September 13, 2010, Nvidia announced a new product in its Fermi-based line-up. It was the GF106 graphics processor and the GeForce GTS 450 graphics card based on it. Priced at $129, the new card was supposed to make Nvidia competitive in the entry-level segment. We discussed it in our review called Fermi Lite: Nvidia GeForce GTS 450 and were quite satisfied with both the new GPU and the new graphics card. The GeForce GTS 450 proved to be as fast as the Radeon HD 5770 and would even occasionally win some tests against the Radeon HD 5830.
Nvidia still had one gap in its product line-up, though. It was in the below-$100 category which, according to Nvidia’s own data, accounts for about 19% of the total graphics card market. Nvidia didn’t have competitive solutions there because its GT215, GT216 and GT218-based cards traced its origin back to the old G80/G92 chips and didn’t support modern technologies like DirectX 11 and Protected Audio Path. The GeForce GT 220 and GT 240 did their job well enough in their time, but had become quite outdated by the end of 2010, especially as AMD offered a full range of competing DirectX 11 compatibles. So, Nvidia needed a new product in that segment and they rolled it out officially on October 11, 2010. It is the GF108 processor, the junior model in the Fermi series.
The GF108 is the simplest new-generation solution developed by Nvidia. Just take a look at its flowchart:
Like its senior cousin GF106, the new GF108 contains but one graphics processing cluster (GPC). It incorporates only two stream multiprocessors with 48 ALUs in each. Thus, the GF108 has a total of 96 ALUs, just like the GT215 chip which used to be installed on entry-level GeForce GT 240 cards. Thanks to the polished-off 40nm tech process, the GF108 has a higher frequency, but it has overall become less complex than the GT215. Why? Let's check out the specifications of the GeForce GT 430 card.
Alas, the new card has only 16 texture-mapping units and 4 raster back-ends, which is too few even compared to the Radeon HD 5570. Despite the increased clock rates and better computing resources, we cannot expect the GF108-based solution to be superior to the GeForce GT 240. On the other hand, we must keep it in mind that $100 and cheaper graphics cards are not designed and bought for serious gaming. Their purpose is different.
They are meant to serve instead of integrated graphics cores and endow the computer with a full set of multimedia capabilities such as HD video support. They may also be used for running simple games at low resolutions. From this standpoint, the GF108 is blameless. It has inherited the best traits of the GF106 and GF104, including their comprehensive support for multichannel HD audio formats Dolby TrueHD and DTS-HD Master Audio by means of Protected Audio Path. Power consumption being an important factor for HTPCs, the GF108 is an improvement over the GT215 in this respect, too. Nvidia says its TDP is only 49 watts (the Radeon HD 5570 consumes even less, though).
Officially priced at $79, the Nvidia GeForce GT 430 completes the Fermi series and puts an end to the outdated G80 architecture. We can’t expect it to break any records in benchmarks, but we’ve got an interesting version of it from Gigabyte. So, we are going to compare the Gigabyte GV-N430OC-1GL with a couple of Radeon HD 5500 products that feature GDDR5 memory.