    900-9X766-003N-SQ01XNDR400G IB/EN OSFP Gen5x16 NoCrypto CX7
AQM-PCIE5-SXM5-Q-P   PCIe Gen5 Transition board for Redstone-Next Universal s
CBL-MCIO-1253RM5FB1-X MCIOX8(RAto STR)FFC,Fold,46-53,5cm,30AWG,RoHS
CBL-MCIO-1253RM5FB2-X 4 MCIOX8(RAto STR)FFC,Fold,46-53,5cm,30AWG,RoHS
CBL-PWEX-1096-43      SurIok 5.7MM and RA-PD 3.6MM hamess,Left Exit,4AWG,43
CBL-PWEX-1131-60      MicroHi(2x2 to 2x2),PH3.0,60CM,9.5A/P,18AWG,RoHS
MCP-240-45810-0N      Sheet-metal tray for Redstone-Next UBB assembly in sc458g,Ro
MCP-240-45807-0B      CSE-458G 3pcs mylar air block set for Redstone-Next,RoHS
SNK-P4010PV           3.5U Passlve GPU 3D-VC HS for NVIDIA Redstone-Next SXM5-Q-P
SKT-1424L-001B-FXC    SOCKETEE1B CARRIER LGA 4677 W/SHIM SP MCCLCC,DG1.0.RoHS

Item                Description
SSD                 Samsunng  PM9A3 1.9TB NVMe PCIex4 U.2 7mm  1DWPD
/////////////////////////////////////////////////////////////////////////////////

SSD Samsunng Description
MZ1LB960HAJQ-00007	Samsung PM983 960GB NVMe PCI-E 3.0 x4 V4 M.2 22x110mm (1.3 DWPD)
MZ1LB1T9HALS-00007	Samsung PM983 1.92TB NVMe PCI-E 3.0 x4 V4 M.2 22x110mm (1.3 DWPD)
MZ1LB3T8HMLA-00007	Samsung PM983 3.84TB NVMe PCI-E 3.0 x4 V4 M.2 22x110mm (1.3 DWPD)

/////////////////////////////////////////////
Samsung PM983a 2.5" 7.68TB SSD(PCIe Gen3)
MZQLB7T6HALA-00207  PM983a 2.5" 7.68TB SSD(PCIe Gen3)

Item                Description
SSD                 Samsunng  PM9A3 1.9TB NVMe PCIex4 U.2 7mm  1DWPD
/////////////////////////////////////////////////////////////////////////////////

SSD Samsunng Description
MZ1LB1T9HALS-00007	Samsung PM983 1.92TB NVMe PCI-E 3.0 x4 V4 M.2 22x110mm (1.3 DWPD)

/////////////////////////////////////////////
SSD Samsunng Description
MZ1LB960HAJQ-00007	Samsung PM983 960GB NVMe PCI-E 3.0 x4 V4 M.2 22x110mm (1.3 DWPD)
MZ1LB1T9HALS-00007	Samsung PM983 1.92TB NVMe PCI-E 3.0 x4 V4 M.2 22x110mm (1.3 DWPD)
MZ1LB3T8HMLA-00007	Samsung PM983 3.84TB NVMe PCI-E 3.0 x4 V4 M.2 22x110mm (1.3 DWPD)

NVIDIA HGX H100 4-GPU
Dual 5th/4th Gen Intel Xenon Scalable Processor
5 Hot-Swap Heavy Duty Cooling Fans
32 DIMM Slot DDR5
Baseboard Area  For GPUs

执行结果中各检查项均显示OK状态，表示GPU诊断无异常。 
time="xxxxxxxxxxxxxxxxx" level=info msg="gpu_check output:"
    GPU PCI Slot: xxxxxxxxx, Diagnosis result: xxxxxxxxxxxxxxxx
    * Double Bit Error Check – OK
    * Info Rom Corrupted Check – OK
    * eRDMA Incorrect Check – OK
    * Kernel Upgrade Check – OK
    * Fabricmanager running Check – OK
    * Power Cable Error Check – OK
    * GPU Device Lost Check – OK
    * GPU Driver Install Check – OK
    * GPU Xid Error Check – OK
    GPU PCI Slot: xxxxxxx, Diagnosis result: xxxxxxxxxxxxxxxx
    * Power Cable Error Check – OK
    * GPU Device Lost Check – OK
    * GPU Driver Install Check – OK

执行结果中某个或多个检查项显示Failed状态，表示GPU诊断出现异常。
time="xxxxxxxxxxxxxxxxxx" level=info msg="Begin to gpu health check"
time="xxxxxxxxxxxxxxxxxx" level=info msg="gpu_check output:"
GPU PCI Slot: xxxxxxxxxxxxxxxxxx, Diagnosis result: xxxxxxxxxxxxxxxxxx
 * Power Cable Error Check – OK
 * Double Bit Error Check – OK
 * GPU Driver Install Check – Failed
 * Kernel Upgrade Check – Failed
 * GPU Device Lost Check – OK
 * Info Rom Corrupted Check – OK
 * eRDMA Incorrect Check – OK
 * GPU Xid Error Check – OK
 * Fabricmanager running Check – OK
time="xxxxxxxxxxxxxxxxxx" level=info msg="Detailed logs are located in /var/log/check_gpu.tar.gz"
time="xxxxxxxxxxxxxxxxxx" level=info msg="Complete GPU health check"

大模型应用技术服务
API、SLD、云监控、资源编排、算力及管理、弹性GPU云主机/容器
训练数据高速读写、NVME SSD
训练数据归档、存储。

$#RV64I ALU + W-variants for 32-bit ops. 4-bi alu_control, isword picks word path

大模型应用技术服务 成功案例：
某算力中心 H100 部署安装成功案例。
某算力中心 H200 部署安装成功案例。
某算力机房 RTX4090 部署安装成功案例。
某算力机房 RTX PRO6000 部署安装成功案例。

某算力 训练数据高速读写NVME SSD 硬盘升级、部署安装成功案例。
某AI训练数据归档系统的存储NVME SSD硬盘跟换成功案例。

某算力中心 NVIDIA H100 GPU Failed  专业技术服务成功案例
某算力中心 NVIDIA H200 GPU Failed  专业技术服务成功案例

某算力大模型的 NVIDIA H100 GPU0 Failed  专业技术服务成功案例
某算力大模型的 NVIDIA H100 GPU1 Failed  专业技术服务成功案例

某算力大模型的 NVIDIA H200 GPU0 Failed  专业技术服务成功案例
某算力大模型的 NVIDIA H200 GPU1 Failed  专业技术服务成功案例

某算力大模型的 NVIDIA H200 运算很慢、卡顿、运算不通畅、专业技术服务成功案例
某算力大模型的 NVIDIA H100 运算很慢、卡顿、运算不通畅、专业技术服务成功案例

某AI算力服务商NVIDIA H100 部署、安装技术服务成功案例。
某AI大型企业NVIDIA H200 部署、安装技术服务成功案例。
某超算中心运营商NVIDIA H100部署、安装技术服务成功案例。
某AI算力服务商NVIDIA H100 的GPU 调试drive软件成功案例。