AMD EPYC 9654 vs AMD Ryzen AI 7 345 Comparison

AMD
AMD

AMD EPYC 9654

CORE STATE Genoa
CORE SPECS 96 Cores / 192 Threads
CLOCK SPEED 2.4 Base / 3.7 GHz Turbo
CACHE 384 MB (shared)
MAX TDP 360W
ARCHITECTURE Zen 4
nm
PROCESS 5 nm
LAUNCH DATE 2022
VS
AMD
AMD

Ryzen AI 7 345

CORE STATE Krackan Point
CORE SPECS 6 Cores / 12 Threads
CLOCK SPEED 2 Base / 4.6 GHz Turbo
CACHE 4 MB
MAX TDP 28W
ARCHITECTURE Krackan Point
nm
PROCESS 4 nm
LAUNCH DATE 2025

PERFORMANCE BENCHMARKS

cinebench_cinebench_r15_multicore
10,248
1,712
cinebench_cinebench_r15_singlecore
1,446
271
cinebench_cinebench_r20_multicore
42,703
N/A
cinebench_cinebench_r20_singlecore
6,028
N/A
cinebench_cinebench_r23_multicore
101,675
11,461
cinebench_cinebench_r23_singlecore
14,354
1,818
passmark_data_compression
N/A
237,484
passmark_data_encryption
N/A
11,814
passmark_extended_instructions
N/A
17,003
passmark_find_prime_numbers
N/A
62
passmark_floating_point_math
N/A
42,621
passmark_integer_math
N/A
63,475
passmark_multithread
N/A
19,927
passmark_physics
N/A
1,089
passmark_random_string_sorting
N/A
25,435
passmark_single_thread
N/A
3,875
passmark_singlethread
N/A
3,875

Analysis: AMD EPYC 9654 vs AMD Ryzen AI 7 345

# Head-to-Head Benchmarks

The data presents a stark contrast between these two AMD processors. In every single benchmark where both chips were tested head-to-head, the AMD EPYC 9654 dominates the AMD Ryzen AI 7 345 by a massive margin. The largest gap appears in Cinebench R23 multicore, where the EPYC scores 101,675 against the Ryzen AI's 11,461 — a delta of -88.7% from the Ryzen's perspective. That means the EPYC delivers roughly 8.9 times the multicore throughput in this particular test. The Cinebench R15 multicore result tells a similar story: EPYC scores 10,248 versus 1,712, a -83.3% delta.

Even in single-core tests, where mobile chips often close the gap, the EPYC 9654 remains decisively ahead. In Cinebench R23 single-core, the EPYC posts 14,354 versus the Ryzen AI's 1,818, a -87.3% delta. Cinebench R15 single-core shows 1,446 for the EPYC against 271 for the Ryzen AI, a -81.3% delta. This is unusual — typically, a lower-TDP mobile part with a higher boost clock would at least narrow the single-threaded distance. Here, the EPYC's Zen 4 architecture and higher base clock of 2.40 GHz appear to overcome the Ryzen AI's 4.60 GHz boost advantage in these particular workloads.

The overall average benchmark scores are remarkably close, however. The Ryzen AI 7 345 has an average benchmark score of 29,461, while the EPYC 9654 sits at 29,409 — a difference of only 0.2%. Both chips occupy the 81st percentile among all CPUs tracked. This suggests that while the EPYC wins every shared benchmark test, the Ryzen AI's broader benchmark suite (including PassMark integer math, floating-point math, and encryption tests) keeps its overall average competitive. The nearest rivals for both confirm this: the AMD Ryzen 7 3800X (29,447) and Intel Core i5-13500HX (29,270) are within 0.7% of both chips, while the AMD Ryzen 7 PRO 5755GE (29,656) sits just 0.7% above.

# Where Each One Wins

Based strictly on the benchmark data, the AMD EPYC 9654 wins all four head-to-head tests: Cinebench R15 multicore, Cinebench R15 single-core, Cinebench R23 multicore, and Cinebench R23 single-core. The EPYC's wins are not marginal — they range from -81.3% to -88.7% deltas, indicating a performance gap of roughly 4.3x to 8.9x depending on the test. This makes the EPYC the clear choice for any workload that relies on Cinebench-style rendering, which typically scales with core count and raw compute throughput.

The AMD Ryzen AI 7 345 wins zero head-to-head tests. However, its benchmark portfolio includes results the EPYC was not tested on: PassMark data compression (237,484), data encryption (11,814), extended instructions (17,003), find prime numbers (62), floating-point math (42,621), integer math (63,475), multithread (19,927), physics (1,089), random string sorting (25,435), and single-thread (3,875). These scores, while not directly comparable to the EPYC's absent results, contribute to the Ryzen AI's average benchmark score of 29,461 — slightly higher than the EPYC's 29,409. The Ryzen AI also has integrated Radeon 840M graphics, which the EPYC lacks entirely, making it usable in systems without a discrete GPU.

For use-case segmentation, the data implies the EPYC 9654 is designed for server and workstation environments where multi-threaded rendering, virtualization, and heavy compute are paramount. The Ryzen AI 7 345, with its mobile segment designation, 28W TDP, and integrated graphics, targets portable systems where power efficiency and built-in display output matter more than raw core counts. The EPYC's twelve-channel memory bus and 460.8 GB/s bandwidth versus the Ryzen AI's dual-channel 89.6 GB/s further underscore this split.

# Architecture Differences

The architectural divide between these two chips is substantial. The AMD Ryzen AI 7 345 uses the Krackan Point codename, built on a 4 nm TSMC process, and belongs to the Ryzen AI 300 generation with Zen 5 and Zen 5c cores. It packs 6 cores and 12 threads, with a base clock of 2.00 GHz and boost clock of 4.60 GHz. Its cache layout includes 80 KB L1 per core, 1 MB L2 per core, and only 4 MB L3 total. It supports DDR5 and LPDDR5X memory across a dual-channel bus, with 89.6 GB/s peak bandwidth. ECC memory is not supported. PCIe connectivity is Gen 4 with 14 CPU-only lanes. The integrated Radeon 840M GPU is present.

The AMD EPYC 9654, by contrast, uses the Genoa codename, built on a 5 nm TSMC process, and belongs to the EPYC 9004 series with Zen 4 architecture. It contains 96 cores and 192 threads — 16 times the core count and thread count of the Ryzen AI. Base clock is 2.40 GHz, boost clock 3.70 GHz. The EPYC's L1 cache is 64 KB per core (smaller than the Ryzen AI's 80 KB), but L2 is identical at 1 MB per core. The L3 cache, however, is dramatically larger: 384 MB shared versus just 4 MB on the Ryzen AI. The EPYC also has 78,840 million transistors spread across 12 dies of 72 mm² each. Memory support is DDR5 only, across a twelve-channel bus delivering 460.8 GB/s — over 5 times the Ryzen AI's bandwidth. ECC memory is supported, and PCIe is Gen 5 with 128 CPU-only lanes.

The transistor count difference is staggering: the EPYC's 78,840 million transistors versus no listed figure for the Ryzen AI. The process node favors the Ryzen AI (4 nm versus 5 nm), but the EPYC's sheer scale and 360W TDP (versus 28W) tell the story of a processor designed for maximum throughput at any power cost. The Ryzen AI's generation mixes Zen 5 and Zen 5c cores, while the EPYC uses pure Zen 4. The EPYC's 12-die design with 72 mm² per die suggests a chiplet-based approach, whereas the Ryzen AI's monolithic mobile design prioritizes integration.

# The Verdict

The data presents an unambiguous picture for raw performance: the AMD EPYC 9654 is in a completely different league for multi-threaded and even single-threaded Cinebench workloads, outperforming the Ryzen AI 7 345 by margins of 81% to 89%. Any user whose primary need is maximum compute — particularly in rendering, scientific simulation, or server-side processing — should choose the EPYC 9654 based on these benchmark results alone. Its 96 cores, 192 threads, 384 MB L3 cache, and 460.8 GB/s memory bandwidth provide the foundation for these dominant scores.

The AMD Ryzen AI 7 345, however, is not without justification. Its average benchmark score of 29,461 slightly edges out the EPYC's 29,409, and its PassMark results cover a broader range of workloads. The 28W TDP, integrated Radeon 840M graphics, and mobile socket (FP8) make it the only viable choice for battery-powered laptops or compact systems. The EPYC's 360W TDP and SP5 socket require enterprise infrastructure. The Ryzen AI's 4 nm process and dual-channel memory support DDR5 and LPDDR5X, offering flexibility the EPYC's DDR5-only twelve-channel design cannot match in portable contexts.

The percentile ranking of 81st for both chips is telling — despite the EPYC's massive head-to-head wins, the aggregate database places them at similar overall levels. This is likely because the Ryzen AI's PassMark scores, while lower in absolute terms, are weighted differently across the benchmark suite. For a user who needs Cinebench-class rendering, the EPYC is non-negotiable. For a user who needs a balanced mobile processor with integrated graphics, the Ryzen AI 7 345 is the only option presented here. The verdict depends entirely on the use case. The EPYC's launch MSRP is $11805, stated once for reference.

# FAQ

Q: Which processor has more cores and threads?

A: The AMD EPYC 9654 has 96 cores and 192 threads, while the AMD Ryzen AI 7 345 has 6 cores and 12 threads.

Q: What is the average benchmark score difference between the two?

A: The Ryzen AI 7 345 has an average benchmark score of 29,461, while the EPYC 9654 has 29,409 — a delta of 0.2% favoring the Ryzen AI.

Q: Does the Ryzen AI 7 345 have integrated graphics?

A: Yes, it includes a Radeon 840M iGPU. The EPYC 9654 has no integrated graphics listed.

Q: Which chip supports ECC memory?

A: The EPYC 9654 supports ECC memory, while the Ryzen AI 7 345 does not.

Q: What is the TDP difference?

A: The Ryzen AI 7 345 has a TDP of 28W, while the EPYC 9654 has a TDP of 360W.

Q: In which benchmark does the EPYC show its largest lead?

A: The largest lead is in Cinebench R23 multicore, where the EPYC scores 101,675 versus 11,461 for the Ryzen AI, a delta of -88.7% from the Ryzen AI's perspective.

# Specification Differences

  • Cores: 6 (Ryzen AI 7 345) vs 96 (EPYC 9654)
  • Threads: 12 vs 192
  • Base Clock: 2.00 GHz vs 2.40 GHz
  • Boost Clock: 4.60 GHz vs 3.70 GHz
  • TDP: 28W vs 360W
  • Socket: AMD Socket FP8 vs AMD Socket SP5
  • Architecture: Zen 5 / Zen 5c (Krackan Point) vs Zen 4 (Genoa)
  • Process Node: 4 nm vs 5 nm
  • Transistors: Not listed vs 78,840 million
  • Die Size: Not listed vs 12x 72 mm²
  • L1 Cache: 80 KB per core vs 64 KB per core
  • L3 Cache: 4 MB vs 384 MB (shared)
  • Memory Support: DDR5, LPDDR5X vs DDR5
  • Memory Bus: Dual-channel vs Twelve-channel
  • Memory Bandwidth: 89.6 GB/s vs 460.8 GB/s
  • ECC Memory: Not supported vs Supported
  • PCIe: Gen 4, 14 Lanes vs Gen 5, 128 Lanes
  • Integrated Graphics: Radeon 840M vs None
  • Market Segment: Mobile vs Server/Workstation
  • Release Date: 2025-01-14 vs 2022-11-09
  • Part Number: 100-000002122 vs 100-100000789
  • Series: Not listed vs EPYC 9004 series

DETAILED SPECIFICATIONS

SPECIFICATION
EPYC 9654
AI 7 345
Core Specs
Cores
96
6 -93.8%
Threads
192
12 -93.8%
Base Clock (GHz)
2.4
2 -16.7%
Boost Clock (GHz)
3.7
4.6 +24.3%
Frequency (GHz)
2.4
2 -16.7%
Turbo Clock (GHz)
3.7
4.6 +24.3%
Multiplier
24
20 -16.7%
SMP CPUs
2
1 -50.0%
Cache
L1 Cache
64 KB (per core)
80 KB (per core)
L2 Cache
1 MB (per core)
1 MB (per core)
L3 Cache
384 MB (shared)
4 MB
Power
TDP (W)
360
28 -92.2%
Configurable TDP
320-400 W
15-54 W
Architecture
Architecture
Zen 4
—
Codename
Genoa
Krackan Point
Generation
EPYC (Zen 4 (Genoa))
Ryzen AI 300 (Zen 5 / Zen 5c)
Process Size
5 nm
4 nm
Transistors
78,840 million
—
Die Size
12x 72 mm²
—
Foundry
TSMC
TSMC
Memory
Memory Support
DDR5
DDR5, LPDDR5X
Memory Bus
Twelve-channel
Dual-channel
Memory Bandwidth
460.8 GB/s
89.6 GB/s
ECC Memory
Yes
No
Platform
Socket
AMD Socket SP5
AMD Socket FP8
PCIe
Gen 5, 128 Lanes(CPU only)
Gen 4, 14 Lanes(CPU only)
Intel Hybrid
Hybrid Cores
—
2 + 4
E-Core Frequency
—
2000 MHz up to 3.4 GHz
AMD Multi-Die
IO Process Size
6 nm
—
AI/NPU
NPU
—
Yes / 50 TOPS
Graphics
Integrated Graphics
—
Radeon 840M
Other
Market
Server/Workstation
Mobile
Production Status
Active
Active
Launch Price
$11805
—
Part Number
100-100000789
100-000002122
Package
FC-LGA6096
FP8
Tj Max
—
100°C
View EPYC 9654 Details View Ryzen AI 7 345 Details