The P965 chipset has certainly evoked a lot of different emotions around our test labs as well as around the various hardware sites and forums. This chipset was hyped as the next 440BX in some circles and promoted ad nauseam before, during, and after Computex last year. At first look this chipset seemed to be an instant replacement for the venerable 975X as it costs less, sports an improved memory controller (at least on paper), and supports all Intel processors from NetBurst (Celeron through Pentium D) to Core 2 Duo, as well as the latest Core 2 Quad offerings.

Leading up to Computex 2006 we found that the majority of the motherboard suppliers were introducing at least one product based on this new chipset, while companies like Gigabyte and ASUS would launch several different product lines dedicated to it. The initial product introductions had this chipset being used in every imaginable configuration from the $80 ECS P965T-A to the $250 (at launch) ASUS P5B-Deluxe WiFi-AP. While the high-end boards like the Gigabyte GA-P965-DQ6 and ASUS P5B-Deluxe have dropped in price, the base P965 boards such as the ECS P965T-A and MSI P965 Neo-F continue to hover around the $80 mark.

At times we have even seen these budget P965 boards practically given away for free with the purchase of a processor, but other chipsets continue to thrive in the low end market for people wanting to use a Core 2 Duo processor. In fact, one of our favorite motherboards after the Conroe launch was the ASRock 775Dual-VSTA featuring the VIA PT880 Pro chipset that afforded users the capability to utilize their existing AGP or DDR memory with a Core 2 Duo processor. We still recommend this board and its successor the 4CoreDual-VSTA that features quad core capability and improved performance in the budget Intel market.

Why hasn't the P965 obliterated the other chipsets in the market sector? Obviously one important factor is pricing as it appears there will not be any $60 P965 boards soon (or perhaps ever). That leaves the budget market to Intel's previous generation chipsets and competitors like VIA or SIS. Unfortunately, with the AMD buyout of ATI we will not see competitive chipsets from ATI in the Core 2 Duo market except for the performance oriented RD600 that is currently being offered from DFI. We also see very good 975X boards such as the DFI Infinity and Intel D975XBX2 in the $160 to $199 range now that offer excellent performance for the price.

The other important factor is performance, as Intel's 975X still offers better clock for clock performance when compared to the P965 in most situations. We expect this to change to some degree as the P965 continues to mature, and additional BIOS tweaks such as 1T command rates should be fully working in the near future. The 975X's front side bus will not overclock as high as the P965, but due to relaxed MCH timings and memory strap changes it is difficult to take advantage of the higher P965 front side bus speeds without premium memory and cooling solutions.

The 975X also offers 8x8 CrossFire capability while the P965 has been limited to a 16x4 hybrid solution. We have not discovered much difference in performance at the lower resolutions, but as game engines and future video cards require greater bandwidth then the P965's CrossFire performance will be hampered. The good news is that the upcoming release of the Bearlake chipset family will feature a performance oriented version with true dual x16 capability. We might even see a few upcoming P965 motherboards offer a revised PCIe controller chip that allows 8x8x4 or 16x0x4 operation with CrossFire performance exceeding that of 975X in 8x8 mode.

Probably the most important factor to date has been compatibility. The initial P965 BIOS releases suffered from memory timing and voltage issues, and there were also general performance drawbacks in the areas of overclocking and memory throughput. The majority of these issues have been solved as the BIOS and motherboard designs have matured. The memory suppliers have also worked closely with the motherboard suppliers to ensure the issues we witnessed at launch with memory modules not booting at 1.8V or having SPD settings that greatly affected timings are basically a thing of the past.

However, the one item that still concerns us and continues to be an almost daily issue in both our labs and those systems owned by the readers is drive compatibility. Intel did just about everything right last year, but the one boneheaded decision they made was to introduce the P965/G965 chipsets without native IDE support. While we agree the IDE interface is getting long in the tooth and SATA is the solution for the foreseeable future, it would have been nice to have coordinated this decision with the optical drive manufacturers. Hindsight is always 20-20, but the best solution would have been to defer the change until the next chipset release, as it appears the optical drive suppliers will have primarily switched to the SATA interface by Q3 of this year, just in time for Bearlake.

In the meantime we have been left with a situation where the motherboard manufacturers had to increase the cost of their boards in order to include an additional IDE capable chipset, and with this they opened a can of worms. The additional complexity involved in getting a chipset like the JMicron JMB363 to play nice with Intel ICH8/R chipsets has proved to be difficult. While Biostar wisely chose the VIA VT6410 chipset for IDE availability at the expense of additional SATA ports, the balance of suppliers have gone the JMicron route. We feel like this is to gain an advantage on the feature checklist as you can have two additional SATA ports for either standard 3Gb/s or e-SATA support.

The auxiliary IDE and SATA support has come at an expense. There have been numerous problems getting the IDE drives to operate in modes other than PIO. There have been conflicts between the RAID operation on the JMicron and Intel ICH8R setups. The ability to use IDE optical drives and SATA drives together on the JMicron controller has caused problems numerous times. Even purchasing a SATA optical drive and using it on the Intel ICH8R controller with RAID enabled has been impossible on some boards. One of the most popular scenarios we witnessed and continue to see is to have RAID enabled on the Intel controller and IDE on the JMicron resulting in the RAID array not being seen. The other problem is enabling RAID on the JMicron setup and not being able to use an IDE drive or RAID on the Intel controller. The majority of these problems have been resolved through BIOS, firmware, or driver updates, but you just never know when it is going to be an issue.

With all that said, let's present a quick recap of our review boards and a few boards that stood out during the past four months of testing.

Test Setup and Benchmarks


View All Comments

  • penga - Friday, January 26, 2007 - link

    anandtech tested ds3 and p5b in roundups, in different revisions, when a new bios came out and so on...iam bored and besides i want me a low power - max performance mAtx rig xD Reply
  • penga - Friday, January 26, 2007 - link

    test matx boards.
    think about new ways to meassure power consumption of cpu (conroe, conroe L2, allendale) and mainboard (P965, matx stuff, 945g) seperately.
    thats what interests me.
    not all of us get wet ourselves if we dont get 3.7ghz but 2.7ghz only. some readers like to know about power consumption of different models, savings by undervolting etc.

    we all know ds3 and p5b are good boards...u proved us in how many reviews again? time to move on and be a lil more innovative.
  • Treripica - Friday, January 26, 2007 - link

    I'm interested in doing a mATX build next time around and would like to see Anandtech's take on the G965 boards and how they compare to each other and comparable boards/chipsets. Reply
  • VooDooAddict - Friday, January 26, 2007 - link

    I don't know of any mATX P965 boards.

    There are the mATX G965 boards ... but from what I understand the overclocking is very limited. The big question I have is how limited. I'd like to see them review the mATX Gigabyte and ASUS G965 boards and see how far they can be pushed.

    Can the G965 FSB be pushed from 266 to 333? to 375? to 400?
    I think many mATX people would be happy just to get 333 FSB. On the e6300/e6400 that alone would be around a 500mhz jump in CPU speed. Bringing the e6400 to 2.66ghz which is a great value booster. (2.6 Should keep a 8800gts well fed.)
  • 96redformula - Friday, January 26, 2007 - link

    I just wanted to comment on the Biostar P965PT, I just recieved one this week and have something to note about it. It is extremely picky in the BIOS and has some errors in it that do not make any sense. I am not the only person that was struggling to break 330-360FSB, it would be unstable or reboot itself. I just found out today that despite not being able to hit 330-360, I could jump right up to 400 FSB and it has been Orthos error free and running perfect. So I would highly advise against this board if you want Overclocking to be simple. Reply
  • Avalon - Friday, January 26, 2007 - link

    That happens to most of the P965 boards, and is not a board fault, but a chipset fault. Most of the Asus boards will do this too, but it's not of much concern since you won't be in such a low range of FSB values while overclocking, unless you're going to use an E4300 (in which case I'd simply suggest lowering the multiplier of that chip until you break the FSB hole, then start increasing the multiplier back up, while keeping your FSB higher than your hole). Reply

Log in

Don't have an account? Sign up now