As I said, unlike me who got bored and went for a X2 a couple days ago, I would NOT buy anything for a few weeks, wait for more reviews, and wait for internet reports from the usual suspect sites, xtreme and others. Different memory brands will almost always have different timings as they try to offer better products than the others. They might be significant or not but often a controller has been tuned/designed for a specific flow of data and slight timing differences can have a massive effect. The timings aren't widely known, sometimes ati tool or the like will be updated to be able to read them, sometimes the data specs on the ram chips will be found online somewhere and timings found, though that doesn't really specifically identify what they have to be run at, though they often won't be out of range of design specs.
But the point is, in 3 weeks we might all find out that all new 512mb cards have the newer faster memory chips, or an incredibly slight tweak to the core mem controller, or a new pcb layout which gives identical performance to the "new" 1gb cards and find the 512mb cards are coming in £30-40 cheaper than the 1gb for identical performance. Its as simple as that.
As I've now said repeatedly, a good review site would have commented on what I have here, they would have mentioned the performance increase across resolutions not including those limited by frame buffer, but they ignored all of it, completely. its a very bare review that doesn't draw any deep conclusions from the numbers they themselves freaking got. The conclusion ignores half of their own findings, so it IS a bad review no matter if you like that or not.
The other "evidence" of this is fairly simple. You can make an X2 card run as one gpu, you can run it with just 1gb of memory. It does NOT perform the same as this "new" 4780 1GB card. The X2 doesn't show that same performance increase over the normal 4870, so the 1GB itself did NOT increase performance in the same way across all resolutions. WHich yet again is another indication that the 1GB is not at all responsible for the card being better, in anything except Grid at a resolution less than 1% of the forum can use.
hold your horses, wait a month and maybe save £40, and maybe £80 on a CF setup if we find out the 512mb's can be just as fast with new mem, new bios, new settings or something else.
PS the 512mb vs 256mb links you put up aren't identically clocked cards on either core or memory so you wouldn't expect more "inline" performance anyway, but when the resolution/memory size becomes an issue there is a large drop in performance. For instance Crysis even at low res is crap on the 256, and at medium res even the 512mb card drops down and is over run. Quake wars, without AA you get the exactly as assumed massive drop in performance, then with 4xaa both cards are running low on framebuffer throughout, but due to AA relying heavily on memory speed the 512mb with the faster memory (1/8ghz vs 1.6Ghz) remains a similar way ahead throughout.
. Arguments both ways. Also read the anandtech article fully.
