IMO it's too much in both cases - a 4.8 GHz 2600k + single 7970 scores more than a stock 2600k + CF 7970s in cloud gate, and while the graphics score scales perfectly with CF, in Fire Strike the overall score scaling is about 68% in Fire Strike and 80% in Fire Strike Extreme.
I think the cloud gate result is a bit silly but I understand it, though I think Fire Strike should be even more heavily GPU weighted as it's the only test that's actually demanding on the GPU.
Ice Storm is designed for tablets, phones and very very old DX9 cards. Unsurprisingly it is a CPU test on modern high end desktops.
Cloud Gate is designed for old DX10-only cards, integrated graphics, APUs and low end budget cards (think GT 520, HD 6450 and such). Unsurprisingly it too will rapidly turn into a CPU test on high end setups.
Use the correct test depending on your hardware. For modern gaming PC, that is Fire Strike. And look; it isn't purely a CPU test on such hardware
- now CPU still matters, because it matters also in games, but it is weighted towards GPU. And you get Graphics and Physics scores to compare raw performance in those categories.
The performance range that 3DMark tries to cover is just so massive that the only way to do it is to have separate benchmarks for each device category. You can run the lightweight tests also on high end systems but after certain point the framerates are just so high that the CPU is the only thing that matters there - you can still compare, but you have to realize that such comparison is basically "yeah, this modern system is WAY faster, so much faster that the measurement accuracy is harmed by it".
Also the cloudgate CF vs no CF is showing the inefficiencies of Crossfire in driver code (eats more CPU).
Also those having issues with AMD cards - any driver prior to 13.2 beta 3 has known issues (does not render everything, can crash). Use latest betas or wait for next set of WHQLs.
.
?
