Data Sheet
Graphic Bus/ GPU
GPU Memory Memory Max Power Interconnect Slot Auxiliary
Brand Model System Height/ Workload1
Memory ECC Bandwidth Consumption Bandwidth Width Cable
Interface Length
AMD Infinity
AMD MI300X OAM 192 GB HBM3 Y 5.3 TB/sec 750W 896 GB/sec N/A N/A N/A AI / HPC
Fabric Links
PCIe Gen4x16/
64 GB/sec HPC/Machine
AMD MI210 64 GB HBM2e Y 1638 GB/sec 300W Infinity Fabric DW FHFL CPU 8 pin
(PCIe 4.0) learning training
Link bridge8
Max1550 128 GB
Intel Y 3276.8 GB/sec 600W Intel XeLink - N/A N/A N/A AI / HPC
OAM HBM2e
PCIe Gen5x16/ 128 GB/sec5
Intel Max1100 48 GB HBM2e Y 1228.8 GB/sec 300W DW FHFL PCIe 16 pin AI / HPC
XeLink bridge8 (PCIe 5.0)
32 GB/sec HHHL/
Intel Flex 140 12 GB GDDR6 Y 336 GB/Sec 75W PCIe Gen4 x8 SW N/A Inferencing/Edge
(PCIe 4.0) FHHL
H100 SXM5
Nvidia 80 GB HBM3 Y 3 TB/sec 700W NVIDIA NVLink 900 GB/sec N/A N/A N/A AI / HPC
(x8)
H100 SXM5
Nvidia 80 GB HBM3 Y 3 TB/sec 700W NVIDIA NVLink 900 GB/sec N/A N/A N/A AI / HPC
(x4)
PCIe Gen5x16/ 128 GB/sec5 HPC/AI/Database
Nvidia H100 80 GB HBM2e Y 2000 GB/sec 300-350W DW FHFL PCIe 16 pin
NVLink bridge8 (PCIe 5.0) Analytics
64 GB/sec5 AI/Performance
Nvidia L40S 48 GB GDDR6 Y 864 GB/sec 350W PCIe Gen4 x16 DW FHFL PCIe 16 pin
(PCIe 4.0) graphics/VDI
PCIe Gen4x16/ 64 GB/sec5
Nvidia A30 24 GB HBM2 Y 933 GB/sec 165W DW FHFL CPU 8 pin mainstream AI
NVLink bridge8 (PCIe 4.0)
64 GB/sec Performance
Nvidia L40 48 GB GDDR6 Y 864 GB/sec 300W PCIe Gen4 x16 DW FHFL PCIe 16 pin
(PCIe 4.0) graphics/VDI
PCIe Gen4x16/ 64 GB/sec5 Performance
Nvidia A40 48 GB GDDR6 Y 696 GB/sec 300W DW FHFL CPU 8 pin
NVLink bridge8 (PCIe 4.0) graphics/VDI
64 GB/sec
Nvidia A16 64 GB GDDR6 Y 800 GB/sec 250W PCIe Gen4 x16 DW FHFL CPU 8 pin VDI
(PCIe 4.0)
64 GB/sec Inferencing/Edge/
Nvidia L4 24 GB GDDR6 Y 300 GB/s 72W PCIe Gen4 x16 SW HHHL N/A
(PCIe 4.0) VDI
64 GB/sec Inferencing/Edge/
Nvidia L4 24 GB GDDR6 Y 300 GB/s 72W PCIe Gen4 x16 SW FHFL N/A
(PCIe 4.0) VDI
32 GB/sec Inferencing/Edge/
Nvidia A2 (v2) 16 GB GDDR6 Y 200 GB/sec 60W PCIe Gen4 x8 SW HHHL N/A
(PCIe 4.0) VDI
32 GB/sec Inferencing/Edge/
Nvidia A2 (v2) 16 GB GDDR6 Y 200 GB/sec 60W PCIe Gen4 x8 SW FHFL N/A
(PCIe 4.0) VDI
64 GB/sec mainstream
Nvidia A10 24 GB GDDR6 Y 600 GB/sec 150W PCIe Gen4 x16 SW FHFL PCIe 8 pin
(PCIe 4.0) graphics/VDI
32 GB/sec Inferencing/Edge/
Nvidia T4 16 GB GDDR6 Y 300 GB/sec 70W PCIe Gen3 x16 SW HHHL N/A
(PCIe 3.0) VDI
32 GB/sec Inferencing/Edge/
Nvidia T4 16 GB GDDR6 Y 300 GB/sec 70W PCIe Gen3 x16 SW FHFL N/A
(PCIe 3.0) VDI
1. Suggested ideal workloads, but can be used for other workloads
2. Different SKUs are mentioned because different platforms might support different SKUs. This sheet doesn’t specifically call out platform-SKU associations
3. Upto 100GB/sec when RTX NVLink bridge is used, RTX NVLink bridge is only supported on T640
4. Structural Sparsity enabled
5. Upto 600GB/sec for A100 and H100 when NVLink bridge is used, upto 200GB/sec for A30 when NVLink bridge is used, upto 112.5GB/sec for A40 when NVLink bridge is
used, upto 400GB/sec for A800 when NVLink bridge is used
6. Peak performance numbers shared by Nvidia or AMD for MI100
7. Refer to Max#GPUs on supported platforms tab for detail support on Rome vs Milan processors
8. A100 w/Nvlink bridge is supported on R750XA and DSS8440; A40 w/Nvlink bridge is supported on R750XA, DSS8440 and T550; A30 w/NVLink bridge is supported on
R750XA, DSS8440 and T550; ; MI210 w/Infinity Fabric Link bridge is supported on R750XA; H100 and A800 w/Nvlink bridge will be supported on R750XA; Max1100 w/XeLink
bridge is supported on R760XA
DW - Double Wide, SW - Single Wide, FH- Full Height, FL - Full Length, HH - Half Height, HL - Half Length
PowerEdge Server GPU Matrix Data Sheet
© 2023 Dell Inc. or its subsidiaries.
NVIDIA
PLATFORM H100 H100 H100
80GB SXM5 SXM5 L40S L40 L4 A40 A30 A16 A10 A2 T4
PCIe (X8) (X4)
XE9680 Shipping
XE9640 Shipping
XE8640 Shipping
R760XA Shipping (4 )
3
Shipping (43) Shipping (43) Shipping (83) Shipping (43) Shipping (43) Shipping (43) Shipping (123)
R760 Shipping (2) Shipping (2) Shipping (2) Shipping (4) Shipping (2) Shipping (2) Shipping (2) Shipping (6)
R760xs Shipping (2)
R760xd2 Shipping (2) Shipping (1) Shipping (2)
R660 Shipping (3) Shipping (2)
R7625 Shipping (2) Shipping (2) Shipping (2) Shipping (4) Shipping (2) Shipping (2) Shipping (2) Shipping (6)
R7615 Shipping (2) Shipping (3) Shipping (3) Shipping (4) Shipping (3) Shipping (3) Shipping (3) Shipping (6)
R6625 Shipping (2) Shipping (2)
R6615 Shipping (2) Shipping (2)
R960 Shipping (4)
T560 Shipping (2) Shipping (5) Shipping (2) Shipping (6)
C6620 Shipping (2)
XR7620 Shipping (5) Shipping (2) Shipping (5)
XR5610 Shipping (2) Shipping (2)
XR8000 Shipping (2)
HS5620 Shipping (2)
AMD Intel
PLATFORM
MI210 MI300X OAM (X8) Flex 140 Max 1100 Max 1550 OAM (X4)
XE9680 Ready to Quote
XE9640 Shipping
XE8640
R760XA Shipping (43) Shipping (103)* Shipping (43)
R760 Shipping (6)* Shipping (2)
R760xs
R760xd2
R660 Shipping (2)*
R7625 Shipping (2)
R7615 Shipping (3)
R6625
R6615
R960
T560
C6620
XR7620 Shipping (2)
XR5610
XR8000
HS5620
2 | PowerEdge Server GPU Matrix Data Sheet
© 2023 Dell Inc. or its subsidiaries.
1. XE8545, DSS8440 are set configs
2. subject to change
3. R760XA, R750XA at a minimum require 2GPUs to be installed at the factory(qty) - max number of GPUs allowed, maximum number of GPUs
allowed might differ in different configurations on the same platform
* Currently available as Customer Install only
** 3 FI, 3 additional T4s can be added through Customer Install
Learn more about Contact a Dell View more resources Join the conversation
Dell solutions Technologies Expert with #HashTag
© 2023 Dell Inc. or its subsidiaries. All Rights Reserved. Dell and other trademarks are trademarks of Dell Inc. or its subsidiaries. Other trademarks may be trademarks of
their respective owners.
