We have already said a lot about the impressive features of the new Intel® Xeon® platform, but at the end of the day, we all know that performance numbers speak louder than words. When a new CPU family arrives in our lab, we make a point of running a suite of demanding benchmarks to get a clear picture of how the new hardware performs. So how does the new v3 Xeon processor stack up against the last generation?
Benchmarking Against the v2 Platform
Every processor line has many product variations, such as core count, frequency, and cache. It was important that this comparison be apples to apples. If we could not create some parity with the hardware in the two test systems, the performance difference between the two processors would not have a lot of meaning. With this in mind, ultimately we decided to use the v2 and v3 versions of the E5-2620 CPUs for these tests.
Why did we go with the Xeon E5-2620 CPUs? Though it may not be top-of-the-line, the 2620 has consistently been one of our best-selling processors. It lives in the coveted sweet spot at the confluence of price and performance. This means the test will be relevant to a huge portion of our customer base. Sure, it would be fun to test the all-new 18 core Xeon 2699’s, but those aren’t exactly flying off the shelves.
The second reason is price. In real-world applications, how the specs change from one processor to another are not as important as how much performance you get for your dollar. While the v2 and v3 versions of the 2620 have different specs, they are extremely similar in cost from one generation to the next. This is key, because it lets us compare two systems of almost identical price which makes differences in performance much easier to analyze.
Speaking of pricing, there is one important side note about these two systems. They are similar in price but the v3 system is still slightly more expensive. Didn’t we just get through saying that the processors cost the same? Well, with the move to DDR4 there is currently a slight premium on the cost of the newer memory. This should subside quickly but can make a difference overall. It really depends most on how much memory is in the system. Our test boxes used 32GB each and the overall price difference was less than 2%, so unless you need a vast amount of memory, the price delta should be small.
Benchmark System Specs
Test Results and Synopsis
For these tests, we used the industry standard Linux-Bench live CD. The results were not surprising, with the v3 platform consistently showing an improvement over the v2. The three different C-Ray tests showed a solid gain of between 20%-30%. Hard Info and Redis each showed gains of 12%-25% and 5%-22% respectively. The standout result was OpenSSL with a huge increase in performance from one generation to the next.
What does all this mean from a price per performance metric? When factoring in the prices of the systems we tested, the Iris 228-12 (v3 system) costs a mere 2% more than the older Iris 226-12. With such an improvement in performance with every benchmark we ran, the new Xeon 2620 V3 is undisputedly the performance-per-dollar winner. Individual test results are in the infographic below:
If you have any questions about these servers, the benchmark tests, or the test results, please call us at 888-828-POGO, or email email@example.com.