Single page Print

Memory subsystem performance
Ok, let's look over this guy's results and see if there's really anything interesting going on here.

Yeah, there's the X79's claim to fame: memory bandwidth. Looks like that bump up to 1866MHz memory paid off for the new Ivy Bridge-E chip, too. This thing more than doubles the throughput of the Haswell Core i7-4770K.

Heh, ouch. Ivy-E gets pwned by Haswell in L1 cache bandwidth, even with six cores to Haswell's four. I believe Intel doubled the L1 cache's throughput in Haswell because AVX2 needs the bandwidth. Having 50% more Sandy or Ivy Bridge cores doesn't give you enough to keep up.


We have any more of that diet Bawls soda, Jamaal? I need something to wash down these ranch Corn Nuts.

Some quick synthetic math tests

Haha, yikes. The 4770K beats the new Ivy-E 4960X in a couple of these tests. You've gotta figure that's AVX2 kicking in. Doubles the floating-point throughput in a lot of cases. The Ivy-E is never slow, but it's kinda embarrassing to get caught by a quad-core with under half the memory bandwidth.