| BIOS Frequency, Voltage and Timings | |||||
|---|---|---|---|---|---|
| ASRock Fatal1ty 990FX Professional | Asus Sabertooth 990FX | ECS A990FXM-A | Gigabyte 990FXA-UD7 | MSI 990FXA-GD80 | |
| Reference Clock | 150-500 MHz (1 MHz) | 100-600 MHz (1 MHz) | 190-400 MHz (1 MHz) | 200-500 MHz (1 MHz) | 190-690 MHz (1 MHz) |
| CPU Multiplier | 8.0-31.5x (0.5x) | 4.0-35.0x (0.5x) | 2.5-23.5x (0.5x) | 7.0-35.0x (0.5x) | 4.0-32.5x (0.5x) |
| DRAM Data Rates | 800-1866 (266.6 MHz) | 800-2400 (266.6 MHz) | 667-1866 (266.6 MHz) | 667-1866 (266.6 MHz) | 800-1866 (266.6 MHz) |
| CPU Vcore | 0.60-2.00V (12.5 mV) | 0.68-2.08V (6.25 mV) | +0 to +500 mV (50 mV) | -0.06 to +0.60V (25mV) | 1.08-2.05V (~10.5 mV) |
| CPU NB | 0.60-2.00V (12.5 mV) | 0.50-1.90V (6.25 mV) | Not Adjustable | -0.06 to +0.60V (25mV) | 1.00-1.82V (~11mV) |
| 990FX Voltage | 1.11-1.66V (10 mV) | 0.80-1.51V (6.25 mV) | +0 to +500 mV (10 mV) | 0.87-1.98V (5 mV) | 0.96-1.39V (~5.5 mV) |
| DRAM Voltage | 1.25-2.07V (10 mV) | 1.20-2.50V (6.25 mV) | -30 to +600 mV (10 mV) | 1.03-2.14V (5 mV) | 1.20-2.45V (~7 mV) |
| CAS Latency | 5-14 Cycles | 5-19 Cycles | 5-14 Cycles | 5-14 Cycles | 5-14 Cycles |
| tRCD | 5-19 Cycles | 2-19 Cycles | 5-19 Cycles | 2-19 Cycles | 2-19 Cycles |
| tRP | 5-19 Cycles | 2-19 Cycles | 5-19 Cycles | 2-19 Cycles | 2-19 Cycles |
| tRAS | 15-40 Cycles | 8-40 Cycles | 10-35 Cycles | 8-40 Cycles | 8-40 Cycles |
All enthusiast-grade motherboards provide a wide enough range of voltage controls to fry our processor, and a wide-enough range of frequencies to make our processor completely unbootable, even before we have a chance to fry it. Yet, our tests target long-term stability, and that’s why we choose a maximum 1.40 V at the CPU core for our overclocking efforts.

Gigabyte leads Asus in maximum CPU overclock, but we still preferred overclocking with the Asus Sabertooth 990FX due to its ability to set a target voltage in firmware. The 990FXA-UD7 forces us to chase the appropriate voltage by increasing core offset and retesting until the desired voltage level is achieved.
ECS appears to beat ASRock, but we were not able to do this with firmware. Because AMD Turbo Core was always enabled, getting a consistent clock required the use of AMD Overdrive. We further had to make adjustments in Overdrive at every reboot, making the 4.30 GHz clock rate unacceptably cumbersome to use.
MSI’s low score was caused by voltage drop that resulted from MSI’s voltage compensation mechanism removed from its first AMD Bulldozer firmware.

The Sabertooth 990FX achieves the highest CPU reference clock, which is a setting that achieves its greatest significance when used to overcome the limits of locked multipliers (which, incidentally, none of the FX processors suffer from). MSI’s apparent misfortune is caused by boot failures whenever we attempted to manually lower its CPU-NB ratio, which climbed past the edge of stability in “Auto” mode as we pushed up the CPU base clock.

Asus must not have gotten the message that AMD’s FX has no DDR3-2133 ratio, as this is the setting it uses to achieve a class-leading DDR3-2205 data rate. Gigabyte forces us to make our push from the DDR3-1866 ratio by increasing reference clock, and both Gigabyte and MSI even drop that capability with four modules installed.
- 990FX: AMD Leads The Chipset Game
- ASRock Fatal1ty 990FX Professional
- Fatal1ty 990FX Professional Firmware
- Asus Sabertooth 990FX
- Sabertooth 990FX Firmware
- ECS A990FXM-A
- A990FXM-A Firmware
- Gigabyte GA-990FXA-UD7
- 990FXA-UD7 Firmware
- MSI 990FXA-GD80
- 990FXA-GD80 Firmware
- Test Settings And Benchmarks
- Benchmark Results: 3D Games
- Benchmark Results: Audio And Video Encoding
- Benchmark Results: Productivity
- Power, Heat, And Efficiency
- Overclocking
- Which 990FX Board Should You Buy?
So, x58 is irrelevant, because SB beats it. Except AMD's offering is somehow relevant even though both x58 and SB beat it. What?????
If you ignore x58 because SB offers better performance, you ignore anything AMD has because a SB setup offers better performance. If you want 36 or less lanes, x58 still offers better processors than you can hope to get from AMD. Bizarre logic.
Not that AMD is irrelevant, just the logic is badly flawed.
but great chipsets cant offset poor CPU's.
Secondly, I would really like to see a piece on extreme CFX/SLI configurations on rigs like this. It seems an article with reliable information on this would be beneficial to gaming enthusiasts, IT professionals, and HPC builders alike!
Hope to see an article along these lines soon!
So, x58 is irrelevant, because SB beats it. Except AMD's offering is somehow relevant even though both x58 and SB beat it. What?????
If you ignore x58 because SB offers better performance, you ignore anything AMD has because a SB setup offers better performance. If you want 36 or less lanes, x58 still offers better processors than you can hope to get from AMD. Bizarre logic.
Not that AMD is irrelevant, just the logic is badly flawed.
Originally it referred to AMD's insistence of comparing its FX-8150 to the 990X to prove that the FX-8150 had far better value. The original version of the paragraph referred to that comparison method a sham, and THEN referred to the SB vs BD debate. I guess it's neither nice nor necessary to call the 8150/990X price/performance comparison a sham, so the paragraph was altered to improve it's tone
Please do a Tri-Sli review with 580's in it.
Compare the 8150 @ $279 vs the 2500K @ $215, who would you recommend?
Hint: http://www.hardocp.com/article/2011/11/03/amd_fx8150_multigpu_gameplay_performance_review/1