I don't know if the problem is the "Old" 2010 5,1 Mac Pro with PCIe 2.0 vs the card's PCIe 3.0 but until Apple decides to upgrade the Mac Pro Line, $500 for a little faster video card doesn't work for me.
I ran some Xbench and Geebench tests and the numbers are almost identical. Even some video tests are actually faster on the 3 year old 5770 with just 1GB.
Your tests mirror MysticalOS's tests in that some were worse on the new card than the old card. It seems that the new card's drivers, while "good", aren't all that optimized yet. Performance in games on anything but the higher clocked CPU Mac Pro 4,1 and 5,1 results in performance that's worse than the GTX 680, which is already worse on those machines (i.e. not 4,1/5,1) than the 5870 is.
So right now this card is a mixed bag, especially with the inability to boot into 32-bit mode despite the drivers having 32-bit code in them and EBC in the Mac-specific ROM on the card.
My Mac Pro has the fastest SSD in the market, the 1TB OWC accelsior
There's a "catch" to that "fastest SSD" bit - I'll lay them out for you (though it is decently fast for what it's intended to do)
1) It's not a single SSD. It's two SSDs, each with its own controller, in a
RAID. Stick two Vertex 4s in a RAID on any SATA3/SAS2 HBA and it'll smoke that Accelsior. I know - I have just such a setup of 2 x 256 GB Vertex 4s on the NewerTech 6G-1i1e card (basically a rebadged RocketRAID 2721 with special Legacy passthrough mode).
Even on my Mac Pro 1,1 I can reach 1 GB/sec on
incompressible data. Good luck with that on the Accelsior since it's using SandForce controllers, which are absolutely
craptacular with incompressible data (they drop to about half their rated speeds in that scenario).
As a plus though, the Accelsior is bootable, and even on a 1,1 Mac Pro's PCIe 1.0 slot it'd smoke the internal SATA bays any day.
2) It's a PCIe 2.0 card, but has a flat ceiling of 500 MB /sec max throughput when encountering incompressible data. That means that even if OWC gets better SSD blades that when combined can reach >1 GB/sec, that card will never get those speeds. PCIe 2.0 link width maximums are 500MB/sec, and that card's connector is an x1 link width. All of its speed comes from
compressible data. That works for a boot drive, but not so much for games, whose data is almost always incompressible (as in it is already in a compressed format).
Barefeats had the Accelsior sitting at the top of its charts in most tests, but that was with compressible data in a synthetic environment. It loses nearly all its advantages when dealing with incompressible data in the real world, and would handily lose to an HBA whose link width is at least x4 and two good SSDs like the Vertex 4 (remember, the Accelsior is a RAIDed pair, so comparing it to a single SSD is apples to oranges and only serves to skew the data in favor of the Accelsior, which becomes obvious when comparing to a
like setup of real world SSDs).
It's bootable though, which is nice.
I'll be impressed once OWC moves away from SandForce controllers
and has something with better than x1 link width for bandwidth.