28

I am doing video rendering these days and one thing I am totally confused about is if someone uses a cheap laptop for rendering videos.

  1. Does video rendered on a high-end i7 laptop look better than video rendered on a dual-core laptop? (Does Intel HD graphics matter, which is used in both?)

  2. Does video rendering degrade the processor's performance after a period of time (going 100% for minutes)?

9
  • A more powerful CPU can actually make the video output (slightly) worse if the built-in GPU's video rendering is used rather than a pure software solution. Jul 4 '18 at 8:57
  • 2
    @IgnacioVazquez-Abrams So...where's the difference between your HD4400 in a i5-4210U and the HD4400 in an i7-4600U? Answer: The speed/date of the CPU does not have to do anything with whether there is a GPU integrated (making it an APU, effectively) or not. E.g. my i7-5820k does not offer an integrated GPU - and it certainly is not less powerful than my i5-4210U.
    – flolilo
    Jul 4 '18 at 15:11
  • @flolilolilo: Can, not does. Obviously this isn't a consideration when an integrated GPU isn't present. Jul 4 '18 at 18:02
  • 2
    @IgnacioVazquez-Abrams well, then it is like saying "A less powerful CPU can actually...." What this is about: The release date of the CPU and/or how powerful it is has nothing to do with a software's use of the (sometimes) available on-chip GPU, and the availability of a GPU on the CPU's die has nothing to do with whether the software uses it or not. (Well, except from that if no GPU is there, then it can't be used. But then again, this does nothing to answer the question whether an Atom CPU provides higher quality rendering than a Core X CPU or vice versa.)
    – flolilo
    Jul 4 '18 at 18:08
  • 1
    @IgnacioVazquez-Abrams: I think you're talking about hardware video encoding, e.g. h.264 with Intel Quick Sync instead of a high-quality software encoder like x264. Last I checked, there was a noticeable quality-per-bitrate difference between HW encoders and x264 -preset slower; for video that will be compressed once but sent over the Internet many times, or kept around on disk forever, spending extra CPU time / electricity up front should pay off in the long run. But that's sep from 3D rendering. Jul 6 '18 at 22:00
43

Does i7 render better picture quality than dual core (Does Intel HD graphics matter which is used in both)?

No, it doesn't. They both render in the quality you tell them to, however, rendering is a really computationally-heavy task, so rendering with an i7 will be a lot faster than rendering with a low-end dual-core processor.
And no, the Internal Graphics Processor (Intel HD Graphics in this case) will not matter, since rendering uses only the CPU. However, some applications for rendering might use your IGP (Intel HD) or GPU (your discrete graphics card, if any is present) to render an image, which will lead to a completely different result. Most consumer-grade CPUs do better at rendering than regular GPUs, and a lot, a lot better than IGPs (both in quality [because of better computational algorithms] and in speed, however, this does not apply to this case). So you should keep this in mind, as it is varies from application to application. (credit to @CliffArmstrong for the suggestion)


Does processor degrade after a short amount of time because I use them to render videos? (videos use 100% CPU for minutes)

No, processors do not degrade. They are manufactured so you don't have to change them regularly. Check this answer for more detailed information.


If the application which is currently rendering makes use of multithreading, then newer processors which also have a higher core count would be able to perform the same task a lot faster.
For example, let us say we have a newer 8-core i7 processor and one older regular dual-core processor and let us say that each core has 2 threads. That makes them a processor with 16 threads and a processor with 4 threads. Theoretically, if the application made use of all the cores and we specify the image to be of quality 1080p (Full HD), the i7 processor would theoretically render the image 4 times faster than the dual-core processor (if all cores work at the same frequency in both processors). However, the image quality would still be 1080p, so they will render the same quality image, but in different time.

And while processors are assigned such heavy tasks, they start producing a lot more heat, which is what can be dangerous. Proper cooling is a must-have when performing such tasks, as @Tetsujin mentioned in his answer, or else your CPU could begin to throttle itself down in order to reduce heat.

10
  • 5
    Processing with a high-end processor will be faster, so you may be able to get higher quality within the same time. If you have a tight deadline, that might mean a faster processor translates to better quality. As for the processor "degrading", don't newer Intel processors basically overclock themselves as long as the heat room allows it? Running at 100 % for minutes wouldn't actually damage the processor, but it would slow the processor back down after a while to prevent issues, which is a sort of a performance degradation compared to a shorter task.
    – ilkkachu
    Jul 4 '18 at 10:53
  • 3
    if the comparison is 10-core vs 1-core, then the time reduction could theoretically be at most tenfold (usually considerably smaller). Just nitpicking : )
    – Agent_L
    Jul 4 '18 at 11:06
  • 16
    All doped semiconductor materials suffer from thermal degradation over time but it should take a long time until a CPU fails due to it even if running at maximum load continuously (within its electrical and thermal specification), certainly longer than it takes for the CPU to become on-topic on Retrocomputing. Jul 4 '18 at 13:03
  • 3
    It's worth it to mention that, depending on the transcoding/rendering software used, it may switch to using integrated transcoding capabilities of the GPU/IGP on hardware where this is available... which does come with a loss of visual quality. Most good rendering/transcoding software avoids this by default... as such hardware rendering features are intended for performant playback... not rendering. Jul 4 '18 at 17:05
  • 2
    Out perform, yes. But most consumer grade CPUs with IGPs and discrete GPUs use less precise math for their calculations. It will render the frames faster... but they will lose quality in that color precision, deblocking, etc will be less correct than if using a more precise algorithm. Not noticeable to the average user... but meaningfully for content creators. Jul 4 '18 at 19:22
22

So long as the machine can keep itself cool enough, the only difference will be the time taken.

When rendering video even on a 12-core Xeon, I intentionally ramp the fans up to maximum. Even though the machine is perfectly capable of keeping itself cool, it considers "cool enough" to be 1°C under 'procHot' which is Intel's specified maximum temperature for the processor [98°C for this particular processor, you'd have to check Intel's figures for your own].

I just like to give it a bit more headroom, but maybe that's just me being a little paranoid.

On the other hand, if it can't keep itself under procHot, it will eventually cause short-term crashes/BSODs or even long-term damage.

Cooling is prime when doing intensive tasks.

5
  • 3
    The CPU will reduce its clock frequency and voltage (“throttling“) if it reaches dangerous temperature levels. Of course that’s bad for performance but won’t harm the CPU. Generally it’s really hard to kill modern CPUs and GPUs with overheating. Theoretically high temperatures are bad for the lifetime of silicon but I haven’t seen any numbers on how bad it actually is (i.e. are we talking 3 weeks or 3 years until failure while running at e.g. 90°C continuously?).
    – Michael
    Jul 4 '18 at 15:06
  • 3
    I'd rather not risk it, tbh. I know the theory, but fans are cheap, Xeons are not ;-) I've never actually lost a fan or PSU either, in at least 10 years.
    – Tetsujin
    Jul 4 '18 at 15:09
  • 4
    @Michael The rule of thumb I've heard for integrated circuits (not CPUs specifically) is that you can cut the lifetime in half for each extra 10C. But only Intel knows what temperature/lifetime number we should be starting at.
    – mbrig
    Jul 4 '18 at 17:52
  • The lifetime in half business is specific to electrolytic capacitors, they have a liquid inside that can dry up or otherwise degrade. Electrolytic caps are only used on the PCB where a large amount of low cost capacitance is needed, they are not in or even very near the CPU. CPU life isn't changed significantly until you reach a high temperature threshold(generally over 80-90c, depends on model) that makes the electrons in the insulating materials dance faster than some critical level, or if higher than design voltage is used. (either triggers electrical "breakdown", usually irreversible)
    – Max Power
    Jul 4 '18 at 18:58
  • @Michael I burnt a core on my i7 6700k after one year, so dont think it's impossible. Intel was great with RMA though! Also, the OS must be aware and throttle the CPU, ESXi doesn't do this always for example which is why i moved to KVM. Jul 6 '18 at 8:41
13

When running the exact same software encoder (program) with exactly the same options and configuration on two different processors you will get the exact same result. The only difference will be time taken to do the encoding.

Using an exact same program with exact same configuration with exact same input should give the same output quality when run on a Xeon, an i7, an i3 or even a Celeron processor.

If you use the built-in hardware video encoders or decoders then you may get different results as they might be set up or optimised differently between processor generations and newer hardware may support newer features. In the same way that using a 5 year old copy of ffmpeg might be slower or yield slightly different results for a given configuration than a newer version, the different hardware video encoders can be thought of as equivalent to different versions of the "software", albeit versions that cannot be upgraded without replacing the hardware.

The processor itself will not likely degrade but as the processor runs hotter the fans will run harder, the power supply will work harder and overall the system will work harder and hotter than it otherwise would if you weren't doing the encoding. In theory this extra work could be thought to be putting an extra strain on your system but in practice your system should be designed well enough that the difference between you using it in this fashion and not using it at all should mean that the working lifetime of the system will be as near the same as makes no difference.

If you have a power supply or cooling system that is not designed or specified well enough to match the load of your system then you might cause a failure sooner than they otherwise should.

Running demanding tasks on an underpowered PSU may cause it to overheat and burn out components within the PSU, or it might "brown out" causing system instability. Unless you bought a bargain basement pre-built machine or built it yourself with the smallest supply you could find this should not be the case.

3
  • Hello @Mokubai, I am using ffmpeg's compile binary from ffmpeg.org on windows plateform, Please check if running command in ffmpeg use hardware encoding. Jul 4 '18 at 12:33
  • 2
    Unless your command line specifically mentions "nvenc" (nvidia), libmfx (Intel) or one of the other hardware encoders mentioned at trac.ffmpeg.org/wiki/HWAccelIntro then I believe it will use the software libx264 encoder. Ffmpeg defaults to software encoding I believe in which case you should get the same result across platforms. Once you start enabling hardware encoders the encoding may be a lot faster, but your results will vary.
    – Mokubai
    Jul 4 '18 at 12:59
  • Not really; x264, for example, might produce non-deterministic output when using more than 1 thread, and there's also a chance of a very tiny quality/size degradation when using more than 1 thread, which increases with more threads, but it's just theory, not worth worrying about. still, the files won't be exactly the same at the bitstream level. So if one's using a deterministic encoder, that's true, but not all of the multithreaded encoders are deterministic. Jul 8 '18 at 9:19
3

On a laptop, generally no. However many laptops are not built to last. The CPU may not degrade but something will. It is abusive to use a laptop in this way, even a "gaming" laptop.

Running over voltage, and over nominated clock rates will shorten the life of many workstation and desktop components. This applies to graphics cards too. This is not necessarily a degradation for a CPU, but a failure.

A GPU can leave the factory with latent issues and working them hard can reveal the faults. this is why we have ECC Ram in graphics cards now. I'm not going to mention any brands but there is a reason there is a warranty.

2
  • Out of curiosity, are these "latent issues" related to "error 43"?
    – user541686
    Jul 6 '18 at 8:05
  • Actually the reason for ECC RAM is that people do mucho GPU compute these days, and unlike in graphics where nobody notices a bit error (nobody hardly even notices when floating point math is several bits short in precision, or when derivatives are calculated incorrectly), when doing compute tasks, you want a little more reliability. Though I agree with your general stance that there certainly is wear and tear on processors when they run at 100% load, especially on laptops.
    – Damon
    Jul 6 '18 at 9:01
-2

This will depend on how much of the computation happens on the CPU and how much happens on the GPU.

In general, CPUs will do more of the serial work where a lot of branching happens and GPUs will do more of the work that's performing the same operation on a large amount of data (i.e. on every pixel).

Also, the amount of cores only help if the rendering makes use of multiple cores. A lot of applications don't fully utilize all cores. So an 8-core processor (or quadcore with hyperthreading) will almost never give an 8-fold increase in speed.

An application that's not optimized at all for multithreading will not even get a speedup at all.

To answer your questions:

  1. No, it will look exactly the same, as the exact same operations are performed.

  2. Considering what I wrote above, it depends on if the temperature of your CPU increases above a certain threshold, which will cause the CPU to tune itself down to not increase the temperature any further. So if the CPU is doing a lot of work it will slow down after some time of full load, especially in laptops (small case, bad cooling). If by degrading you mean long-term degradation, then refer to the answers above (tl dr; they don't degrade by much).

1
  • 2
    your answer does not answer OP's question, in particular will there be a loss in quality. It's not a question about speed. Jul 5 '18 at 5:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.