eddard | 18 August, 2008 22:10
The three 9800 GTX+ cards - the one with the radiator connected to it is the Zone edition from Zotac.
A quick run-through of the videocard’s form factors and features is in order: The 9800 GTX+ is a dual-slot card, utilizing a massive cooler similar to the one found on high-end GTX200s and the older 8800 series cards. Two golden SLI fingers peek from around the black shroud of the cooler assembly, and two 6-pin power connectors stand by to power this mid-range card. While much of the PCB is covered by the shroud, we can easily surmise that the core itself is much smaller than previous generations due to the smaller manufacturing process. A small 2-pin connector can be found to the left of the power connectors – this is for the HDMI-through audio. The description above applies only to two cards thou: the XFX and the regular Zotac 9800 GTX+s. The third card in this roundup is the “Zone” edition of Zotac’s 9800 GTX+, which was overclocked and temperature measured as part of the testing.


The Zotac 9800 GTX+ Zone deserves special mention, as this is only the second card that has passed throught the hands of this editor that sports a built-in water-cooling contraption, specifically a single fan (120mm) radiator mounted with an orange-tinted case fan that is meant to hang on a casing’s rear exhaust port. Tubes connect the radiator assembly to the main water block on the 9800 GTX+ core, with additional heavy-duty fins surrounding it and covering the memory, making for a much more massive card than a regular GTX+. Cooling performance and overclock levels will be explored a bit later.


The XFX 9800 GTX+ and the Zotac 9800 GTX+ are both “vanilla” cards running at stock factory settings, although all three represent something revolutionary for the green team. The release of the 9800 GTX with the “plus” sign at the end is a watershed event, although it will mostly go unnoticed by most buyers, as most of the changes on the GTX+ from the plain old GTX is on the inside. The 9800 GTX+ is the first Nvidia part built on a 55nm process, and is currently the highest (factory-clocked) card in the Nvidia lineup, with a core running at 738MHz, shaders at 1836MHz and a comparatively normal memory clock of 1100MHz (dual pumped to 2200MHz). Even the top-of-the-line GTX280 runs at stock clocks of 602, and 1296MHz – nowhere near that of the 9800 GTX+. Comparing to the old GTX, which sports 675, 1690, and (the same) 1100MHz speeds respectively, clock speeds have been increased, with testing and the products we’ve received showing that the GTX+ can get up a lot more speed during overclocking.

Other details in the dry technical jargon of microarchitecture features remain pretty much the same as its predecessor, with the same number of shaders (128), the same memory bus width (256-bit), and the same number of transistors (754 million), but all on a much smaller chip. This die-shrink theoretically nets better efficiency directly due to cooler operation and lower power consumption, although Anandtech indicates in its tests that power consumption during idle and load for the GTX+ is actually higher than for a GTX, probably due to the higher clocks on the new GTX+.

So what does all these new details and improvements mean for Nvidia and its buyers? For one, Nvidia now has a product that can take on ATI’s surprisingly effective 4850 in performance, power efficiency, and bleeding-edge tech, but unfortunately not in the price department based on SRP. At launch the 9800 GTX+ was priced at $229 (~P10000), while the regular GTX matched up to the Radeon 4850’s starting price of $199 (~P8700). Savvy shoppers (or well-connected readers) will of course be able to score these cards for a lower price, and a difference of $30 (P1300) shouldn’t matter much to dedicated Nvidia and ATI fans.

The XFX follows Nvidia's reference designfor the 9000, 8000 and GTX200 series.
This is a critical point, as during our testing of the three cards mentioned above, and during comparison to our two previous tests of the a stock 4850 and Sapphire’s overclocked 4850 Toxic, our benchmarks got tossed this way and that, with no clear winner after all of the dust had settled. To wit: while a stock 9800GTX+ loses out to a stock 4850 by the smallest of margins, (3DMark06 delta 20 plus minus for 1260x1024 4xAA 8xAF, and around 220 3DMarks at 1600x1200 and the same settings, (7620 for the 9800GTX+ versus 7846 for the Gecube 4850), tables are turned when the overclocked Zotac 9800GTX+ Zone is added into the mix.

Dual 6-pin power connectors feed the 9800 GTX+.

The rear of the card follows conventional design with two DVI and a single s-video connector.
The two Zotacs were overclocked during testing to simulate the typical tinkerings of a card owner on their new purchases – especially for the Zone edition, which features water-cooling and should prove to be more tolerant of high clocks and the resulting high temperatures. A little bit of disappointment was mixed up with the satisfaction of reaching high clocks while attaining lower temperatures than the regular Zotac and XFX 9800 GTx+’s, as the Zone could only reach another 30MHz clock speed higher than the vanilla 9800 GTX+. While both the Zotac and the XFX were able to attain stable clocks of 830 core, 2059 shader, and 1160 (2320 effective), achieving a 3DMark Vantage score of 5587 @ a temperature of 65 degrees Celcius, the Zone edition reached a core clock of 860, 2133 shader, and 1180 memory (2360 effective) for a Vantage score of 5651 while maintaining a max temperature at load of only 59 degrees Celcius. The Toxic edition Radeon from Sapphire, which is an overclocked version of the 4850, got a Vantage score of 5622 – not quite reaching the level of the water-cooled, overclocked 9800 GTX+.

The contenders.
If all of the numbers above made you a bit confused, the simple conclusion of the testing would be: the two cards from the warring sides of the graphics landscape are dang close when it comes to performance. While the overclocked 9800GTX+ edges out the overclocked 4850 based on 3DMark Vantage scoring, the small difference can almost be attributed to margin of error. The effort (or perhaps fun – depending on the user) of overclocking the 9800GTX+ should also be taken into consideration. The normal XFX 9800 GTX+ (air-cooled and manually overclocked) is neck-and-neck with the Sapphire Toxic 4850, but is literally at the limit of its overclocking capability (higher clocks produced artifacts and hangs in Crysis benchmarking). As it is, the water-cooled Zone is also at the limits, albeit with much cooler (and safer) temperatures, both at idle and at load.

The 9800GTX+ at stock settings for establishing a baseline.
Fans of either company, be it Nvidia or AMD, are all winners in this battle for supremacy. The 9800GTX+ is, as mentioned above, a watershed release that pushes technological boundaries further outwards, at a faster pace, than ever before. So eager PC enthusiasts get their high-tech hardware so much sooner. Further improvements, and other, more capable overclockers than I will be able to push the 9800 GTX+ to performance better than the competition – its potential is yet unexplored – its up to you Nvidia fans out there to discover where the limits are. More pictures to follow.
The 9800GTX+ during benchmarking.