Hey guys, I spent quite a while looking this up and couldn't really find a straight answer or any real tests. I was going to do benchmarks and post up my results, but after my findings with the 4890 crossfired with a 4870 1GB, I didn't really see a point in doing all the work (you'll see why). Also I am running on 2GB since one of my ram sticks is bad, but I don't think it is holding me back much, if at all.
Anyways, I was able to get a 4890 new for only $10 more than a 4870 so I decided to do it. Some people (I'm not sure if they had any experience with this combination) said it would be clocked down, some said they would run at their own clocks and show improvement. However with my findings on my system they're both wrong, to an extent (C).
Q9550 @ 3.7ghz 1.25v
G-Skill DDR2-1066 2x2GB (using 1 right now)
Abit QuadGT X38 LGA 775
WD 640GB Black (which is going to be returned soon so i'm stuck with a slower 250GB SATA atm)
Sapphire 4890 1GB
Sapphire 4870 1GB
Using 3dmark06 (which I get 21.6k on now compared to my single 4870 1GB, hehe), a stand-still in Farcry 2, and some Crysis benchmarks I was able to conclude that;
A: The clocks on the 4890 were not clocked down to the 4870's.
B: Performance with the 4890 clocked above the 4870 showed no FPS increase (maybe .1-1). I tried all the way up to 925/1100 with the 4870 on its stock of 750/900.
C(extent): After looking at other benchmarks I kind of feel like it is slightly stronger than a 4870x2 or 4870 1GB crossfire. Probably due to the 4890's superior 'tech' other than higher clocks. Also of course in single GPU games I'm quicker. And I guess it could also be newer drivers *shrug.
I'm running my 4870 at 800/975 and leaving my 4890 at its factory stock of 850/975 and am loving it compared to what I had (a single 4870 1GB). I will probably end up selling my 4870 for $100 and getting another 4890, if 5870/50s haven't dropped in price by then (where are you GTX300? =P [yes I know, Q1 '10, if I remember right]).
Thanks,it looks a good review but if you could post some game benchmarks it would be even better.
About "A" I think that's because one is HD 4890 and the other one is HD 4870,however even if they were both the same card,the higher speed model "Should" downclock itself to reach the speeds of the other one,but users (Like mousemonkey)reported that for their cards,the faster model didn't downclock
I paid $250 for it 371 days ago. I probably won't be ready to buy a new 4890 (hurray for holidays) for at least another month or so, I'll send you a PM when the time comes if you are still interested by then.