Is LGA1366 the most future-proof choice?
Posted on 16th Jul 2009 at 10:29 by Clive Webster with 46 comments
Let me just expand on a few of those points before inviting your feedback. As far as upgrade potential goes, LGA755 isn't going to take you that far, unless you originally boarded the train at Pentiumville or Core 2 Street. The jump from one of these cheap CPUs to a Q9650 is a huge one, but the jump from even a Q6600 to the Q9650 isn't so significant.Only if you're strapped for cash at the moment is LGA775 a sound long-term investment, and only then as long as Intel keeps making Q9650s for the next few years.
Then there's Socket AM3 - again, it's true that if you get on the AM3 bus at the low-end you can ride it all the way to a fast Phenom II X4 in a year or two's time (This metaphor has now been stretched too far, so I'll give up on it). However, cheap dual-core AMD processors aren't as overclockable as Intel's, and AMD's top-end CPUs cease at a level of performance far in the shadow of Intel's Q9650.
Socket AM3 CPUs are unlikely to push the Q9650 for performance either, as AMD is already up to 3.2GHz with its Phenom II X4 955 BE. There are rumours of a 3.4GHz CPU coming soon, but that will likely be as fast as AM3 Phenom II goes.
This is because a Phenom II X4 is essentially four slightly modified Athlon 64 cores bolted together, and AMD has always struggled to get this architecture much above 3.2GHz without having silly consequences such as a 140W TDP (much to the annoyance of motherboard manufacturers). AMD should be loathe to do the same thing again, and will likely try to ensure all future Phenom II CPUs require the same power and produce the same heat as the current range.
It's likely that AMD will release a sextuple-core Phenom for AM3 (as with Istanbul core Opertons) but if this is to fit within AM3's power and heat budget, the frequency probably won't be that high. This is the Achilles heel of the six-core Opteron range. It'll be interesting to see whether the trade-off in clock speed to accommodate two more execution cores will actually constitute an upgrade at all over a Phenom II X4 955 BE, especially as games and consumer apps aren't particularly multi-threaded.
Then there's Bulldozer (the first Fusion CPU, that will have an integrated GPU), but it's highly likely that the extra power requirements and the need for extra communication channels to let the GPU output a screen will mean that this CPU will need a new socket..
So, at the moment, the only option if you're really concerned about an long upgrade path is LGA1366. Intel has already confirmed that it will release new (6-core) CPUs for this socket and that LGA1156 won't kill it. So there you are - anyone who hopped onto LGA1366 early can feel pretty smug as it's the only current socket that seems to have any future.
Of course, with LGA1156 and Lynnfield on the way, anyone who isn't in desperate need of a new PC should wait the couple of months until it's released - just to see what it offers if nothing else. However, it's a pretty odd state of affairs that there's only one choice of socket and CPU for anyone needing to upgrade now.
Of course, the objection that could undermine my point is that you consider the length of an upgrade path to be based on price on not on longevity - in this case LGA775 is still a great choice as you can get a cheap yet fast dual core CPU now and know that there is a Q6600 or Q9650 waiting for you when you finally save enough cash. So, how do you define an upgrade path, and how happy are you with your choice of CPU socket?