All of the motherboards in today’s comparison provide frequency ranges that exceed by far any realistic expectations of stability and most voltage ceilings are far beyond any realistic expectations of reliability. Experienced overclockers know not to shoot for the limit at first, but instead determine the maximum supported speed and required voltage by making small changes and retesting repeatedly.
| BIOS Frequency And Voltage Settings (For Overclocking) | ||||
|---|---|---|---|---|
| Asus | DFI LANParty DK | Gigabyte | MSI |
| CPU Ref Clock | 200-600 MHz (1 MHz) | 200-700 MHz (1 MHz) | 200-500 MHz (1 MHz) | 200-600 MHz (1 MHz) |
| CPU Multiplier | Yes | Yes | Yes | Yes |
| DRAM Data Rates | 800/1066/1333/1600 | 800/1066/1333/1600 | 800/1066/1333/1600 | 800/1066/1333/1600 |
| PCIe Clock | 100-150 MHz (1 MHz) | 100-250 MHz (1 MHz) | 100-200 MHz (1 MHz) | 100-150 MHz (1 MHz) |
| CPU Vcore | 0.80-1.70V (0.0125V) | -800-+775mv (25mv) | +25-600mV (25mV) | 0.98-1.93V (0.01V) |
| IMC Voltage | 0.80-1.55V (0.0125V) | +3.57, 7.14, 10.71% | +25-600mV (25mV) | 0.85-1.83V (0.01V) |
| 790FX Core | 1.10-1.40V (0.02V) | 1.12-1.50V (0.04V) | +0.1, +0.2, +0.3 V | 0.75-1.83V (0.01V) |
| SB750 Core | 1.20-1.35V (0.15V) | 1.20-1.50V (0.20V) | +0.1, +0.2, +0.3 V | 0.85-1.83V (0.01V) |
| DRAM Voltage | 1.50-2.50V (0.02V) | 1.50-2.49V (0.22V) | +50-750mV (50mV) | 1.19-2.40V (0.01V) |
| CAS Latency | 4-12 Cycles | 4-11 Cycles | 4-12 Cycles | 4-12 Cycles |
| RAS To CAS Delay | 5-12 Cycles | 5-11 Cycles | 5-12 Cycles | 5-12 Cycles |
| Row Precharge | 5-12 Cycles | 5-11 Cycles | 5-12 Cycles | 5-12 Cycles |
| tRAS | 15-30 Cycles | 16-30 Cycles | 15-30 Cycles | 15-30 Cycles |
The two boards with the “smallest” CPU voltage regulators also have the most aggressive voltage levels, but we limited our CPU core to 1.45 V to assure longevity.
The two motherboards with 10-phase power regulation finished first and last, with Gigabyte providing the highest stable CPU clock at our chosen voltage levels.
MSI has the best reference clock stability. A combination of first-place reference clock and second-place CPU stability could make the 790FX-GD70 a top overclocking choice for multiplier-locked processors.
Dropping timings all the way back to 9-9-9-28 at 2T wasn’t enough to get most configurations beyond DDR3-1600, though Asus did break that barrier with four modules installed.
The two five-phase power regulators placed first and last in thermal performance. MSI’s win is probably due to its enormous VRM sink.
All of today’s motherboards appear energy efficient, yet none stand apart. Chipset and CPU manufacturer AMD gets all the credit for an overall good showing.
- 790FX Is Still The King
- Socket AM3 Motherboard Features Comparison
- Asus M4A79T Deluxe
- DFI LANParty DK 790FXB-M3H5
- Gigabyte GA-MA790FXT-UD5P
- MSI 790FX-GD70
- Test Settings
- Benchmark Results: 3D Games
- Benchmark Results: A/V Encoding And Productivity
- Benchmark Results: Synthetic
- Overclocking, Power, And Heat
- Conclusion





(2.80 GHz, 86.0 MB Cache)
I never knew any AMD processors had that much cache!
Super CPU!
1.66% better than the worst performer in the tests. :-)
Not much point in throwing that much GPU horsepower at a motherboard review; the card Thomas used is our current reference--and it's more than ample for showcasing the differences between these boards.
If you want to see quad-CrossFire for any specific reason, feel free to let us know and we can put together a story idea!
Consistent would have been better if not for the fear of people pointing out ever tiny inconsistency. Just remember, if you're 5'11" tall it's clear to just about everyone that you're not 6'!!!
Could you clarify that a bit Chris? Are you saying you would not expect any difference in crossfire between the MBs? If so, that's OK. However, if there's going to be a difference, that is the main reason to buy a 790FX. If you aren't going to crossfire might as well get a GX.
I suspect the chipset isn't the whole story and the various manufacturers could still screw up the PCI-E voltages or something
It depends on the numbers you're looking for. I interpreted the OP's request to mean he wanted performance results with 4870 X2s, which are overkill for comparing these boards. If it's a matter of comparing PCI Express scaling (the reason for going FX instead of GX), check out this piece: http://www.tomshardware.com/reviews/crossfire-pci-express,2095.html. It isn't on the AMD platforms, but it still illustrates fairly well how much/little performance you'd lose with PCIe 2.0 x8 vs. x16.
Granted my request isn't based on the real needs of the consumer, it's more of a scientific curiosity behind it. But I remember a while back tom's had an article about nvidia's SLI with 2, 3 or 4 cards in different configurations and that was an interesting read.
PS: Just to clear something up in the 2 x 4870X2 vs. 4 x 4870 or other versions. I know that 4x4870 are more $$ then 2x4870X2 but if you buy a 4 slot PCIe motherboard what if you start out with a single 4870 or 4890 and just add in cards over the period of a year since prices change and people might not have all the cash to get 4 cards at the same time.
"We guys" just made a suggestion since Chris stated he's opened to ideas. The asrock X58 article a few days back is another good example of a story that can be looked at from a more technical perspective. Since that board and these two AMD ones share 4 PCIe ports and ATI & Nvidia cards support ATI STREA/CUDA one would be inclined to have a look beyond the "can it run crysis" comment. And I'm sure finding 4 identical cards isn't an issues since the time that takes to right an article wasn't up for debate.