Question / Help Serious comparision of 1080p Handbrake transcoding : CPU vs Haswell QuickSync (with screenshots)

[Sanjiv Dutta]

Dear all members,

For the last 2 weeks, I am thoroughly following the reviews, articles, benchmarks related to HD transcoding. Actually I am planning to get a new i5 4570 (or 4670) system for 1080p transcoding. I am not interested in GPU assisted transcoding. My focus is on haswell based QuickSync assisted Handbrake, because I heared that quality of Haswell QS is BETTER than or same as CPU based Handbrake conversion. But never got to know what quality settings were used.

I have observed that most benchmarks were not transparent about certain points -

1) What performance settings were used? I had tested a i5 2500k using QRF of 21, 'slow' preset and high profile. Found about 8.2 FPS!! So I guess these benchmarks either use 'faster' or even 'very fast' settings in their 'CPU only' tests.

2) Whether it is 1080i or p? And what was the target size. Sometimes they use iphone profile. But as a general rule, one should state the worst case, I mean, state the FPS for 1080p, without any scaling or resizing (even 720p would be fine, but should be clearly mentioned. The FPS for 1080p would be approx 3 times less).

3) What QuickSync performance settings were used? I heard that haswell supports 7 performance vs quality settings. Also never found if QS supports any other parameters.

4) What was the clip length? Many graphs depict the time in seconds, but forget to mention the clip length! So whats the point? Mentioning FPS rather could resolve this confusion.

5) A VITAL POINT part most reviewers miss is the SAMPLE IMAGE or CLIP. Without sample screenshots, how could one be convinced? Specially, when we are hearing that Haswell QS is BETTER that CPU based Handbrake conversion.

6) Again on processor side, most reviewers just use 4th gen. i7 or other latest CPUs. Reviews should focus on worst case. If not, at least on medium case. So 3rd Gen i5 or i3 should also be included.

I request you to pl. provide us with such an comprehensive comparison with will help many users like me, to settle all doubts for good. Thx so much in advance...​

 

[Jack0r]

I am really wondering why you post this in the OBS forum? I don't know of any benchmark that has been posted in here either. Then you should probably read up on the difference of live encoding and encoding. Benchmarks only check encoding while for live encoding you target different settings.
From the OBS side of things we either look at live streaming or live recording. Both use an input at a certain framerate and want to save the file at this exact frame rate, live. Compared to an encoding benchmark where the encoder can just encode as fast or slow as its settings allow it, this is quite different.

You should maybe read up on the differences of QuickSync in the different generations and between the different graphics chips. Your statement about the 4th gen i7 compared to i3/i5 shows you are lacking some knowledge there.
Last but not least, where in gods name did you read that Haswell QS is better or same as your so called "Handbrake conversion", whatever that means exactly. But if I understand it correct you are probably talking about the quality of a QuickSync encoded Video compared to an x264 encoded Video. In this case, my earlier recommendation to read up on the QuickSync generations will help. QS in most generations was completely targeted towards speed, not quality. While the last Haswell Generation increased the quality a bit.

At the moment, some people might compare Haswell Quick Sync to the veryfast or fast preset of x264, but this is not totally true, as QuickSync has some quirks and can look severly worse in bad situations where x264 is just more optimized to handle everything.

To point you in a direction: http://jack0r.com/high/quicksync_x264.html
http://tetrachromesoftware.com/q264Test1Analysis/q264test_1.html (I dont like this "test", but it gives you some values)
http://en.wikipedia.org/wiki/Intel_Quick_Sync_Video

 

[Sanjiv Dutta]

My sincere apology for posting it in PBS forum. I posted the question in few fora related to video transacoding, that were listed by Google, without much thought. Sorry for that again.

Thx anyway for your reply. I am sorry for any confusion I created, however, I am fully aware of what I was writing e.g. differences in Sandy, Ivy, Haswell and in which platform they were available. That's why I have specifically mentioned "3rd Gen i5 or i3" which have Ivy. Further, I have also mentioned "Haswell QS is BETTER that CPU BASED Handbrake conversion" meaning handbrake without any H/W acceleration.
BTW, I had actually checked these links earlier but they have the problems / ambiguities I menaioned.

Sorry again... and pl. don't bother to reply as this is not the proper platform for my query.