CPU Comparison

The Intel Xeon E5-2628L v3 and the Intel Xeon E-2226G occupy different corners of the server and workstation market, yet their benchmark results place them in the same performance tier. The E5-2628L v3 is a 10-core, 20-thread Haswell-EP part designed for dense, multi-socket servers, while the E-2226G is a 6-core, 6-thread Coffee Lake-S workstation chip with integrated graphics. Despite the architectural gap of roughly five years and a significant difference in core counts, their Cinebench scores are nearly identical. The data shows the E5-2628L v3 wins all six head-to-head benchmarks, but the margins are razor-thin, ranging from 0% to 0.5%. This makes the choice between them less about raw performance and more about platform features, memory bandwidth, and specific workload characteristics.

FAQ

Q: Which processor has more cores and threads?

A: The Intel Xeon E5-2628L v3 has 10 cores and 20 threads, while the Intel Xeon E-2226G has 6 cores and 6 threads. The E5-2628L v3 offers more than double the thread count.

Q: How do their multi-core Cinebench R23 scores compare?

A: The E5-2628L v3 scores 9417, and the E-2226G scores 9368. The E5-2628L v3 leads by 0.5%, a difference of only 49 points, which is within run-to-run variance.

Q: Do they support ECC memory?

A: Yes, both processors support ECC memory. This is a key requirement for server and workstation reliability.

Q: What are the socket requirements for each?

A: The E5-2628L v3 uses Intel Socket 2011-3, while the E-2226G uses Intel Socket 1151. They are not interchangeable.

Q: Which processor has a higher single-core clock speed?

A: The E-2226G has a base clock of 3.40 GHz and a boost clock of 4.70 GHz. The E5-2628L v3 has a base clock of 2000.00 MHz (2.00 GHz) and a boost clock of 2.50 GHz. The E-2226G has significantly higher clock speeds.

Q: What is the difference in memory bandwidth?

A: The E5-2628L v3 supports quad-channel DDR4 with a theoretical bandwidth of 59.7 GB/s. The E-2226G supports dual-channel DDR4 with a theoretical bandwidth of 42.7 GB/s.

Architecture Differences

The E5-2628L v3 is built on the Haswell-EP architecture, fabricated on Intel's 22 nm process node. It features 2,600 million transistors on a 356 mm² die, reflecting the larger, more complex design of a server-class chip. Its cache layout includes 64 KB of L1 and 256 KB of L2 per core, with a substantial 25 MB of shared L3 cache. This large cache is typical of multi-core server processors designed to handle large datasets and heavy multi-threaded workloads. The E5-2628L v3 also offers 40 PCIe Gen 3 lanes (CPU only), enabling high-bandwidth connectivity for multiple GPUs or NVMe storage devices.

In contrast, the E-2226G uses the Coffee Lake-S WS architecture, built on Intel's 14 nm process. It has a smaller die size of 154 mm², and while it shares the same 64 KB L1 and 256 KB L2 per-core configuration, its shared L3 cache is only 12 MB. This is half the L3 capacity of the E5-2628L v3, which can impact performance in cache-sensitive workloads. The E-2226G provides only 16 PCIe Gen 3 lanes (CPU only), limiting expansion options compared to the E5-2628L v3.

A critical difference is the integrated graphics. The E-2226G includes HD Graphics P630, while the E5-2628L v3 has no integrated graphics. This means the E-2226G can drive displays without a discrete GPU, a significant advantage for compact workstations. The E5-2628L v3 requires a discrete graphics adapter for any visual output. Both processors support ECC memory, but the E5-2628L v3 uses a quad-channel memory bus, whereas the E-2226G uses a dual-channel bus, directly impacting theoretical memory bandwidth.

The Verdict

The benchmark data presents a unique situation: the E5-2628L v3 wins every single head-to-head Cinebench test, but by the smallest possible margin. In Cinebench R23 multi-core, the E5-2628L v3 scores 9417 versus 9368 for the E-2226G, a 0.5% difference. The single-core R23 scores are 1329 versus 1322, again a 0.5% lead for the E5-2628L v3. In fact, the only test that is not a 0.5% margin is Cinebench R15 single-core, where both chips score exactly 133, resulting in a 0% delta.

Given this, the data suggests that neither chip has a meaningful raw performance advantage in Cinebench workloads. The E5-2628L v3 is the nominal winner with a 6-0 record, but the practical performance is effectively identical. The decision must be made on platform and feature differences. The E-2226G is the sensible choice for a single-socket workstation requiring integrated graphics, higher clock speeds (4.70 GHz boost versus 2.50 GHz), and a more modern architecture. The E5-2628L v3 is the logical pick for a multi-socket server environment where the quad-channel memory bus, larger 25 MB L3 cache, 40 PCIe lanes, and higher thread count (20 threads) are more valuable than clock speed or integrated graphics. The data indicates the E5-2628L v3 is the better choice for memory-bandwidth-intensive and highly parallel server workloads, while the E-2226G is the choice for a cost-sensitive, single-socket workstation with GPU-less display needs.

Specification Differences

| Specification | Intel Xeon E5-2628L v3 | Intel Xeon E-2226G |

|---|---|---|

| Cores | 10 | 6 |

| Threads | 20 | 6 |

| Base Clock | 2000.00 MHz | 3.40 GHz |

| Boost Clock | 2.50 GHz | 4.70 GHz |

| TDP | 75 W | 80 W |

| Socket | Intel Socket 2011-3 | Intel Socket 1151 |

| Architecture | Haswell | Coffee Lake |

| Process Node | 22 nm | 14 nm |

| Die Size | 356 mm² | 154 mm² |

| Transistors | 2,600 million | Not specified |

| L3 Cache | 25 MB (shared) | 12 MB (shared) |

| Memory Bus | Quad-channel | Dual-channel |

| Memory Bandwidth | 59.7 GB/s | 42.7 GB/s |

| PCIe | Gen 3, 40 Lanes (CPU only) | Gen 3, 16 Lanes (CPU only) |

| Integrated Graphics | None | HD Graphics P630 |

| Release Date | 2014-09-07 | 2019-05-28 |

| Launch MSRP | $1364 | $255 |

Head-to-Head Benchmarks

The head-to-head data shows a consistent pattern of the E5-2628L v3 winning by a hair. In Cinebench R15 multi-core, the E5-2628L v3 scores 949 versus 944 for the E-2226G, a 0.5% advantage. The single-core R15 test is a tie at 133 points for both, marking the only dead heat.

Moving to Cinebench R20, the multi-core results are 3955 for the E5-2628L v3 and 3934 for the E-2226G, with the E5-2628L v3 winning by 0.5%. The single-core R20 scores are 558 versus 555, again a 0.5% lead for the E5-2628L v3. The most recent Cinebench R23 results continue this trend: multi-core scores are 9417 versus 9368, and single-core scores are 1329 versus 1322. In both cases, the E5-2628L v3 wins by 0.5%.

These results are notable because the E-2226G has significantly higher clock speeds (4.70 GHz boost versus 2.50 GHz) and a newer architecture. The fact that the E5-2628L v3 still manages to edge ahead in single-core tests suggests that the additional cores and threads, combined with the larger 25 MB L3 cache, compensate for the lower clock speeds. The E5-2628L v3's 10 cores and 20 threads provide more parallel processing capacity, which is reflected in the multi-core scores, but the single-core wins are surprising given the clock speed deficit.

Where Each One Wins

The E5-2628L v3 wins in every benchmark category recorded, but the margins are so small that they are unlikely to be perceptible in real-world use. The data shows its real advantages lie outside Cinebench. Its quad-channel memory bus offers 59.7 GB/s of theoretical bandwidth versus 42.7 GB/s for the E-2226G, a 39.8% advantage. This makes the E5-2628L v3 the better choice for memory-bandwidth-bound tasks such as large-scale virtualization, in-memory databases, and high-performance computing workloads that stream data through the memory controller.

The E5-2628L v3 also wins on platform scalability. With 40 PCIe Gen 3 lanes, it supports far more expansion cards, storage controllers, and GPUs than the E-2226G's 16 lanes. Its 20 threads and 25 MB of L3 cache are better suited for heavily threaded server applications like database servers, web servers, and multi-tenant cloud workloads. The E5-2628L v3 is the winner for dual-socket or multi-socket server configurations where its 2011-3 socket and higher core count are essential.

The E-2226G wins in areas not captured by the Cinebench benchmarks. Its integrated HD Graphics P630 eliminates the need for a discrete GPU, making it the winner for headless workstations or systems that require basic display output. Its higher clock speeds (3.40 GHz base, 4.70 GHz boost) provide lower latency for lightly threaded applications, even if Cinebench does not show a benefit. The E-2226G also uses the 1151 socket, which is common in mainstream motherboards, and its 14 nm process node is more modern than the 22 nm node of the E5-2628L v3. For a single-socket workstation with light to moderate multi-threading needs and a requirement for integrated graphics, the E-2226G is the practical winner, despite losing the benchmark tally.