Front
Delidded
Connectivity
AMD Socket SP3
The AMD EPYC 7473X is a server/workstation processor with 24 cores, launched in March 2022, at an MSRP of $3900. It is part of the EPYC lineup, using the Zen 3 (Milan) architecture with Socket SP3. Thanks to AMD Simultaneous Multithreading (SMT) the core-count is effectively doubled, to 48 threads. To further increase overall system performance, up to two EPYC 7473X CPUs can link up in a multi-processor (SMP) configuration. EPYC 7473X has 768 MB of L3 cache and operates at 2.8 GHz by default, but can boost up to 3.7 GHz, depending on the workload. AMD is building the EPYC 7473X on a 7 nm production process using 33,200 million transistors. The silicon die of the chip is not fabricated at AMD, but at the foundry of TSMC. The multiplier is locked on EPYC 7473X, which limits its overclocking capabilities.
With a TDP of 240 W, the EPYC 7473X is extremely power hungry, which means you need top-notch cooling. AMD's processor supports DDR4 memory with an eight-channel interface. The highest officially supported memory speed is 3200 MT/s, but with overclocking (and the right memory modules) you can go even higher. ECC memory is supported, too, which is an important capability for mission-critical systems, to avoid data corruption. For communication with other components in the system, EPYC 7473X uses a PCI-Express Gen 4 connection. This processor lacks integrated graphics, you might need a graphics card.
Hardware virtualization is available on the EPYC 7473X, which greatly improves virtual machine performance. Additionally, IOMMU virtualization (PCI passthrough) is supported, so that guest virtual machines may directly use host hardware. Programs using Advanced Vector Extensions (AVX) will run on this processor, boosting performance for calculation-heavy applications. Besides AVX, AMD is including the newer AVX2 standard, too, but not AVX-512.
Physical
Socket: |
AMD Socket SP3
|
Foundry: |
TSMC |
Process Size: |
7 nm |
Transistors: |
33,200 million |
Die Size: |
8x 81 mm² |
I/O Process Size: | 12 nm |
I/O Die Size: | 416 mm² |
Package: |
FCLGA-4094 |
Processor
Market: |
Server/Workstation |
Production Status: |
Active |
Release Date: |
Mar 22nd, 2022 |
Launch Price: |
$3900 |
Part#: |
100-000000507WOF 100-000000507 |
Performance
Frequency: |
2.8 GHz |
Turbo Clock: |
up to 3.7 GHz |
Base Clock: |
100 MHz |
Multiplier: |
28.0x |
Multiplier Unlocked: |
No |
TDP: |
240 W |
Configurable TDP: | 225W-280W |
Architecture
Codename: |
Milan-X |
Generation: |
EPYC
(Zen 3 (Milan)) |
Memory Support: |
DDR4
|
Rated Speed: |
3200 MT/s
|
Memory Bus: |
Eight-channel
|
ECC Memory: |
Yes
|
PCI-Express: |
Gen 4, 128 Lanes (CPU only) |
Core Config
# of Cores: |
24 |
# of Threads: |
48 |
# of CCDs: | 8 |
Cores per CCD: | 3 |
SMP # CPUs: |
2 |
Integrated Graphics: |
N/A
|
Cache
Cache L1: |
64 KB (per core) |
Cache L2: |
512 KB (per core) |
Cache L3: |
768 MB (shared) |
Features
- MMX
- EMMX
- SSE
- SSE2
- SSE3
- SSSE3
- SSE4.1
- SSE4.2
- SSE4A
- AVX
- AVX2
- ABM
- BMI1
- BMI2
- FMA3
- AES
- RdRand
- SHA
- ADX
- CLMUL
- F16C
- Real
- Protected
- SMM
- FPU
- NX bit
- SMT
- AMD-Vi
- AMD-V
- SME
- TSME
- SEV
- SenseMi
- Boost 2
- 3D V-Cache
|
Notes
Level 3 cache arranged as 256MB 2D cache plus 512MB 3D V-Cache. |