Intel: Takes The Performance Crown
The fastest Intel processor at the moment is the Core 2 Extreme QX9770, which is manufactured on a 45 nm process and consists of two 3.20 GHz dual-core processors. Each of the two CPUs accesses a 6 MB shared L2 cache. So, in total, the processor offers four computing cores and 12 MB of L2 cache. The front side bus of this CPU has been raised to 400 MHz (1600 MHz effective) and is thus faster than all other "Extremes." Officially, this processor is only supported by the X48 high-end chip set.
If the performance of a QX9770 isn’t enough for you, you can always try the Skulltrail platform, which has a dual-socket motherboard that houses two of these processors. Since Skulltrail requires the workstation Socket 771, the identical CPUs go by another name. A Skulltrail configuration gets you a total of eight cores and 24MB of L2 cache.
Despite it’s new 45nm manufacturing process, Intel’s flagship models are anything but power-friendly. A QX9770 has a 136 watt TDP. The fastest model, the QX9775 tailored to dual-socket boards, needs as much as 150 watts.
Switching focus to the older 65 nm models, which Intel still sells, its dual-core Core 2 Duo E6600 is considered very attractive. It has 4 MB L2 cache and is clocked at 2.40 GHz. The quad-core model Core 2 Quad Q6600 runs at the exact same frequency. Both processors can be overclocked easily to high levels. After the introduction of its 45 nm processors, Intel’s aforementioned 65 nm CPUs fell in price considerably, which makes them something of a bargain due to the excellent cost/performance ratio.
Within the 45 nm segment of Intel’s product lineup, the most popular models are the Core 2 Duo E8400 at 3.00 GHz and the Core 2 Quad Q9450 at 2.66 GHz. These can even be overclocked better than the 65 nm processors.
All 45 nm processors have SSE4.1 capabilities, but these extensions do not yet have the full range of commands Intel has mapped out—that will only happen in the coming generation of Nehalem-based processors. Opinions regarding the extensions’ effect on acceleration of video processing are divided. Some software developers already offer support, whereas others are yet to see a performance benefit.
Intel carries out stepping updates to its processor families from time to time. In some cases, we found considerable improvements to energy consumption. For example, in the Pentium dual-core series with the M0 stepping, Intel was able to reduce the energy requirements by as much as 29%.
Especially the new Athlon 6500@3GHz would be interesting.
By the way: the Intel-system uses DDR3-1333, AMD DDR2-1066; that makes some difference in the price (just like the mainboard).
Yannis,
Thank you for the feedback.
It does not make sense to handicap the Intel platform simply because AMD's infrastructure is currently priced to compete with Intel's mid-range. This would completely eliminate the scores for higher-end configurations like Skulltrail and the Extreme Edition CPUs, providing an incomplete picture of the current processor landscape.
In the final charts the details "3.33 GHz, DDR3-1333 (Wolfdale)" could be supplemented by the price for the entire system. Then it would be fair (you might even include Intels with DDR2 and Athlon X2s as well -- that would make a great list!).
Also, could you please give the exact command line for the LAME benchmark? And why do you keep on benchmarking it in CBR mode, even with the version bump, when all the work by its developers in recent years has essentially been on VBR mode? VBR is also highly recommended over CBR.
Why ... well it make the little green guys look even worse.
Bert you don't really need to cheat on the benchmarks to prove the Intel CPU's are generally better.
We do know that.
You have ordered the scores with fastest at the top, but what about order by price? Wouldn't it make more sense? If it's apples-apples comparison, then put apples against apples, not oranges. If it's price/performance comparison, then mention price differences also. How much both systems TCO is.
You guys make it seem like not getting the fastest mobo and ram would make a real difference in these benchmanrks.
If they didn't mismatch, what would you have wanted them to do? Only show the intel procs/setups that are closer in perfomance to AMD? If they did that, the Intel fans would whine about the lack of good intel setups.
However I still believe you made a mistake. Instead of using the M3A32 for AMD, you should have used the M3A79 or another SB750 based bored due the fact that many benchmarks have already proven the SB750 dominance over its earlier counterpart.
All in all, good article.