Intel Haswell-E Processor Details Leaked

Add Your Comments

Details of upcoming Intel processor, codenamed Haswell-E, have been leaked by VR-Zone. According to a few slides, Intel plans to release a new high-performance platform in the second half of 2014. While Intel plans to release Ivy Bridge-E later this year for the well-established X79 socket, Haswell-E will come with its own platform, codenamed Wellsburg, and the new 2011-3 socket. The all-new high-performance desktop platform is said to implement some advances geared toward greater performance. Notably, a quad-channel DDR4 memory configuration for enhanced bandwidth over current DDR3.

Alleged Intel Roadmap Leaked

Image Courtesy: XBit Labs

Image Courtesy: XBit Labs

The processor itself will be an evolution of newly released Haswell mainstream processor. According the same slides, it would be built on Intel’s current 22nm fabrication process. Haswell-E is expected to pack up to 8 cores with support for HyperThreading, for a total of up to 16 threads, effectively doubling the core/thread count of mainstream Haswell. The leaked graphics claims a 55% performance increase over the 4 core parts, which doesn’t really add up as doubling the number of cores alone should yield a 100% performance increase. Increases over IVB-E are said to be in the 30% range. Never mind the guy doing the math there.

Apart from the number of cores and DDR4 support, the new processor should implement 40 third generation PCI-E lanes for 2×16 or 4×8 SLI/CrossFire discrete graphics configuration. The CPU itself will not feature a built-in GPU like mainstream parts.

Read the full report at Xbit Labs

Leaked slides also point to an Haswell refresh in mid-2014, which would show up just before Haswell-E in the same manner as Haswell showed up before IVB-E is expected to be released.

– dostov

Leave a Reply

Your email address will not be published. Required fields are marked *

Discussion
  1. The 55% performance gain for doubling # of cores could make sense if not all the cores can work at full speed simultaneously (e.g., due to power or heat envelope), or if you can't get enough data to all cores simultaneously. Or it could just mean that existing benchmarks and software can't fully utilize the extra cores. (Parallelization usually experiences decreasing returns as you add more threads or cores.)