Intel Lynnfield Core i5 and Core i7 Processors

Intel today launches its new Lynnfield processor based on its tremendously dominant Nehalem architecture. New Lynnfield processors bring with them a new socket that commands a new series of motherboards and both are very much on track to capture a huge share of the enthusiast computer hardware market.

continued...

Synthetic Benchmarks

Our regular readers know that we do not put much weight into synthetic benchmarks like those shown below. The fact of the matter is that most of the deltas you see below in performance using these synthetic benchmarks will not relate to actual performance differences you will see in real world applications.

That said, the overclocking and computer hardware enthusiast communities are full of members that use synthetic benchmarks as a measuring stick when it comes to who has the "fastest" computer system. We have moved to using synthetic benchmarks for other reasons. Synthetic benchmarks are the easiest way of telling if you have a problem with your hardware quickly and easily. Additionally, synthetic benchmarks are the easiest way to tune your system for performance gains as well. Regardless, synthetic benchmark numbers are almost always of interest when we see new processors launched and fuel the fire of many conversations of what is best.

Sandra Memory Bandwidth Buffered Integer

Article Image

At 2.66GHz / 1066MHz clock speeds the new Lynnfield processors look to be making good use of its new integrated memory controller. Intel has been very specific in telling us that the new Lynnfield memory controller is "more robust" and it seems to be showing us so here.

As our memory bus is pushed to 1600MHz, we see the i5 fall a bit behind. This is likely due to the HyperThreading on the i7 processors being able to push more data, so effectively, we are seeing dual channel DDR3 at 1600MHz begin to "starve" a bit in this benchmark using our i5 processor.

Additionally when we populate our i7-920 processor’s third channel we see our memory bandwidth soar to the numbers you likely associate with Core i7.

Obviously at all clock speeds and memory speeds we see great increases over Intel’s Core 2 and AMD’s Phenom II processors.

Sandra CPU Dhrystone ALU (2009 v1542)

Article Image

First and foremost, we see our Core i5-750 slaughter AMD’s Phenom II and Intel’s own Core 2 scores even with those two processors have a huge MHz advantage. As we have shown in the past, Phenom II still only manages to nip at Core 2’s heels at equal clocks.

At 2.66GHz and at 3.2GHz we see our Core i7 processors come in nearly dead even, with actually our Core i7-870 having a slight scoring advantage. Given these i7 processors are both Nehalem cores with HyperThreading, we should expect exactly what is recorded. As for the i7-870 having a slight advantage over the Core i7-920, this is likely due to the new memory controller of the Lynnfield.

Again Phenom II is nipping at the heels of Core 2.

Hiper Pi v 0.99B

Article Image

We use Hiper Pi in a single threaded mode. Contrary to the synthetics above, we see our scaling reversed in terms of our Core iX family of processors, albeit by minuscule amounts. I am left to wonder why our Core i7-870 is actually a tad slower than our Core i7-920 this time.

Worth mentioning is that our Core i5 and i7 processors at 2.66GHz manage to beat out our Core 2 3.2GHz processor.

Phenom II at 3.2GHz is left in the dust.

wPrime v2.00

Article Image

wPrime is a our synthetic benchmark that will run a thread for every thread available on the processor. So with our Core i7 processors, we will run 8 threads and with our Core i5, Core 2, and Phenom II processors we will run 4 threads.

HyperThreading immediately shows its dominance in this test by shaving off almost 25% off the calculation time when you compare our Core i7 processors to our Core i5 processors at equal clocks.

Giving credit where credit is due, we see AMD’s current generation Phenom II weigh in with its first victory….over Intel’s Core 2 architecture that was introduced three years ago.