MSI 790FX-GD70 AM3 Motherboard Review

[lemonlime]

Overclocking Results

Overclocking Results

First and foremost, if you are completely new to Phenom II overclocking we highly recommend that you check out our article on the subject, aptly titled "Hardware Canucks: Benchmarkers Guide to the Phenom II". Although it is limited to the AM2+ versions, the fundamentals remain the same as with the new Phenom II AM3 processors.

For our overclocking tests today, we’ll be sticking with our Phenom II X2 550 dual-core processor. Based upon the same silicone as the popular top-end X4 955, we’re expecting that it should have quite a bit of overclocking headroom in it.

Let’s see how it does in the GD70.


Maximum CPU Core Overclock

Click on image to enlarge​

Our X2 550 sample has proven to be a very decent overclocker. We had no difficulty achieving a maximum core clock speed of about 3.96GHz. We were oh-so close to reaching that magical 4GHz mark, but we’re afraid that we couldn’t quite keep the processor 100% stable despite our best efforts. We’re confident that with better cooling – like mid-high end watercooling – 4GHz wouldn’t be an issue. Although this is the first board we’ve tried this processor in, we have no reason to believe that it is the motherboard and not the processor holding us back. None the less, a 28% overclock is certainly not too shabby.


Maximum Bus Speed Overclock

Click on image to enlarge​

The MSI 790FX-GD70 did not disappoint when it comes to Bus overclocking. Ability to hit a reference clock speed of 350MHz in our testing means that the GD70 is an excellent board for pushing standard – i.e. not Black Edition – Phenom II processors to their limits. We should note that the board has difficulty posting at about the 310MHz mark, but once in the operating system, the reference clock could be increased quite a bit more using the OC dial or a software application such as AMD Overdrive. At one point we were pushing beyond 365MHz, but couldn’t quite get the system 100% stable. Amazingly, with a reference clock this high, our memory was pushing 700MHz using the 400MHz divider!
To achieve this high frequency, we ran about 1.4V into the 790FX chipset.


Maximum Memory Overclock

Click on image to enlarge​

You may be wondering why we’re showing you a screenshot of a super tight 6-6-5 memory configuration running at 716/1432MHz. Isn’t this test supposed to be about maximum clock speeds and not about tight timings? You bet. The problem is that this is the maximum frequency. If we loosened the timings to 9-9-9, we couldn’t do any better than about 735MHz. Increasing CPU-NB voltage to as much as 1.43V didn’t help and neither did any combination of other system voltages. Reducing the NB frequency and HTT frequency also didn’t seem to help. Talk about a serious memory frequency wall. Given that the memory runs at the maximum frequency with super tight 6-6-5 timings only leads us to the conclusion that our X2 550 possesses an especially weak integrated memory controller. This type of hard wall isn’t very uncommon unfortunately. We saw a similar issue – albeit at a higher frequency of 830MHz – during our ASUS M4A78T-E review with an X3 processor. This is an unfortunate reality of AMD’s relatively new entry into the DDR3 market. We’re hoping that as the platform matures, higher memory clock speeds will become a reality.

Although the motherboard does need to provide stable power to the DRAM modules and CPU while maintaining appropriate timing/frequency parameters in the BIOS, the CPU is ultimately what makes or breaks a good memory overclock. Thankfully, tight timings and lower frequencies appear to play nice with most Phenom II processors. Keeping memory at 1333MHz and shooting for CAS6 timings is a good way to keep performance high without stressing the system with higher memory frequencies.

 

[lemonlime]

AMD Deneb Core Unlocking

AMD Deneb Core Unlocking

AMD didn’t just stop at making unlocked multipliers available to enthusiasts in their “Black Edition” processors, but also “accidentally” – please note the quotations there – allowed the disabled cores of the X3 and X2 Phenom II processors to be enabled using a special combination of BIOS settings. Having the ability to unlock the disabled cores can drastically increase the CPU power on-tap; potentially much more so than overclocking alone.

Unlocking a Phenom II X2 or X3 processor requires three things: An unlockable CPU, disabled cores that can operate with stability and last but not least, a motherboard and BIOS that supports it. We should definitely note up front that core unlocking is luck-of-the-draw. Not all Phenom II X2 and X3 CPUs can be unlocked as various steppings produced at varying times will have varying results. Just like getting a CPU that overclocks well, finding an unlockable one is luck of the draw.

One thing that is very controllable however is selecting a motherboard that allows core unlocking. Thankfully, most enthusiast grade – and quite a few value offerings – include support for this undocumented feature. Some support this right out of the box, and others may require a BIOS flash to work.

We just so happen to have an “unlockable” Phenom II X2 550 in our possession with the popular 0922APMW stepping code. Not only has this processor tested to be a very decent overclocker, but also fully stable with all four cores unlocked. We’ll be putting the MSI 790FX-GD70 through its paces to see just how it fairs with our 550 sample.

Unlocking the X2 550

• BIOS Version 1.0 Retail: No
• BIOS Version 1.3 Retail: No
• BIOS Version 1.42 Beta: No
• BIOS Version 1.43 Beta: Yes, but not stable
• BIOS Version 1.46 Beta: Yes, but not stable
• BIOS Version 1.51 Beta: Yes, stable.

Our board shipped with BIOS revision 1.0, which appeared to be a stable and fully functional BIOS revision. Unfortunately, we didn’t have any luck getting the cores unlocked. Although the “ACC” feature could be enabled and set to “Auto” or various values, none succeeded in unlocking the processor. The most recent – at the time of writing – official 1.3 BIOS revision also didn’t allow core unlocking. We then moved on to some of MSI’s unreleased beta BIOS revisions and finally came across one that worked dubbed 1.4b3 or 1.43 Beta.

Click on image to enlarge​

A new option was added to the BIOS called “ECC Firmware”. Along with setting the ACC or “Advanced Clock Calibration” to “Auto”, the “ECC Firmware” mode had to be changed from “Normal” to “Special”. We had no issues booting into windows with all four cores enabled, but unfortunately couldn’t get the system stable. Whenever any heavy load was placed on the system – such as Prime95 for example – it would hardlock and had to be restarted.

Click on image to enlarge​

Thankfully, a new beta BIOS dubbed 1.5b1 was released only a few days ago that corrected this issue. The only outstanding issue that we could find with it was that HTT frequency had to be kept a bit low to ensure that the system didn’t suffer cold-boot problems. Usually the HTT frequency and NB frequency are both set to 2000MHz. Dropping the HTT multiplier and reducing HTT to 1600MHz seemed to correct the cold-boot issue.

Given that the GD70 is targeted squarely at enthusiasts, it seems unfortunate that buyers of Gigabyte, DFI, Biostar and even ASRock boards have been able to enjoy unlocked X3 and X4 chips while the high end GD70 couldn’t until only days ago. Better late to the party than never we suppose. Word on the street is that an official 1.5 BIOS release will hit the web in the not too distant future, providing unlocking capability to the masses.

Overclocking the “new” Phenom II X4 B50

Click on image to enlarge​

Our 0922APMW processor is certainly no slouch, and it doesn’t appear that the extra disabled cores have held the overclock back much at all. We were able to hit an impressive 3.8GHz without much effort at about 1.55V.

Talk about extra value. Who wouldn’t love to get four cores at 3.8GHz instead of two at 3.1GHz. Given the X2 550’s price tag of only around $120, you certainly can’t beat that bang for the buck.

 

[lemonlime]

Memory Benchmarks

Memory Benchmarks

Lavalys Everest Ultimate v5.02

Everest Ultimate is the most useful tool for any and all benchmarkers or overclockers. With the ability to pick up most voltage, temperature, and fan sensors on almost every motherboard available, Everest provides the ability to customize the outputs in a number of forms on your desktop. In addition to this, the memory benchmarking utility provides a useful tool of measuring the changes to your memory sub-system.

As you can see, increasing CPU-NB frequency has the largest impact on memory bandwidth. We saw some significant improvement going from 7-7-7 to 6-6-6 timings as well. It’s pretty clear that with the frequency limits we were faced with, the best memory performance strategy is to keep timings as tight as possible and crank the CPU-NB as high as it will go to maximize efficiencies in the memory subsystem. It should also be noted that the L3 cache in the Phenom II processor operates at the same frequency as the CPU-NB. So not only does this improve main memory access efficiency, but also L3 efficiency.

We see a similar pattern emerge when it comes to memory latency. CPU-NB frequency is key to maximum memory performance on the Phenom II platform.


ScienceMark v2.0

Although last updated almost 3 years ago, and despite its rudimentary interface, ScienceMark v2.0 remains a favorite for accurately calculating bandwidth on even the newest systems.

ScienceMark appears to favor the “unlocked” quad configuration, and shows the highest bandwidth. Our guess is that memory “copying” plays a significant role in this benchmark as the quad configuration leads in this regard with Everest as well. None the less, high CPU-NB frequencies once again make the most significant improvements.

 

[lemonlime]

System Benchmarks

System Benchmarks

SuperPi Mod v1.5

When running the SuperPI 32MB benchmark, we are calculating Pi to 32 million digits and timing the process. Obviously more CPU power helps in this intense calculation, but the memory sub-system also plays an important role, as does the operating system.

As we’re accustomed to seeing, SuperPI prefers raw computational power above all. Not very significant gains were realized with memory subsystem performance improvements.


PCMark Vantage X64

The latest iteration of the popular system benchmark is PCMark Vantage from the Futuremark crew. The PCMark series has always been a great way to either test specific areas of a system or to get a general over view of how your system is performing. For our results, we simply run the basic benchmark suite which involves a wide range of tests on all of the sub-systems of the computer.

In our PCMark Vantage testing, the quad “unlocked” configuration did best as these tests are multi-threaded. Small but measurable gains were seen across the various memory configurations.



Cinebench R10 64-Bit

Developed by MAXON, creators of Cinema 4D, Cinebench 10 is designed using the popular Cinema software and created to compare system performance in 3D Animation and Photo applications. There are two parts to the test; the first stresses only the primary CPU or Core, the second, makes use of up to 16 CPUs/Cores. Both are done rendering a realistic photo while utilizing various CPU-intensive features such as reflection, ambient occlusion, area lights and procedural shaders.

Wow, talk about a massive boost thanks to the two unlocked processor cores. Cinbench scales very well in multi-core scenarios. Memory performance alone didn’t seem to impact Cinebench scores much. It appears that it prefers raw computational power.


X264 HD Benchmark

Tech Arp’s recent development of the x264 HD Benchmark takes a 30 second HD video clip and encodes it into the x264 codec with the intention of little to no quality loss. The test is measured using the average frames per second achieved during encoding, which scales with processor speed and efficiency. The benchmark also allows the use of multi-core processors so it gives a very accurate depiction of what to expect when using encoding application on a typical full length video.

We see a similar pattern emerge in x264. The extra processing cores made a significant improvement in performance and memory performance boosts did little to improve the frame rates.


WinRAR 3.80

The last of our real-life tests will be with the highly popular & multi-threaded WinRAR 3.8.0 tool. WinRAR is a compression and decompression tool with a built in benchmark. We simply run the benchmark up to 500MB of processed data and record how long it takes.

WinRAR realizes some significant performance improvements with higher memory efficiency and loves the quad configuration due to its multi-threaded nature. We found it pretty surprising that there was a whopping thirty eight second delta between 7-7-7 and 6-6-6 timings.

 

[lemonlime]

Gaming Benchmarks

Gaming Benchmarks

Company of Heroes

We benchmarked Relic’s Company of Heroes using the built in performance test benchmark. Image quality settings were set to a mixture of medium and high and the resolution was set to 1680x1050.

To see a full display of our Company of Heroes configuration settings, click here.

Although Company of Heroes appears to be fairly GPU limited in these configurations, it is pretty evident that the minimum framerate benefitted the most from improved memory efficiency and clock speed improvements.


Crysis 1.2

We benchmarked Crysis using the “Crysis Benchmark Tool 1.0” and chose a “benchmark_cpu” timedemo to ensure that the GPU was not bottlenecking our results. To further ensure a healthy CPU/platform bottleneck, the tests were conducted in ‘DX9’ mode, with all image quality settings set to “Medium”. The resolution used was 1280x1024.

To see a full display of our Crysis configuration settings, click here.

Crysis is definitely not as GPU limited at 1280x1024 and “Medium” IQ settings. As you can see, additional boosts in clock speed made a very significant framerate improvement and it appears that Crysis prefers four cores over two even if they are clocked a bit slower. Memory subsystem efficiency improvements didn’t seem to do a whole lot for Crysis.


Half-Life 2: Episode 2

It’s pretty hard to believe that it’s been over three years since the release of Half-Life 2. Since then, Valve has made numerous graphical improvements to the source engine and included them in the Episode Two release. Although Episode 2 is much more graphically intensive than the original 2005 release, it is still very CPU limited on modern graphics platforms and makes a great CPU/Memory performance benchmark. To test, we created a custom timedemo with plenty of physics and mayhem in the “Outland” portion of the game. The timedemo was played back three times and an average framerate taken. All visual details were set to maximum quality, and the timedemos run at a resolution of 1920x1200.

To see a full display of our HL2:EP2 configuration settings, click here.

It looks like we were pretty GPU limited in this configuration even though we were pushing framerates in excess of 100FPS. We’d argue that all of these results were within the margin of error.


Futuremark 3DMark06

The Futuremark 3DMark series has been a part of the backbone in computer and hardware reviews since its conception. The trend continues today as 3DMark06 provides consumers with a solid synthetic benchmark geared for performance and comparison in the 3D gaming realm. This remains one of the most sought after statistics, as well as an excellent tool for accurate CPU comparison, and it will undoubtedly be used for years to come.

We see a pretty familiar trend with 3DM06. Being a multi-threaded benchmark, the quad core configuration did best, but small gains were realized with the various memory configurations.

 

[lemonlime]

Voltage Regulation/Temperature Testing/Power Consumption

Voltage Regulation

Most people don’t realize that stable system power is the responsibility of more than just the PSU. A motherboard is essentially a complex power supply on its own, converting 12V and other supplied voltages to the required outputs for various system components. We’ll be validating the GD70’s “dialed in” voltages in the BIOS using a digital multimeter.

Click on image to enlarge​

MSI was kind enough to label many voltage read points across the GD70. Unfortunately, quite a few of them are obstructed by the heatpipe cooling solution. None the less, it certainly takes the guess work out of the process. For our tests today, we’ll be measuring three of the most critical system voltages; CPU core voltage, CPU-NB voltage and memory voltage. For grounding purposes, we used the ground lead on a fan header toward the bottom of the motherboard.

As you can see the GD70 does very well when it comes to voltage set accuracy. For the most part, specified values in the BIOS are within 0.025V in actuality. The vCORE tends to undervolt very slightly and does droop by about 0.01V under load, but that is about the extent of it. The CPU-NB tends to overvolt a little more than we’d like to see. A value of 1.40V set in the BIOS equated to 1.447V in actuality. In our opinion, a 0.05V variation is a bit too much and overclockers will need to cognizant of this. DIMM voltage on the other hand is extremely accurate. Every single step up was right on the money with only 0.007V of deviation.

All of the voltage outputs were extremely stable and steady when being examined on the DMM. Overall, we were very pleased with what we saw.


Thermal Performance

To test the effectiveness of MSI’s heavy-duty passive cooling solution, we conducted some basic temperature measurements using a digital thermometer. Although there are two built-in temperature probes on the GD70, we were unable to determine their exact location and questioned their accuracy. We attached our probe to the heatsink on the Southbridge as well as the main heatsink close to the MOSFETS and farther down toward the 790FX.

Full system load was achieved using Prime95 “Large FFT” testing while “Furmark” was run to get the GPU all toasty. For the best effect, we wanted to measure temperatures in a less than ideal cooling environment. To accomplish this, we disconnected our Tech-Station side fan, and let the motherboard heatsinks run entirely passively. Our Noctua NH-U12P provides a small amount of collateral airflow to the MOSFET/790FX heatsink, but very little in comparison to a case with good circulation.

As you can see the heatsinks can get quite toasty, but contrary to popular belief, this is actually a very good thing. This means that a great deal of heat load is being transferred up from the components to the base of the heatsink. We were also very pleased to see that actively cooling only the upper portion of the heatsink resulted in a quick reduction in Southbridge temperatures. Clearly the heatpipe joining the two heatsinks is very functional and not just a cosmetic addition to the board.


Power Consumption

Since its not every day that we run into a motherboard with active phase switching, we thought we should conduct some power consumption tests. Full system power consumption – minus the monitor – was measured using a UPM EM100 power meter. Full system load was achieved using Prime95 “Large FFT” testing.

Tests were conducted with AMD’s Cool’n’Quiet enabled and disabled, and then repeated with “GreenPower” phase switching set to “Auto” in the BIOS.

Although it’s not terribly significant, we were consistently able to see a savings of about 3W with “GreenPower” enabled. With the system at idle, both the memory and CPU phases were chopped down to a single phase. As soon as significant load was put on the processor and memory, the phases would jump back up to their default values. Since the system needs full-phase power at load, there were no improvements realized in any of the configurations.

We were pleased to see that GreenPower worked seamlessly and caused no amount of system instability.

 

[lemonlime]

Conclusion

Conclusion

We must admit, the MSI 790FX-GD70 has left us with a very positive impression.

Beginning with physical aspects of the board, we’d have to say that the overall layout of the GD70 is very good. Considering that it is stacked with expansion slots, MSI did a great job of keeping most connectors and components out of the way. The chipset position and large passive cooling solution is also top-notch. Not only does it look great, it is actually very functional and kept the toasty components nice and cool.

From an overclocker’s perspective, the GD70 is an excellent board. Not only can the latest 1.5b1 BIOS release unlock both X2 550 and X3 720 processors, but it can also provide plenty of juice and stability to take them to their limits. Buyers with “standard edition” Phenom II processors will also be pleased to know that reference clock frequencies beyond 300MHz are very realistic with the GD70. Our memory overlocking experiences were not pleasant, but this has more to do with physical limitations of Phenom II processors than the board or memory modules themselves.

Other enthusiast features, such as the “OC Dial” are extremely handy. We can’t tell you how much easier that little dial made our overclocking tests. Along with the convenience switches and labeled voltage read points, the GD70 is a great bench-board. It would have been nice to see these labeled voltage points moved along the edge of the board and away from heatpipes and other obstructions, but we’re not going to complain as most boards don’t include this feature at all.

Although a great feature that works well, the “Active Phase Switching” seems a little out of place on the GD70. A fair bit of extra “tech” was added to this board to make that feature work, and we can’t help but feel that it belongs on a “less enthusiast targeted” motherboard. A total savings of 3W at idle on a higher-end 790FX based motherboard that’ll be packed with power hungry video cards seems just a little unnecessary. We’d much rather see this feature implemented on 790GX based boards and other HTPC type products in exchange for a lower price tag. None the less, it certainly doesn’t hurt and the feature works seamlessly.

The final strong point the GD70 has going for it is its value. Now although we have been calling this board a “higher end” model throughout this review, its reasonable $200CDN price tag makes it a whole lot of board for not a whole lot of money. To be perfectly honest, we would have expected a board of this stature to price in at least fifty dollars higher.

Overall, we were impressed with the MSI 790FX-GD70 and we’re pleased to award it with Hardware Canucks’ “DAM GOOD” award!



Pros:

- Four PCI-E 16X Slots! (8x/8x/8x/8x bandwidth)
- Good physical layout
- Oversize and functional passive cooling solution
- 8 SATA, eSATA and and seven USB 2.0 ports
- Excellent bus-overclocking headroom
- Very feature rich BIOS with 13 individual voltage adjustments
- OC Dial feature and convenience switches
- Stable voltage output
- Can unlock X2/X3 disabled cores
- Fair price


Cons:

- Bottom PCI-E 16X slot only accepts single-slot cards in most enclosures
- Included software on utility CD is buggy. Definitely be sure to download the latest versions from MSI’s website.

Our thanks to MSI for making this review possible!​

http://www.hardwarecanucks.com/foru...70-am3-motherboard-review-comment-thread.html​