Intel Improves Performance by 80%!!! There’s a Catch Though
In a world where every regime is trying to topple the current controlling empire, even small performance gains can become big news. Typically, a 10-15% performance gain on a new generation of CPUs is expected, but if that performance is increased by 20-25%, then excitement goes through the roof! Intel isn’t bragging about a mere 25% though. In their newest line up of Xeon processors, they are claiming up to an 80% increase in performance! At first glance, this is nothing short of astronomical, but upon further investigation, the implications of this move may not be the best thing for the PC gamer.
Intel’s new family is the Xeon E7 V2. Depending on whether businesses want power or cores, they have a 6 core chip at 3.4 GHz and a 15 core chip at 2.8 GHz, as well as plenty of chips to fill in everywhere else. They also sport up to a 37.5 MB cache. Another point of interest in these specs is the die size. While the 6 core chips run a fairly standard 257mm2 die size, the 15 core chips are a massive 541mm2! To put this in perspective, an AMD FX-8350 has a die size of 315mm2. Unfortunately, this carries some of the same vibes that Haswell had when it released. While Haswell did increase performance, it felt very much like a factory overclocked Ivybridge chip. To initially sum up these new server chips, it feels like Intel simply decided that bigger is better. But what about that 80% performance increase?
Here’s where In-Memory Analytics comes into play. For giant businesses like eBay (who handles data sets of more than 50 petabytes), when they have software analyzing this, the CPU has to constantly pull the info from the hard drives to analyze. What In-Memory does is take the whole data set, store it in the memory, and then proceed to analyze it from the much faster DRAM. Hence the “up to 80%” increase in performance. So while this family of CPUs is better than the last gen, the real innovation is by getting the processor to handle 3x more memory capacity while having 4x the I/O Bandwidth. This is great news for those giant data-crunching conglomerates, but this has a possible negative connotation for the home user.
This move by Intel looks to form an even wider gap between enterprise use and consumer use. A while ago, it was easy to see what was hitting the server world and watch that technology influence the enthusiast world. Unfortunately, the level of data-crunching this family is meant to handle doesn’t look like a practical technology to implement in a typical gaming environment. With Intel focusing this much effort into enterprise, and constantly endeavoring to slide more into the mobile market, there could be a chance that the enthusiast end of CPUs will be overlooked. But before you get too upset, this is merely a possible observation, and even I have doubts that Intel would actually do that.
Perhaps the best implication of this release is that this is just another sign of technology eventually reaching the limits of silicon. While In-memory Analytics is a great innovation that the Xeon E7 v2 Family exploits rather well, the majority of the architecture is nothing new. The concept of 15 cores may sound like a huge leap for Intel, but more than doubling the chip size to reach that core count isn’t exactly impressive. As that limit edges closer and closer from the horizon, new materials will have to start stepping in to allow the growth of technology. I, for one, have high hopes for graphene!
Written by JoshJ