Intel Core i9 9900K vs. AMD Ryzen 9 3900X Linux Gaming Performance

[torsionbar28]

Most corporations only buy Intel. It doesn't matter how it performs. The admins often hate AMD hardware.

The racks and racks of AMD Opteron servers in my employer's three datacenters beg to differ. So do the US $4M in AMD EPYC servers we just ordered to replace them. All running RHEL.

I would also add that the litany of intel-only hardware flaws discovered in recent years, makes your last sentence even more laughable.

[Spooktra]

I know everyone is riding AMD's jock when it comes to their current 12C/24T offering but people need to keep a level head about them. Those cores are only useful in workloads that can actually use them and I hate to break it to people but most workloads do not scale all that well.

I currently own a i3 7100 with 8GB ddr4 2400 and a 32 GB Optane drive, a R5 1600 with 8 GB ddr4 2400 and a 4790 based Xeon with 16GB of ddr3 1600 and I use this last machine as my pc of choice because in many real life instances it's as fast or faster than the R5.

If I run a pure encoding benchmark, with an uncompressed source, using x264 and x265, the R5 1600 is about 20%-25% faster. However, if I use a high quality compressed source, like a Prores Lossless or JPEG2000, then due to the massive decoding bottlenecks associated with these 2 formats, the difference between the R5 and the Xeon, becomes negligible, thanks to the Xeon's higher clock speed.

If I add filters, for instance a 3 chain filter setup with Saturation, Contrast and Sharpening, then the Xeon is faster because this chain of filters doesn't fully saturate even a 4C/8T cpu, much less a 6C/12T cpu.

If you guys look at these link:

https://techgage.com/news/benchmarking-magix-vegas-pro-17/

https://techgage.com/article/magix-vegas-pro-17-cpu-gpu-performance/

You will see that a 9900K proved twice as fast as a 18-core Core i9-9980XE in one benchmark despite the i9 having all it's cores maxed out, and in any event, it took it 1hr20min to finish a 5 second clip, so if working on a real project, it doesn't matter how fast the system is it will still take days to finish.

Bottom line is that one needs to profile their workload before they spend big bucks on a high end multicore cpu; my next system will probably be a 3400G with 16Gb of ddr 4 3000 and tight timings, coupled with 2 fast NVMEs or if I can hold out I will wait for the Zen2 APUs to come out.

[max0x7ba]

Like some people pointed out: The resolution tested are too high to show how good the CPUs are at gaming as almost all of these tests are gpu bound.
I agree with pinguinpc that 1080p or even 720p should be used if comparing CPU performance.

2560x1440 is the sweet spot nowadays in terms of value metric defined as ((visual-quality * fps) / hardware-cost). So, I couldn't care less about gaming benchmarks at other than 2560x1440 resolution.