Why is it in the "internet era", everyone seems to think there's always a magic "link" to any and all information?? Maybe it's born from the same thinking that everything should be free on the internet.
Sorry, can't point you to a link - most benchmark attempts are published by magazines and they're not likely to running a high load datafeed/charting/analysis workload.
However, some of us have actual hands on experience. So all I can tell you is what is obvious from having done about 20 years of internals development on a spectrum of OSes and having designed and integrated dozens of very large scale, high performance systems over the years. Take it or leave it - your choice.
However, it's axiomatic that if nitro's workload is routinely running at 85% composite CPU utilization on a dual processor - that a single processor machine isn't going to run that same workload with the same throughput unless you jack up the processor speed comparably - it's not especially complex arithmetic.
Average rules of thumb for CPU intensive application mixes (note, this of course does not apply to running a single, single threaded application):
dual = 1.85-1.9x of a single
quad = 3.6-3.8x of a single
octo = 6.9-7.3x of a single
Since multi-processor configs are more complex, it might be less expensive to buy a faster single processor config (if such a processor exists) - e.g., instead of using dual 1.5 Ghz processors, use a single 2.8-3 GHz processor - but if you needed four 2.4 GHz processors for your workload, you're not going to find a 9-10 GHz single processor anyway.