Gigabyte MA790GP-DS4H Motherboard
|
Settings
Jumpers and switches |
Clear CMOS jumper |
To be more exact, the motherboard has two contacts, which should be closed with a jumper. |
Award BIOS 6.00PG |
Allows to disable specific CPU functions |
+ |
K8 Cool'n'Quiet Virtualization Away Mode |
Memory timings |
+ |
Unganged/Ganged Mode, 1T/2T Memory Timing, CAS Latency, RAS to CAS Delay, Row Precharge Time, Min RAS Active Time, RAS to RAS Delay, Write Recovery Time, Precharge Time, Row Cycle Time, TwTr Command Delay |
Memory frequency selection |
+ |
400, 533, 667, 800, 1066 MHz, you actually specify a multiplier to the FSB frequency |
HT bus setup |
+ |
Frequency: 200, 400, 600, 800, 1000, 1200, 1400, 1600, 1800, 2000, 2200, 2400, 2600 MHz |
Integrated GPU frequency control |
+ |
200-2000 MHz at 1 MHz steps |
Peripheral bus frequency control |
+ |
PCI-E = 100-200 MHz at 1 MHz steps |
PCI IRQ manual assignment |
+ |
|
FSB frequency setup |
+ |
200-500 MHz at 1 MHz steps |
CPU multiplier |
+ |
from x5 at 0.5x steps for the cores and memory controller) |
CPU core voltage control |
+ |
+0.025-0.600 V at 0.025 V steps (two independent settings: for the cores and memory controller) |
Memory voltage control |
+ |
+0.05-0.55 V at 0.05 V steps |
Chipset voltage control |
+ |
-0.200-+0.300 V at 0.025 V steps (Northbridge) +0.1-0.3 V at 0.1 V steps (Southbridge) +0.1-0.3 V at 0.1 V steps (SidePort video buffer) |
We used BIOS F2A, the latest release at the time of our tests. The mentioned BIOS parameters are available in this version, but the viability of non-standard settings hasn't been tested.
The motherboard automatically read data from the EPP profile in our memory modules (reduced timings and raised voltage) and worked with these parameters correctly. The choice of overclocking settings is close to maximum, it will please even hardcore users. As usual with Gigabyte motherboards, you can open full settings by pressing Ctrl+F1 in the main BIOS Setup menu.
Overclocking
In order to evaluate motherboard and its BIOS, we overclock our testbed processor to a maximum stable level. We use all features of the motherboard in this test, including raising CPU voltage and adjusting multipliers and frequencies of system and peripheral buses, if necessary. But if, for example, reducing Hyper-Transport frequency does not improve overclocking, we leave the default multiplier. Memory is set to the standard frequency for a given memory module (multiplier correction), if a manufacturer does not publish any ways to improve memory overclocking. Otherwise, we analyze their efficiency as well. From now on, this review included, we'll be using AMD OverDrive stability test to evaluate stability of an overclocked system. We run all its tests for 5 minutes; the load is a tad higher than that of the previously used WinRAR test (Tools -> Benchmark and hardware test), but it usually reveals instabilities much faster, and what's even better, it does not freeze the system as the frequency grows (the test just signals about errors and stops). As overclocking potential is an individual property of a given motherboard sample to some degree, we don't set the task to determine overclocking potential to within a single MHz. In practice, we are to find out whether CPU overclocking will be limited by a motherboard as well as to evaluate its behavior in non-standard modes, including automatic restoration of a correct frequency after a failed overclocking attempt, etc.
Test |
Clock, MHz |
FSB clock (multiplier), MHz |
Core voltage (according to BIOS), V |
HT bus frequency (multiplier), MHz |
CPU NB frequency (multiplier), MHz |
Note |
Athlon X2 4850e (2.5 GHz) |
3189 |
255 (x12.5) |
1.47 |
1325 (x5) |
|
|
Phenom X4 9550 (2.2 GHz) |
2695 |
245 (x11) |
1.26 |
2205 (x9) |
2205 (x9) |
|
2717 |
247 (x11) |
1.26 |
1976 (x8) |
1976 (x8) |
CPU NB and HT multipliers are reduced |
Phenom X4 9850 Black Edition (2.5 GHz) |
2950 |
236 (x12.5) |
1.49 |
2320 (x10) |
2320 (x10) |
Bus-overclocked |
3000 |
240 (x12.5) |
1.51 |
1920 (x8) |
1920 (x8) |
Ditto; CPU NB and HT multipliers are reduced |
2958 |
204 (x14.5) |
1.49 |
2040 (x10) |
2040 (x10) |
Multiplier-overclocked (ACC OFF) |
3297 |
206 (x16) |
1.54 |
2060 (x10) |
2060 (x10) |
Multiplier-overclocked (ACC +2%) |
Overclocking results are quite good for all three processors, although they don't break any records. And we managed to overclock the integrated graphics core to 1000 MHz raising voltage by 0.175 V. So far it's the best result among motherboards on AMD 790GX. The motherboard does not require any input from a user to restore default parameters after a failed overclocking attempt. As a rule, it starts up with default values after restart and opens BIOS for a user to adjust settings.
In order to dot all the 'i's about the effect of Advanced Clock Calibration on overclocking standard Phenoms (with the locked multiplier), we reduced multipliers of HT bus and the built-in Northbridge to eliminate all theoretical bottlenecks. The effect from this procedure is minimal (30-50 MHz of gain), so ACC is just as useless here. That was to be expected, because computing cores are the most complex elements of a CPU in terms of their architecture and manufacturing. Apart from some anomalies, their overclocking potential must be the shaping factor. To make the CPU NB multiplier matter, we should test overclocking with reduced CPU core multiplier. Sometimes it's done in motherboard tests with a vague objective to determine maximum reference frequency of the on-board clock generator. In this artificial case it's necessary to decrease the multiplier to the 'fifth core' as well, lest its frequency should really be unnaturally higher than the frequency of computing cores. But this approach has nothing to do with practical overclocking (a higher result obtained with this method for a given motherboard does not guarantee that all processors will overclock just as well here). In practice, the real bottleneck is not in the ability of a clock generator to work at higher frequencies in artificial conditions, but in the circuit design in general.
By the way, speaking of the fact that Advanced Clock Calibration is efficient only for processors with the unlocked multiplier (Black Edition), it should be added that it does not mean the necessity to keep the reference clock strictly at 200 MHz for these processors. It's quite possible to adjust this frequency to overclock the system as high as possible. It makes sense -- when you raise the multiplier, you increase the frequency at larger steps, and reference clock adjustments act as finetuning. Moreover, this finetuning does not interfere with ACC, at least within the range we used in our case.
Performance
Testbed configuration:
- CPU: AMD Phenom X4 9550
- Memory: 2 x Corsair CM2X1024-6400C4 (2 GB, DDR2-800, 5-5-5-15-2T)
- HDD: Seagate Barracuda 7200.10 (SATA, 7200 rpm)
- Graphics card: ATI Radeon HD3870, 512 MB GDDR4
- Power supply unit: AcBel ATX-550CA-AB8FB
- OS: Windows XP SP2
We decided to compare our motherboard with Foxconn A7DA-S on the same chipset with the same type and volume of video memory: 128 MB DDR3-1333.
Test |
Int. graphics |
Discrete graphics |
Foxconn A7DA-S |
Gigabyte MA790GP-DS4H |
Foxconn A7DA-S |
Gigabyte MA790GP-DS4H |
Archiving with WinRAR, min:sec |
1:38 |
1:35 |
1:38 |
1:35 |
MPEG4 (XviD) encoding, min:sec |
4:45 |
4:40 |
4:43 |
4:39 |
Unreal Tournament 2004 (Medium@800x600), fps |
55.6 |
54.0 |
59.2 |
59.0 |
Unreal Tournament 2004 (High@1024x768), fps |
49.4 |
49.2 |
59.1 |
59.2 |
Unreal Tournament 2004 (Highest@1600x1200), fps |
29.8 |
29.8 |
58.4 |
59.6 |
FarCry (Medium@800x600), fps |
105.9 |
109.7 |
135.0 |
136.3 |
FarCry (High@1024x768), fps |
76.2 |
76.5 |
125.9 |
137.3 |
FarCry (Highest@1600x1200), fps |
35.0 |
35.1 |
125.4 |
137.9 |
DOOM III (Medium@800x600), fps |
71.2 |
70.8 |
162.4 |
157.2 |
DOOM III (High@1024x768), fps |
50.8 |
50.6 |
161.3 |
160.9 |
DOOM III (Highest@1600x1200), fps |
24.5 |
24.5 |
156.6 |
158.1 |
These motherboards perform on a par in modes that depend on performance of the integrated graphics core. What concerns computing tests, Gigabyte enjoys a minimal advantage plus more optimal configuration of the graphics interface in this motherboard (results with a discrete graphics card correspond to maximum results possible with this graphics card).
Power consumption (entire system unit)
Phenom X4 9550 (Cool'n'Quiet OFF) |
Int. graphics |
Discrete graphics |
Foxconn A7DA-S |
Gigabyte MA790GP-DS4H |
Foxconn A7DA-S |
Gigabyte MA790GP-DS4H |
Text editing, W |
54 |
43 |
61 |
71 |
Playing FarCry, W |
78-95 |
76-96 |
102-126 |
100-124 |
When idle, the motherboard from Gigabyte consumes less power with the integrated graphics enabled. But the situation changes to the contrary, when a graphics card is installed. Both motherboards demonstrate identical power consumption in 3D mode.
Conclusions
This motherboard combines functionality of Gigabyte MA78GPM-DS2H and Gigabyte MA78G-DS3H based on AMD 780G. Besides, it now supports processors with TDP up to 140 W and enjoys advantages of the 790GX chipset: higher performance of the integrated graphics core, symmetric mode of graphics ports for CrossFire, Advanced Clock Calibration and RAID 5.
However, unlike the above motherboards (exclusive AMD 780G models to some degree), 790GX-based motherboards from other manufacturers also offer high functionality and are rigged up to the reasonably maximum level. So users can afford to be choosy now. In particular, this motherboard offers one of the most powerful CPU voltage regulators (it's an economic model at the same time -- owing to field-effect transistors of the new generation) and an imposing cooling system, we've ever seen in motherboards of this class. But it has some drawbacks as well: relatively low characteristics of the integrated analog outputs and the lack of eSATA (which is available in Gigabyte MA78GPM-DS2H, by the way).
The motherboard provided by the manufacturer, ATI Radeon HD 3870 graphics card provided by PowerColor.
Write a comment below. No registration needed!
|
|
|
|
|