CVData/CVCX are multi-threaded, with the exception of index generation. Index generation does not multi-thread well. So, multi-cores doesn’t help.
As for simulation, multi-cores helps enormously. However, both processors support eight threads, due to hyper-threading in the I7. Hyper-threading does not give you an entire additional set of cores. In fact, the original hyper-threading wasn’t that good. But, the last three generations have shown enormous improvement in HT. OTOH, the AMD chip isn’t really eight full cores either. It’s four pairs of cores with each pair sharing functionality.
On overclocking, the 4790K does not overclock well. Above 4.4Ghz, you must increase CPU voltage at a ridiculously high rate to keep it stable. I’m only eking out another 3%, instead of the 20% I got with the I7-2700K.
Then there is the chip architecture. Haswell added additional load/store and integer units. This is great for simulation. The CPU speed increases for the last three generations have all benefitted simulation more than then normal benchmarks would suggest.
You really can’t compare Ghz between two entirely different sets of architecture. The old Pentium got up to 3.8Ghz, and was a fraction of the speed of new CPUs. The complexity of these tiny chips is incredible. 1.4 billion transistors in the 4790.
Having said all that, I haven’t tested the AMD chip. (And, I wouldn’t until the building air conditioning is turned on.
)
Bookmarks