One of my friends wants a video card because he wants to play Battlefield 3 in high setting with a resolution of 1280*1024. Any sugestion?
You could also check out the 7850 as well the 1GB is about $180 the 2GB is about $210 and that card when overcloocked can perform around the level of a 570 or 580. http://www.newegg.com/Product/Product.aspx?Item=N82E16814102999&name=Desktop-Graphics-Cards And here's proof he could max BF3 out at his res. on this card http://www.guru3d.com/articles_pages/amd_radeon_hd_7850_and_7870_review,21.html
HD7870 are great cards and often can be found at $200. but ppl should try avoid HD7xxx cards for BF3 because of the black screen sound loop problem.
Never underestimate the used market. Going a gen or two older will bag you a cracker of a card for a bargain price. Though having said that, 680's seem to be really low atm on ebay, at least where I am. Anyway, my method for buying a good card on a budget is to check benchmarks, the Passmark ones are generally a good rule of thumb : http://www.videocardbenchmark.net/high_end_gpus.html , then cross-reference that with prices for cards on Ebay etc. and see exactly how much bang you can get for your buck. It seems like a lot of work, but its well worth it when you get your ideal card.
PassMark is NOT a good way to determine performance. It should blacklisted/censored here and in any other hardware forum. It's bullsh!t. Use proper real-world benchmarks.
The reason that occurs is because the benchmarks are user submitted. Any person with half a mind will know a 690 is OBVIOUSLY better than a 680, but the nature of user submitted scores rely a lot upon numbers of scores submitted (a bad 690 will affect the group a lot more than a bad 680, because there are less people submitting 690 scores). Passmark is good because it provides a comprehensive list of pretty much every GPU, and a realistic view of how it performs (more or less) Remember, a lot of review samples used in other website's benchmarks will be silicon cherrypicked (sadly), or will have small samples numbers, and therefore will be more affected by the silicon lottery. *EDIT* Sorry if I'm not being clear, I've done an all-nighter and I'm like 80% asleep.
Yeah, you should go to sleep. Synthetic benchmarks isn't an accurate way to determine real-world performance. Especially if they're as crappy as PassMark (1, 2). 3DMark has some better tests, but they're not accurate either. It's simply not enough to run one or few types of benchmarks. More different tests means more accurate results. Handpicked GPUs don't even affect performance. Only overclocking potential. And most samples are not handpicked. This can be seen by usually low overclocking results and higher end-user results. If you want to know how video card actually performs in games, there are plenty of HW websites. It's just that some are more comprehensive than others. TPU, for example, runs 20 benchmarks (18 games and 2 synthetic). This makes their performance charts more accurate than most others.
Sapphire 7870 for $200 http://www.newegg.com/Product/Product.aspx?Item=N82E16814102983. Good deal at that price, plus with the new 12.11 you get a performance boost and they overclock well too!