Just to weigh in on the GTX 275 numbers. Quickly canvassing reviews from Xbit, Anand, Tech Power Up and Hexus the breakdown looks like:
1680............58 benchmarks ( GTX275 = 38, HD 5830 = 18, tied = 2)
1920............65 benchmarks (GTX 275 = 45, HD 5830 = 18, tied = 2)
2560............29 benchmarks (GTX 275 = 23, HD 5830 = 6)
Locally at least the GTX 275 is still in stock-
$NZ399 ($US278). The HD 5830 isn't listed here, although as a rough pricing guide the cheapest HD 5850 is $US341, and the HD5770 $US195.
I've checked through quite a few online reviews of this card as it is a likely upgrade for many of my customers once the HD 4890 and GTX 275/260 OC parts run down. As the card is a salvage part of the HD 5850 (which is itself a salvage part of the HD 5870) I'm more interested in the uniformity of these cards regarding the defects the dies carry and how wide a discrepency the fusing off has affected the cards concerned. Are all HD 5830's created equal you might say. Given that testing setups and regimes will differ from site to site, probably the best indicator of relative uniformity within the cards would be power draw. Comparing the power draw to the HD 5850 it seems that there is quite a large variable in how efficient these cards are:
[H]OCP for instance had power consumption at 20 watts less than the HD 5850, while Tweaktown tested the HD 5830 at 36 watts more than the HD 5850 * -quite a large variation
considering both cards were AMD/ATI "reference" cards presumably both assembled by PC Partner (AMD's reference card builder and parent company of Sapphire) both cards could be considered "equal"-this could lead to wild variations in standard and overclocked performance and heat output/temps in addition to power draw if repeated in production.
Anyway, that's my 2c on the matter, for what it's worth.
* Other sites power draw HD 5830/5850 analysis :Xbit recorded identical power draw, bit-tech, Hexus, TechPowerUp and Rage 3d at 1-5 watts more for the 5830, Firing Squad at 12w and Anand at 15w more for the HD 5830 respectively.