There aren't "benchmark values" for ASE on different platforms because there isn't a "benchmark" workload to measure. Unless of course you consider the TPC (http://www.tpc.org) set of tests and other similar exercises. But those aren't real workloads, rather workloads used to measure configurations relative to each other.
There are far too many factors that go into a hardware platform decision that aren't mentioned here. How many connections, engines (threads), memory, database size, workload profile etc. Even the fact whether the box is dedicated to ASE, or running other apps as well. Virtualization comes into play, network topology (10G nics/switches, or multiple bonded 1G nics) etc.
In the past, storage performance was the dominant factor in the "speed" of a system. Mostly because physical io's were so dominant in the workload. With larger memory configurations though, other limitations such as spinlocks, and lock contention are more common challenges.
Also, with multi-processor, multi-core, multi-threaded processing, context switching and cpu/core binding come into play sometimes as well.
I have customer now who chose a T4-2 box, with 2 chips, 16 cores, 8 threads per core to run a 16-engine (16-thread) ASE on. That's 128 logical CPU's all to run a 16 engine ASE. It hosts ASE as well as the application. While this isn't the best setup for ASE, much can be done by a competent DBA, and Unix team to separate processing and contention to make it work well.
And that's probably my point is that most of the obvious platform choices (Oracle Solaris, IBM AIX, HP-UX, and to some degree the comercial Linux choices) can all be equally capable solutions. You have to know how to manage OS resources, kernel tuning specifics, driver configurations, ASE tuning, etc all specific to your workload requirements. Not even to mention what you expect from your hardware vendor for support. Many can argue and offer opinions and experiences from one vendor to another about that aspect (me included).