x264 and Intel's Media Engine / Sandy Bridge

[CruNcher]

Configurable RD optimization (fastest speed, best quality, balanced quality and speed).


If that is usable on QuickSync also :) it would be already much more advanced then any GPU Encoder did someone saw something in the Media SDK 2.0 documentation if the RD modes can be changed also for the Hardware Encoding with QuickSync ?

@aegisofrime
see my post about this restriction matter second part http://forum.doom9.org/showthread.php?p=1470631#post1470631

[frenchfries]

The cards shaders are a key part of the "quicksync" experience. Quote from this (http://www.anandtech.com/show/4083/the-sandy-bridge-review-intel-core-i7-2600k-i5-2500k-core-i3-2100-tested/8) Anandtech article for those too lazy to read the article

"Motion search, the most compute intensive part of the transcode process, is done in the EU array. It's the combination of the fast decoder, the EU array, and fixed function hardware that make up Intel's Quick Sync engine."

Therefore, without a way to kick off the onchip video, quicksync wont be happening.

[SeeManRun]

The cards shaders are a key part of the "quicksync" experience. Quote from this (http://www.anandtech.com/show/4083/the-sandy-bridge-review-intel-core-i7-2600k-i5-2500k-core-i3-2100-tested/8) Anandtech article for those too lazy to read the article

"Motion search, the most compute intensive part of the transcode process, is done in the EU array. It's the combination of the fast decoder, the EU array, and fixed function hardware that make up Intel's Quick Sync engine."

Therefore, without a way to kick off the onchip video, quicksync wont be happening.

I think this is not correct. The act of transcoding involves multiple steps. First, decode the content, then analyze the content and adjust content (resize or downscale or whatever), and finally encode it to a new format. This discussion is really about encoding and not decoding or transcoding. There is no reason if you were using quicksync that you couldn't use your dedicated video card to decode the video and feed it into the quick sync functionality.

The advantage of the bigger chips with more of the EU's is because the decoding can be done on the video card part of the CPU; the less of these units you have the slower the decoding, and slower overall transcoding would be.

[Mixer73]

There is no reason if you were using quicksync that you couldn't use your dedicated video card to decode the video and feed it into the quick sync functionality.


Yes there is, Quicksync will only be available if you are using the onboard GPU, I find this feature very paradoxical for this reason.

[frenchfries]

@SeeManRun

Unless I am gravely misinformed motion estimation is most definitely a part of the H.264 encoding process. It's supposedly the most strenuous part of the process as well.

[SeeManRun]

Yes there is, Quicksync will only be available if you are using the onboard GPU, I find this feature very paradoxical for this reason.

I am aware of this. I meant no technical reason other than Intel's decision to limit this.

@frenchfries
While that is correct, all the info I have read on QuickSync indicate that it does not use the dedicated graphics functionality to do any part of the encoding. Much like Cuda or openCL, the encoder would have to be written to do that, and if Intel didn't artificially limit the hardware, could be done on your nVidia or ATI card and then sent back to QuickSync.

I could be off on this as I haven't read as much as I should since learning the feature is useless unless you have integrated video and h67, so please forgive me if I am continuing to spew fud.

[CruNcher]

Jesus i never read so much strange things i mean we see opinions here drifting far away what is fixed function and what is not in SB

The advantage of the bigger chips with more of the EU's is because the decoding can be done on the video card part of the CPU; the less of these units you have the slower the decoding, and slower overall transcoding would be.

Surely not it makes absolutely no sense and it wouldn't be possible to get so low CPU utilization the Decoding part is like Nvidias/ATIs fixed function as well most probably its the PowerVR VXD core, Intel also used it on the GMA500 doesn't mean they also use it in their Intel HD GPU but why should they have changed it ? :) anyways we don't know for sure what Intel is doing especially the Encoding part but definitely the EUs play no role for the Decoding part (also watch this (http://www.intel.com/technology/visualtechnology/snb_demo/latin/Module_C.flv) to understand how it works and realize where the problems on discrete GPUs come from this way of Shader based postprocessing and what influences its performance) on Fusion and SB it is much more efficient even with such low shader GPU cores you can gain more of the end performance then you would with a discrete card still Motion Estimation on these wouldn't be as efficient as on the CPU or a specialized part (same amount of shader performance).
They have absolutely no impact on the Decoding Performance of a 6 EU SB (HD2000) or a 12 EU SB (HD3000) you can see the heaviest impact on Games as EUs (Shaders) the more the better in today's heavy Shader based 3D Game World. Also reviews indicate that the EUs have absolutely nothing todo with Motion Estimation (in Intels Quicksync Encoder) as a 6 EU SB is as fast as a 12 EU SB so the EUs aren't doing the Motion Estimation (so in that sense it's not done by the GPU like in a GPU Encoder). Also Francois Piednoel (http://www.linkedin.com/in/francoispiednoel) didn't talked about anything of using EUs for accelerating X264 he talked about something different it didn't sound like a GPU more like a specialized Hardware part. In Intels Release Material its not really visible what that is it just gets a overall term but nothing specific what it is, though most probably no one ever will get low level access to it except Intel or some partner under heavy NDA. So yeah we can see the work its doing in Quicksync but that's all for the moment :)
And it seems Francois was how should i say it to enthusiastic with his idea that he could be able to get his boss to allow x264 to get low level access to it he must have run into some heavy resistance on this idea or was actually never heard by anyone (hello anyone here *echo*) even with his background as a engineer @ Intel only business counts Intel currently demonstrates this in a very heavy way with all the restrictions layer they developed and currently enforcing on consumers.

[Sharktooth]

jeez, it took me almost 6 minutes to parse the text on your post and elaborate it to understand it. please, use punctuation AND short periods.
however, you're right *BUT* it seems QS cant work without GPU. that sounds weird and maybe there are some kind of "relations" that makes it impossible to enable QS without enabling the GPU too.
still, i think that that silicon could have been used to add some more interesting stuff instead of a random black-box that no one knows how it works except intel.

[CruNcher]

Sharktooth currently it's very confusing how and when it works some reviews say you just need to connect a Monitor on 1 of the onboard outs so Dual Monitoring on a H67 Motherboard with a discrete Card to activate it. Would also fit to whats written in some Motherboard Manuals, you can find the information that if you don't explicitly set the shared memory value the VGA output on a H67 Board gets disabled though no word about DVI and if that might be disabling QuickSync also. Though still it holds what i wrote earlier the Intel guy said (http://www.youtube.com/watch?v=gui_6sNc7Eo) hes running a P67 board and their you saw QuickSync working with Lucids Virtu GPU Virtualization and officialy that shouldn't be possible on a P67 Motherboard (especially as a P67 Motherboard has absolutely no onboard outs) so i suspect the disabling of the QuickSync Detection in different scenarios is mainly controlled by the Intel Driver so the chain for this could be Bios->Intel Driver->Media SDK API->3rd Party Application :)
Lucids Virtu copys the Framebuffer to the Onboard out that has absolutely nothing todo with the QuickSync Enable/Disable (can be seen by a 3rd party app/ cant be seen) so something smells iffy here, especialy on a P67 board. All that gets even more confusing if you read Anandtechs article about the same Intel Presentation where suddenly a H67 board is in that machine (that's also why i made the joke, that earlier someone seems to not understood correct about the new intel innovation of a automatically switching chipset ;) ).

[popper]

carriage returns are good CruNcher, please use them more :cool:.
Sharktooth currently it's very confusing how and when it works some reviews say you just need to connect a Monitor on 1 of the onboard outs so Dual Monitoring on a H67 Motherboard with a discrete Card to activate it.

Would also fit to whats written in some Motherboard Manuals, you can find the information that if you don't explicitly set the shared memory value the VGA output on a H67 Board gets disabled though no word about DVI and if that might be disabling QuickSync also.

Though still it holds what i wrote earlier

the Intel guy said (http://www.youtube.com/watch?v=gui_6sNc7Eo) hes running a P67 board and their you saw QuickSync working with Lucids Virtu GPU Virtualization and officialy that shouldn't be possible on a P67 Motherboard (especially as a P67 Motherboard has absolutely no onboard outs)

so i suspect the disabling of the QuickSync Detection in different scenarios is mainly controlled by the Intel Driver

so the chain for this could be
Bios->Intel Driver->Media SDK API->3rd Party Application :)

Lucids Virtu copys the Framebuffer to the Onboard out that has absolutely nothing todo with the QuickSync Enable/Disable (can be seen by a 3rd party app/ cant be seen)
so something smells iffy here, especialy on a P67 board.

All that gets even more confusing if you read Anandtechs article about the same Intel Presentation where suddenly a H67 board is in that machine

(that's also why i made the joke, that earlier someone seems to not understood correct about the new intel innovation of a automatically switching chipset ;) ).

it's clear by now that a BIOS update will perhaps be the quickest fix here, so as to set the required control "bit" for the activation of the ASIC and related circuit's.

it would also seem to me now, that upper Intel board members
(just like AMD and their UVD decode, and as odd as it might seem to doom9 readers and the like),
didn't really consider the Media Engine ASIC as a mass market want/need, but rather just a way to grow their new vertical on-line video protected video markets etc, and the ability to PR match the Encode/Decode ability's as found on all ARM cortex A8/9 SOC today...OC

only now that the mass web presence and bad reviews bringing up the flaws of disabling their Media Engine ASIC if there's no ability to use/plug in an on-board monitor is it becoming clear to them we do want to try and use it as its available on all SB as a generic standard, and gives 3rd party vendors something new to advertise and make profit on.

it seems like good luck rather than good management choices saves Intel in this, as they dont seem to include the DRM parts inside the same ASIC as AMD did with their UVD, and so Intel can release the Encode/Decode info required later for embedded Linux use etc, whereas AMD dont and cant without a signed NDA, even today 3 UVD versions in.

(and AMD still didn't think to separate out that DRM from the decode UVD ASIC to date, never mind put some form of Encode in there for 2011 use to match ARM cortex capabilities etc).

[7oby]

There is no status. Nothing has happened. I haven't talked to him in weeks.
News:
I don't know about Francois, but our team is definitely working on it [bring Intel’s Quick Sync Encode/decode ASIC to general linux use], keep an eye on the mailing list for patches and status updates.
http://phoronix.com/forums/showthread.php?29110-Intel-s-Linux-Sandy-Bridge-Graphics-Still-Troubling&p=167917#post167917

But maybe this just refers to the DEcoding part (boring).

[kolak]

Quite interesting:

Intel provided a technological preview of the Grass Valley EDIUS 6 nonlinear editing (NLE) software, at the Intel Forum 2011 held in Japan, on January 18th 2011. At the Tokyo press conference, Intel highlighted its newly announced 2nd Generation Intel Core processor, embedded with Intel Quick Sync Video technology.
Grass Valley presented a technology preview of a forthcoming update to EDIUS 6, which has been fully optimised for the new Intel processor for hardware encoding support for H.264 videos, allowing for “faster than realtime” encoding of full HD (1920×1080) H.264 videos on a broad range of computers from notebook PCs to desktop PCs, using the new 2nd Generation Core Processor. More than 500 laptop and desktop PC platform designs are expected from all major computer manufacturers worldwide based on this new processor family. The EDIUS 6 update allows users to export to Blu-ray, or to the AVCHD format, faster than real-time directly from the EDIUS timeline.
“With our second generation Core processors, video editors using Grass Valley’s EDIUS 6 will be able to complete their work faster than ever,” said Fumihisa Shimono, Director of Software Marketing of Intel. “We know that in video, editing is all about performance and speed, and the combination of our second generation Core processors and EDIUS 6 deliver just that.”
“Working closely with Intel, we have fine-tuned EDIUS 6 to deliver the best and fastest post-production solutions to the market,” said Jeff Rosica, Executive Vice President of Grass Valley. “With H.264 and AVCHD becoming a key format for distribution, providing a faster solution is exactly what video editors need.”


Andrew

[Mug Funky]

“With H.264 and AVCHD becoming a key format for distribution, providing a faster solution is exactly what video editors need.”

ha!

the LAST thing an editor needs is their client whingeing about poor quality video and demanding more time in the suite!

when i switched to x264encoder on the mac, clients were gobsmacked that their pictures looked so good at email-able sizes. they kept ringing after the fact trying to figure out why their files always came out so much bigger and looked terrible.

a good solution in the long run will be faster than a fast solution that is not as good. editors don't want to be pulled out of sessions to re-do things that were out the door a week ago.

[kolak]

AVCHD became very popular, because of such a big use in cameras and compatibility with BD players, but it's a very bad format for editing.

EDITORS have no clue about quality and codecs. All old school AVID editors understand only one word TAPE. For them something like DNxHD, AVC-I is a black magic:) They just do editing and at the end need export button.
When I said I can export 2h HD timeline in 15min, my AVID friend was confused- 2H in 15min? It's impossible- hehehe :)

Edius users want ok looking AVC as fast as possible. They complain about time, which is needed to export AVC (Edius uses Mainconcept engine) and being told about x264, but for them to use it is also black magic. They still use MPEG2 for BD. They just don't wan't to hear about avisynth, cmd line, scripting etc :) GV knows about it and because they have a good programmers their implementation may be ok and fast. That's all what their customers need.
Why they just don't make x264 plugin- heh, good question, but for different debate :)

I was just quite surprised that they so quickly jumped into new technology. Good for Intel and they made a story of it :)
Grass Valley has quite good reputation for their products.

You would be surprised how many times I received 5Mbit AVC file (encoded with QT) for use on BD- because it's small- buahha :)

Andrew

[Audionut]

@kolak

I hope you spend as much time training these people to not be so spastic as you do voicing your opinions on this forum.

[Lyris]

You would be surprised how many times I received 5Mbit AVC file (encoded with QT) for use on BD- because it's small- buahha :)

I've heard of some smaller studios sending assets on DVD and asking if the compressionists can just re-encode those - quality? Nahhhh!

But 5mbit AVC (especially with Apple's implementation of AVC!), what are they smoking?

[kolak]

Lest wait till end of Q1 and we will see what (or rather how fast) Quick Sync will be able to deliver from Edius timeline.

Andrew

[weasel_]

said Fumihisa Shimono, Director of Software Marketing of Intel. “We know that in video, editing is all about performance and speed, and the combination of our second generation
This is absolutely wrong. Its about quality /efficiency.
I rly cant understand how he can said that.

[Mixer73]

This is absolutely wrong. Its about quality /efficiency. I rly cant understand how he can said that.

Well it depends, if they are accelerating realtime functions while editing and alowing them for rushes/preview I think its OK.

Especially if they just use the decoding side then use the Canopus sw codecs for encoding, this could be a very good boost to the productivity.

[laserfan]

This is absolutely wrong. Its about quality /efficiency. I rly cant understand how he can said that.
Look up his word "performance"--there are numerous definitions which include the word "quality" e.g. this one:

performance [pəˈfɔːməns]
n
1. the act, process, or art of performing
2. an artistic or dramatic production last night's performance was terrible
3. manner or quality of functioning a machine's performance
4. Informal mode of conduct or behaviour, esp when distasteful or irregular what did you mean by that performance at the restaurant?
etc..

When I read the quote from Shimono I did not question that when using "performance" he meant it to include a measure of "quality".

[weasel_]

If hi meant that then ok, my bad.

[Rodger]

Hi guys,

so I got my Sandy Bridge here now ;)
Vcore set @1,3V my 2500K does a rock stable 4.560Mhz.

Boy I can tell this new system makes my old Core2Quad system look ooold :D

If I can do any tests for you just send me a short PM as I may be busy the next days.

x264 HD BENCHMARK 3.0 RESULTS

Please do not compare it with older versions of the benchmark!
Please copy/paste everything below the line to to report your data

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Results for x264.exe r1342
--------------------------
encoded 1442 frames, 125.25 fps, 3900.68 kb/s
encoded 1442 frames, 125.09 fps, 3900.68 kb/s
encoded 1442 frames, 125.59 fps, 3900.68 kb/s
encoded 1442 frames, 125.60 fps, 3900.68 kb/s
encoded 1442 frames, 37.64 fps, 3971.09 kb/s
encoded 1442 frames, 37.27 fps, 3971.60 kb/s
encoded 1442 frames, 37.67 fps, 3971.38 kb/s
encoded 1442 frames, 37.64 fps, 3970.90 kb/s

[kolak]

Hi guys,

so I got my Sandy Bridge here now ;)
Vcore set @1,3V my 2500K does a rock stable 4.560Mhz.

Boy I can tell this new system makes my old Core2Quad system look ooold :D

If I can do any tests for you just send me a short PM as I may be busy the next days.

Try new x264- this one uses old build- much slower (at least on my X5355)

Andrew

[Rodger]

Sorry, I´m afraid I didn´t get you.

You´re talking about the benchmark, or usual encoding?
When encoding I use the current release from komisar (http://komisar.gin.by/).

The benchmark should be the current release too.
Used the only link to be found on that page to download the prog.
You got a link where to download anything newer?

[kolak]

Sorry, I´m afraid I didn´t get you.

You´re talking about the benchmark, or usual encoding?
When encoding I use the current release from komisar (http://komisar.gin.by/).

The benchmark should be the current release too.
Used the only link to be found on that page to download the prog.
You got a link where to download anything newer?

Replace x264.exe with latest one. It's 1.5 faster on my 8 core machine.

Andrew

[CruNcher]

My Sandy Bridge is still stucking @ the Shop the Memory has run out i ordered with so i have to wait before it arrives @ the seller :(
So still some time before im gonna be able to compare Nvidia (the real Nvidia Encoder not what Annandtech compared with) vs Quick Sync

btw this is really amazing on the Desktop (and their seem to be still reserves left to save off,without hurting much other usage scenarios):)

http://img262.imageshack.us/img262/6802/power5.png

that's the same power consumption the boxee box uses (according to a talk with a boxee engineer, CE4100 @ Full HD complete system) for 1080p playback :)

it really looks like the same Video Decoder (1080p) core is working inside Westmere and Sandy Bridge that works inside the GMA500 (CE4100) :)

i cant wait to see where CoreAVC and DiAVC stay (also in feature comparison like x264 lossless support) and how Nvidias VPX compares to that :)

Being able to utilize this for Decoding/Encoding (fully utilizing Media SDK 2.0) instead of Nvidias nvcuvid could show some benefits with the lower overhead being non discrete as well much lower overhead then shuffling data around from the CPU/GPU via PCI-E especially nice for Flash 10.2 :)

No review site so far released any Power Consumption results @ Encoding :(

[kypec]

Benchmarking such ancient revisions of x264 (r1342) on new generation CPUs is especially meaningless!
Please stop posting those benchmark results. They serve no other purpose but confusion.

[nm]

that's the same power consumption the boxee box uses (according to a talk with a boxee engineer, CE4100 @ Full HD complete system) for 1080p playback :)

it really looks like the same Video Decoder (1080p) core is working inside Westmere and Sandy Bridge that works inside the GMA500 (CE4100) :)

You can't draw such conclusions from system power consumption alone when the entire architecture is different. The hardware decoder in Sandy Bridge is much faster than the one in GMA500: according to QuickSync benchmarks, SB MFX can decode 1080i30 video at over 10x realtime speeds. That's about 4x faster than other widely available hardware decoders.

I think this is also a contributing factor to why QuickSync outperformed fast GPUs and software encoders (Ati) used in the reviews -- the competition was bottlenecked by much slower hardware decoders.

[kypec]

The hardware decoder in Sandy Bridge is much faster than the one in GMA500: according to QuickSync benchmarks, SB MFX can decode 1080i30 video at over 10x realtime speeds. That's about 4x faster than other widely available hardware decoders.
Is there any chance that QS video-decoder engine be used in players/source filters (MPC-HC, VLC, ffdshow, ffmpegsource) in near future?

[nm]

Is there any chance that QS video-decoder engine be used in players/source filters (MPC-HC, VLC, ffdshow, ffmpegsource) in near future?

It can be used now through DXVA2 in MPC-HC, VLC and ffdshow-tryouts. To use it more efficiently and to get the decoded video back to AviSynth like DGDecNV does, somebody would need to write a source filter that uses Intel Media SDK.

[Sharktooth]

or OpenCL 1.1, so that will work on nvidia, ati and SB (according (http://news.cnet.com/8301-13924_3-20016302-64.html) to Thomas Piazza - Intel director of graphics architecture)

[Rodger]

Here are the results:

x264 HD BENCHMARK 3.0 RESULTS

Please do not compare it with older versions of the benchmark!
Please copy/paste everything below the line to to report your data

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Results for x264.exe r1342
--------------------------
encoded 1442 frames, 181.25 fps, 3911.51 kb/s
encoded 1442 frames, 181.61 fps, 3911.51 kb/s
encoded 1442 frames, 180.88 fps, 3911.51 kb/s
encoded 1442 frames, 179.82 fps, 3911.51 kb/s
encoded 1442 frames, 34.25 fps, 3962.88 kb/s
encoded 1442 frames, 38.63 fps, 3963.13 kb/s
encoded 1442 frames, 38.72 fps, 3962.83 kb/s
encoded 1442 frames, 38.63 fps, 3962.43 kb/s


System Details
--------------
Name Intel Sandy Bridge
product System Product Name
Codename
Specification Intel(R) Core(TM) i5-2500K CPU @ 3.30GHz
Specification PC3-12800
Specification PC3-12800
Core Stepping
Technology 32 nm
Stock frequency 3300 MHz
Core Speed 3343.5 MHz

Northbridge Intel ID0100 rev. 09
Southbridge Intel ID1C46 rev. 04

Command Rate 2T
Command Rate 2T
Memory Type
Memory Size 8192 MBytes

Windows Version Microsoft Windows 7 (6.1) Home Premium Edition (Build 7600)

Number of processors 1
Number of threads 4
Number of threads 4 (max 16)
L2 cache 4 x 256 KBytes, 8-way set associative, 64-byte line size
Instructions sets MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, EM64T, VT-x, AES, AVX
Package (platform ID) (0x1)

Temperature 0 33°C (91°F) [0x21] (SYSTIN)
Temperature 1 60°C (139°F) [0x188] (CPUTIN)
Temperature 2 35°C (94°F) [0x46] (AUXTIN)
Temperature 0 42°C (107°F) [0x38] (Core #0)
Temperature 1 44°C (111°F) [0x36] (Core #1)
Temperature 2 47°C (116°F) [0x33] (Core #2)
Temperature 3 43°C (109°F) [0x37] (Core #3)


For the clever ones of you! I PERSONALLY exchanged the x264 with the current one from komisar.
So it´s NOT r1342 but r1882 from komisar! The release-Version is obiously a FIX VALUE in the result.rtf file!!!
So don´t bug me anymore with wrong sugguestiones!
As there are more things wrong in the log-file.
CPU-Speed is 4560Mhz !!!
(RAM-Speed is 811Mhz)
Command Rate for RAM is 1T !!!

[CruNcher]

So @ 4.5 + 1.1 Ghz (Nominal clock ?) you reach +10 fps more you can say 1 Ghz = 10 so + 500Mhz could be + 5 fps @ what are each core clocks for you when running the benchmark and did you tried how long it sustains stable (turbo boost, at that temperatures i guess it can hold pretty long) ?

The difference isn't that big in the build it stays nicely comparable with the gathered Review results maybe -+1 fps (nothing really to care
about) :)

PS: Sandy Bridge is on its Way :)

You can't draw such conclusions from system power consumption alone when the entire architecture is different. The hardware decoder in Sandy Bridge is much faster than the one in GMA500: according to QuickSync benchmarks, SB MFX can decode 1080i30 video at over 10x realtime speeds. That's about 4x faster than other widely available hardware decoders.

I think this is also a contributing factor to why QuickSync outperformed fast GPUs and software encoders (Ati) used in the reviews -- the competition was bottlenecked by much slower hardware decoders.

True wow which review showed that 10x with deinterlacing on the 6 (HD2000) or 12 (HD3000) EUs ?
and i hope its faster then the CE4100 SGX, it seems really perfect suited for high efficient HTPCs with the Power for other Tasks available on Demand finally we go that direction :)
Lucidlogix just has to release Virtu and we can mix that with discrete 3D performance and then Power switch like on Optimus on the Desktop that will rock not to think about the other possibilities like using both DSPs :D and CUDA while using Quick Sync and mix and match :)
Platform wise it would also rock for a new Game Console it would blast the Xbox 360 into oblivion @ the same power consumption not sure about the PS3 though :)

Combining Nvidias/ATis DSP with CUDA/AAP and Quick Sync, AVX could make some crazy frameworks possible (has to be evaluated, working on their own or combined) :)

[Sharktooth]

+1 fps on the second half of the test.

[Rodger]

Here are some screenshots for the interested:

Idle..............................................................................OC...............................................................................RAM
http://i55.tinypic.com/r074ts.jpghttp://i52.tinypic.com/oardoo.jpghttp://i51.tinypic.com/2zrdr3m.jpg

Overclocking SANDY BRIDGE:
There are some new things that pop up as unclear every now and then.
CPU 2500 without K = NO OVERCLOCKING!
CPU 2500K = Overclockable!
Intel P67 Chipset = Overclocking but NO GRAPHICS ONBOARD! (You need discrete graphics!)
Intel H67 Chipset = NO OVERCLOCKING but you can use integrated HD Graphics
So it´s pretty clear to go like 2500K CPU + Board with P67 Chipset OR 2500 without K + Board with H67 Chipset!
Thanks you Intel you can´t have both !!!
Turbo-Mode on the other side is active with both chipsets and so it seems turbo mode is no overclocking to Intel ?! (crazy guys!)
Overclocking the "K"-CPUs is done via the Turbo-Mode multiplier wich is unlocked here. All other CPUs have a fixed Turbo-Mode multiplier!
So if you want to overclock you CPU, you´d better go with a i7 2600K (TOP MODEL!) or the i5 2500K.
Only the top model 2600K and 2600 without K (;)) do support HT (HyperThreading)!

Additional you can´t or hardly can overclock a CPU via FSB alias BCLK! At the most you should set that value to 106 since you seriously can damage attached Hardware! Intel even WARNS to use anything but clean 100Mhz!
OC Profile on Asus-Motherboards uses 103Mhz! This just as a sidenote.

If I got cruncher right this is the answer:
Turbo-Mode is not limited! It´s only limited if your CPU-temperatures are bad! In that case the CPU reduces or disables the turbo-mode to prevent overheating.
To me - having a watercooled system :D - the Turbo-Mode is never ending ;) So as soon there is enough CPU load the CPU switches to turbo-mode.

[CruNcher]

Here are some screenshots for the interested:

Idle..............................................................................OC...............................................................................RAM
http://i55.tinypic.com/r074ts.jpghttp://i52.tinypic.com/oardoo.jpghttp://i51.tinypic.com/2zrdr3m.jpg

Overclocking SANDY BRIDGE:
There are some new things that pop up as unclear every now and then.
CPU 2500 without K = NO OVERCLOCKING!
CPU 2500K = Overclockable!
Intel P67 Chipset = Overclocking but NO GRAPHICS ONBOARD! (You need discrere graphics!)
Intel H67 Chipset = NO OVERCLOCKING but you can use integrated HD Graphics
So it´s pretty clear to go like 2500K CPU + Board with P67 Chipset OR 2500 without K + Board with H67 Chipset!
Thanks you Intel you can´t have both !!!
Turbo-Mode on the other side is active with both chipsets and so it seems turbo mode is no overclocking to Intel ?! (crazy guys!)
Overclocking the "K"-CPUs is done via the Turbo-Mode multiplier wich is unlocked here. All other CPUs have a fixed Turbo-Mode multiplier!
So if you want to overclock you CPU, you´d better go with a i7 2600K (TOP MODEL!) or the i5 2500K.
Only the top model 2600K and 2600 without K (;)) do support HT (HyperThreading)!

Additional you can´t or hardly can overclock a CPU via FSB alias BCLK! At the most you should set that value to 106 since you seriously can damage attached Hardware! Intel even WARNS to use anything but clean 100Mhz!
OC Profile on Asus-Motherboards uses 103Mhz! This just as a sidenote.

If I got cruncher right this is the answer:
Turbo-Mode is not limited! It´s only limited if your CPU-temperatures are bad! In that case the CPU reduces or disables the turbo-mode to prevent overheating.
To me - having a watercooling system :D - the Turbo-Mode is never ending ;)

Though its not entirely true also the H67 boards allow to change the Turbo multiplier just very calm not in the range of where you are x45 most is x37 for H67 but for me using every resource possible is more important so P67 was no choice and Z67 not available so i set my hopes into Lucids Virtu for H67 and if that should be a wrong guess i gonna switch over to Z67 later.
If im right and didn't missunderstand i can boost the 2400 i took to a max x37 even on the H67, though i wont use that as im not interested in overclocking rather undervolting hehe ;)

[Rodger]

I have never ever heard of changeable turbo-mode multipliers on ANY H67 board!
Which one does support it???

Of course you may set the turbo mode multiplier to fixed value of the certain CPU which will be CPU stock + 4 Steps.
Example Core i5 2500 (without K): 3,3Ghz + 4x100Mhz Turbo = 3,7Ghz

So in your case you are wrong!
Core i5 2400: 3,1Ghz + 3 Steps (3x100Mhz) = 3,4Ghz MAX!!! (Multiplier is LOCKED by Intel!!!)
Only 2500 + 2600 have 4 Steps for Turbo-Mode! All others I think only 3 Steps.

- GERMAN -
In unserer Sprache hier zitiert aus z.B. dem Mindfactory-Board:
Wie ich soeben erfahren habe lassen sich die Multiplikatoren für die frei übertaktbaren K-Prozessoren bei Mainboards mit Intel H67 Chipsatz nicht frei einstellen.
Es macht also keinen Sinn ein H67 Mainboard mit einer ****K CPU zu bestücken.
- GERMAN OFF -

Cooling Advice:
In order to use the turbo-mode as much as possible you should look out for a CPU-cooler capable of coping with the heat from the CPU.
The stock cooler is said to be OKAY for system with a lot air flow. But systems aiming on silence need a better/bigger CPU cooler!
Watch out! Since there is a lot to be read, that coolers collide with parts on mainboard because of their sized. You better read about positive opinions on which cooler is workling fine with the mainboard of your choice.

[CruNcher]

Thats what i meant fixing it on x34 in my case though im confident that someone can brake this restriction anyway in the future ;)
lol just realized they canceled the Motherboard on my order also MF, locally i find only P67 boards geez :(

[nm]

True wow which review showed that 10x with deinterlacing on the 6 (HD2000) or 12 (HD3000) EUs ?

This one with HD3000: http://www.tomshardware.com/reviews/sandy-bridge-core-i7-2600k-core-i5-2500k,2833-5.html

But I doubt the deinterlacing is anything fancy at that speed.

Here's 11x decoding speed for a 12 Mbps 24p source: http://www.anandtech.com/show/4083/the-sandy-bridge-review-intel-core-i7-2600k-i5-2500k-core-i3-2100-tested/9 (see the last figure on the page)

[CruNcher]

It depends what quality Intels Deinterlacing is @ :) but yeah Yadiff on the SB should be super fast as well but the more you can push on the DSPs the more you can free up the CPU for x264 encoding, the last HQV test though showed that Intel didn't reached Nvidia nor ATI yet @ Deinterlacing Quality but maybe they can blow it @ speed also, though hard to say with the current reviews :)
Now i have a SB system without Motherboard :p seems the demand is still heavy

[Sharktooth]

Here are some screenshots for the interested:

Idle..............................................................................OC...............................................................................RAM
http://i55.tinypic.com/r074ts.jpghttp://i52.tinypic.com/oardoo.jpghttp://i51.tinypic.com/2zrdr3m.jpg

Overclocking SANDY BRIDGE:
There are some new things that pop up as unclear every now and then.
CPU 2500 without K = NO OVERCLOCKING!
CPU 2500K = Overclockable!
Intel P67 Chipset = Overclocking but NO GRAPHICS ONBOARD! (You need discrete graphics!)
Intel H67 Chipset = NO OVERCLOCKING but you can use integrated HD Graphics
So it´s pretty clear to go like 2500K CPU + Board with P67 Chipset OR 2500 without K + Board with H67 Chipset!
Thanks you Intel you can´t have both !!!
Turbo-Mode on the other side is active with both chipsets and so it seems turbo mode is no overclocking to Intel ?! (crazy guys!)
Overclocking the "K"-CPUs is done via the Turbo-Mode multiplier wich is unlocked here. All other CPUs have a fixed Turbo-Mode multiplier!
So if you want to overclock you CPU, you´d better go with a i7 2600K (TOP MODEL!) or the i5 2500K.
Only the top model 2600K and 2600 without K (;)) do support HT (HyperThreading)!

Additional you can´t or hardly can overclock a CPU via FSB alias BCLK! At the most you should set that value to 106 since you seriously can damage attached Hardware! Intel even WARNS to use anything but clean 100Mhz!
OC Profile on Asus-Motherboards uses 103Mhz! This just as a sidenote.

If I got cruncher right this is the answer:
Turbo-Mode is not limited! It´s only limited if your CPU-temperatures are bad! In that case the CPU reduces or disables the turbo-mode to prevent overheating.
To me - having a watercooled system :D - the Turbo-Mode is never ending ;) So as soon there is enough CPU load the CPU switches to turbo-mode.
you can get more performance from your RAM rising the CAS latency or maybe overvolting the RAM modules.
check this table: http://www.webalice.it/f.corriga/stuff/memlatency.jpg

[Didée]

Just that RAM performance does hardly ever matter. It's mostly important for e-peniz signatures. Overall performance stays almost unaffected. L3 cache is where the party's happening.

[Rodger]

Yeah you´re right and I´m not done with my RAM yet :D
I bought a 8GB Corsair Vengence KIT. Here is a Test of the CL9 Kit if you like http://www.overclockersclub.com/reviews/corsair_vengeance_8gb/
But I chose the CL8 Kit ;) So AT LEAST I´m going to be able to set the 1833er divider. And very important to me...without giving the RAM absurd voltages. 1.65V will definetly be MAX!
Sandy Bridge doesn´t scale so much with the RAM timings but at least a little with the clocks of RAM.
So it´s better to choose clocks over timings.
1300Mhz 8-8-8-21 won´t be as fast as 1600Mhz 9-9-9-24
And even more important in my point of view is the 1T Command Rate. A whole lot of performance is unleashed with that setting.

[woah!]

+1 fps on the second half of the test.

i'm actually not that impressed by it. i get that fps from my 1055T at 3.9ghz. and the cpu is much cheaper than the 2500K is.

and it is the 2nd pass that really matters for most here, as the 1st pass has the fastest settings you can on x264 set.

bulldozer here i come..

[Rodger]

You are comparing a 4 Core CPU with a 6 Core CPU???
And even better...the 2500K does this with 95W power consumption...will be better not to tell how much the AMD CPU sucks right out of the line :rolleyes:

The 2600K with HT will easily knock down any AMD system. Bulldozer hopefully will be able to catch up but it seems and everything points to it...it will be a huge disappointment.

[Sharktooth]

comparing different generations CPU is not so fair. however AMD x6 CPUs are still on track with performance per dollar/euro ratio.
bulldozer is not out yet, but it will be surely an interesting CPU.

[woah!]

what does it matter how many cores it has if its $60 cheaper? it wasnt long ago that intel used 3.2ghz to match a 2.5ghz amd chip, but intel actually sold more because more mhz must means faster right?

basically i want a encoding box for cheap, and well 6 core / 12 core whatever, if its as fast and cheaper how does it matter?

[Rodger]

In THAT CASE wait for the soon released 10core CPUs from Intel based on the new architecture.

But anyway I need a full package to be done by my computer and this is where current AMD CPUs just suck big time. As soon as all cores are used AMD CPUs do become interesting and useful. But that is just not much the case...

[woah!]

i am not trying to come across as amd bias here, i have had intel QX6700 / Q9550 for years. i was looking to upgrade to a SB thats all. yes a westmere might do the trick if amd bulldozer fails.

[Rodger]

Me too...I don´t want to sound like an Intel fan guy or something. I´ve had AMD for so many years myself. I like the idea AMD stands for but they didn´t do very well for quite a while now and insider reports really don´t sound nice when it comes to bulldozer.
It´s said bulldozer won´t even catch the performance of the first generation Core i series, which would be devastating. AMD can´t possibly be able to stand that price dumping of their CPUs for much longer. Be honest...the 6 Core CPUs would have to be much more expensive if AMD would have to make money with selling them.

No! No need to upgrade for you, if you really do have your total preferences on encoding.
I based my decision on a huge test from a site here in germany. So it´s in german language but the numbers speak volumes. http://www.computerbase.de/artikel/prozessoren/2011/test-intel-sandy-bridge/28/#abschnitt_x264_hd_benchmark_319
In this article the AMDs don´t stand a chance. At stock speed the AMD1075T and the Intel 2500K are this one time at same speed when going for the second pass. But just this once.