New Intel E5-2600v3 processor is coming, how fast is it?
- Details
- Category: Our Blog
- Created on 06 August 2014
Last week we spent testing the next-generation on NOVATTE HPC systems based on the upcoming Intel Haswell architecture processors. We ran not only generic tests, but also several scientific applications (loaded with real-world scientific models) and are really excited to share our findings with you.
77% performance increase for optimized apps
The next-generation NOVATTE HPC will be 77% faster than the current generation! We did not see that kind of performance jump since year 2011. Even when Intel change it’s Xeon architecture from Sandy Bridge to a current Ivy Bridge a year ago, the performance increase was only 25-35%.
The heart of our next-gen HPC systems is E5-2600v3 processors that are built on a Haswell microarchitecture. One of the key features of E5-2600v3 is a new AVX2 instructions set that allows to gets potential x2 times speed up over an Ivy bridge (E5-2600v2) processors. L1/L2 cache bandwidth is increased too to support x2 FLOPS. Some applications that are already optimized for the E5-2600v3 (e.g. latest versions of Intel Linpack benchmark) will show 70-80% performance gain with just a simple HPC hardware upgrade.
15-25% performance increase for non-optimized apps
Applications that are currently not optimized for E5-2600v3 yet, such as Gromacs and others, will only get 15–25% performance boost mainly due to the increased number of CPU cores, increased processor cache and wider QPI bandwidth.
In summary, E5-2600v3 is a great processor, however you need to optimize your applications to take full advantage of the new processor instructions and hence maximize the efficiency and performance.
HPC
