QUOTE(mruzian @ Nov 27 2006, 04:10 PM)
the big diff is the socket
s939 got 939 pins
meanwhile am2 got 940 pins
hehe the major diff is AM2 supports DDR2
but S939 DDR1
performance wise dun know ... but
i think there's not much diff
System 1 - 939
DFI LanParty NF4 SLI-DR Expert (939)
AMD Athlon64 X2 3800+ (939)
2x1GB OCZ PC4000EL Gold DDR1 @ 3-4-8-4
ATI X1900XT 512MB PCI-E @ 625/650 (stock speed), using Omega 6.7 drivers
Creative Audigy2 ZS PCI sound
80GB Hitachi SATA II
OCZ GamerXStream 600w psu
NEC 3550A DVD-RW
System 2 - AM2
DFI Infinity NF4 Ultra II-M2 AM2 (AM2)
AMD Athlon64 X2 3800+ (AM2)
2x1GB OCZ Gold GX PC2-8000 (DDR2 1000Mhz) @ 5-6-15-6
ATI X1900XT 512MB PCI-E @ 625/650 (stock speed), using Omega 6.7 drivers
Creative Audigy2 ZS PCI sound
250GB Seagate 7200.10 SATA II
OCZ Powerstream 600w psu
NEC 3550A DVD-RW
Everything was ran first at stock speed, then again @ 250x10 (2500Mhz cpu speed)
Here's the first comparison chart, with just the end-result scores from the benchmarks:


Well, what it looks like to me is that in gaming (synthetic tests but it is good indication of performance differences since the tests are identical, and the platforms are virtually identical), as we already know, the gpu is the factor these days, as the game devs as well as Microsoft (with their DX9.0c API) have moved the majority of the load off the cpu and into the gpu.
This is good in a way, but is in no way an indication of a full-bore game like Ghost Recon Advanced Warfighter for example, which has a large amount of particle and regular phsysics, as well as some pretty advanced AI.
The only problem with testing real games...is that there's no standard test. Time Demos that we are used to are not really good indicators of exact performance since it is a recorded event.
Live gaming is a perfect performance measurement, but each game plays differently, with the player and usually the AI and physics engines also doing something different each time, as well as my 30 second run through Half Life 2 for a benchmark might be totally different than your 30 second benchmark run, even if we started in the exact same spot.
The point is, there's very little difference in these cpu's, whether stock or overclocked. You can see up to a 10% difference in pure cpu tests, but to me, that is still very little difference.
The other important factor right now is the price of DDR2. Good DDR2-1000Mhz modules for overclocking, like the OCZ modules I have, are up around $400 for a 2GB kit...pretty outrageous considering the DDR500 kit for the 939 is right around $170-$180.
Since the 3800+ cpu is the same price, and the motherboards, video cards, power supplies, etc are all the same price regardless of platform, that means a pretty large $200 gap between the "old" DDR1 and the "new" DDR2.
Right now I would easily recommend a 939 platform over the new AM2 for a couple of reasons:
1. price of RAM, as I just mentioned
2. not a big performance gap...yet
3. 939 platforms are old...which means they are mature and most of the bugs have been ironed out. AM2...very new, we still don't know how many bugs we will find!
4. Stock Speed Database + Overclocking Database...there's 1 entry in the AM2 OCDB, and 2 in the SSDB...and a hundred in the 939 OCDB and 2 pages' worth in the SSDB.I'm not bagging on the AM2 at all. I've found myself very impressed with the DFI Infinity NF Ultra II-M2 board and my DDR2 1000Mhz OCZ modules as well as my 3800+ X2 cpu. The Infinity AM2 is also quite easy to overclock so far....if you look in the Overclocking Database you will see my entry, and how I've not really touched much. On top of that, right now I am shooting for 260x10, and after that I will check of course 270x10 and 280x10 etc until it fails, then I will start dropping dividers on it and see just how far she will go.
That should tide me and you over until the new AM2 Nvidia SLI & ATI Crossfire boards arrive
So there you have it. I've tested pretty thoroughly I think, though of course some of you will cry that I didn't do enough, didn't do it right, didn't do something that you would have.
Too bad, I did what I always do, and for 99% of you, this information is exactly what you are looking for, not some nit-picky anal-retentive nonsense from some guy who just absolutely has to have the very last half-mhz of performance squeezed out of his rig.
Source