Intel Announces Early 8th Gen Core Processors, Coffee Lake

[starshipeleven]

The -only- way benchmarking a game is valid as a real world benchmark is when it's benched at the settings you use. In all other circumstances they are too complex to isolate specific bottlenecks and -cannot- be used as synthetics.

You are confusing troubleshooting and with pure hardware benchmarking here.

[Spooktra]

No patches are needed, just install the appropriate drivers and compile with the appropriate flags, I do it all the time, choose here's my config for my current ffmpeg build:

--enable-nonfree --enable-gpl --enable-version3 --enable-libass --enable-libmp3lame --enable-libopus --enable-libfdk-aac --enable-libvpx --cpu=native --enable-vaapi --enable-vdpau --enable-libx264 --enable-libx265 --prefix=/apps/ffmpeg/git

You can also enable nvenc functionality by adding --enable-nvenc to the above parameters and enabling cuda is also failry simple.

If you need a compile guide, let me know.

Thanks.

[duby229]

You are confusing troubleshooting and with pure hardware benchmarking here.

Troubleshooting what? The trouble with that benchmark showed is it represents results -nobody- is going to see. Ever.

[starshipeleven]

Troubleshooting what?

Finding bottlenecks is troubleshooting.

The trouble with that benchmark showed is it represents results -nobody- is going to see. Ever.

It shows how the hardware has improved over time, which is what most synthetic benchmarks are for.

It's not showing how well the game runs, but how well the hardware runs vs the older hardware, since both ran the same test.

[duby229]

Finding bottlenecks is troubleshooting.
It shows how the hardware has improved over time, which is what most synthetic benchmarks are for.

It's not showing how well the game runs, but how well the hardware runs vs the older hardware, since both ran the same test.

It absolutely does not show improvement over time (or any bottlenecks at all). First of all you have to disable most of the games functionality and then reduce data to process by lowering resolution. (You have to increase data to process in order to hit most types of bottlenecks, not reduce them) It indicates a CPU's internal IO scalability with no provable facts at all. (And even then -only- when compared against real world settings on the same hardware) At best it -un-provably indicates- that AMD's architecture scales internal IO better than Intel's. Because games are so complex you cannot isolate what, why or how. You need specifically designed synthetics to do that. The best you can do is make un-provable inferences with it.

[Sidicas]

It absolutely does not show improvement over time (or any bottlenecks at all). First of all you have to disable most of the games functionality and then reduce data to process by lowering resolution. (You have to increase data to process in order to hit most types of bottlenecks, not reduce them) It indicates a CPU's internal IO scalability with no provable facts at all. (And even then -only- when compared against real world settings on the same hardware) At best it -un-provably indicates- that AMD's architecture scales internal IO better than Intel's. Because games are so complex you cannot isolate what, why or how. You need specifically designed synthetics to do that. The best you can do is make un-provable inferences with it.

Agreed. Been known in gaming communities for over a decade now that low resolution benchmarks are nothing more than RAM benchmarks. Where RAM performance is never a bottleneck in a gaming pc at normal and higher graphics settings.

Looking forward to the 4k and 1080p results.

[GI_Jack]

4k is at least more relevant than 720p. Please show me a desktop system with a 720p screen nowadays. Many sites test at 720 to show the world something about the cpu that nobody will experience in their day to day work / game (besides laptop-gamers, but those are different cpus)

Here is the thing. Many people, if not most people don't view video or web pages on a full screen. 720p is great for general internet video to save bandwith if people view it in a window anyway.

[duby229]

Agreed. Been known in gaming communities for over a decade now that low resolution benchmarks are nothing more than RAM benchmarks. Where RAM performance is never a bottleneck in a gaming pc at normal and higher graphics settings.

Looking forward to the 4k and 1080p results.

It's not even possible to say that even, because games are not purpose built RAM benchmarks. You can un-provably infer that at low settings and resolutions games mostly hit RAM, but with a single bar on a graph representing overall performance there is no data proving that. If you want to benchmark RAM then you are far better off using a synthetic RAM benchmark. Those results would be provable at least.

[GreatEmerald]

No patches are needed, just install the appropriate drivers and compile with the appropriate flags, I do it all the time, choose here's my config for my current ffmpeg build:

--enable-nonfree --enable-gpl --enable-version3 --enable-libass --enable-libmp3lame --enable-libopus --enable-libfdk-aac --enable-libvpx --cpu=native --enable-vaapi --enable-vdpau --enable-libx264 --enable-libx265 --prefix=/apps/ffmpeg/git

You can also enable nvenc functionality by adding --enable-nvenc to the above parameters and enabling cuda is also failry simple.

If you need a compile guide, let me know.

Thanks.

It's set like that by default in openSUSE (Packman). Yesterday I got both NVENC and QuickSync working that way.

[gururise]

Please include Price/Performance graphs in next reviews would be very helpful!