you mean 1280x800

I think he tried this but the load of scaling was greater than the reduction in compression

Is it true that you need a special software in a cd to boot from it and do the calibration ?

@@eduardocarrasco8830 yes, but it is easy to get on internet

I'd even go ahead and replace the old crusty fans with some Noctuas or something with a little less whine

@@RobotnikPlays not possible, because apple's fan pins are different from pc's

It's highly unlikely such a GPU with PowerPC Mac support exists.

Can't you do any amd card with the Foss drivers?

@@golemnety Power Mac cards had a different bios to the Intel Macs, I doubt we'd get past the POST screen for a driver to work

@@golemnety I'd be really surprised if any of those modern drivers are included in the PPC tree, but I guess it's possible.

@@raspberry1440kb don't forget he's running Linux. The BIOS might not even need to touch the card.

Same

You know you feel like a scumbag when you have 2 gaming PC one with 16gb and the other with 32gb also one has a 1060 3gb and the other has a R9 390 8gb. Yes I will be selling one of them but Wich one? Hmm?

It's not that big of a deal, modern PowerPC (since renamed OpenPOWER) can get up to 2TB DDR4 and was the first to support PCIe 4.0.

@@BringMayFlowers why is there modern power PC? What's the application for using it vs x86 or arm?

@@bland9876 Mostly for datacenters, supercomputers, and servers, but there's also a niche in libre computing because the desktop computers using it have entirely open-source firmware, no Intel Management Engine analogue, and the CPUs themselves are manufactured in Europe. Plus, owing to their main uses, they're stupid powerful, with four threads per core, a more efficient ISA than amd64 but more powerful than most ARM chips, and can handle ridiculous specs no issue. It's also an open ISA, you don't need to license it from the OpenPOWER Foundation, IBM, or NXP to make your own.

Might the distributed threads actually work faster than the ultra-fast encoding on a single CPU core?

I'd additionally recommend rebuilding ffmpeg with CPU specific optimizations.

This is something that was brought up on the streams, and it was a deliberate compromise to not use all the system resources on generating the stream itself to leave some to spare for other tasks. A stable 10fps with headroom was preferable to not having anything to spare.

@@Megabobster If that became a problem, presumably one could set threads to 3, leaving one core to do other things. Also, see my later comment on using another PPC Mac as an X Terminal -- probably not for the Quad G5, but in the event of earlier PPC shenanigans.

@@Megabobster You can set a cap on resources for ffmpeg and lower the process priority so its giving more priority to other tasks.

My brain still runs at 68000

@@ActionRetro That explains a few things...😉

if firefox was compiled with cpu optimisations for that CPU it would also perform alot better. | although i think midori or firefox fork would of been a better choice.

It for all intents and purposes is STILL very much obsolete, but for a machine a quarter century old, it's pretty impressive.

Oh yeah I do really want one of those

@@ActionRetro I'd recommend you build rather than get it prebuilt, it'll save you thousands and you can get a larger case that fits a GPU.

Oh awesome!

Wasn't the 2.7 dual CPU not dual core?

@@virtualtools_3021 That's a good point.. IDK then

Those 2.7 machines are only 2 single core CPU’s, unlike this 2.5 machine with 2x 2 core CPU’s.

@@thegeforce6625 Yeah I realized that after some quick googling... Oof.. Oh well... it was wishful thinking.

@@ShiggitayMediaProductions it is probably still faster in a lot of applications since multi threading wasnt so big back then, and it has a higher bus speed

FFmpeg's asm structure uses AltiVec or VSX as needed. Trying to micro-optimize FFmpeg's non-asm code with -mcpu/-mtune flags generally isn't a *huge* gain.

@@EscurelEsquire I have to wonder if 'generally' doesn't mean x86, ie. generally, FFmpeg is compiled and run on x86 (x64) architecture. Also, I wonder if this use case may be one of those where any little improvement might make the difference between barely usable and marginally acceptable.

@@joelavcoco By 'generally', I mean that the compiler translating the C/C++ code into SIMD doesn't often improve speeds much when the main code paths being used have dedicated asm devoted to them (and FFmpeg has a **lot** of dedicated asm, including for AltiVec and VSX), so the critical pieces aren't being sped up that much. For parts of the codebase that *don't*, sure, whatever the compiler does would be the only thing you can rely on, and it might be the thing that can push it over the line into usability, but it highly depends on the workload and input/output formats. And separately from the machine tuning, there's also the influence (or potential influence) of the -O flags, which depending on the size or complexity of the code, can make a pretty huge difference. It wouldn't surprise me if distro maintainers tend to err on the side of caution and compile everything at the software's defaults, which is often (usually?) -O0 - or in other words, not turned on at all. It is somewhat trickier dealing with that, though, since some software will just break at higher -O levels. Somewhat of a case-in-point, FFmpeg-git (or 4.x, 5.x) can do ~100fps (or more; IIRC, the 100fps number was actually under Leopard, not Void) decode of 1080p H.264 on the G5 Quad. VP9 @ 1080p is even watchable, hovering somewhere in the 30-45fps range. And all of that is on the CPUs, thanks to the AltiVec asm it includes (the FX4500 only has an MPEG-2 decoder, so the graphics card is doing nothing regarding that). Granted, these were somewhat short tests, and it was at the profile/level/bitrate constraints that KZfaq uses for its videos, but the H.264 number shouldn't be surprising at all given that machines like - and including - the Quad would have been getting used in media production right around the time of the widespread adoption of HDTV and the launch of Blu-ray, and even for several years after that. But then again, FFmpeg has had 17 years to get AltiVec into at least the common decoders (I didn't really try and benchmark any of the encoders).

Next up: replacing the EXPLODED PSU on my PowerMac G5

Haha true, I have no idea what San Francisco is like!

Ok

Whoops! Thanks