Core i3 vs. Core i5 Performance Impact On OpenGL/Vulkan Linux Gaming

[Bram]

The case where the i3 is beating the i5 cannot be explained by the minute difference in base-frequency.
The performance lead is much larger than the small increase in base-clock.

What is happening here, could very well be a case where an attempt to accelerate with multi-threading, has back-fired.
And instead of higher performance on 4 threads over 2 threads, the performance is actually lower.

Would it in any be possible to disable two cores on the i5? If you did, I expect the i3 performance lead over the i5 to disappear in that particular test.

[DanL]

Michael, do Kaby Lake chips significantly improve Radeon gaming performance relative to older generations like Sandy and Ivy Bridge?

I'm not an OC'er or very much of a gamer, but I've seen plenty of them that are satisfied with their quad-core Sandy/Ivy CPU's and didn't feel compelled to upgrade to Haswell or Skylake.

[tg--]

These results aren't terribly surprising.
The 2 benchmarked titles are based on engines that actually have a ancient core (UE3 from 2006 and Source from 2004 and earlier) and really haven't taken well to modern hardware.
This was the time when dualcores started to appear and it took both engines many years to actually make use of it.

Source to this day cannot utilize more then 2 cores properly, so it just won't scale to quadcores and above.

In the end this means what you're testing is basically the IPC of up to 2 cores; so it is not at all surprising that the i3 with similar IPC will have the same result as a i5 with double the cores and the same IPC/core. Had the i3 run at 4.2 GHz, the i5 wouldn't have seen any gains.

Also not that surprising are side effects of "turbo boost". There are many scenarios where the turbo features are counter-productive and a temporarily over-saturated GPU might well cause the CPU clock and power-management controls to oscillate and cause performance drops.
When the GPU is in overload, the CPU has nothing to do for a while and will reduce clocks and possibly disable internal components (like caches) to save power. If the GPU is then done, the CPU-heavy engines like Source will suddenly demand a lot of work from the CPU until the next GPU overload.

[cRaZy-bisCuiT]

I would be interested to see Benchmarks in CPU limited games like City: Skylines and Civilization V and so on...

[Passso]

I remember that such a benchmark was run some years ago by some Windows games benchmarking website. An overclocked single-core CPU is quite competitive in games.

You know, Pentium have 2 cores since years now
Atm my i5 cannot beat my Pentium G3258 @ 4Ghz in 90% of games...

[ResponseWriter]

That Furry that you are so hot for*

* And by "hot", I'm talking about the almost irresponsible power consumption levels for a rather pedestrian level of performance.

Oh.

[cj.wijtmans]

You know, Pentium have 2 cores since years now
Atm my i5 cannot beat my Pentium G3258 @ 4Ghz in 90% of games...

that is because they arent programmed well to do work loads in a scalable fashion.

[Passso]

that is because they arent programmed well to do work loads in a scalable fashion.

that is because games arent programmed well to do work loads in a scalable fashion.