It depends what you are looking for, but you should probably go with the socket 754, will be much cheaper, and you can keep your ram. The A64 3200+ w/ socket 754 has a higher clockspeed but uses normal ddr ram, so you can keep your ram. The socket 939 uses dual ddr so your ram would be no good, and has a lower clock speed. The idea is the faster ram makes up for the loss in clock speed. Also, I don't know from expierence, but most people say go with the nforce 3 over the via chipset.
The socket 939 uses dual ddr so your ram would be no good
You are mistaken, A64s on Socket 939 will use the same DDR modules as Socket 754 and 462 CPUs. The biggest difference between 754 and 939 A64 is that 939s have a Dual Channel On Die Memory Controller that have twice the theorical memory bandwidth of the Single Channel On Die Memory Controller found on Socket 754 CPUs.
Considering the matched pair of Low Latency memory Boz already have, going for a Socket 754 platform would be a waste of ressources, beside 90nm Winchester cores are'nt available for Socket 754 and they out-perform the older 130nm Newcastle core when it comes to overclocking.
@ Boz :
I suggest that you go for a 3000+ or 3200+ Socket 939 CPU based on the 90nm Winchester core, those are great overclockers and will reach frequencies beyond 2.5GHz on stock cooling and voltage, better cooling and a slight voltage bump will yeild even better results than what can be expected from 130nm Newcastle cores.
The motherboard is up to you, read reviews and go for the one that have all the options you need at a price that you consider acceptable, visit the different Manufacturer's technical support website to get a feel of how serious and in-depth they are, broken links, slow downloads, complicated and innacurate search tools are a bad sign.
Watercooled Mobile Barton 2500+ @ 2.6GHz (200MHz x 13)
Abit NF7-S V2.0
2x 512MB of Samsung TCC4
Sapphire Radeon 9700 128MB @ 360/310
2x Maxtor 40GB 7200RPM RAID-1
How about cranking ur XP2500 into an XP3200. U have a board that can do it. That way you don't have to buy anything for awhile.. unless maybe you're planning another use for what you have now or planning to sell it?
Change the CPU voltage to about 1.825V and put the memory at 2.7V and let Prime95 rip and see what it does. change fsb to 200mhz and there u go
Wow, I didn't realize that. Something made me think that the ddr2 used a different memory than normal ddr. Thanks for the info, that makes me rethink my decision to go with socket 754 on my build. So normal ddr400 ram will work as dual channel ram as well?
Nforce4 (with Pci-exp) is starting to trickle out. We are stuck with it, though AGP will be available for a while yet. I'm waiting for a board from Abit, but dont look forward to having to drop a lot of money on a new graphics card. Oh well, the real problem is I want to replace my 9600pro now, but wont be able to get 2 cards in as many months.
What is the difference between AMD A64 and AMD A64"Winchester" ???
Newcastle (130nm) and Winchester (90nm) are codenames for CPUs manufactured using different processes, same basic architecture but the transistors and traces were "shrinked" so AMD could fit more cores onto the same waffer to increase their manufacturing capacity, decrease the heat output and ramp up the clockspeed.
Take the K7 architecture for example, the original Athlon was manufactured using a 250nm process with 512<font color=red>K</font color=red>B of off-die L2 cache (Slot A) and maxed out at 700MHz, AMD switched to a 180nm process for the K75 which maxed out at 1GHz.
Then came the Thunderbird for Socket A with 256KB of on-die cache, also manufactured on 180nm technology, it reached up to 1400MHz but ran extremely hot for its time.
The first Athlon XP was based on the Palomino core and was the last 180nm CPU, it featured most of Intel's SSE instructions and the 2100+ ran at 1733MHz. It was replaced by the short lived and rather lame Thoroughbred-A. Something went wrong with the shrink from 180nm to 130nm and that first Thoroughbred just would'nt run any faster than 2GHz, that core was a big deception among the AMD overclocking community.
AMD took a few months to fix their 130nm process and relased the Thoroughbred-B, the 2800+ ramped up all the way to 2250MHz. Unlike the A's, B's were able to reach much higher frequencies than what they were designed for, a 1.4GHz 1600+ could be overclocked to 2.3GHz with high end cooling and 2.0V !
AMD tweaked the Thoroughbred core again and introduced the Barton core with 512KB of L2 cache, it is not uncommon to see 1.8GHz 2500+ running at 2.5GHz with good cooling and a moderate voltage bump.
It seems like the K8 architecture will follow a similar path, the 90nm Winchester is AMD's most recent core and its overclocking potential is well known, 2.6GHz on stock cooling and voltage is nothing to sneeze at.