Benchmarks used
There is be a set of system benchmarks and the other half is gaming oriented. Bioshock requires a custom run since it does not have a benchmarking tool. For this one, Fraps will be used to record the framerates, while for all the other gaming tests, the built-in benchmarking tool will be used. For all of these, the tests will be ran at three different resolutions : 1280x1024, 1680x1050 and 1920x1200.
- SiSoftware Sandra 2009
- HandBrake
- POV-Ray
- WinRAR
- Cinebench
- HD Tune
- Futuremark PCMark Vantage
- Crysis Warhead
- Bioshock
- Far Cry 2
- Left 4 Dead
First of all, I looked at what I could get from the Easy OC Switch. After having loaded the Optimized default settings in the BIOS, in other words the current settings when you first turn on the board, I set the switch to the second position, boosting the HT link reference clock by 10%. This allowed me to run the Phenom II X4 at 3741MHz with the IMC and HT link at 2200MHz - simply by turning on a switch. That's great, however it would not go anywhere with the 15% and 20% overclock without a bump in voltage.
Next, I tested up to what HT link reference clock the NF980-G65 would go. I was really eager to know, as the ASUS M4N82 Deluxe did not do very well on that. I was hoping to reach 300MHz as many motherboards with ATI chipsets are capable. Unfortunately the testing ended at a stable 262MHz. Anything over would cause a BSOD at Windows startup, or would simply not POST. Obviously, to reach that I had decreased the various multipliers. Also, increasing the HT link or northbridge voltage did not allow me to go further. Also, notice that in fact the HT link reference clock is at 263.6MHz, while I had it set at 262MHz.
Finally was the time to get a real overclock. Knowing that that new revision Phenom II X4 965 did 4048MHz on the Gigabyte MA790FXT-UD5P, I immediately went for the 4GHz mark at the beginning. At 1.506V, that frequency was stable. I then slowly worked my way up using different reference clock and multipliers combinations to get the clocks I wanted, since the steps in frequency caused by the even-only reference clock are too large. I finally settled for a 4022MHz overclock:
- HT link reference clock of 206MHz
- CPU multiplier of x19.5, resulting in 4022MHz
- CPU voltage of 1.496V
- CPU-NB multiplier of x13, resulting in 4681MHz
- CPU-NB voltage of 1.171V
- HT link multiplier of x12, resulting in 2475MHz
Overall, the NF980-G65 is a good overclocker. 262MHz on the reference clock is decent, however not as good as I had hoped. Also, its inaccuracy on the reference clock is not always fun to work with; a 1.6MHz offset represents 27.2MHz when used in conjunction with the stock multiplier. It did reach a decent overclock though; only 26MHz short of the MA790FXT-UD5P on the cores, but at these settings I could overclock the IMC a little further. Finally, the Easy OC switch is a great tool for people who are afraid of BIOS tuning.
This program includes benchmarks for most hardware. The CPU arithmetic and multi-core efficiency benchmark will be run as well as memory bandwidth and latency.
If we average the GIPS and Gflops measurement for each board, the MSI NF980-G65 gets last. The Gigabyte MA790FXT-UD5P gets first, with the ASUS M4N82 Deluxe being in the middle. As for multi-core efficiency, our subject was almost on par with the CrossfireX enabled board, while the DDR2 board was far behind. The memory bandwidth benchmark gave the advantage to the DDR3-equipped NVIDIA board, and obviously the M4N82 finished last. Surprisingly, all boards tied for a 78ns memory latency, except obviously the overclocked NF980-G65, which beat everything by a big margin.
HandBrake is an application that converts sound and video files to other formats. Compared to TMPG we used in previous reviews, this one is multi-threaded so it can use the processor to its full potential.
POV-Ray, for Persistence of Vision Raytracer, is a 3D rendering software that has impressive photorealistic capabilities.
Both of these benchmarks showed the same results. The MSI board looks to be a tad better than the ASUS one based on the same chipset, but it is behind the Gigabyte MA790FXT-UD5P. These two benchmarks also showed the great power added by overclocking.
Everyone knows WinRAR, so no need to explain what it is. I will compress our custom 10MB, 100MB and 500MB using the greatest compression setting, in the ZIP format.
Cinebench 10 is another rendering program. I will run both the single-threaded benchmark as well as the multi-threaded.
WinRAR put both NVIDIA boards at the same level, whereas the ATI one score a mere one second lower in the 1GB test. Cinebench showed the same trend as HandBrake and POV-Ray on the previous page. Overclocking once again greatly reduced the completion time of both benchmarks.
HDTune is a benchmarking program for hard drives. Their speed also depends on the chipset so this is why I run such a benchmark.
In case PCMark is unknown to you, it is pretty much the same as the 3DMark suite except the fact that it includes many other tests like hard drive speed, memory and processor power, so we consider it as a system benchmark and not just a gaming benchmark.
HDTune showed the three boards to be very close in terms of average speed, with the MSI board first. When it came down to burst time, the ASUS flew miles away from the competition. The NF980-G65 is still ahead of the ATI board though. PCMark put the NF980-G65 in first place, followed by the MA790FXT-UD5P. Overclocking allowed it to get closer to the M4N82 Deluxe in terms of burst speed, and also increased its PCMark score a tad.
Crysis Warhead is a standalone expansion pack of the original Crysis. It uses an enhanced version of the same engine.
Bioshock is a creepy first person shooter. It is the oldest of the games in our benchmarking suite, hence the high FPS.
The NVIDIA-based boards scored the same in Crysis, however they were 1 FPS lower than the Gigabyte board at 1280x1024 and 1920x1200. Bioshock put the MSI in second position, behind the same top performer. The overclocking had weird effects on that game; our subject got 4 more FPS at the highest resolution, but lost 4 at the lowest. In Crysis, it allowed it to gain an extra 2 FPS.
Far Cry 2 is another first person shooter. It has been developed by Ubisoft compared to the first one who has been made by Crytek. The story takes place in Africa, where the ultimate goal is to assassinate an arms dealer.
Left 4 Dead is a first-person shooter developed by Valve. It uses the Source Engine. Four survivors must fight against infected people in order to reach a safe area.
The MSI board pulled ahead of the ASUS one in the two smallest resolutions tested in Far Cry 2, however it was owned by a solid 2 FPS at each resolution by the MA790FXT-UD5P. In Left 4 Dead, it's a tie between the NF980-G65 and the ATI-based board, which has one more FPS at 1280x1024 but one less at 1920x1200. The ASUS pulled ahead of them. Far Cry 2 got 3 more FPS in total when overclocked, and Left 4 Dead got a whopping 4 FPS at the lowest resolution.
Since this board has an integrated GeForce 8300 - and video outputs -, I will compare it to the last competitor's IGP I tested. The shared video memory size is set on default, which is Auto. Here are some quick benchmarks against the HD 4200 with a 128MB sideport memory, and without.
As can be seen, the GeForce 8300 is clearly inferior to AMD's IGP. The difference is not that big against the HD 4200 alone, but if this one is paired with a sideport memory, what the GeForce 8300 cannot do, it just gets owned.
Power Usage
To measure power usage, I used a Kill A Watt P4400 power meter. Note that the above numbers represent the power drain for the entire benchmarking system. For the 'idle' readings we measured the power drain from the desktop, with no applications running; for the 'load' situation, I took the average reading during the first few minutes of the OCCT power supply test.
The MSI NF980-G65 is not the greenest board, but it is not too bad either. It is a bit better than the M4N82 Deluxe, but still far from the MA790FXT-UD5P. Also, the Active Phase Switching decreased the power consumption by 4W at idle. At ful load, all power phases were active so there was no gain.
Conclusion
As expected, the MSI NF980-G65 performed better on average than the ASUS M4N82 Deluxe. It won in two games, lost in Left 4 Dead only, and tied for Crysis. Furthermore, it won in the big majority of the system benchmarks. However, it could not match the top of the line MA790FXT-UD5P based on the high-end 790FX chipset from AMD. It did beat it in a few instances, for example in PCMark Vantage or HDTune. In gaming, it lost in three out of four benchmarks against it. The NForce 200 might add some extra latency in the PCI-E communication, resulting in slightly lower FPS, however I can't really confirm that. What this chip does though is allow the user to run a 3-Way SLI configuration - without any hacking, as I mentioned in the introduction.
One thing I disliked about the NF980-G65 is that it has only four USB ports at the back. Yes it has four internal headers, but that means a USB bracket must be used to get more USB ports at the back. Many motherboards have six or even eight of them already, so four is not much; they can quickly become occupied by different peripherals. The optical and coaxial outputs are not present either, and an eSATA would have been great. The IGP output is a big plus though. Running on the IGP can always prove useful, especially for troubleshooting. It can also work well as an HTPC; the GeForce 8300 is able of 1080P decoding and the Blu-Ray audio can be output via the HDMI port. However, at $160, it is a bit pricey for that particular use; a 785G motherboard is much more affordable, and as we have seen, the HD4200 is more powerful, especially when paired with a sideport memory.
As for overclocking, the NF980-G65 did not deceive me. Although it is not a top notch overclocker, it yielded a good 4022MHz. Its HT link reference clock overclocking performance is on the average, but it suffers from a bit of inaccuracy on it. The Easy OC Switch is a great tool for a quick overclock without going in the BIOS, however it has its limitations; the voltages are not adjusted automatically and in some cases, some multipliers need to be decreased.
Overall, MSI did much better using the NForce 980a chipset than ASUS with the M4N82 Deluxe. If one shops for an SLI-licensed AMD board, I have no choice but to recommend the NF980-G65. I have also noticed recently that DDR2 prices actually went up, since DDR3 has now taken a good chunk of the market. So why not switch to DDR3? The NF980-G65, along with its little brother based on the 750a chipset, are the only boards sporting DDR3 slots along an NVIDIA chipset. If one feels confident enough about enabling SLI on an AMD motherboard, that could also be a solution. That's something I will soon have to test myself though.
Update: I have tested an SLI configuration on the NF980-G65. The article can be found here.
Views: 7605 
