This release is more oriented toward the mainstream and HTPC market though, so this is a little bit different than the high-end video card reviews we often publish. In fact, none of these motherboards support CrossfireX setups, even though the ASUS M4A785TD-V EVO and the Gigabyte MA785G-UD3H have two PCI-E x16 slots. The others only have one PCI-E x16 slot since they are mATX-sized. These are the MSI 785GM-E65, the ECS A785GM-M and the Gigabyte MA785GMT-UD2H. All of them support Hybrid Crossfire, which makes the integrated HD 4200 working with the low-end dedicated card for a substancial boost.
Furthermore, the dedicated graphics card can be disabled to save energy when performance is not needed. The HDMI output also makes them very suitable for an HTPC build; they can be hooked to an HDTV when running on the HD 4200.
They all feature an AM3 socket ready for the latest Athlon II & Phenom II processors and the faster DDR3 memory, except the ATX board from Gigabyte, which still has an AM2+ socket, meaning DDR2 memory must be used.
So I will put these five motherboards in direct comparison to determine which one is the best option for a given use.

Test Setup
- Phenom II x4 965 Black Edition processor
- ASUS M4A785TD-V EVO motherboard
- MSI 785GM-E65 motherboard
- Gigabyte MA785G-UD3H motherboard
- Gigabyte MA785GMT-UD2H motherboard
- ECS A785GM-M motherboard
- NVIDIA GTX 260 videocard with Forceware 190
- ATI HD 4200 IGP with Catalyst 9.8
- Cooler Master V8 heatsink
- Seagate 7200.11 750GB with 32MB cache SATA2 drive
- Mushkin XP-800AP power Supply
- 2x2GB Mushkin DDR2-1066 5-5-5-15-1T (Only for GA-MA785G-UD3H)
- 2x2GB Mushkin DDR3-1600 9-9-9-24-1T
Benchmarks



At stock, the Gigabyte boards won the arithmetic tests since they overclock the processor a few MHz at stock. It's the DDR3 version that also won the memory bandwidth test, both at stock and overclocked. The UD3H with DDR2 lost both memory benchmarks, obviously. As for the multi-core efficiency, ECS' solution won the benchmark. Since the M4A785TD-V EVO yielded the highest frequency, it won the overclocked CPU arithmetic. The 785GM-E65 was the only board able to get the memory latency under 70ns, thanks to the good integrated memory controller overclock.

The MSI 785GM-E65 performed very well when it came down to data storage; it finished second and first in the stock and overclocked testing phases, respectively. Also, somehow I could not get normal burst speeds on the A785GM-M.

The scores were pretty much even in Crysis Warhead, at stock. The memory sub-system seemed to have a great effect in the overclocked testing phase, in which the DDR2 board finished last and the MSI board got first, most probably due to its high overclock on the IMC.

In Far Cry 2, one more time it was ahead at stock, on par with the MA785GMT-UD2H, which won the overclocked phase, followed by the MSI 785GM-E65.
It can be interesting to see which motherboard performs best when running on the HD 4200. I expect the ASUS M4A785TD-V EVO and the MSI 785GM-E65 to be much stronger than the three others since they have a 128MB DDR3 sideport memory. That's what we will see. For this test, I am using the Catalyst drivers release 9.8.
Street Fighter IV is a game made by Capcom. It has a free benchmarking tool that I will be using to test the HD 4200 capabilities. World in Conflict is developed by Massive Entertainment and it also has a benchmarking tool, accessible in the game settings. Both games will be tested at 1280x1024. For World in Conflict, the graphics details will be set as "Very Low", and for Street Fighter IV they will be at the lowest settings.


As I had expected the boards featuring a DDR3 sideport memory just smashed the competition. It's the MSI 785GM-E65 which pulled ahead in both tests, with the ASUS board not far behind. In third position was the 785GMT-UD3H. Last was the board featuring DDR3.
Overclocking
Both Gigabyte motherboards performed the same when it came to overclocking, and in fact they did as well as the MA790FXT-UD5P. They all maxed the processor at 3914MHz, with the only difference being the voltage needed. For the MA785GMT-UD2H, I had to set the voltage to 1.50V. For the others, I set it to 1.475V in the BIOS, and then used AMD Overdrive to increase it another 0.0125V since Gigabyte BIOSes only have 0.025V increments. The UD2H has only a 4+1 phase power design so that might be why it required the extra voltage.
- Baseclock of 206MHz
- CPU core voltage of 1.4875V / 1.50V
- CPU multiplier of x19, for a clock of 3914MHz
- CPU-NB voltage increase of +0.1V
- CPU-NB multiplier of x12, for a clock of 2472MHz
- HT multiplier of x12, for a clock of 2472MHz
Next was the MSI 785GM-E65. Sincerely, I did not know what to expect from this board also equipped of a 4+1 phase power design. I would qualify its overclock as being medium to good. It was quite a few MHz short of the Gigabyte and ASUS boards, but still it achieved a very reasonable frequency. I did not really like though the unpreciseness of the reference clock. It was pretty much always half a MHz higher than what is set in the BIOS. However, it achieved a good integrated memory controller frequency.
- Baseclock of 203MHz, resulting in 203.5MHz
- CPU core voltage of 1.49V set in BIOS
- CPU multiplier of x19, for a clock of 3867MHz
- CPU-NB voltage increase of +0.2V
- CPU-NB multiplier of x13, for a clock of 2646MHz
- HT multiplier of x12, for a clock of 2442MHz
With the ASUS M4A785TD-V EVO, I reached the highest stable frequency I ever did on the Phenom II X4 965. If we recall the article about it, I was able to go up to 3914MHz on the core. Anything higher than that would cause a blue screen during the OCCT stress testing test. Well, on the ASUS board, I was able to go 5.5MHz higher, for a final speed of 3919.5MHz. However, when I overclocked and overvolted the integrated memory controller afterwards, the processor run hotter. You probably know that Phenom IIs are affected very much by temperatures. Thus, its cores were not stable anymore. For this reason, I limited my IMC overclock so that I could keep my extra 5.5MHz. I must however share something else with you. At first, when I overclocked the CPU-NB, there seemed to be no change applied. So I let ASUS be aware of my issue, and very quickly, I was provided with an updated BIOS that fixed my issue. A few days later, that version 0405 made its way on ASUS' website, so anyone having this issue can fix it, although I was the first to report it. Moreover, by the time Neoseeker published this article, I see another version, the 0410, has been published and is supposed to increase performance when overclocking. That means one could possibly reach even better frequencies than I did.
Here is a summary of the settings changed:
- Baseclock of 201MHz
- Loadline Calibration set at 0%
- CPU core voltage of 1.4825V set in BIOS
- CPU multiplier of x19.5, for a clock of 3919.5MHz
- CPU-NB voltage increase of +0.0375V
- CPU-NB multiplier of x12, for a clock of 2412MHz
- HT multiplier of x12, for a clock of 2412MHz
Furthermore, ASUS provides a utility called GPU NOS which allows altering the frequency and voltage of the IGP, for a substancial performance boost. If we recall the article on the 785G IGP video performance, the M4A785TD-V EVO reached a whole 1Ghz easily.
Lastly, for the ECS A785GM-M, I had a few issues. First of all, at the beginning, I could not even get the RAM to run at 1600MHz. When that frequency was set, the board would just post and boot into Windows, running it at a lowly 800MHz. I was then provided a new BIOS which fixed this issue, so I would be able to run the benchmarks at stock, since 1600MHz DDR3 is part of our testing platform. The main overclocking issue was not resolved though. As soon as I would enable manual control over the frequencies, the board would not post, no matter the settings I put. The numerical LED display on the board would just hang on error code "d0". I was not provided with a solution yet, except overclocking via AMD Overdrive, which anyway would not even let me increase the HT reference clock, voltages and memory controller multiplier. That is why there are no overclocked scores of the A785GM-M, but as soon as I am provided with a solution, I will update this article.
Update 09/10/04: As requested by a reader, I tested which board would be better for pure HT link reference clock overclocking. Whereas the ASUS M4A785TD-V EVO could not go over 240MHz.
Power Consumption
The lowest power usage at idle was won by the MSI board, although it did in the average at load. It's the ASUS M4A785TD-V EVO which was the best at load, and by far. Oddly enough it was also the loser at idle. The MA785GMT-UD2H I think has a good equilibrium, but it really depends what one intends to do with it. For example, in a folding rig, I would wholeheartedly go with the ASUS one since the computer would be loaded practically 100% of the time. Moreover, it has two PCI-E x16 slots so it can run the GPU clients on two graphics cards, and a maximum of five GPUs, considering the IGP.

Conclusion
All of the tested motherboards have their pros and cons. First of all, if one plans to run on the integrated graphics processor, I highly suggest the MSI 785GM-E65 or the ASUS M4A785TD-V EVO since they have a 128MB sideport memory which substantially boosts their performance. The ASUS is ATX-sized and more expensive, whereas the MSI, which cheaper and did provide the best IGP performance, will fit in small mATX cases used for HTPC builds.
Next, if one is upgrading his already existing computer, the Gigabyte MA785G-UD3H might be the right choice. Since it has DDR2 slots, it will allow the user to keep his DDR2 memory sticks, while having the latest IGP on the market, and using either the new Phenom IIs/Athlon IIs or the older ones. The other motherboards all require DDR3 memory sticks and the latest processors from AMD. However there are many other 785G boards on the market using DDR2 so it really comes down to what features and price one is searching for.
However this DDR2 also has its cons. As seen in the benchmarks, the MA785G-UD3H had a hard time beating the other motherboards. In fact, it did not win any benchmark. The top dogs were the ASUS M4A785TD-V EVO and the Gigabyte MA785GMT-UD2H, followed not far behind by the MSI 785GM-E65.
The ECS A785GM-M is great for having an extra SATA controller. The JMicron chip allows it to have two eSATA connectors at the back while still having six internal SATA ports. This can be useful if one needs many SATA ports as well as eSATAs, for example in a home server. The SB710 supports RAID 0, 1 and 10. And who knows, maybe it will be able to overclock higher than the M4A785TD-V EVO once I am provided with a solution to the issue I have!
Overall, it is great to see all the different flavors available on the market built around the 785G chipset. It really comes down to what the user wants. Is power consumption important for you? What about performance? Size? Cost? Everyone can find what he needs. There are many great boards in the bunch, but in the ones I have tested here, the MSI 785GM-E65 and ASUS M4A785TD-V EVO is my very favorite. There is also its mATX counterpart available for those in need of a smaller board.