discussion Value Meets Performance: HD 4890 Cards from Gigabyte and MSI

[SKYMTL]

Crysis: Warhead (DX10)

Crysis: Warhead (DX10)

To benchmark Warhead, we recorded a 5 minute timedemo on the Ice level which included ranged and hand to hand combat. The sequence was then played back using the Crysis Benchmarking Tool from HOC

1680 x 1050

1920 x 1200

2560 x 1600

 

[SKYMTL]

Dawn of War II

Dawn of War II

Even though Dawn of War II has its own in-game benchmarking tool, we decided not to use it. Instead, we played through approximately 10 minutes of the final Sapce Marine mission “Last Stand” while recording framerates with FRAPS. With a massive amount of enemies rushing your position and additional troops deployed on your side, this mission features some of the largest battles in the single player game. In addition, since the Tyranid swarms appear at regular intervals, this mission is also perfect for comparison testing.

1680 x 1050

1920 x 1200

2560 x 1600

 

[SKYMTL]

Fallout 3

Fallout 3

For these Fallout 3 benchmarks we decided to use one of the more graphically intensive portions of the game: a gameplay session which starts at the exit from Vault 101, through Springvale, over a small hill towards Washington and then back towards Megaton. Within this session two ants were fought near Springvale. All in all, the gameplay time is about 6 minutes.

1680 x 1050

1920 x 1200

2560 x 1600

 

[SKYMTL]

Far Cry 2 (DX9)

Far Cry 2 (DX9)

Even though Far Cry 2 has its own built-in benchmarking tool with some flythroughs and “action scenes”, we decided to record our own timedemo consisting of about 7 minutes of game time. It involves everything from run-and-gun fights to fire effects. The built-in benchmarking too was then set up to replay the timedemo and record framerates

1680 x 1050

1920 x 1200

2560 x 1600

 

[SKYMTL]

Far Cry 2 (DX10)

Far Cry 2 (DX10)

Even though Far Cry 2 has its own built-in benchmarking tool with some flythroughs and “action scenes”, we decided to record our own timedemo consisting of about 7 minutes of game time. It involves everything from run-and-gun fights to fire effects. The built-in benchmarking too was then set up to replay the timedemo and record framerates

1680 x 1050

1920 x 1200

2560 x 1600

 

[SKYMTL]

Left 4 Dead

Left 4 Dead

For benching Left 4 Dead, we used a pre-recorded 5 minute timedemo taken on the No Mercy campaign during The Rooftop Finale mission. Framerates were captured with FRAPS.

1680 x 1050

1920 x 1200

2560 x 1600

 

[SKYMTL]

Tom Clancy’s Hawx (DX9)

Tom Clancy’s Hawx (DX9)

Even though HawX has a built-in benchmarking tool, we found that it did not provide accurate enough results needed to reflect what actual in-game performance would be. Thus, we played through the first 6 minutes of a mission and recorded the results with FRAPS.

1680 x 1050

1920 x 1200

2560 x 1600

 

[SKYMTL]

Tom Clancy’s Hawx (DX10 / 10.1)

Tom Clancy’s Hawx (DX10 / 10.1)

Even though HawX has a built-in benchmarking tool, we found that it did not provide accurate enough results needed to reflect what actual in-game performance would be. Thus, we played through the first 6 minutes of a mission and recorded the results with FRAPS. All ATI cards have DX10.1 enabled.

1680 x 1050

1920 x 1200

2560 x 1600

 

[SKYMTL]

Heat, Acoustics & How to Fix the MSI Cooler

Heat, Acoustics & How to Fix the MSI Cooler

For all temperature testing, the cards were placed on an open test bench with a single 120mm 1200RPM fan placed ~8” away from the heatsink. The ambient temperature was kept at a constant 22°C (+/- 0.5°C). If the ambient temperatures rose above 23°C at any time throughout the test, all benchmarking was stopped. For this test we use the 3DMark Batch Size test at it highest triangle count with 4xAA and 16xAF enabled and looped it for one hour to determine the peak load temperature as measured by GPU-Z.

For Idle tests, we let the system idle at the Vista desktop for 15 minutes and recorded the peak temperature.

Before we get into the disaster that is the performance of the MSI card, let’s talk about Gigabyte’s little miracle worker. The Zalman cooler that adorns the HD 4890 Ultra Durable pays for itself in spades with some downright impressive numbers. Its idle temperature is particularly impressive when compared to the reference based-cards’ numbers. The best part about this however is that the fan on this card stayed extremely quiet throughout the test.

MSI, what can we really say? Here we have a massive heatsink that is rendered absolutely useless by a first-rate screw up of an installation. As we mentioned before in this review, due to MSI using the reference backplate, the Cyclone heatsink isn’t able to make proper contact with the GPU core. The result is some stratospheric temperatures which would have actually gone higher if the core wouldn’t have begun throttling its performance in a pointless effort to keep things under control. Remember, when the core on the HD 4890 gets hot enough it will begin throttling and WILL NOT artifact. As such, if your card is exhibiting lower than expected performance, check the temperatures. If you want to see how bad things got, here is the Temperature Log (.txt file).

So what did we do to improve the temperatures and complete the testing regime?

​

In order to increase the contact pressure we installed a small rubber washer between the backplate and the head of the screw. This is the type of washer that can be found at any hardware store but it should be at least 1/8” high.

As you can see in the chart above, this simple modification brought down the temperatures to where they should be considering the size of the Cyclone cooler.

We have seen a ton of customer reviews (ironically, no professional review picked up on the mounting issue) stating that the fan on this card runs at ultra high levels. Since the heatsink’s fan is directly tied to the core temperature, our mod also allowed the fan to spin at a much more leisurely pace.

All in all, we would call this an absolute failure on MSI’s part. They have an amazing heatsink on their hands but shoddy mounting meant both decreased gaming performance as well as dangerously high temperatures.

 

[SKYMTL]

Power Consumption

Power Consumption

For this test we hooked up our power supply to a UPM power meter that will log the power consumption of the whole system twice every second. In order to stress the GPU as much as possible we once again use the Batch Render test in 3DMark06 and let it run for 30 minutes to determine the peak power consumption while letting the card sit at a stable Windows desktop for 30 minutes to determine the peak idle power consumption. We have also included several other tests as well. Please note that after extensive testing, we have found that simply plugging in a power meter to a wall outlet or UPS will NOT give you accurate power consumption numbers due to slight changes in the input voltage. Thus we use a Tripp-Lite 1800W line conditioner between the 120V outlet and the power meter.

Power consumption is exactly what we would expect with both cards consuming slightly more electricity than a stock unit. However, Gigabyte’s Ultra Durable design looks to be paying some dividends as their card’s idle power consumption is noticeably lower than that of a reference card.