Dual Core or single does it really matter for captures. Which mainboard?

jagabo · Member Joined: 09 Dec 2005 Location: none

CPU benchmarks: http://www23.tomshardware.com/cpu.html

gll99 · Renegade Joined: 31 May 2002 Location: Canadian Tundra

redwudz · Mod Neophyte Joined: 07 Sep 2002 Location: AZ, USA

I get about 16.959 sec with my HTPC, 5200+ Athlon X2 AM2 OC'd to 3Ghz. My HP laptop with a AMD 1500+ 1.33Ghz CPU took 86.112 secs.

When I reboot in a while, I'll see what difference the HTPC does with no OC.

Thanks, Soopafresh, that's a handy little program.

Soopafresh · Dismember Joined: 01 Jan 2004 Location: United States

Thanks.

The script will provide us with very rough values as to transcoding speed of different procs. Hardly scientific, but good enuff fer us city folks

Dual Xeon 2.8 Proc : 22 secs

chesterfield · Member Joined: 04 Jan 2006 Location: United States

ok, this thread has got me curious. i just built a core 2 duo E6600, 4 gig ram, vista ultimate, ati 2600xt video card, win tv 150 cap card, gigabyte MOBO, 2 250 gig WD HD's in RAID 0, three dvd burners, two 500 gig storage drives.....and my time was 13.604 secs---see screen shot---is this good??? can it be better???

gll99 · Renegade Joined: 31 May 2002 Location: Canadian Tundra

There is no direct comparison between my system and the E4400 but comparing a E4400 to a P4 3Ghz shows that the E4400 is almost twice as fast.

http://www.vertexmonkey.com/benchmark_scene_203_modoLogoMany.php

The Opteron 275 is about 30 to 35% faster than the E4400 on 2 tests.

This site shows similar differences using the same os.

http://www.eofw.org/bench/

The numbers don't quite match up for video compression as shown by the script benchmark here.

Suppose it's true that the Opteron 275 is 30% faster than the E4400 then that would mean that the Core2 duo E4400 would take about 22 seconds on this test. But the E4400 shows to be twice as fast as a P4 3Ghz so I'd expect the P4 to take over 40 seconds to complete yet my P4 1.6 (clocked at 2159Ghz) comes in around 29 seconds. According to those number my P4 1.6 should take 50 to 70 seconds. So that part doesn't make sense.

This would mean that running the Core 2 E4400 at stock ~~1.8Ghz~~it 2.0Ghz would be about 30 to 35% faster than my oc'd P4. All things considered running a 3.2Ghz P4 (30% more Ghz) and encoding xvid/divx should achieve the same numbers as the E4400. Yet that seems to go against expectations. The only way the numbers would line up is if the E4400 and a P4 3.2Ghz results were about the same at 30% slower than the Opteron 275 which would mean roughly 22 seconds and my P4 1.6 somewhere between 29 to 33 seconds.

Unfortunately on the site jagabo provided I couldn't find an Opteron 275 to compare with the E4400 for xvid/divx encode.

If anyone has a Intel Core 2 even if its a E2160 or E6x00 can you run the script to see what you get some idea. I'd really like to see the E4400 results but can probably extrapolate from a similar cpu by adjusting for the difference in speed.

Edit:

Good just saw the E6600 after posting. I went to eat before hitting send

gll99 · Renegade Joined: 31 May 2002 Location: Canadian Tundra

@chesterfield
Is that stock speed or are you overclocking?

jagabo · Member Joined: 09 Dec 2005 Location: none

This is a really poor benchmark. I'll tell you what I got on my Core 2 Duo E6300 anyway: anywhere from about 12 to 24 seconds.

The benchmark creates a file full of raw YUV video (UYVY I think), converts it to an uncompressed RGB AVI file, then converts the RGB AVI file to an MSMPEG4 compressed MP4 file. Only the last conversion is timed.

The reported conversion time is very erratic because the two intermediate files are still being to the hard drive (Windows' write behind caching) as the timed step is being run! It's as if you were performing file copies with Explorer while converting. Performance will vary greatly depending on how fast your drives are, how fragmented they are, any other disk activity that might be going on, how much free memory you have, etc.

If you run the benchmark repeatedly, without deleting the intermediate video files, the YUV file is recreated but the second program warns you that the AVI file already exists. If you press ESC it skips the conversion (the next step gets the existing AVI file), if you press ENTER it recreates the AVI file. Runtime of the benchmark varies depending on which of these options you choose and how long you wait before pressing ESC or ENTER (because the prior file is still being written to the drive).

During the last step, the step which is timed, ffmpeg converts the uncompressed RGB AVI file to MPEG 4. Given the size of the AVI file, about 316 MB, people with more than ~512 MB of memory will probably have the AVI file completely cached in memory while the benchmark is being run. People with ~512 MB or less will end up with disk thrashing and get much slower results.

There is something wrong with the AVI file. It doesn't render properly in any program I could find.

There is something wrong with the MP4 file. It doesn't render properly in anything either.

I don't trust a benchmark which cannot generate proper results. If it doesn't generate something verifiable, who knows what problems it may have internally. I suppose we could all get together and run binary comparisons of the output results to verify that each processor generated the same results.

The benchmark uses the MSMEG4 codec. This is a very old, single threaded encoder. It takes no advantage of multiple processors. If you add "-threads 2" to the command line it will abort with an error message saying the selected codec doesn't support multithreading.

If you run just the last step, the timed step, the timing is very repeatable. I get about 12.1 +/- 0.1 seconds. I can run two at the same time and get about 13 seconds each.

aedipuss · aBigMeanie Joined: 25 Oct 2005 Location: 666th portal

my old 32 bit athlon 3200+ averages 30 seconds.
_________________
--
"a lot of people are better dead" - prisoner KSC2-303

chesterfield · Member Joined: 04 Jan 2006 Location: United States

@gll99---no over clocking at all and a ton of programs running in the background...

budz · Posted: Sep 03, 2007 20:32

gll99 · Renegade Joined: 31 May 2002 Location: Canadian Tundra

@budz
That would make it 2 times faster than my p4 at almost the same cpu speed. I have one stick 512 ddr333 (pc2700) Both my drives are ide 7200 (300, 200) with room to spare. Our cpu's are oc'd almost at the same Ghz. If you're results are normal with the E2160 at this speed then that would mean that a core 2 cpu would have twice the processing speed of a comparably set (Ghz) older single core P4 cpu. That might even improve with the higher E6x00 models because of the cache and different chip-set but it is still what I was hoping to see. Even when I reran mine I deleted the in between yuv, avi and mp4 files so except for the first time my results were very similar at roughly 28.5 to 29 seconds.

chesterfield's E6600 runs at 2.4GHz and he got 13.6 seconds so again that's consistent

The one that falls out of line is jagabo with such a wide variation. My runs have been quite similar to each other. So it's a bit puzzling.

At least there is something to build on here. If someone else gets similar results with a core 2 then, in spite of other possible differences, that will tend to confirm the pattern. If people start getting erratic results then I'm back to square one and I won't know what to buy.

jagabo · Member Joined: 09 Dec 2005 Location: none

Obviously you should get a Core 2 Duo. Just about every modern video encoder is multithreaded and can take good advantage of two cores. If you can afford a little more you should get a Core 2 Quad (the Q6600 is selling for under US$300 now).

budz · Posted: Sep 03, 2007 21:45

This one is from my Intel Conroe 2.13ghz overclocked to 2.60ghz.

=== ffmpeg -i file.avi -vcodec msmpeg4 -an -s 1280x720 -y output.mp4 ===
FFmpeg version SVN-r6373, Copyright (c) 2000-2004 Fabrice Bellard
FLV VP6 mux version. Patch by Benjamin Larsson, compile by Klaus Post
configuration: --enable-memalign-hack --enable-avisynth --enable-swscaler --e
nable-mp3lame --enable-gpl
libavutil version: 49.0.1
libavcodec version: 51.16.0
libavformat version: 50.5.0
built on Sep 29 2006 19:38:14, gcc: 3.4.5 (mingw special)
Input #0, avi, from 'file.avi':
Duration: 00:00:08.8, start: 0.000000, bitrate: 292751 kb/s
Stream #0.0: Video: rawvideo, bgr24, 720x576, 29.41 fps(r)
Output #0, mp4, to 'output.mp4':
Stream #0.0: Video: msmpeg4, yuv420p, 1280x720, q=2-31, 200 kb/s, 29.41 fps(c)

Stream mapping:
Stream #0.0 -> #0.0
[msmpeg4 @ 00723E30]removing common factors from framerate
[mp4 @ 0072F4F4]Warning, using MS style video codec tag, the file may be unplaya
ble!
frame= 260 q=24.8 Lsize= 33085kB time=8.8 bitrate=30660.2kbits/s
video:33082kB audio:0kB global headers:0kB muxing overhead 0.010199%

Execution time: 14.236 s

gll99 · Renegade Joined: 31 May 2002 Location: Canadian Tundra

@budz what model is that?

redwudz · Mod Neophyte Joined: 07 Sep 2002 Location: AZ, USA

I redid the test. I agree, It's 'quick and dirty', but Soopafresh seemed to make that clear up front.

I made sure nothing was running and ran it from my secondary drive. 5 passes overclocked and 5 at stock settings. Delete the output files each time. The overclock was just by raising the FSB from 200 to 230Mhz. No other changes:

OC'd . at.. 3.0Ghz Average time: 12.208 secs

Non-OC at 2.6Ghz Average time: 13.595 secs

So if my math is correct, I got about a 10% increase in speed between Non-OC and OC for about a 15% increase in CPU speed. It's difficult to equate that with real-world encoding, but in only this example, that would mean a one hour encode would speed up by about six minutes. Nothing spectacular.

I'm not planning to do any other benchmarks because the PC 'is what it is' and works just fine. The OC is very stable and my temps are low enough, so I'll leave it in place.

budz · Posted: Sep 03, 2007 21:55

budz · Posted: Sep 03, 2007 22:17

jagabo · Member Joined: 09 Dec 2005 Location: none

If you're anti Intel, the Athlon 64 X2 6000 (~US$170) is another popular choice these days.

gll99 · Renegade Joined: 31 May 2002 Location: Canadian Tundra

@Budz

When compared to your E2160, the E6400 cpu has a 1066MHz FSB, double the cache and you upped it to 2.6Ghz but got only a modest increase in the test. Based on specs you would have thought the difference would be greater. I suppose there's enough variations in the Benchmark to account for this. Give or take a margin of error, this result still fits in the general pattern of the other core 2.

I'm trying to keep costs down so I don't buy more than I need.

This week it's down to crunch time. Unless one of 3 local sellers shows me otherwise I plan to go with a Asrock mobo. Not great for high overclocking but fine in the zone I plan to OC and I need the other flexibility it offers.

As far as a cpu, I was thinking of the E4400 but I can't ignore the results you got with the E2160. It's not much but is the E4400 worth the extra $20?

If I understood right:

The E2160 is Conroe core while the E4400 is Allendale core. How well do these cores compare. Does it matter?
The E2160 L2 cache is 1M per core while the E4400 has a 2 MB cache per core
The E2160 is clocked at 1.8Ghz and the E4400 is rated at 2.0Ghz
The E2160 has a 9X multiplier and the E4400 has a 10X multiplier
They are both 800MHz FSB

The frequency difference matters little if you OC them to the same frequency so one difference is the cache. The biggest factor may be the core type and that's why I highlighted it. Any thoughts anyone?

budz · Posted: Sep 03, 2007 23:46

gll99: the E2160 is a Allendale core same as the E4400.

jagabo wrote:

Soopafresh · Dismember Joined: 01 Jan 2004 Location: United States

The E2160 L2 cache is 1M per core while the E4400 has a 2 MB cache per core

That makes a big difference in the encoding speed, and that's what gives the newer generation of procs an even greater advantage with the recently upgraded 4MB of L2 cache.

stiltman · Hi Ladies Joined: 10 Jul 2003 Location: Unleashed