VOGONS


Reply 20 of 48, by Tertz

User metadata
Rank Oldbie
Rank
Oldbie

It's better to use video examples available externally so other people could add their results too.
There may to be free video clips (advertisings, made by you, public property). Some countries may allow to distribute for non-commercial exploratory/study use short clips from movies (up to X seconds).

DOSBox CPU Benchmark
Yamaha YMF7x4 Guide

Reply 21 of 48, by Standard Def Steve

User metadata
Rank Oldbie
Rank
Oldbie

-Added MPEG2 @ 1080p

Azarien wrote:

You could also add 640x360, as this was a typical resolution in PIII/P4 heyday. (480p was considered high quality back then)

I probably will do a lower resolution/lower frame rate Xvid test to give the K6 processors a better chance. Something sourced from a 24fps movie DVD instead of a 30fps NTSC DVD.

Tertz wrote:

It's better to use video examples available externally so other people could add their results too.
There may to be free video clips (advertisings, made by you, public property). Some countries may allow to distribute for non-commercial exploratory/study use short clips from movies (up to X seconds).

I'll soon be posting VP9/YouTube results. 😀

P6 chip. Triple the speed of the Pentium.
Tualatin: PIII-S @ 1628MHz | QDI Advance 12T | 2GB DDR-310 | 6800GT | X-Fi | 500GB HDD | 3DMark01: 14,059
Dothan: PM @ 2.9GHz | MSI Speedster FA4 | 2GB DDR2-580 | GTX 750Ti | X-Fi | 500GB SSD | 3DMark01: 43,190

Reply 22 of 48, by Standard Def Steve

User metadata
Rank Oldbie
Rank
Oldbie

Added a chart for 1080i60 H.264. This is kind of an ultimate 1080 directshow test for these CPUs. Far more punishing than the regular 1080p24 stuff. It features very high bit rates (35 mb/s), software deinterlacing and decoding, higher frame rates, and of course lossless audio decoding.

P6 chip. Triple the speed of the Pentium.
Tualatin: PIII-S @ 1628MHz | QDI Advance 12T | 2GB DDR-310 | 6800GT | X-Fi | 500GB HDD | 3DMark01: 14,059
Dothan: PM @ 2.9GHz | MSI Speedster FA4 | 2GB DDR2-580 | GTX 750Ti | X-Fi | 500GB SSD | 3DMark01: 43,190

Reply 23 of 48, by rick6

User metadata
Rank Member
Rank
Member

Great work.
I remember back in 2001 or so when i was quite a bit into Unreal Tournament and still rocking a AMD k6-2 450Mhz, and was unable to play smoothly the mythical "Fooze Frag movie" without video lagging behind audio! I really hated that situation and was the first time i felt i was loosing touch with the gaming community because of my aging machine!

My 2001 gaming beast in all it's "Pentium 4 Williamate" Glory!

Reply 24 of 48, by Standard Def Steve

User metadata
Rank Oldbie
Rank
Oldbie

Added YouTube/VP9 results!

P6 chip. Triple the speed of the Pentium.
Tualatin: PIII-S @ 1628MHz | QDI Advance 12T | 2GB DDR-310 | 6800GT | X-Fi | 500GB HDD | 3DMark01: 14,059
Dothan: PM @ 2.9GHz | MSI Speedster FA4 | 2GB DDR2-580 | GTX 750Ti | X-Fi | 500GB SSD | 3DMark01: 43,190

Reply 25 of 48, by falloutboy

User metadata
Rank Member
Rank
Member

Thank you for doing all this tests! 😀

result 2160p with i7-720QM

Despite less than 100% CPU usage, motion was not at all smooth. There appears to be a hidden bottleneck here.

My theory is
100% load on 4 real cores and 0% on the HT cores will show you an overall CPU load of 50% (I think).
Using HT cores doesn't mean it will be twice as fast as before.
In worst case, an overall load of 50% can be the maximum for what a CPU with HT is able to perform.
I don't know how useful the HT cores are for video playback but it would be interesting to see some results without them (maybe only for one CPU).

Reply 26 of 48, by Standard Def Steve

User metadata
Rank Oldbie
Rank
Oldbie

Added two new CPUs to the VP9 chart: the AMD A10-8700P mobile APU and Atom X7 Z8700. Both are quad core parts. The A10 runs at 1.8GHz with a max turbo frequency of 3.2GHz; the Atom runs at 1.6GHz with a max turbo speed of 2.4GHz.
I thought for sure that the A10 would crush the Atom, but it was actually a pretty close race. A10 was only slightly faster than the Atom, and both were outperformed by a Core 2 Duo E7500.

A10 and Atom both perform significantly better in IE/Edge, thanks to hardware accelerated video decoding (YouTube streams standard H.264 to IE/Edge).

falloutboy wrote:
Thank you for doing all this tests! :) […]
Show full quote

Thank you for doing all this tests! 😀

result 2160p with i7-720QM

Despite less than 100% CPU usage, motion was not at all smooth. There appears to be a hidden bottleneck here.

My theory is
100% load on 4 real cores and 0% on the HT cores will show you an overall CPU load of 50% (I think).
Using HT cores doesn't mean it will be twice as fast as before.
In worst case, an overall load of 50% can be the maximum for what a CPU with HT is able to perform.
I don't know how useful the HT cores are for video playback but it would be interesting to see some results without them (maybe only for one CPU).

You might be right there. The virtual cores may indeed be running out of steam before Windows thinks they should be. However, HT definitely seems to helps with video decoding. It helped the P4-520 @ 3.73 jump ahead of the other single-core CPUs in all of the DirectShow tests.

P6 chip. Triple the speed of the Pentium.
Tualatin: PIII-S @ 1628MHz | QDI Advance 12T | 2GB DDR-310 | 6800GT | X-Fi | 500GB HDD | 3DMark01: 14,059
Dothan: PM @ 2.9GHz | MSI Speedster FA4 | 2GB DDR2-580 | GTX 750Ti | X-Fi | 500GB SSD | 3DMark01: 43,190

Reply 27 of 48, by amadeus777999

User metadata
Rank Oldbie
Rank
Oldbie

Awesome list!
As for hyper threading - Agner Fog has done some extensive research regarding how it behaves... although not in direct relation to video decoding.

http://www.agner.org/optimize/blog/read.php?i=6

Reply 28 of 48, by SPBHM

User metadata
Rank Oldbie
Rank
Oldbie

regarding VP9, you can force youtube on chrome to use h264 (and you would want to do that if you have a GPU with fixed function acceleration and slow CPU, but h264 on chrome is a lot more demanding than running it on a program like MPCHC) by using a simple extension called h264ify, since only the GTX 950/960 (GM206) with forcerware 361 I think actually supports VP9, if you don't have a GPU capable of H264 acceleration it might be better to use VP9 anyway (it got a little lower CPU usage on the limited test I've done with a t4400 laptop with software only)

also, all the Radeons that I have (4670, 5570, 5850) can't handle youtube 1080P 60FPS with h264 acceleration (720p60 works well, same for 1080p30), so it might be worth using VP9 (with the CPU) for those.

VP9 4K60FPS on youtube can be brutal my 3Ghz sandy Bridge i5 can't handle it well enough, I suppose an i7 2600 would handle it well enough.

and yes, for the older hardware it's probably best to use older software with better support, trying to decode DVD with MPCHC gave me some not so great result with slower CPUs,

I thought CoreAVC would have a larger advantage, I remember it doing some miracles back in 2007-2008 with my A64 X2 trying to run some demanding 1080P samples, but once I started using GPU acceleration (from a 8600GT) it was no longer needed, also h264 when it works is pretty amazing, I was running XBMC and MPCHC at some point with a p3 650 and the 8400GS PCI and it could run some HD h264 videos quite fine with low CPU usage due to the GPU (but it was totally useless for web browser videos, even with fast CPU, I think because of the PCI bandwidth)

Reply 29 of 48, by washu

User metadata
Rank Newbie
Rank
Newbie
Standard Def Steve wrote:

Added two new CPUs to the VP9 chart: the AMD A10-8700P mobile APU and Atom X7 Z8700. Both are quad core parts. The A10 runs at 1.8GHz with a max turbo frequency of 3.2GHz; the Atom runs at 1.6GHz with a max turbo speed of 2.4GHz.
I thought for sure that the A10 would crush the Atom, but it was actually a pretty close race. A10 was only slightly faster than the Atom, and both were outperformed by a Core 2 Duo E7500.

The reason the A10-8700P didn't crush the Atom is that it is really only a dual core chip while the Atom is a true quad core. The A10 only has 4 integer units and data L1 caches, the rest of the key components like the instruction decoder, branch predictor, and FPU/SIMD unit only have 2 each. Video decoding is pretty FPU/SIMD heavy which puts the A10 at a disadvantage.

Reply 30 of 48, by Standard Def Steve

User metadata
Rank Oldbie
Rank
Oldbie

I didn't even know YouTube had 2160p/60 video already. I'm going to have to give that a shot!
Yeah, CoreAVC used to be much faster than FFMPEG, but the gap is much smaller now.
Expansion bus bandwidth is definitely a thing with newer versions of Windows. I remember testing a PCI Radeon x1550 on a 3.33GHz Core 2 Quad. That card struggled with full screen DVD playback (at 2560x1440)!

washu wrote:
Standard Def Steve wrote:

Added two new CPUs to the VP9 chart: the AMD A10-8700P mobile APU and Atom X7 Z8700. Both are quad core parts. The A10 runs at 1.8GHz with a max turbo frequency of 3.2GHz; the Atom runs at 1.6GHz with a max turbo speed of 2.4GHz.
I thought for sure that the A10 would crush the Atom, but it was actually a pretty close race. A10 was only slightly faster than the Atom, and both were outperformed by a Core 2 Duo E7500.

The reason the A10-8700P didn't crush the Atom is that it is really only a dual core chip while the Atom is a true quad core. The A10 only has 4 integer units and data L1 caches, the rest of the key components like the instruction decoder, branch predictor, and FPU/SIMD unit only have 2 each. Video decoding is pretty FPU/SIMD heavy which puts the A10 at a disadvantage.

Yeah, I guess the A10 isn't fully a quad core, but damn... I guess I just expected AMD's top mobile APU to distance itself a little more from one of Intel's slowest processors. Especially with such a high Turbo frequency.

P6 chip. Triple the speed of the Pentium.
Tualatin: PIII-S @ 1628MHz | QDI Advance 12T | 2GB DDR-310 | 6800GT | X-Fi | 500GB HDD | 3DMark01: 14,059
Dothan: PM @ 2.9GHz | MSI Speedster FA4 | 2GB DDR2-580 | GTX 750Ti | X-Fi | 500GB SSD | 3DMark01: 43,190

Reply 32 of 48, by washu

User metadata
Rank Newbie
Rank
Newbie
Standard Def Steve wrote:

Yeah, I guess the A10 isn't fully a quad core, but damn... I guess I just expected AMD's top mobile APU to distance itself a little more from one of Intel's slowest processors. Especially with such a high Turbo frequency.

There's a few more things at play here than just core marketing BS.

1. The current batch of Carrizo APUs including your A10 have variable TDPs from 12 to 35W. If the laptop you are using is built towards the lower end of that range then there is no headroom to use Turbo.

2. Despite it being AMD's current top mobile APU, it is still just a revision of the terrible bulldozer architecture. It's per clock performance is lower than many older architectures. If it can't clock up because of the laptop design then you are left with a low clock low IPC chip. Just like the P4 which AMD should have learned from, bulldozer derived chips need high clocks to perform at all well.

3. The Atom may be one of Intel's slowest current processors, it's not the same old Bonnell based chips that deservedly got a reputation for being slow. The Silvermont/Airmont Atoms like yours are much faster even if they are the current bottom of the barrel from Intel.

You basically have a very poorly designed, slow hot chip shoehorned into a laptop where it will be thermally constrained to lower than its potential. No wonder it doesn't perform that great.

I think a far more interesting comparison would be a Carrizo-L VS the Atom. Unlike the Carrizo, the "L" version is based off the Jaguar architecture and is designed for low end laptops and tablets like the Atom. They are true quad cores as well.

Reply 33 of 48, by Standard Def Steve

User metadata
Rank Oldbie
Rank
Oldbie
SPBHM wrote:

I've used this video for 4k60
https://www.youtube.com/watch?v=SXZ3qWAvbVs

(it goes up to 4K60 with VP9, with h264 only 1080P60)

Thanks for the link!
Added 2160p60 to the VP9 chart.

washu wrote:

1. The current batch of Carrizo APUs including your A10 have variable TDPs from 12 to 35W. If the laptop you are using is built towards the lower end of that range then there is no headroom to use Turbo.

2. Despite it being AMD's current top mobile APU, it is still just a revision of the terrible bulldozer architecture. It's per clock performance is lower than many older architectures. If it can't clock up because of the laptop design then you are left with a low clock low IPC chip. Just like the P4 which AMD should have learned from, bulldozer derived chips need high clocks to perform at all well.

Good point. I was watching CPU frequency via Win10's task manager during video playback. I'm not sure how accurate it is, but saw the A10 hit 3+GHz only a few times. It spent most of its time running at around 2.3-2.5. The Atom seemed comfortable sustaining 2.2GHz, which isn't bad at all for being stuck in a fanless Surface 3.

But as poor as the A10's CPU performance/watt is, the IGP is a big step up from the Intel HD graphics in the Atom.

P6 chip. Triple the speed of the Pentium.
Tualatin: PIII-S @ 1628MHz | QDI Advance 12T | 2GB DDR-310 | 6800GT | X-Fi | 500GB HDD | 3DMark01: 14,059
Dothan: PM @ 2.9GHz | MSI Speedster FA4 | 2GB DDR2-580 | GTX 750Ti | X-Fi | 500GB SSD | 3DMark01: 43,190

Reply 34 of 48, by Standard Def Steve

User metadata
Rank Oldbie
Rank
Oldbie

Added two more CPUs to most of the charts:

-An older Atom N470 (1.83GHz Pine Trail, single-core w/HT, in-order Bonnell architecture). Man was this CPU weak.
-A Phenom II X6 1055T, slightly overclocked to 3.1GHz

P6 chip. Triple the speed of the Pentium.
Tualatin: PIII-S @ 1628MHz | QDI Advance 12T | 2GB DDR-310 | 6800GT | X-Fi | 500GB HDD | 3DMark01: 14,059
Dothan: PM @ 2.9GHz | MSI Speedster FA4 | 2GB DDR2-580 | GTX 750Ti | X-Fi | 500GB SSD | 3DMark01: 43,190

Reply 35 of 48, by Standard Def Steve

User metadata
Rank Oldbie
Rank
Oldbie

Added two 4GHz CPUs to most of the charts:
-Core 2 Quad Q6700 overclocked to 4GHz on 1600MHz bus. Running on P45 based motherboard with 8GB dual channel DDR3-1600.
-Replaced the previous Phenom II X6 1055T @ 3.1GHz with a Phenom II X6 1090T running at 4070MHz with 16GB of DDR3-1760 memory. This CPU had no problem handling 30fps 2160p VP9 YouTube video! But even with six cores running at 4GHz, it couldn't handle those brutal 2160p/60fps clips!

For the HTML5 VP9 tests, these new CPUs were tested with Chrome 50. However, the HTML5 video performance of the latest version of Chrome seems to be identical to that of v47.

P6 chip. Triple the speed of the Pentium.
Tualatin: PIII-S @ 1628MHz | QDI Advance 12T | 2GB DDR-310 | 6800GT | X-Fi | 500GB HDD | 3DMark01: 14,059
Dothan: PM @ 2.9GHz | MSI Speedster FA4 | 2GB DDR2-580 | GTX 750Ti | X-Fi | 500GB SSD | 3DMark01: 43,190

Reply 37 of 48, by Standard Def Steve

User metadata
Rank Oldbie
Rank
Oldbie

Added an MPEG1/VCD chart!

P6 chip. Triple the speed of the Pentium.
Tualatin: PIII-S @ 1628MHz | QDI Advance 12T | 2GB DDR-310 | 6800GT | X-Fi | 500GB HDD | 3DMark01: 14,059
Dothan: PM @ 2.9GHz | MSI Speedster FA4 | 2GB DDR2-580 | GTX 750Ti | X-Fi | 500GB SSD | 3DMark01: 43,190

Reply 38 of 48, by Standard Def Steve

User metadata
Rank Oldbie
Rank
Oldbie

Added the first 4K DirectShow chart: 24fps H.264 at 58 mb/s.

Will also be adding the following 4K DirectShow tests in the near future:
-60fps H.264 at 100 mb/s
-24fps HEVC
-60fps HEVC

P6 chip. Triple the speed of the Pentium.
Tualatin: PIII-S @ 1628MHz | QDI Advance 12T | 2GB DDR-310 | 6800GT | X-Fi | 500GB HDD | 3DMark01: 14,059
Dothan: PM @ 2.9GHz | MSI Speedster FA4 | 2GB DDR2-580 | GTX 750Ti | X-Fi | 500GB SSD | 3DMark01: 43,190

Reply 39 of 48, by Standard Def Steve

User metadata
Rank Oldbie
Rank
Oldbie

-Added the first HEVC chart and it's a brutal one: 10-bit 2160p at 60fps with an average bitrate of 62.5 mb/s.

Not even the 4GHz Phenom X6 could handle this one. Software HEVC decoding appears to benefit hugely from SSE4 support. I wish I had some LGA1366 equipment I could test. I remember the first-gen i7s being not that much faster than Penryn clock-for-clock, at least when they were first released. However they did have 8 threads and full SSE4 support. It Would be interesting to see how well they'd do at this test.

P6 chip. Triple the speed of the Pentium.
Tualatin: PIII-S @ 1628MHz | QDI Advance 12T | 2GB DDR-310 | 6800GT | X-Fi | 500GB HDD | 3DMark01: 14,059
Dothan: PM @ 2.9GHz | MSI Speedster FA4 | 2GB DDR2-580 | GTX 750Ti | X-Fi | 500GB SSD | 3DMark01: 43,190