Looking for CPU benchmarks

I’m on the hunt for some actual CPU benchmarks—specifically focusing on multi-threaded performance—that demonstrate how these processors handle various real-world workloads by providing results in seconds rather than just abstract numbers.

It seems like every single benchmark I stumble upon online just spits out those generic, arbitrary scores that honestly don't mean a thing to me in any practical sense.

http://www.tomshardware.com/charts/c...marks,187.html
http://anandtech.com/bench/CPU/39

Your points are being compared against everyone else's. 😉 Besides, just like you already figured out, it wouldn't make any sense anyway.

What’s your end goal here? The real headache is that finding solid CPU benchmarks for actual professional software—stuff people truly rely on—can be incredibly difficult, if not impossible. Meanwhile, you can find dozens of benchmarks for every trivial little game out there...

That said, synthetic tests aren't entirely useless. For instance, running Cinebench multithreading would give you a pretty accurate look at how your system handles a heavy video rendering hierarchy.

We’re talking about an application here that handles some pretty heavy-duty statistical computations involving massive datasets. On my current processor—an Intel Core i7-5930K—a single calculation can drag on anywhere from a single minute to several dozen minutes, even when I'm running it at low precision. On average, though, we're looking at a few minutes per run. For instance, when I'm running this specific simulation to benchmark speed without any overclocking, it usually clocks in at about seven minutes; once I bump it up to 4.1 GHz, it drops to six. Ideally, I’d love to see that same calculation crushed down to a minute or even less. But, considering I'm already rocking a top-of-the-line i7, it looks like the only way to see that kind of performance leap would be to jump to a dual-CPU setup, maybe something like two Intel Xeon processors with 16 cores each, which just doesn't make financial sense for me right now. The only other avenue I've looked into that might actually be cost-effective would be an Intel Xeon Phi solution, but even that is on hold for a few months since I'm expecting some algorithmic improvements in the software itself.

Oof, you’d be looking at something about five times faster than a 5930K... I mean, if it's for work, sure, but you're probably right about those dual 16-core Intel Xeon setups—that's really the only way to get the performance you actually intended.

The real question is whether saving roughly five minutes per render will actually offset that massive upfront cost anytime soon.

By the way, I'm curious why anyone would grab a 5930K? For $667 less, you could just go with a 5820K (it's still 6 cores, just a lower clock speed, though it's unlocked anyway), or if you want to spend $1000 more, you've got the 5960X (8 cores/16 threads, also unlocked).

When I originally pulled the trigger on this one, it stood out from the other three options as having the absolute best bang for my buck; I had my eye on the 5960X initially, but I quickly realized that for what I actually needed at the time, it would have been a complete waste of money, especially since I had no idea I’d eventually require something this much more powerful. And as for the difference in processing speeds... we're talking five times faster... which is just massive 😁
I can simply feed the software a hundred tasks to churn through overnight, and if a single task takes about five minutes, you're looking at roughly eight hours of work. But once you divide that by five, suddenly you've got everything wrapped up in an hour and a half.