Last week, Intel unveiled its brand new Core microarchitecture under the guise of Core 2 Duo. I spent many, many man-hours testing four of the new processors, concluding that Core 2 Duo simply kicks ass
in every processor-intensive task I've tested.
With the exception of gaming performance - which is limited by the video card nowadays - Core 2 is simply so much faster than AMD's equivalent processors, leaving little reason to consider buying a new desktop machine with an AMD processor at the heart.
Just hold on for a minute before we get too carried away, because happenings during Intel's quarterly conference call
have changed the economics of upgrading to a new Core 2 Duo system next week.
Yesterday, Intel's Chief Executive Paul Otellini said that "We notified customers we're pulling in both the desktop and server [launch] of the first quad-core processors into the fourth quarter of this year from the first half of 2007."
"It now appears to make little sense to spend money on a Core 2 Duo too, unless you're planning on using the chip as a stop gap."
So, Intel has just launched its new dual-core processors based on the Core microarchitecture, and they are set to be replaced by quad-core processors at the end of this year. This announcement will make Intel's Core 2 Duo processors obsolete before they've even had the chance to really penetrate the market in any substantial volumes.
If it makes little sense to spend money on an Athlon 64 - unless you already own a Socket 939 or Socket AM2 motherboard - it now appears to make little sense to spend money on a Core 2 Duo too. That is unless you're planning on using the chip as a stop gap until the quad-core processors are available.
Last week, the E6400 looked to be the pick of the Core 2 Duo processors. While it will still be a great purchase next week when it's available, it seems Intel has already moved the goalposts.
The other Core 2 Duo processor that really caught our eye was the E6600. Intel's list price for the chip is US$316 - this is a little more than the Athlon 64 X2 5000+, which looks set to sell for around US$282. However, the advantages to spending an extra $35 are simply massive - the 12% price increase is worth over 20% in terms of performance. Considering this, the E6600 would be my choice when it comes to D-Day on the 24th... if it wasn't for Otellini's revelation yesterday.
In light of this, if you're desperate to build a machine based on Intel's Core 2 Duo processors, I recommend that you spend as little as possible, saving your money for the quad-core upgrade later this year.
Intel's Core 2 Duo launch was the first major change in the metrics of the processor industry for over a year and a half, back when Intel announced the first dual-core Pentium D processors. AMD then followed later in the year with its Athlon 64 X2 chips and beat the competition into a corner. As they say: what goes around comes around
. This time it is AMD's turn to take the back seat and work on making something that will take the fight back to Intel.
I guess we're catching up for lost time, as 2006 will see yet another change of metric in the desktop processor industry when Intel launches its quad-core Kentsfield processor. As if Intel wasn't already far enough ahead of the competition, it is continuing to push AMD harder and harder.
Back to top
There has been some discussion about AMD's 4x4 architecture in our forums, which combines a pair of dual-core processors on one motherboard in a multi-socket configuration. This is something that both AMD and Intel have been trying to avoid in the past and it is a bridge too far away from many consumers due to the 'massive performance at massive cost'
nature of the architecture.
AMD says that you will be able to upgrade from two dual-core processors to a pair of quad-core processors when they launch. However, current reports suggest that AMD's quad-core architecture isn't due to launch until mid-2007.
While the relationship between Kentsfield and Conroe is exactly what Presler was to Cedar Mill - i.e. Kentsfield is a pair of Conroe chips on the same package, plugging into one socket - Kentsfield looks more promising because of the way that Intel has designed its Core microarchitecture.
"Kentsfield should work in current mobos, providing they support Core 2 Duo."
Intel's Smart Memory Access and Smart Cache are the main reasons behind why I believe that Kentsfield will be a greater success than Presler. Presler saturated the front side bus and suffered in many multi-tasking scenarios because of the increased cache coherency traffic and overheads involved.
The same is true for Kentsfield, because each Conroe core is going to have to talk to the other via the front side bus. However, unlike Cedar Mill, Conroe's memory access algorithms are optimised and intelligent, making the best use of the available memory bandwidth.
I believe that these architectural features will allow Kentsfield to succeed in the short term against both the 4x4 platform and AMD's own quad-core architecture when it arrives later in the year. By that time though, I would hope that Intel's architects were well on their way to completing a native quad-core version of Conroe with all four cores sharing the same cache - it may be needed to compete with AMD's native quad-core chip at that time.
One great thing about Kentsfield is that it should work in current motherboards, providing they support Core 2 Duo processors. I've seen early reports with an engineering-sample Kentsfield chip running in Intel's D975XBX (rev. 304) motherboard with the latest official BIOS.
Obviously, support for quad-core is going to be flakey for a while, but BIOS engineers have plenty of time to get things right. This should mean that you're not going to have to buy yet another motherboard to support the upcoming processors.
Core 2 Duo will go on sale next week and if you are looking to build a system now, you should look no further than Intel's new processors as a starting point. However, with Kentsfield looking like a drop-in replacement, it might be worth making a conservative upgrade now in readiness for quad-core.
Bring it on!
Update 21/07/2006 18:00GMT
I think that some people have misunderstood exactly what I've said here, so I have attempted to clarify my thoughts in this forum post