It depends on the architecture, I believe.
Personally, I think these cards are close enough other features that I'd go for the one with the higher memory.
From 772Mhz to 816Mhz core clock speed there's just not a whole lot of gain, while the memory is double.
UPDATE for comment:
Because the stats are layed out differently for the different cards, it took me a bit to find the different clock speeds.
Most modern GPUs, because they are purpose built for the types of calculations in graphics applications, do not need to be nearly as high frequency (Mhz, Ghz) as desktop CPUs, which are general purpose and thus not quite so "good" at one particular thing. So, the 4008Mhz right away could not have been the GPU core speed.
The 4008Mhz (effective) rate is the memory frequency. This is a bit of a misnomer though. GDDR5 is a very high bandwidth DDR (Double-Data-Rate) memory architecture that, due to it's capability of handling two instructions per clock cycle plus a wider bus is labeled or advertised at four times it's actual core speed. I may be off here. Someone please correct me if I'm wrong.
Thus, 4008Mhz is really only a more pedestrian 1Ghz with multitasking, essentially, which isn't all that great when it comes to GDDR. The fastest GDDR currently clocks in at more /than 7Ghz effective (1.75Ghz real).
You'll find the specs on the ASUS Matrix card show this:
In the specifications tab of the product page, the Memory clock is "4008Mhz (1002Mhz GDDR5)".
The EVGA only notes the first value, which is identical at 4008Mhz and adds the label "(effective)" to clarify.
In looking for the GPU speed, look for terms like "Core clock", "GPU", "Engine clock", and so forth. These will indicate the speed at which the GPU on the cards run.
In this case, the EVGA product says "772Mhz GPU" and the ASUS "Engine Clock 816Mhz".
Standardization would be great, but so long as you're careful and know what to look for it's not too hard to find out which numbers ought to be compared.