H.265 - Are all existing IGPs and GPUs about to be rendered obsolete?

[blastingcap]

4k is passe.

8k for me.

http://www.latimes.com/business/technology/la-fi-tn-ces-sharp-8k-tv-20130110,0,5741879.story

 

[chibimike]

I'm not familiar with H.265, but it being a more efficient standard, won't it make older GPUs and iGPUs more relevant? My ignorance apart, I can only see a positive here.

The "efficiency" they are referring to is bandwidth, not compute power. It may likely be much more demanding on processors. Only time will tell.

 

[Mark Rejhon]

4k is passe.

8k for me.

http://www.latimes.com/business/technology/la-fi-tn-ces-sharp-8k-tv-20130110,0,5741879.story

8K 120Hz FTW.

Actually, let's make it 8K 1000Hz sample-and-hold.
-- 1000Hz so we don't need strobe backlights (bad for flicker)
-- 1000Hz so we don't need motion interpolation (input lag; bad for computer/games)
-- 1000Hz so we don't have input lag. (only 1ms input lag)

Obviously, would require a really, really _insane_ GPU.

Might never happen. But there's obviously benefit well beyond 120Hz, as proven by motion tests (PixPerAn, DisplayMate Motion Bitmaps, etc). And all the people complaining about PWM flicker even at 360Hz. Probably not until the 2050's (or beyond) before we get 1000fps@1000Hz to say goodbye to silly motion interpolators & silly strobe backlights, and keep the 100% full CRT-style motion benefit in an eye-comfortable sample-and-hold flicker-free display. The ONLY way to do this without interpolation and without strobes is 1000fps@1000Hz.

(Note: LightBoost Zero Motion Blur (2ms strobes) has a measured motion equivalence to an approximately ~480fps@480Hz sample-and-hold LCD. People, including myself, are still noticing the difference (Science & References, scroll down). However, LightBoost strobe backlight 2ms strobe flashing 120 times a second, is still CRT-style flicker, plain and simple, and not everyone likes CRT flicker. There are points of diminishing returns. But benefits keeps going _well_ beyond 120 due to indirect effect such as eye-tracking-based motion blur, stroboscopic effects, eyestrain problem (if impulse-driven instead of sample-and-hold), and other indirect factors, as you can see in the scientific information.

Thanks,
Mark Rejhon
BlurBusters.com Blog -- Eliminating Motion Blur on LCD Technologies -- Lightboost HOWTO

 

[A5]

Maybe h.265 will have a higher license fee, and will force down the price of the license fee for h.264. So maybe everyone will keep using h.264, and h.265 will fall by the wayside and be a niche codec for 4K stuff only? So we enjoy lower prices on existing things, but will pay when we adopt 4K.

That's not how this works. Not to mention H.264 is already ubiquitous for all kinds of video content.

It depends on what the compression is at the cost of, if they're sacrificing quality to gain better compression then not much will change, but if the standard is for a compression type that uses significantly more CPU/GPU time to encode/decode then older hardware may struggle.

H.265 is a new set of algorithms that give you the same quality with half the bandwidth. It's pretty much the same relationship that MPEG-2 and H.264 had.

 

[Piroko]

There is still untapped performance in h.264 like 10bit color accuracy, better motion detection and smarter bitrate distribution to reduce size. Most of this is almost free from a performance perspective. From what I currently know h.265 is not much more than a new money grab to boost sales.

 

[taltamir]

There is still untapped performance in h.264 like 10bit color accuracy, better motion detection and smarter bitrate distribution to reduce size.

Some do make use of those features

 

[A5]

There is still untapped performance in h.264 like 10bit color accuracy, better motion detection and smarter bitrate distribution to reduce size. Most of this is almost free from a performance perspective. From what I currently know h.265 is not much more than a new money grab to boost sales.

The room left in H.264 isn't going to cut bitrates in half with the same quality. What you "know" about H.265 is wrong.

 

[fuzzymath10]

Also things like 10-bit are generally not GPU accelerated anyway so you still have software decoding to deal with because of profile compatibility.

 

[biostud]

With more online video services the need for bandwidth increases. With h.265 they can service double amount of customers with same hardware. For PC users it won't be a problem as software will be able to decode the codec. Tablets and phones may have trouble if they don't support it in hardware.

 

[Piroko]

The room left in H.264 isn't going to cut bitrates in half with the same quality. What you "know" about H.265 is wrong.

H.265 isn't going to do that either in its current state. For the most part it's improving on h.264, removing bottlenecks and advancing some algorithms. Part of this stuff has been done in newer h.264 encoders as well and could have been easily integrated into the current h.264 standard via a new profile. That would have saved us from the marketing shitstorm which is starting to form in the distance...

 

[VulgarDisplay]

H.265 isn't going to do that either in its current state. For the most part it's improving on h.264, removing bottlenecks and advancing some algorithms. Part of this stuff has been done in newer h.264 encoders as well and could have been easily integrated into the current h.264 standard via a new profile. That would have saved us from the marketing shitstorm which is starting to form in the distance...

Translation: I didn't read the article and I am making things up.

 

[Piroko]

There wasn't any information whatsoever in that article except for the claim that h.265 would - and I quote - send HD video at half the bit rate of H.264. Which is not a technical term because current codecs work with variable bitrate and variable target quality. Nothing will stop you 'sending a HD video at half the bitrate of mp3', it wouldn't look pretty though.

I don't disagree that it will be a performance jump in percieved quality per bitrate over the current bluray h.264 profile. But so will a maxed out x.264. Or Main Concept.

 

[VulgarDisplay]

There wasn't any information whatsoever in that article except for the claim that h.265 would - and I quote - send HD video at half the bit rate of H.264. Which is not a technical term because current codecs work with variable bitrate and variable target quality. Nothing will stop you 'sending a HD video at half the bitrate of mp3', it wouldn't look pretty though.

I don't disagree that it will be a performance jump in percieved quality per bitrate over the current bluray h.264 profile. But so will a maxed out x.264. Or Main Concept.

They didn't say bit rate. They said bandwidth. h265 will send a same quality video as h264 using half the bandwidth, which is amazing.

 

[Piroko]

Bitrate is proportional to bandwith. And so far there have only been very few blind tests even though h.265 test encoders have been around for months.
H.265 will probably scale better with high resolutions (as most of the work was done in that field) but currently its encoder is still in a rough state. A minute of SD video can take more than an hour to encode and so far PCs have a hard time doing full hd playback.

H.265 will take over from h.264 sooner or later, but it's not as fundamentally different as h.264 was to DivX/XviD.

 

[Chiropteran]

4k is passe.

8k for me.

640k should be enough for anyone.

 

[Lonbjerg]

640k should be enough for anyone.

That myth/lie needs too die...and especially on techforums.

You do know it's a false statment right?

 

[Mopetar]

No.

To begin with, most video decoders are built to run on a CPU. Even if you have a more efficient GPU decoder, it doesn't matter since a CPU decoder is more than good enough. Second, anything that doesn't have a good enough CPU (ARM SoCs, etc.) has a dedicated hardware decoder that's more efficient than either a CPU or GPU could hope to be.

H.265 is going to equate to lower bandwidths required for streaming video content or a better stream quality at the current bandwidth rates. GPUs don't even factor into the discussion.

 

[Kippa]

They say you can half the bandwith of h264 video using h265 and still retain similar quality. If that is true then at a guess it is going to be computationally more agressive than h264 for both encoding and decoding. Encoding only really matters to people who create videos, decoding on the other hand, it might require more cpu processing power to decode h265 video.

It will be interesting to see what different cpu usage is between the same video decoding in h264 and h265.

 

[WhoBeDaPlaya]

I was actually thinking of the HTPC crowd, who depends on HW accelerated playback to keep everything smooth. Adoption of a new codec, means that existing IGPs and GPUs with hardware codec support are immediately rendered obsolete, requiring users to buy a new GPU with support for H.265.

That's why I like my custom ffdshow solution. Plays and upscales everything (mpg, mpg2, wmv, mjpeg, dv, huffyuv, indeo, cinepak, real video, etc.). A little CPU heavy, but what else am I going to use quad cores for?

 

[Lepton87]

1080p h.264 needs at least core 2 duo at 2GHz/Athlon X2 2.5GHz to decode in software. I wonder what CPU power it takes to decode 1080p h.265 stream. Is it a known information yet? I won't be surprised if it needs 8x more processing power.

 

[sxr7171]

The Wikipedia entry has quite a bit of detail and will answer some of the questions here.