Unlike socket 754, socket 939 is regarded as a” prestigious” platform when you have access to all the high end Athlon 64 models, where you can build your dream SLI rig and can enjoy even highly demanding DirectX 9 games in all their glory. The first socket where you are no longer encumbered by the limitations of the AGP bus – although you can still find AGP on the low end, size of the video card as it was the case with the only PCI-express board for socket 754 that I used, low amount of memory or single core processors.

It was released back in 2004 and it really felt like a step forward, despite using the same single core CPU’s as its brother, socket 754, at launch. Overtime AMD released far stronger single core processors, culminating with the Athlon FX-57, the highest clocked socket 939 Athlon. But what really stood out was the release of dual cores Athlon 64 X2, a monumental achievement back in the day; it was really the next step in processors evolution and the first sign of things to come.

Socket 939 is the bridge between the past and the future: it still has chipsets with AGP support but also some with PCI-express; it has SATA first generation on some platforms but also second generation SATA on others; it still supports IDE interface natively for those who want to build a Windows 98 system on one side; on the other side you can build a very high performance system that can shred trough era correct games. For the first time the platform’s components are powerful enough to play games released after the socket was pushed into the sunset by AMD with the release of socket AM2; previously socket 754 or socket A cannot play 2005 or later games at maximum details because of either CPU, RAM or video card limitations.

Because of the variations in the chipsets and processors available I have decided to split the testing in 2 CHAPTERS, each with 3 parts:
1. AGP with single core processors. Because the most powerful AGP card in existence, the radeon 3850 which I do not have, is so weak it can’t be used for testing beyond year 2004 I see no reason to use a top of the line CPU’s such as Athlon 64 X2 4800 for testing AGP platforms. You can not play 2005 games with an AGP card and full details anyway so it would be pointless to use a dual core Athlon 64 X2. You do not get a performance boost in 2004 and earlier games with a dual core CPU so a strong single core is all you need!
However, I have a very good reason to use a single core CPU’s: clock frequency and cache size IDENTICAL to socket 754 counterparts which is ideal for a direct comparison between the single channel socket 754 and dual channel socket 939.

I truly admire your thoruoughness in the endeavours of testing these platforms! It's really impressive!

939 is my favourite platform. I have an AGP board and a PCI-E board that I swap out depending on what I feel like running. I also picked up an insanely good 3700+ San Diego CPU that easily does 3GHz on only a small overvolt.

All in all, a really nice daily driver retro PC.

I love So939. I have an Opteron 180 (185 ready to swap) with 4 Gigs of RAM and a PCIe Radeon GHz Edition. It is my Media feeder rig running dozens of drives on IDE, SATA and SCSI. Main OSses are FreeDOS & Win7 on SSD. I also wanted ReactOS but it didn't work and Linux Mint either. May give it another try soon when I moved the system to another Tower (more drive Space 😁).

H3nrik V! wrote on Yesterday, 12:09:

I truly admire your thoruoughness in the endeavours of testing these platforms! It's really impressive!

Thank you sir!
What really stands out on socket 939 is the incredible amount of choice you got: it has no less than 7 revisions of single core Athlon 64; no other AMD socket from the classic era - socket A - AM2+ - has that!

Lostdotfish wrote on Yesterday, 12:59:

939 is my favourite platform. I have an AGP board and a PCI-E board that I swap out depending on what I feel like running. I also picked up an insanely good 3700+ San Diego CPU that easily does 3GHz on only a small overvolt.

All in all, a really nice daily driver retro PC.

San Diego is the last and best revision, also that one has 1mb of L2 cache , it should be one of the best!

AncapDude wrote on Yesterday, 13:52:

I love So939. I have an Opteron 180 (185 ready to swap) with 4 Gigs of RAM and a PCIe Radeon GHz Edition. It is my Media feeder rig running dozens of drives on IDE, SATA and SCSI. Main OSses are FreeDOS & Win7 on SSD. I also wanted ReactOS but it didn't work and Linux Mint either. May give it another try soon when I moved the system to another Tower (more drive Space 😁).

I did install Windows 7 on the best socket 939 system I got but it still feels slow; in all benchmarks and games XP is faster although I confess I have only 4gb of RAM and used Windows 7 X64.

In part 2 I will be comparing socket 754 and 939 with single core Athlon 64 at the same frequency using PCI-express video cards.
In the last part of chapter 1, part 3, I will be comparing different revisions of single core CPU's with the same frequency and amount of L2 cache.

Just like in the previous battle of the platforms, you will see a head on comparison between the main chipsets for the socket. Before anyone will say: but integrated memory controller makes all chipsets equal!! I have already demonstrated that chipset choice still matters for socket 754 as for example subpar AGP performance relegated nforce3 to the last place whilst reference PCI-express implementation put nforce4 at the top of the charts. We shall see if this still stands true.

In chapter 2, part 1, I will investigate dual cores processors using pci-express GPU's.
PCI-express allows us to use extremely powerful video cards to unleash the potential of the platform together with dual cores Athlon 64. Because period correct video cards are so weak, one can build a socket 939 system with later PCI-express cards so he can play all the late Windows XP era games in all of their glory.
Part 2 will consist of a direct comparison with its successor: socket AM2 with Athlon 64 X2 using DDR2 can be compared with the identical clocked/cache size CPU on socket 939 to see if DDR2 brings any speed benefits!
Because the test suite will change a lot due to availability of powerful pci-express cards it is most likely that results from chapter 1 with single cores Athlon's and chapter 2 will not be directly comparable.

In chapter 2, part 1, I will investigate dual cores processors using pci-express GPU's.
PCI-express allows us to use extremely powerful video cards to unleash the potential of the platform together with dual cores Athlon 64. Because period correct video cards are so weak, one can build a socket 939 system with later PCI-express cards so he can play all the late Windows XP era games in all of their glory.
Part 2 will consist of a direct comparison with its successor: socket AM2 with Athlon 64 X2 using DDR2 can be compared with the identical clocked/cache size CPU on socket 939 to see if DDR2 brings any speed benefits!
Because the test suite will change a lot due to availability of powerful pci-express cards it is most likely that results from chapter 1 with single cores Athlon's and chapter 2 will not be directly comparable.

Just like in the previous battle of the platforms, you will see a head on comparison between the main chipsets for the socket. Before anyone will say: but integrated memory controller makes all chipsets equal!! I have already demonstrated that chipset choice still matters for socket 754 as for example subpar AGP performance relegated nforce3 to the last place whilst reference PCI-express implementation put nforce4 at the top of the charts. We shall see if this still stands true.
Again I am going to use only Abit motherboards, so expect to see quite a few of them 😀!

CHAPTER 1; PART 1:
In part one we will investigate the advantages that dual channel brings to the Athlon 64 in a clock for clock comparison between socket 754 Athlon 64 and socket 939 Athlon 64. In addition to dual channel memory socket 939 brings a higher hyper transport frequency! The jump from 800 MHz to 1.000 will also be investigated.
Because of the integrated memory controller, it is very easy for board manufacturers to build socket 754 and 939 motherboards with the same chipset: socket 754 nforce3 is renamed to nforce3 ultra for the socket 939; nforce4: as you have seen in the pictures in the previous battle of the platforms for socket 754 where Abit NV8 officially has nforce4 4X chip but actually has nforce4 SLI, nforce4 is the same chipset regardless of the board implementation; VIA did not even bother to rename the K8T800 PRO chipset; ULI also kept the same M1689 name for both socket 754 and 939 boards.
Part 1 will feature AGP only so no nforce4!

While there are boards with SIS chipsets out there Abit did not manufactured a single one so you will not see it! It seems that SIS was not loved at all as there are very few socket 939 SIS boards with AGP.
Abit chose to not manufacture a socket 939 board with nforce3 which, in my opinion, was a good thing because, as you have seen on socket 754, nforce3 is the lowest performance chipset due to poor AGP implementation. So nforce3 is out of the picture.
VIA is present on both socket 754 and 939 with the K8T800 PRO chipset: KV8 PRO is the socket 754 board while AV8 is the socket 939 one.
ULI – the outsider, also with very few board partners – is also present with the M1689: an AGP chipset. The socket 754 board is KU8 and the socket 939 one is the UL8 - a very rare and little known motherboard!