Introduction
Today Intel® finally unveiled their latest generation of the Intel® Xeon® processor E5-2600 product family with the launch of their v4 SKUs. The core, codenamed Broadwell offers several enhancements over the existing Haswell core featured in the current v3 range, yet remains compatible with the existing set of servers and mainboards in Supermicro's X10DP range.
To upgrade, simply update your existing X10DP motherboard to the latest BIOS from Supermicro's® website before installing your new processor to enable support. To identify systems which are Broadwell ready, Supermicro have made this very simple, any compatible system with BIOS 2.0 or greater can take the new processor.
To make it even easier, we at Boston have compiled a simple to follow BIOS upgrade guide to help you through the process which you can view by clicking here.
If in doubt however, our technical team are standing by to help, simply open a ticket on our portal and we'll get back to you with helpful advice as soon as we can.
Improvements with v4
The differences between the various models are outlined in the below table, but can be simply summed up as, more cores and faster / improved memory compatibility for the same power consumption.
| Feature | Xeon® processor E5-2600 v2 (Ivy Bridge-EP) |
Xeon® processor E5-2600 v3 (Haswell-EP) | Xeon® processor E5-2600 v4 (Broadwell-EP) |
|---|---|---|---|
| Cores per Socket | Up to 12 | Up to 18 | Up to 22 |
| Threads per Socket | Up to 24 threads | Up to 35 threads | Up to 44 threads |
| Last-level Cache | Up to 30MB | Up to 45MB | Up to 55MB |
| QPI Speed (GT/s) | Up to 8.0GT/s | 2 x QPI 1.1 channels 6.4, 8.0, 9.6 GT/s | |
| PCIe* Lanes/Controllers/Speed (GT/s) | 40 / 10 / PCIe* 3.0 (2.5, 5, 8 GT/s) | ||
| Memory Population | 4xDDR3 Channels, RDIMM, UDIMM, LRDIMM | 4 channels of up to 3 RDIMMs or 3 LRDIMMs | + 3DS LRDIMM |
| Max Memory Speed | Up to 1866 | Up to 2133 | Up to 2400 |
| TDP(W) | 150 (Workstation), 130 | 160 (Workstation only), 145, 135, 120, 105, 90, 85, 65, 55 | |
A tick in the Intel® tick/tock release schedule, Broadwell brings a smaller / newer manufacturing process based on 12nm technology which naturally enhances power consumption and allows for better clock frequencies at equivalent or lower TDP's than its predecessor.
Comparing two popular SKU' across the v3 and v4 range in the table below, we can see that whilst frequency drops slightly, power consumption remains the same, even though the number of cores and cache increases.
| E5-2650 v3 | E5-2680 v3 | E5-2650 v4 | E5-2680 v4 | |
|---|---|---|---|---|
| CPU Base Frequency (GHz) | 2.3 | 2.5 | 2.2 | 2.4 |
| Cores | 10 | 12 | 12 | 14 |
| Threads | 20 | 24 | 24 | 28 |
| Level 3 Cache (MB) | 25 | 30 | 30 | 35 |
| Memory Channels | Quad | Quad | Quad | Quad |
| TDP (Watts) | 105 | 120 | 105 | 120 |
Utilising these additional cores, performance naturally increases between generations, however for those applications which are single threaded or prefer to use fewer cores Intel® still offers a frequency optimised set of processors, the flag ship of which is the E5-2637 v4, a 4 core part with a base frequency of 3.5GHz, which can turbo to up 3.7GHz when conditions allow.
Preliminary Benchmarks
To give an idea of the kind of performance benefits which you might see in your real world applications, we have been busy in Boston Labs running through a series of benchmarks to help you choose the right SKU for you.
All benchmarks are collected from the same system with default BIOS settings and fully updated drivers for all hardware. If you'd like to know more about the system in question or our results, please drop us a line.
Our results are colour coded to show newer v4 processors in green, against v3 which are coloured in blue.
Cinebench 15 - Windows 64-bit
A test suite designed to test your systems performance in relation to Maxon's 4D Cinema modelling software. It measures Open GL, CPU (multi-threaded) and CPU (single thread) performance. The test itself renders a photorealistic scene by breaking it up into tiles, then sending one of those to each core to be rendered - this makes it an excellent benchmark for multi-core operations with real world applications.
Results measured in points - the higher the result, the better the performance.
Intel® Optimised Linpack - Linux 64-Bit
Intel® Linpack is well known in the industry as being the de facto benchmark for analysing the performance of processors for mathematical / scientific performance. It's often used as a measure of how compute clusters compare and is even used as a yardstick to measure the coveted Top 500.
Results measured in GFLOPS - the higher the result, the better the performance.
Stream - Linux 64-Bit
A benchmark that exclusively looks at memory bandwidth, Stream is widely accepted as the standard for measuring transfer speed. With the improvement in DDR-4 memory performance of approximately 12% raw clock speed, it is interesting to see how this benchmark realises that difference.
Results measured in MB/s - the higher the result, the better the performance.
Power Consumption Linpack Load - Inline Power Meter
We measured the power consumption of a complete system including all other components at the wall, then stress tested the system with the Intel® Linpack benchmark and recorded the results. Small gains in power consumption can be seen as a result of the new process - great news; especially with the enhanced performance delivered in tandem.
Results measured in Watts - the lower the result, the less power is consumed
These results are just a small selection of those that we have been working on here at Boston Labs
For more information about this latest range of Intel Xeon Processors, contact our sales team.