Hyper-threading: is it worth it?

[andrewk89]

Description of problem or question
I have an Intel Core i5. Feeling a lot of processor envy. Wondering if I swap it out for a Core i7, would hyper-threading cut the encode time in half?

Typical encodes use AVC x.264 / HEVC x.265 and libav AAC / FDK-AAC. I sometimes use decomb, denoise, deblock, too. Encodes on the Core i5 load all 4 cores to 99%. So faster encode times come down to faster clock speeds or more cores (real or fake, I mean, virtual).

This pointer to anandtech seems to suggest hyperthreading gain is only 30% faster, not 2x faster viewtopic.php?f=6&t=26788&p=123076&hili ... ng#p123076

I was wondering if that is still accurate with the HB 1.0.7 builds or if anyone has different experience.

Steps to reproduce the problem (If Applicable)


HandBrake version (e.g., 1.0.0)
1.0.7

Operating system and version (e.g., Ubuntu 16.04 LTS, macOS 10.3 High Sierra, Windows 10 Creators Update)
Windows 10 x64 Fall Creator's

Error message text or screenshot (If Applicable)

Provide a URL from any image hosting site.

HandBrake Activity Log ***required*** (see How-to get an activity log)

Use any pastebin website and provide the URL it generates
- or -

Code: Select all

Paste log text between these markers.
[code][/i]

[rollin_eng]

It’s pretty much still the same, so hyper threading will get you some advantage but it also depends on your settings.

[s55]

This pointer to anandtech seems to suggest hyperthreading gain is only 30% faster, not 2x faster

That's about right and not necessarily even guaranteed at that.

Put it this way, unless your going to jump 3~4 generations of processors at the same time, or are getting an i7 for free, I wouldn't bother.

[rollin_eng]

And I would probably wait until all this Meltdown/Spectre stuff is sorted out.

[mduell]

For some things HT is a big advantage like 30%. For video encoding it's a much smaller percentage gain.

[andrewk89]

I thought about that. One estimate is chips designs are 3-5 years away which don't have those vulnerabilities. So buying new right now isn't any more or less secure than what I'm running today.

And I would probably wait until all this Meltdown/Spectre stuff is sorted out.

[andrewk89]

Ivy Bridge (3xxx) to Coffee Lake (8xxx) would be that jump. CPU frequencies aren't much higher after 5 generations (25%?). Thinking about getting an i5-8xxx when they go mainstream later this year, but wondering if the extra money for an i7 is well spent or money-down-the-drain.

On the flip side, Coffee Lake i5 mainstream desktop will be 6 core, so that should be a 50% improvement straight-up.?.?.?.?.

Put it this way, unless your going to jump 3~4 generations of processors at the same time, or are getting an i7 for free, I wouldn't bother.

[rollin_eng]

I thought about that. One estimate is chips designs are 3-5 years away which don't have those vulnerabilities. So buying new right now isn't any more or less secure than what I'm running today.

And I would probably wait until all this Meltdown/Spectre stuff is sorted out.

Yeah, but there might be better offers 6-12 months from now if Intel sales fall, or rebates from class actions. Who knows?

[mduell]

I thought about that. One estimate is chips designs are 3-5 years away which don't have those vulnerabilities.

Intel says end of year.