Rendering - Old CPUs vs New

[Shephard]

Summary

Quad Cores:
i7-3700 @ 3.4ghz w/ hyper-threading - 4 minutes
i7-3570k @ 3.4ghz - 8 minutes 40 seconds
A10-5800k @ 3.8ghz - 13 minutes 17 seconds
Core 2 Quad Q8200 @ 2.3ghz - 17 minutes 26 seconds.

Dual Cores:
Core 2 Duo T7200 (mobile CPU) @ 2.0ghz - 28 minutes
AMD X2 4000+ @ 2.1ghz - 45 minutes
Core Duo T2500 (mobile CPU) @ 2.0ghz - 52 minutes

Avidemux/Handbrake Settings:
Output - Mpeg4 (x264) - Quality @ 18.
Audio - Copy
Container - .Mkv
Priority - Normal

- encoding 13 minute long 17gb raw Starcraft 2 footage.

I said I was going to make this topic and do some comparisons so here it is!

There is no debate that more cores help with multi-threaded tasks like video rendering. How big are improvements from a dual core to a quad core and architecture improvement..? Time to do some real world tests!

I will be comparing some old and newer CPUs to see how the cores and architecture really help for render times.

For the tests I am using...

Windows 7 64 bit, Mac OS X, and Linux.
Avidemux or Handbrake
My own footage
No other programs running to disrupt tests.

---

The first battle is between the i5 3570k @ stock clocks vs an AMD X2 4000+ at stock clocks.

The 3570k is recording to a 7200 RPM SATA 3 WD Black 1TB.

The AMD X2 4000+ is recording to a 7200 RPM IDE Samsung 200gb.

 

[Ayah]

Haha, not a fair fight!

 

[Shephard]

3570k finished in 8:40.

 

[Shephard]

AMD X2 4000+ finished in 45:00.

 

[Sequences123]

Highest score wins? ^.^

 

[Durvelle27]

not a fair fight at all

more like

Core 2 Quad vs i5

Core 2 Duo/ AMD x2 vs i3

 

[Shephard]

Obviously not a fair fight. The point is some people may not realize how much the architecture has improved a long with additional cores.

Next up will be just that... Core 2 Quad q8200.

 

[Hulk]

Very interesting! Thanks for doing the legwork on this. It's always fun to look back as the next architecture is about to be released.

 

[SiliconWars]

If the Athlon had four cores and the same 3.4 GHz clock speed they'd be quite close with 100% scaling (which is an unreasonable assumption, but worth noting anyway).

 

[inf64]

I like topics like this since they show the true value of upgrade . Good work and I' looking forward to C2Q vs i5 comparison.

 

[Imouto]

Different quantizer.

Sometimes I wonder why ppl do stuff like this.

 

[skipsneeky2]

Used to love to bench how much faster my q6600@3ghz would encode over my e6750 using crappy ass windows media encoder LOL,i would swap between the chips cause honestly i fell in love with my e6750 and it was a cool running kick ass chip,but in the end the q6600 laid the smackdown and a game or two of 2007 did make use of those 2 extra cores so adios went the e6750.

 

[Blitzvogel]

If the Athlon had four cores and the same 3.4 GHz clock speed they'd be quite close with 100% scaling (which is an unreasonable assumption, but worth noting anyway).

Uh probably not. IIRC Athlon x2 4000s have 64 bit FPUs. The i5 accepts 256 bit AVX in a single pass IIRC.

 

[SlickR12345]

Cool, I'd like to see the results. Maybe do from an older 4 core generation like Q9200?

 

[Headfoot]

Different quantizer.

Sometimes I wonder why ppl do stuff like this.

Because this is the performance difference real people actually see when they upgrade.

 

[wlee15]

Uh probably not. IIRC Athlon x2 4000s have 64 bit FPUs. The i5 accepts 256 bit AVX in a single pass IIRC.

x264 doesn't use FP data, it's all SIMD integer.

 

[Vectronic]

I have no useful statistics, but regarding rendering in 3D, my 3570K even at stock speeds, is about 3 to 5 times faster than my Q6600 2.6GHz, and (very) roughly about 16 to 20 times faster than my P4 3.0GHz.

Time varies depending on software and scene rendered. The main difference that I enjoy possibly more than the increase in speed, is that I can also do other things at the same time... browse YouTube, play a "lite" game, edit something in Photoshop while it's rendering with no noticeable impact on those, and only about a 10% impact on rendering speed.

With the P4, or Q2.6... rendering was rendering, best find something "in real life" to do until it's done.

Edit: by the way, P4 and Q2.6 are still running, and running the same software... not some anti-rose-colored-glasses "back in the day hills both ways" kinda thing. I say "was" because... it's just not practical to use them for that purpose anymore.

 

[Imouto]

Because this is the performance difference real people actually see when they upgrade.

Different settings for the benchmark. That's not something you do when you try to compare 2 devices.

 

[Headfoot]

never mind I see now

 

[Shephard]

Forgot to update the topic here. Got some more comparisons for everyone.

A10-5800k at stock clocks. Recording to a 7200 rpm SATA 6 Seagate.

Finished just a little over the 13 minute mark.

 

[Bearach]

Different settings for the benchmark. That's not something you do when you try to compare 2 devices.

This. You've invalidated your results using different settings.

 

[Piroko]

Been a while since I did encodes myself, but if I remember correctly the quantiser doesn't have that much of an impact (higher quantiser might take a bit longer, again iirc). You can't draw definite conclusions from that test, but it does give a general idea of the performance you can expect.

And yeah, that is quite the progress. It shows that x264 is very well suited for IB, more than five times the performance is definitely a feat.

 

[Shephard]

I am using the exact same settings for all the rendering. Screenshots are there as proof.

 

[Bearach]

Been a while since I did encodes myself, but if I remember correctly the quantiser doesn't have that much of an impact (higher quantiser might take a bit longer, again iirc). You can't draw definite conclusions from that test, but it does give a general idea of the performance you can expect.

And yeah, that is quite the progress. It shows that x264 is very well suited for IB, more than five times the performance is definitely a feat.

Lower number is better, and it does have an impact. Not much but some. If the settings the OP is using is variable then that may explain the different quantiser.

 

[Shephard]

i7-3770 with hyper-threading @ stock settings running Mac OSX Mountain Lion. This is the top of the line iMac you can buy.

It took 20 minutes to render. The A10 and 3570k beat it.

So why did it take so long? A 3770 with hyper-threading should demolish the 3570k, nevermind the A10. Maybe the OS is to blame?