Core 2 Extreme Benchmarks, Round 2

6/8 - At the last IDF in the spring, Intel took the rare step of allowing certain attendees to benchmark their pre-release silicon. In the meantime, others have been benching the Core microarchitecture and Core 2, as well.

One of the ones doing the benchmarking at the last IDF was AnandTech. AnandTech recently got its hands on additional Intel silicon, for an additional round of benchmarking. At Computex, Anandtech staff were "faced with the choice of testing Conroe or sleeping".

There's still lots more to be done, when Core 2 is officially released. AnandTech staff were under severe time contraints. The hardware had to be returned in a timely manner, "as to not alert anyone that it was missing".

Test results are still preliminary, much like the benchmarks done last spring: much like the testing done by others since then. Why benchmark at all, one might ask. For one, certain low level characteristics of the chip were tested this time. For another, it was nice to benchmark on systems "without the help of Intel".

There's good news, and there's bad news, for AMD folks.

First the good news. Surprisingly, given all the hype surrounding Intel's new chip, AMD won the memory tests. AMD performed especially well on the memory bandwidth test. The reason is that Intel uses a front side bus (FSB) to communicate with main memory. There are limitations to this approach. That's right. Intel uses a FSB, AMD does not.

The memory results are less surprising, when one reflects on the fact that AMD puts the memory controller directly on the CPU die. Given AMD's on-die memory controller, what is surprising, is how close Intel comes to matching AMD in the memory latency test. Very close. The reason this is impressive is because Core 2 "has no on-die memory controller".

Intel likes to improve upon its previous generation hardware when it releases a new chip. Core 2 sports a faster FSB, lower CPU clock. Compared to the Pentium D, "memory access latency is reduced tremendously".

Now for the bad news, for AMD. Memory is just a subsystem. The Core 2 Extreme chip won every other contest, sometimes by alot, sometimes by a little.

In the cases in which the race was close, sometimes this was due to the inherent virtues of the Athlon 64 FX-62 chip. The "Athlon 64 FX-62 can still be pretty competitive in some areas".

At other times, it may simply mean that the bottleneck for the application being tested lie somewhere besides the CPU.

So, who is going to want a Pentium D, when they can have a Core 2? And let's not forget the even older, single core Pentium 4. Intel intends to cut prices on its previous generation chips in order to help move inventory. The news of Intel's impending price cuts was so big that not only Core 2, but also the price cuts, were "the talk of the town in Taiwan".

Cutting prices may help sell older chips, but it also eats into (profit) margins, and lower margins in turn knock down the price of your stock. Ouch! When will it end?

Even so, when Core 2 Duo is released, we should have dual-core for under $100! AMD shall probably follow suit and cut prices, too. "AMD's cheapest dual core offering is still hovering around the $300 mark".

Remember, you can have winners and losers with computer chips. The Athlon 64 was a winner. Some would consider NetBurst a failure. With this latest round of benchmarks, it looks like "Intel really has a winner on its hands".