Will Penryn CPUs greatly increase encoding times?

[ellesshoo]

I don't know much about how H.264/x.264 use a CPU, but will SSE4 and the Radix-16 divider (does both floating point and integer at 4 bits per cycle as opposed to the usual 2 bits) increase speeds at the same clock frequency as todays dual/quad cores?

Any clue whether AVC encoders will be able to see gains right away or whether they will have to be recompiled/recoded to take advantage?



Speed vs quality is a major issue I'm dealing with (as one might be able to tell from my other post)... I'm still on a single core cpu but am hesitant to get a new rig so close to a new CPU family release.

Lastly, are there any advantages for AVC encoding by going the Xeon route vs the normal Core route? Do most AVC encoders make use of dual Xeon machines?

[Dark Shikari]

I don't know much about how H.264/x.264 use a CPU, but will SSE4 and the Radix-16 divider (does both floating point and integer at 4 bits per cycle as opposed to the usual 2 bits) increase speeds at the same clock frequency as todays dual/quad cores?x264 uses basically no floating point math, and basically no integer division, since both are operations generally avoided in any CPU-intensive program whenever possible.

SSE4 is basically completely useless for encoding; the only "benefit" was in a specially contrived DivX build designed solely to get maximum benefit from it. x264's sequential elimination ESA algorithm is actually faster than SSE4's ESA operation.

Penryn's will likely be at least 10% faster overall but not because of SSE4.
Lastly, are there any advantages for AVC encoding by going the Xeon route vs the normal Core route? Do most AVC encoders make use of dual Xeon machines?x264 can easily use up to 8 cores or more with the latest threading algorithm, so no worries there.

[St Devious]

so x264 cannot take advantage of latest penryn processors ?

can other H.264 encoder's like the one from mainconcept take advantage of new processors ?

[fields_g]

so x264 cannot take advantage of latest penryn processors ?

can other H.264 encoder's like the one from mainconcept take advantage of new processors ?

I don't know if you are getting it. More or less... SSE4 optimizes processes that are not optimally designed. The result is still can't compete with a well designed process that doesn't use SSE4.

That being said... a penryn processor is better, even for x264. Due to processor changes in efficiency, cache, and FSB, etc.... You will get about 10% better preformance per clock than current core2 chips. Being that penryn will move us to higher frequencies, more cores, and runs cooler, many of us will end up with these chips... that is, when availability and pricing is reasonable.

[Dark Shikari]

so x264 cannot take advantage of latest penryn processors ? The new instruction set has nothing worth taking advantage of.

[Yoshiyuki Blade]

What's more important to me is the improvement in speed by going from dual core to quad core. Penryn also has more cache than last gen's Kentsfield. There's 12MB of cache on most of the Yorkfields (those are the quad core Penryns; dual core Penryns are Wolfdale). Will that help improve performance clock-for-clock over last gen also?

I can't wait for the mainstream Yorkfields to come. These can overclock like beasts. We're talking easily over 4.00 GHz on air cooling, though there probably won't be a motherboard to push them to their limits for a while. Anything less than the $1000 model thats out right now will require some serious mobo overclockability to max out.

[Sagekilla]

Performance gains from cache boosts are tricky... Depending on the application there may or may not be a benefit from raising it. I'd imagine with a larger cache, x264 will see an improvement as it can hold several uncompressed frames in the L2 Cache. And since L2 is faster than your main memory, you're looking at almost 10 full uncompressed DVD resolution frames to store in it! :)