CORRECTION: the last paragraph is wrong, where I wrote “SPARC M7 cores are typically 2-3x faster than x86 cores, all the way up to 11x faster”. This is not true. I wrote that paragraph late and was tired. I can not edit so I post an addendum here. The correct version is:

“…And if you really want to talk about core vs core, SPARC M7 cores are typically 1.5 – 2.0x faster than x86 cores. This is proven by looking at all the benchmarks below, where one single 32 core SPARC M7 cpu typically is twice as fast as two E5-2699v3 with 36 cores.

And as we know, x86 cpus and cores are faster than POWER cpus and cores – this must mean that SPARC M7 cores are faster than POWER8 cores, as well. Here are several benchmarks for this new SPARC S7 sonoma scale-out cpu, compared to x86 and POWER8. And if you dig a bit, you can also find 30ish benchmarks of SPARC M7 vs x86 or POWER8…”
https://blogs.oracle.com/BestPerf/entry/20160629_jbb_sparc_s7_2

For raw floating point performance, SPARC M7 holds the world record with 832 SPECfp2006, vs IBM POWER8 reaching 468 SPECfp2006. Of course, x86 reaches 474 SPECfp2006 making IBM POWER8 the slowest in class.
https://blogs.oracle.com/BestPerf/entry/201510_specpu2006_t7_1

Here are 40ish different world records SPARC M7 holds, other than SQL benchmarks. For instance, neural network, big data, SPECcpu 2006, graph traversal, STREAM ram bandwidth, etc etc etc. And regarding business enterprise software such as SAP, databases, virtualization, SPARC M7 crushes, being up to 11x faster than the fastest x86 or POWER8 cpu.

I dont see that SPARC is only fit for running SQL benchmarks? It is fastest in every single benchmark I have ever seen. I have never seen SPARC M7 being slower in any benchmark. Typically it is 2-3x faster, all the way up to 11x faster.

When we talk about the best performing cpu, the only interesting thing is the performance of the… cpu. Not core. In this case, who has the fastest core does not matter. Neither does it matter who has the lowest latency, fastest ALU, etc – or other tiny parts of a cpu. We are comparing cpus here: POWER vs SPARC vs x86. You can not extrapolate from benchmarking a small part of the cpu, to conclude which cpu is fastest. For instance, SPARC M7 has 32 cores, and if you bench one core, you are benching 1 / 32 = 3% of the SPARC M7, and from that tiny number you are trying to reach a conclusion about the whole SPARC M7 cpu. You can not draw a conclusion by looking at less than 3% of the SPARC M7 cpu. That is wrong to do.

Which car is fastest? Corvette, because “it has faster pistons in the engine”. Never mind the top speed of Corvette is 240 km/h, and Lamborghini has 320 km/h. Corvette will still say “our car is the fastest, because our piston is fastest on the market”. This is weird and faulty logic. Some would even call it FUD. If Lamborghini runs 320km/h the car is faster. No matter how fast the Corvette pistons move. This is exactly the situation with IBM POWER. IBM claims their POWER cores are faster so the entire POWER cpu is faster – never mind SPARC cpu scores higher in the benchmarks. Here is an example:
https://ibmadvantage.com/2013/04/29/weblogic-12c-on-oracle-sparc-t5-8-delivers-half-the-transactions-per-core-at-double-the-cost-of-the-websphere-on-ibm-power7/
“…[Oracle SPARC] being “fastest processor in the world” means that such processor must be able to handle the most transactions per second [b]per processor core[/b]……..IBM POWER produced the world record result in terms of EjOPS per processor core – truly a measure of the fastest processor known to men……Since Oracle knew they can not produce the most efficient result in terms of cost or transactions per second, the only way for them to claim world record was to throw large hardware at it and produce the biggest total number of EjOPS. Not a very useful metric I must admit….”

.

And if you really want to talk about core vs core, SPARC M7 cores are typically 2-3x faster than x86 cores, all the way up to 11x faster. And as we know, x86 cpus and cores are faster than POWER cpus and cores – this must mean that SPARC M7 cores are faster as well. Here are several benchmarks for this new SPARC S7 sonoma scale-out cpu, compared to x86 and POWER8.
https://blogs.oracle.com/BestPerf/entry/20160629_jbb_sparc_s7_2

And if you really need maximum raw floating point crunching power, the SPARC M7 is much faster than both the x86 and the POWER8. SPARC M7 is basically four times faster than this SPARC S7 sonoma cpu. SPECint2006 benchmarks can be found in the same link above.

Performance is of course somewhere in between similar and few times better, than best 22(24) core Xeon E5/E7 V4 processors (depends on benchmark) …at least according to Oracle data.

Plus S4 core has big advantage in accelerated encryption, where can run fully encrypted system with minimal performance penalty.

But of course, in other situations, where depends on theoretical raw floating point performance SPARC does not make any sense. In this situations will be better choice Xeon E5 or Xeon Phi (or GPU).

No worries. Sparc machines have several different kinds of virtualization technologies, but VMware is, with the exception of some stuff they did for mobile phones, restricted to the X86 instruction set. Earlier Sparc systems have dynamic domains, a kind of hardware partition that was reconfigurable on reboot. Solaris, of course, supported containers (often called zones), but these are really a type 2 hypervisor and not a container lock is used with Docker or LXC on Linux. (Heavier, with a full shared file system and kernel, but distinct Solaris runtimes that magically looked like full operating systems to the applications. Sparc T series and I presume S series processors also support LDoms or logical domains, which are akin to VMware ESXi virtual machines but are in no way compatible with ESXi.