The two 8800GT graphic accelerators are found with PCB designs different from the P393. Most notable is the use of 3-phase power regulation in both the Eagle and Vvikoo graphics accelerators. The Vvikoo has twice the graphics buffer of the Eagle and is expected to have an unfair advantage at higher resolutions.

They may be pumping the same silicon heart, have the same penchant for speed, but each will have its own talents and handicaps. Graphics acceleration for nothing less than the online heroes of this planet. Pwn those cybervillians with a mutant in your rig!

Eagle's 8800GT is one of a kind, utilising a proprietary PCB layout alongside a customised thermal solution. The one we have with us is an early sample, hence you won't see accessories or bundles for the Eagle in this review.

Early sample in full glory.

3 Phase GPU VRM.

Fujitsu solid polymer capacitors throughout where polarised parts are used.

Heatsinked RAM ICs.

Qimonda 1ns parts.

G92-270, like any other.

Vgpu in 3D.

Vmem in 3D.

The first thing that struck us when we took hold of the Vvikoo (pronounced vee-vee-koo) was the sheer resemblance to the XpertVision 8800GT we reviewed some time back. The two may appear to be of a different race, but the layout of components suggest significant similarities.

Boxed up and ready.

1GB of video memory.

Bundled accessories. The supplied DVI-HDMI adapter is not featured in this picture.

The card in all it's purple glory.

Dual DVI.

1ns Qimonda parts.

G92 shows up.

The clock crystal is a through-hole unit, unlike the SMD part on P393.

3 phases' worth of MOSFETs.

Rubycon and Sanyo low-ESR capacitors throughout.

Richtek RT8802 VRM for the GPU.

Vgpu in 3D.

Vmem in 3D.

Overclocking boils down largely to two two quantifiable aspects: thermal management, and operation/signalling voltages. By utilising proprietary voltage regulation and thermals, both the Eagle and Vvikoo are essentially tuners' dream-come- true. Let's take a look at the thermal solution presented with the two.

Two slot design.

With the exception of two, sample RAM ICs were individually cooled. Air is directed over the entire card in bid of cooling supporting components.

Springed mounts ensure even pressure over the GPU die.

Not the shiniest surface, but flat nontheless. Heatpipes are soldered onto the baseplate, but press-fitted into the fins for our sample.

Two-pin fan used does not allow fan control via software. Cooler was relatively silent though.

All-so-familiar Zalman VF-1000 onboard. Heatpipe is soldered to the fins.

0.14A fan directs air over the entire assembly. Even though the fan runs at full-speed all the time, like the Eagle, there was not much noise to talk about.

RAM heatspreader.

The stamped aluminium profile in contact with the RAM ICs.

Grey-goop. A free performance boost from premium paste is possible.

GPU MOSFET heatsink.

We ran three loops of 3DMark06's GT2 and GT4 with 8xAA for GPU load, with room temperature hovering around 20 degrees Celsius. Temperatures were monitored on RivaTuner. Rather unfortunate was the inability to test for temperatures on the Eagle, which did not a have a onboard thermal sensor detected by RivaTuner. Not to forget, the Eagle is still an early sample, compared to the on-the-shelves Vvikoo.

Maximal load temperature.

Testing graphics accelerators can get really tricky because benchmarks tend to sniff out more irreguarlities than is representative of real-world variations. Tests here were run thrice and averaged. Misfits were thrown out with the dishwater.

For the testbed, we employed the Asus P5K3 Deluxe motherboard based on the Intel P35 chipset. Processor was a Swiftech Apogee GT cooled Intel Core 2 Extreme QX9650 running at no less than 4.4GHz. Part of the ploy was Cosair CM3X1024-1800C7D 2GB memory kit. To power up the monster, we deserved nothing less than a Silverstone Olympia 1200W Power Supply Unit. Dribs-and-drabs included stuff from Seagate, Asus, Logitech and Everglide. A Dell 3007WFP interfaced us to the setup and enabled some pretty nifty HD Gaming. Forceware 174.13 and Catalyst 8.2 drivers were used for comparisons. More on the details below:

CPU-Z data for the testbed

Below is the compiled specifications of the graphics accelerators we're testing the Mutant Duo with. 1GB graphics accelerators are rare and hence you won't see them compared. The Sparkle 8800GT, like the reviewed duo, is overclocked too. Users who are looking at either graphics accelerators should be comparing them with similar factory-overclocked offerings.

LCD prices have been falling steadily over the years, to the extent where humongous panels are actually affordable to many. For some, it is a necessity to drive large panels for work (e.g. media manipulation) in 2D even when the budget does not allow for a highend 3D graphics acceleration. When the graphics accelerator cannot keep up to speed in 3D, users (those who play the occassional game) are forced to run at lower resolutions. LCD panels understandably, do not display images sharply when forced to downsample below native resolution. Larger Video RAM buffers are supposed to alleviate the lag resulting from extra large textures in high resolution gaming. We took the opportunity to test Vvikoo's purported 1GB graphics buffer.

Dell's 3007WFP.

Using Level 5 on both World and Texture detail settings, we ran a timedemo of Heat Ray at 2560x1600 to obtain 60.84FPS on the Eagle and 56.56FPS on the Vvikoo. Here it becomes apparent that clockspeeds still decide the real-world speeds on in-game graphics.

With all but the largest screens, it would be unlikely for anyone to see a performance improvement with a 1GB graphics accelerator at the moment. Considering the lower memory clocks and slower latencies asscociated with more Video Memory, it may not be a wise idea to spend your dollar on those unused MBs.

Overclocking factory-clocked cards have for long, been very much of a taboo issue amongst veteran overclocking enthusiasts. Limited headroom from the preset clocks have always rendered overclocked gaphics accelerators poor value for money amongst those who tweak down to the last stable MHz. Worse of all, many factory overclocked graphics accelerators are in fact not speed-binned at all. Let's see what the Eagle and Vvikoo can squeeze out of their two slots.

Eagle overclocked.

Vvikoo overclocked.

The Eagle has a significantly higher Vgpu than the Vvikoo, or the Sparkle 8800GT and P393 variants. The poorer overclocking on the Eagle boils down to a poorer batch of G92 chipset, and possibly a less capable cooling system. Due to the larger size of memory buffer on the Vvikoo, the memory clocks scaled lower as expected. Memory bandwidth is still sufficient for the core and shader clocks we obtained, however. The Vvikoo is a clear winner in the overclocking department, whilse the Eagle remains undecided until their appearance in the market. Either way, both designs have proven their worth at being better overclockers than any P393 8800GT, with better cooling, and higher chipset voltages.

Watch those numbers, however, for instability could set in!

Driver malfunction.

Warhol plays 3DMark.

The non-P393 8800GT cards in this review proved worthy options for users keen on getting G92 performance nearing the GTS without having to pay the premium. Where design is concerned, engineers from Eagle and Vvikoo took on different approaches - efforts that pay off when the user overclocks their Eagle or Vvikoo.

More interesting, however, was the Eagle's Vgpu voltage. Running a fair bit higher than default designs and Palit/Vvikoo boards, it was surprising that the Eagle did not clock higher. The Eagle had it's flaws, with its premature cooling option, whilst the Vvikoo's money was put into unnecessary buffer size. Special mention goes to the Vvikoo for a solidly built graphics accelerator.

We'd love to give a conclusion on the Eagle, but it is still a little raw in form. For the Vvikoo...

Return Home

Check Hottest Deals

Check Latest Prices