Intel Xeon E3-1585 v5 vs Intel Xeon E5-2628 v3 Comparison

The Intel Xeon E5-2628 v3 and the Intel Xeon E3-1585 v5 are both end-of-life server/workstation processors, but they represent two fundamentally different design philosophies from Intel. The E5-2628 v3 is a high-core-count Haswell-EP part built for multi-socket servers, while the E3-1585 v5 is a compact Skylake-H chip with integrated graphics aimed at embedded and workstation use. Despite their architectural chasms—8 cores versus 4, 22 nm versus 14 nm, a 356 mm² die versus a 171 mm² die—the benchmark data shows they land in nearly the same performance tier, making this a fascinating comparison of how generation and core count trade blows.

Head-to-Head Benchmarks

The most striking takeaway from the head-to-head results is how close these two CPUs are, despite their different core counts. In the Cinebench R15 multi-core test, the E5-2628 v3 scores 723 against the E3-1585 v5’s 720, a win margin of just 0.4%. That pattern repeats across every single benchmark in the comparison. The E5-2628 v3 wins all six tests, but the largest victory is a mere 1% in the R15 single-core test (102 vs 101). In the R20 multi-core test, the E5 leads 3015 to 3001, a 0.5% advantage. The R23 multi-core results show 7179 versus 7146, again a 0.5% gap. Single-core scores in R20 and R23 are similarly tight: 425 vs 423 and 1013 vs 1008, respectively.

What does this mean in practice? The E5-2628 v3’s 8 cores and 16 threads are effectively neutralized by the E3-1585 v5’s much higher clock speeds. The E5 runs at a 2.50 GHz base and 3.00 GHz boost, while the E3 runs at 3.50 GHz base and 3.90 GHz boost. That 0.9 GHz boost advantage on the E3 nearly closes the gap created by the E5 having double the cores. The data shows the E5-2628 v3 has a 0.4% to 1% edge in every test, but these deltas are so small they fall within typical run-to-run variance. For a builder choosing between them, the benchmark scores are effectively a tie. The average benchmark scores confirm this: the E5-2628 v3 averages 2076, while the E3-1585 v5 averages 2067, a 0.5% difference that places both at the 46th percentile among all CPUs.

FAQ

Q: Which CPU has better multi-core performance?

A: The Intel Xeon E5-2628 v3 wins every multi-core benchmark, but by a razor-thin margin. In Cinebench R23 multi-core, it scores 7179 versus 7146 for the E3-1585 v5, a 0.5% lead. The R15 multi-core result is even closer: 723 vs 720, a 0.4% advantage. The E5’s 8 cores and 20 MB of shared L3 cache give it a theoretical edge, but the E3’s higher clock speeds erase most of it.

Q: Is the E3-1585 v5 faster in single-core workloads?

A: No. The E5-2628 v3 wins all three single-core tests, but the margins are minuscule. In Cinebench R23 single-core, the E5 scores 1013 versus 1008 for the E3, a 0.5% difference. The R20 single-core test shows 425 vs 423, and R15 shows 102 vs 101. Despite the E3’s 3.90 GHz boost clock versus 3.00 GHz on the E5, the E5 still edges ahead, likely due to its larger cache architecture.

Q: Which processor supports more memory bandwidth?

A: The E5-2628 v3 has a substantial advantage here. It supports DDR4 memory on a quad-channel bus with 68.3 GB/s of bandwidth. The E3-1585 v5 supports both DDR3 and DDR4, but on a dual-channel bus with only 34.1 GB/s. That is a 2x difference in theoretical memory bandwidth, which matters for memory-intensive server workloads.

Q: Do both CPUs support ECC memory?

A: Yes, both the E5-2628 v3 and the E3-1585 v5 support ECC memory. This is a key server/workstation feature, but it does not differentiate them. Both are classified as Server/Workstation market segments.

Q: What are the PCIe lane counts?

A: The E5-2628 v3 offers 40 PCIe Gen 3 lanes (CPU only), while the E3-1585 v5 offers 16 PCIe Gen 3 lanes (CPU only). This is a major difference for expandability. The E5 can support more GPUs, NVMe drives, or other add-in cards directly from the CPU, whereas the E3 is limited to a single high-bandwidth device or a few lower-bandwidth ones.

Q: Does the E3-1585 v5 have integrated graphics?

A: Yes, the E3-1585 v5 features Iris Pro Graphics P580. The E5-2628 v3 has no integrated graphics at all. This makes the E3 a viable option for systems that need a display output without a discrete GPU, though its primary market is still server/workstation.

Architecture Differences

The two chips come from different eras and design targets. The E5-2628 v3 is built on Intel’s 22 nm process with 2,600 million transistors on a 356 mm² die. It uses the Haswell-EP architecture (codename Haswell-EP) and fits into the Intel Socket 2011-3. Its cache hierarchy consists of 64 KB of L1 per core, 256 KB of L2 per core, and a substantial 20 MB of shared L3 cache. This is a large, power-hungry design meant for multi-socket servers, though the 85 W TDP is moderate for its core count.

The E3-1585 v5 is a newer, smaller part. It uses the 14 nm process with 2,300 million transistors on a 171 mm² die—less than half the die size of the E5. It is based on the Skylake architecture (codename Skylake-H) and uses the Intel BGA 1440 socket, meaning it is soldered to the motherboard rather than socketed. Its cache is smaller: 64 KB L1 per core, 256 KB L2 per core, and 8 MB of shared L3. The E3 also includes Iris Pro Graphics P580, which the E5 completely lacks. The E3 supports both DDR3 and DDR4 memory, while the E5 only supports DDR4. The E3’s memory bus is dual-channel, while the E5’s is quad-channel. The E5 offers 40 PCIe Gen 3 lanes versus 16 on the E3. Both are end-of-life products, but the E5 was released in September 2014, while the E3 arrived later in May 2016.

The Verdict

The data points to a clear, if counterintuitive, conclusion: these two CPUs are performance equals in CPU-bound benchmarks, but they are not interchangeable in a system. The E5-2628 v3 wins all six head-to-head benchmarks, but with deltas between 0.4% and 1%. That is a statistical tie. The E5’s advantage comes from having 8 cores and 16 threads, plus 20 MB of L3 cache and quad-channel memory bandwidth. The E3-1585 v5 counters with much higher clocks (3.50 GHz base, 3.90 GHz boost) and a newer 14 nm process, but its 4 cores and 8 threads can’t overcome the E5’s core advantage in multi-threaded tests.

The real differentiators are platform-level, not benchmark-level. If you need a socketed, upgradeable server CPU with massive memory bandwidth (68.3 GB/s) and 40 PCIe lanes, the E5-2628 v3 is the only choice. If you need a compact, soldered BGA chip with integrated graphics and lower power draw (65 W TDP versus 85 W), the E3-1585 v5 wins. The E5’s quad-channel memory and 20 MB L3 cache make it better suited for memory-heavy virtualization or database workloads, even if the raw Cinebench scores say otherwise. The E3’s integrated GPU and dual-channel memory make it a fit for embedded systems or compact workstations where a discrete GPU is not an option. For pure CPU performance, pick either; for platform features, the choice is dictated by your motherboard and chassis.

Specification Differences

The two processors differ in nearly every hardware specification. The E5-2628 v3 has 8 cores and 16 threads, while the E3-1585 v5 has 4 cores and 8 threads. Base clocks are 2.50 GHz versus 3.50 GHz, and boost clocks are 3.00 GHz versus 3.90 GHz. TDP is 85 W for the E5 and 65 W for the E3. The E5 uses the Intel Socket 2011-3, while the E3 uses Intel BGA 1440. Architecture is Haswell-EP for the E5 and Skylake-H for the E3. The process node is 22 nm for the E5 and 14 nm for the E3. Transistor count is 2,600 million versus 2,300 million, and die size is 356 mm² versus 171 mm². L3 cache is 20 MB shared on the E5 versus 8 MB shared on the E3. Memory support is DDR4 only on the E5 versus DDR3 and DDR4 on the E3. Memory bus is quad-channel on the E5 versus dual-channel on the E3. Memory bandwidth is 68.3 GB/s on the E5 versus 34.1 GB/s on the E3. PCIe lanes are 40 on the E5 versus 16 on the E3. The E3 has integrated graphics (Iris Pro Graphics P580); the E5 has none. The E5’s part number is SR201, and the E3’s is SR2RB. The E3 has a launch MSRP of $556; the E5 has no listed launch MSRP.

Where Each One Wins

The E5-2628 v3 wins on raw core count, cache size, memory bandwidth, and PCIe expandability. Its 8 cores and 16 threads, combined with 20 MB of L3 cache, make it the better pick for heavily threaded server workloads that can use all available cores. The quad-channel DDR4 memory with 68.3 GB/s bandwidth gives it a 2x advantage in memory throughput over the E3, which is critical for applications that stream large datasets. The 40 PCIe Gen 3 lanes allow for multiple GPUs or high-speed NVMe arrays without needing a separate PCIe switch. If you are building a virtualization host or a database server with dense memory requirements, the E5 is the clear winner.

The E3-1585 v5 wins on clock speed, power efficiency, and integrated graphics. Its 3.50 GHz base and 3.90 GHz boost clocks are significantly higher than the E5’s 2.50 GHz base and 3.00 GHz boost, which explains why it nearly matches the E5 in single-core and multi-core benchmarks despite having half the cores. The 65 W TDP makes it easier to cool in compact chassis, and the BGA socket means it is often found in pre-built embedded systems where upgradeability is not a concern. The Iris Pro Graphics P580 is a unique feature that allows the E3 to drive displays without a discrete GPU, making it suitable for kiosks, medical equipment, or other integrated systems. If you need a low-power, high-clock CPU with built-in graphics and do not require massive memory bandwidth or many PCIe lanes, the E3-1585 v5 is the better fit.