which processor for avisynth encoding

[squid_80]

There's still no way those results are accurate.

[squid_80]

I've tried to load the test.avs the same way I loaded the plugins.
I can't get a .txt result when I run Virtual Dub, it just closes after running the test.avs, even with Pookies tweak.
I can't get a result above utime=30.157s
We said load test.avs with virtualdub to make sure the script is working properly and not simply returning an error statement.
You can't run the benchmark with virtualdub.
FFmpeg is used by the batch file to encode test.avs to out.avi; that's where the benchmark figure comes from.
Execute runme.bat and load out.avi with virtualdub, it should look the same as test.avs.

[Jeremy Duncan]

Ok, I deleted everything and started over.

Specification Intel(R) Pentium(R) M processor 1.60GHz
L1 Data cache 32 KBytes, 8-way set associative, 64-byte line size
L1 Instruction cache 32 KBytes, 8-way set associative, 64-byte line size
L2 cache 2048 KBytes, 8-way set associative, 64-byte line size
bench: utime=804.203s

Ha ha. That looks more like it. :D

I'll delete my other posts about this now.

[Terka]

please change the line setmemorymax(128) to higer value
at least setmemorymax(512)

[Jeremy Duncan]

please change the line setmemorymax(128) to higer value
at least setmemorymax(512)

I only have 512 MB ram in my PC.

[Blue_MiSfit]

Hey everyone... I thought I would post another score, this time with heavy overclocking and tight RAM timings. From a stock speed of 2.2 GHz, I am now running at 2.6 GHz, with CAS2.5 and 1T. Memory speed is reduced (divider), but overall performance is higher.

2.6 GHz CAS2.5 1T results:

Specification AMD Athlon(tm) 64 Processor 3500+
L1 Data cache 64 KBytes, 2-way set associative, 64-byte line size
L1 Instruction cache 64 KBytes, 2-way set associative, 64-byte line size
L2 cache 512 KBytes, 16-way set associative, 64-byte line size
bench: utime=243.469s


2.2 GHz, CAS3 2T results:

Specification AMD Athlon(tm) 64 Processor 3500+
L1 Data cache 64 KBytes, 2-way set associative, 64-byte line size
L1 Instruction cache 64 KBytes, 2-way set associative, 64-byte line size
L2 cache 512 KBytes, 16-way set associative, 64-byte line size
bench: utime=277.906s

[Pookie]

What do you have your SetMemoryMax() set to ?

[Blue_MiSfit]

128. I have not modified the avisynth script.

It seems too fast to me, almost 2x faster than terka's results, which seems a bit strange...

~MiSfit

[Adub]

Using Avisynth 2.5.7 MT version by TSP rev 2.0
SinlgeThreaded,128 mb memory
Specification Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz
L1 Data cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L1 Instruction cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L2 cache 4096 KBytes, 16-way set associative, 64-byte line size
bench: utime=356.219s


Multithreaded,128 mb memory
Specification Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz
L1 Data cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L1 Instruction cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L2 cache 4096 KBytes, 16-way set associative, 64-byte line size
bench: utime=326.203s




Mutlthreaded, 1024 mb mem
Specification Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz
L1 Data cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L1 Instruction cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L2 cache 4096 KBytes, 16-way set associative, 64-byte line size
bench: utime=20.953s


Mutlthreaded, no mem restrictions
Specification Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz
L1 Data cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L1 Instruction cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L2 cache 4096 KBytes, 16-way set associative, 64-byte line size
bench: utime=70.313s

Edit: Oh, why does ffmpeg crash when SetMemoryMax(2000) is set? Or Higher?

[Pookie]

You possibly aren't.... Try it with 128MB. If the numbers are the same, the script isn't being read. See above for Jeremy's issue and solution.

[Adub]

Actually, it wasn't loading my plugins. I will load the new results up when they are done.

[Pookie]

I'm going to modify the batch file to determine installation locations so the loadplugin commands are more specific.

edit - On second thought, I decided not to. The only way to do it was to query the registry with a command line util, and that seems too risky and invasive. Yes, it is only a read operation, but why even take the slightest chance....

[squid_80]

I'm going to modify the batch file to determine installation locations so the loadplugin commands are more specific.Maybe add a hashing command on out.avi to determine if it's been created properly?

Specification Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz
L1 Data cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L1 Instruction cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L2 cache 4096 KBytes, 16-way set associative, 64-byte line size
bench: utime=356.219s
That's weird, why is your overclocked e6600 performing worse than my stock system?

[Adub]

Maybe because the memory was limited. were you using limited memory on yours?

I will run a new instance and see if that changes anything.

My memory is not overclocked, only my processor.
When I increase the SetMemoryMax(1024) and keep it single threaded, then I get this;
Specification Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz
L1 Data cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L1 Instruction cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L2 cache 4096 KBytes, 16-way set associative, 64-byte line size
bench: utime=43.219s

[squid_80]

With the original script (setmemorymax(128)): 319.922s
Original script with setmtmode(4): 186.375s
Original script with setmemorymax(896): 42.297s
setmemorymax(896) and setmtmode(4): 23.563s

[BigDid]

Grmbl :mad: :scared: :sly:

After 2 days, finally got it running, the plugins were not loading!
Proc (see my signature) is an athlon64 running @ 2420
Ram is 2*512 DDR3200 Corsair- 2.2.2.5.1 @ 201.6Mhz

With setmemorymax(128)
Specification AMD Athlon(tm) 64 Processor 3000+
L1 Data cache 64 KBytes, 2-way set associative, 64-byte line size
L1 Instruction cache 64 KBytes, 2-way set associative, 64-byte line size
L2 cache 512 KBytes, 16-way set associative, 64-byte line size
bench: utime=584.594s

With setmemorymax(384)
bench: utime=278.047s

With setmemorymax(768)
bench: utime=88.735s

Did

[Adub]

setmemorymax896 and setmtmode(4): 23.047s

This is wierd.

Edit: Wait, what speed are you running your ram at? My is at stock DDR2 667 Mhz.

[foxyshadis]

Edit: Oh, why does ffmpeg crash when SetMemoryMax(2000) is set? Or Higher?

32-bit address limitations, no single process can access more than 2gb without the /3gb startup switch (and proper coding) or segmented addressing.

Avisynth is good and will complain about being out of memory, but if another plugin that doesn't check for allocation failure gets there's first, it goes boom.

[krieger2005]

I'm going to modify the batch file to determine installation locations so the loadplugin commands are more specific.
Is it possible to integrate a more than one encoding (at one time) and take than the average result? I think this would be usefull to get more proper results.

[Pookie]

Modified to run 3 benchmarks - 25% Total RAM, 50% Total RAM, 100% Total RAM-64MB. The results will now be affected by at least 7 variables -> 1) CPU speed 1a) Number of CPUs 2) L2 cache 3) Amount of RAM 3a) Speed of RAM 4) Version of Windows 5) Version/Build of Avisynth. We're going to learn something that's completely profound and original- a fast CPU with lots of fast RAM will give us fast avisynth performance.:D

You'll need to copy the newly modified Runme.bat and test.avs files from the first page. I removed the setmemorymax line, as it is dynamically re-added later in the batch file, depending on the amount of RAM is installed on your system.

I had to hardwire the loadplugin commands at the start of the script. I didn't want to risk doing a registry query with reg.exe.
Change the script to reflect the location of your plugins if it is different. Hopefully, 90% of everyone won't need to touch it.

If you wish to test Dual/Duo Core functionality, make whatever modifications you wish to the script and enjoy. As we all know, mvtools is not compatible with MT in actual use, so the results aren't going to reflect real life performance.

Specification Intel(R) Pentium(R) 4 CPU 2.40GHz
L1 Data cache 8 KBytes, 4-way set associative, 64-byte line size
L2 cache 512 KBytes, 8-way set associative, 64-byte line size

Total Memory Installed=1024MB

256MB memory
bench: utime=317.546s

512MB memory
bench: utime=244.561s

960MB memory
bench: utime=260.415s

[BigDid]

Hi,

I have a script error (when testing test.avs alone) with this line in the new script:

base = last.minblur(2).FluxSmoothT(9)

Beside that the 3time script seems to work...

Did

[Pookie]

Try this version of fluxsmooth.dll 73,728 bytes

http://www.bestsharing.com/files/YbncY203554/fluxsmooth_.rar.html

And add a few carriage returns to the bottom your test.avs file. buzzqw pointed out the batch was cutting off the bottom line, causing the encode to fail.

[buzzqw]

@Pookie

please edit the runme.bat
it will download the ffmpeg 7375 but un7zip another version

BHH

EDIT: when creating 256test.avs the last } is missing

[squid_80]

EDIT: when creating 256test.avs the last } is missing
I got that too. Just add a new line to the end of test.avs.
Also win64 doesn't like lmod - I guess because it's a 16-bit program. I made a file called memory.bat:
set tmem=%2
and modified runme.bat:type cpuinfo.txt | findstr "L1" >>result.txt
type cpuinfo.txt | findstr "L2" >>result.txt
type cpuinfo.txt | findstr "Memory Size">>memory.txt

type memory.txt | findstr "Size" | findstr "Memory" > tmem.bat
call tmem.batand the results:Specification Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz
L1 Data cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L1 Instruction cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L2 cache 4096 KBytes, 16-way set associative, 64-byte line size


512MB memory
bench: utime=104.250s

1024MB memory
bench: utime=41.937s

1984MB memory
bench: utime=41.563s
@Merlin7777: I'm using stock DDR2 PC5300.

[Pookie]

Many thanks, buzzqw and squid_80. Done. :o

Weirdness... I just added a #end remark which gets amputated during the concatenation, but now it preserves the } for the script to run properly.

Specification Intel(R) Xeon(TM) CPU 3.06GHz
Specification Intel(R) Xeon(TM) CPU 3.06GHz
L1 Data cache 8 KBytes, 4-way set associative, 64-byte line size
L1 Data cache 8 KBytes, 4-way set associative, 64-byte line size
L2 cache 512 KBytes, 8-way set associative, 64-byte line size
L2 cache 512 KBytes, 8-way set associative, 64-byte line size

Total Memory Installed= 2048MB

512MB memory
bench: utime=218.922s

1024MB memory
bench: utime=100.156s

1984MB memory
bench: utime=101.234s

Looks like diminishing returns after 50% RAM allocation.

[Moitah]

Specification Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz
L1 Data cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L1 Instruction cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L2 cache 4096 KBytes, 16-way set associative, 64-byte line size

Total Memory Installed= 2048MB

512MB memory
bench: utime=93.766s

1024MB memory
bench: utime=38.140s

1984MB memory
bench: utime=38.266s

Using official 2.5.7, 2x1GB PC6400 DDR2.

[check]

It seems the script uses less than 1gb. It would be interesting to do tests for every 64mb increase :)

[squid_80]

Damn, I've been de-throned!
Using official 2.5.7, 2x1GB PC6400 DDR2.
Which cost more, the RAM or the CPU?

[BigDid]

... And add a few carriage returns to the bottom your test.avs file. buzzqw pointed out the batch was cutting off the bottom line, causing the encode to fail.
Grmbl again
I believe it was the truncated end of script that made the error.
I now have no error but here is the result:
"Total Memory Installed=MB
MB memory
bench: utime=0.063s
MB memory
bench: utime=0.063s
-64MB memory
bench: utime=0.062s"

And here is the test.avs after being run:
"setmemorymax()

ÿÿÿÿÿÿÿÿ"

I have alot of files in the directory I made for the circumstance so I will wipe out anything but what is needed and try again

Edit: wiped out extra files, no good.
Did

[ThEdEaDLiVe]

Specification AMD Athlon(tm) 64 X2 Dual Core Processor 5000+
L1 Data cache 2 x 64 KBytes, 2-way set associative, 64-byte line size
L1 Instruction cache 2 x 64 KBytes, 2-way set associative, 64-byte line size
L2 cache 2 x 512 KBytes, 16-way set associative, 64-byte line size

Total Memory Installed= 2048MB

512MB memory
bench: utime=234.297s

1024MB memory
bench: utime=59.328s

1984MB memory
bench: utime=59.797s

[Moitah]

Which cost more, the RAM or the CPU?
CPU was $315, RAM was $280 - $50 mail-in rebate (which I received) = $230. USD.

[Adub]

It looks like ram speed is highly influential. Moitah is using 800mhz ram while I am only using 667mhz ram. Will post results when finished.

[Moitah]

I'm not sure the RAM speed makes that much difference, I tried underclocking it and here's what I came up with:


http://www.moitah.net/misc/tmem1.png
(underclocked)

1st run:
1024MB memory
bench: utime=37.562s

2nd run:
1024MB memory
bench: utime=37.344s


http://www.moitah.net/misc/tmem2.png
(normal speed)

1st run:
1024MB memory
bench: utime=37.204s

2nd run:
1024MB memory
bench: utime=37.250s


The reason why the time is better than what I posted last night is because I rebooted right before each set of tests and had nothing else running.

At both speeds, the 1984MB always failed at the same place, ffmpeg just exited silently. I tried the script in AVS2AVI and it errored with "AVIReadStream failed", and in VirtualDub with "Run video analysis pass" I got this: The instruction at "0x004bdc4d" referenced memory at "0x0000000d". The memory could not be "read". Only with nothing running though, once I start Firefox to use a bit more RAM, it works fine.

[Pookie]

Moitah & squid_80 - You guys are running different versions of XP (32 vs 64), right ? That could account for the performance difference.

Bigdid - It is almost like you don't have environment space to hold variables. Findstr.exe might be corrupt or missing. It comes with Windows, do a search. It might also not be in your path. Copy it to the same folder that you're running the benchmark in. (Sorry to end a sentence with a preposition):D

ThEdEaDLiVe - Thanks for posting, and welcome to the forums.

[Fizick]

Hm..., why test script is too slow?
may be more simple regular script would be more useful (and fast).

[Moitah]

@Pookie: Yeah, that could be it. I'm running Windows Server 2003 Standard, 32 bit.

[BigDid]

... Findstr.exe might be corrupt or missing. It comes with Windows, do a search. It might also not be in your path. Copy it to the same folder that you're running the benchmark in...
Correct,

It was missing from Win/System32
It is now in both places but result is same as in my previous post
Test incomplete and test.avs being truncated ...

I can live without the test results, but if other ideas, I am willing to try it.

Did

[Pookie]

Fizick - 30 frames only(x3), but very stressful on the cpu. My P4 2.4 with 1GB of RAM is in last place. :D I had to find a script that would be difficult even for the fastest PCs out there, and we're testing machines that are 6 times faster at processing than the slowest one. It is also the slowest result that you can expect on a system (unless you combine it with a h264 1080p transcode), and with 30 frames it is easy to make your own calculations. I think it is a good script to test, because many people want to use the newest functions (mcbob, mvbob, mvdegrain2) and now they can see what kind of performance they can expect if they buy a new system.

Bigdid - Try this one. I tested it and it works for me.



BTW, the Runme.bat I included in that (now deleted) download is bad. I modified the one on the first page to work with more "native" windows apps.

[BigDid]

Bigdid - Try this one. I tested it and it works for me.

http://www.bestsharing.com/files/WL1WJm204176/avstest.zip.html
Thanks,

Worked for me too. I'll wait the actual encode to finish to redo the test and post results.

Did

[Jeremy Duncan]

Fizick - 30 frames only(x3), but very stressful on the cpu. My P4 2.4 with 1GB of RAM is in last place. :D I had to find a script that would be difficult even for the fastest PCs out there, and we're testing machines that are 6 times faster at processing than the slowest one. It is also the slowest result that you can expect on a system (unless you combine it with a h264 1080p transcode), and with 30 frames it is easy to make your own calculations. I think it is a good script to test, because many people want to use the newest functions (mcbob, mvbob, mvdegrain2) and now they can see what kind of performance they can expect if they buy a new system.

Bigdid - Try this one. I tested it and it works for me.

http://www.bestsharing.com/files/WL1WJm204176/avstest.zip.html

This test takes too long. I quit after over running the test more than a hour.

[Pookie]

Jeremy said "This test takes too long. I quit after over running the test more than a hour"

http://frogstar.soylentgeek.com/wav/scoobhuh.wav Modify TEST.AVS from trim(100,129) to trim(100,114) and multiply the results x2




Specification Dual Core AMD Opteron(tm) Processor 275
Specification Dual Core AMD Opteron(tm) Processor 275
L1 Data cache 2 x 64 KBytes, 2-way set associative, 64-byte line size
L1 Instruction cache 2 x 64 KBytes, 2-way set associative, 64-byte line size
L1 Data cache 2 x 64 KBytes, 2-way set associative, 64-byte line size
L1 Instruction cache 2 x 64 KBytes, 2-way set associative, 64-byte line size
L2 cache 2 x 1024 KBytes, 16-way set associative, 64-byte line size
L2 cache 2 x 1024 KBytes, 16-way set associative, 64-byte line size

Total Memory Installed= 3072MB

768MB memory
bench: utime=99.734s

1536MB memory
bench: utime=66.906s

3008MB memory
bench: utime=66.953s



- Edit - Tested in 128MB Increments all the way to 3008MB

128MB
bench: utime=624.109s

256MB
bench: utime=331.000s

384MB
bench: utime=303.172s

512MB
bench: utime=264.140s

768MB
bench: utime=92.375s

896MB
bench: utime=70.047s

1024MB
bench: utime=66.953s

1152MB
bench: utime=66.797s

1152-3008 utimes stayed between 66.79 and 67.79s

[BigDid]

Here are the results :(

I'm afraid something is wrong cause the numbers do not reflect those of the first test: http://forum.doom9.org/showthread.php?p=933583#post933583

Same parameters
"Total Memory Installed=c MB
0MB memory
bench: utime=2881.938s
0MB memory
bench: utime=2878.078s
-64MB memory
bench: utime=2884.218s
"

Will try again with trim(100,114) and multiply the results x2 :rolleyes:

Did

[Pookie]

Nope, you're doing everything right. It was my fault. I'm probably driving you nuts with all of these issues. The "Runme.bat" File I posted in the link I gave you was bad. My apologies. Insomnia.

Replace the one you're using with one:

http://www.bestsharing.com/files/QJjwoO204794/runme.zip.html

Note: this link above is specific to BigDid's issues.

[Jeremy Duncan]

Modified TEST.AVS from trim(100,129) to trim(100,114) and multiplied the results x2

Specification Intel(R) Pentium(R) M processor 1.60GHz
L1 Data cache 32 KBytes, 8-way set associative, 64-byte line size
L1 Instruction cache 32 KBytes, 8-way set associative, 64-byte line size
L2 cache 2048 KBytes, 8-way set associative, 64-byte line size

Total Memory Installed=504 MB

126MB memory
bench: utime=590.438s

252MB memory
bench: utime=275.906s

440MB memory
bench: utime=385.094s

[Jeremy Duncan]

Here's a link to the necessary software. You'll still need to get the plugins yourself though.

Link (http://www.bestsharing.com/files/vd7YK204826/Pookies%20Test.zip.html)

[BigDid]

Hurray, Working,

Specification AMD Athlon(tm) 64 Processor 3000+
L1 Data cache 64 KBytes, 2-way set associative, 64-byte line size
L1 Instruction cache 64 KBytes, 2-way set associative, 64-byte line size
L2 cache 512 KBytes, 16-way set associative, 64-byte line size
Total Memory Installed=1024 MB - about 750Mb free -

With latest revised script:
with 15 frames*2 (trim 100,114) and MVTools 1.4.13.0 :

256MB memory
bench: utime=85.641s *2=171.282s

512MB memory
bench: utime=41.656s *2=83.312s

960MB memory
bench: utime=49.875s *2=99.75s


Complete test with 30 frames (trim 100,129) and MVTools 1.4.13.0 :

256MB memory
bench: utime=271.516s (MVT 1.3.0.0=301.625s, MVT 1.6.2=321.547s )

512MB memory
bench: utime=209.062s (MVT 1.3.0.0=232.828s, MVT 1.6.2=255.329s )

960MB memory
bench: utime=153.640s (MVT 1.3.0.0=160.734s, MVT 1.6.2=102.390s)


Proc (see my signature) is an athlon64 running @ 2420
Ram is 2*512 DDR3200 Corsair- 2.2.2.5.1 @ 201.6Mhz
Avisynth 2.57 final
Plugins:
# MVTools 1.4.13.0
# MedianBlur 0.8.4.1
# MaskTools 2.0.30.0
# TTempSmooth 0.9.4.0
# FluxSmooth ? Pookie's rev
# RemoveGrain/Repair 1.0 SSE2

Did

[zilog jones]

Using files from the link above:
Specification Intel(R) Core(TM)2 CPU 6300 @ 1.86GHz
L1 Data cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L1 Instruction cache 2 x 32 KBytes, 8-way set associative, 64-byte line size
L2 cache 2048 KBytes, 8-way set associative, 64-byte line size

Total Memory Installed=1024 MB

256MB memory
bench: utime=94.266s *2 = 188.532s

512MB memory
bench: utime=32.953s *2 = 65.906s

960MB memory
bench: utime=33.281s *2 = 66.562s

(With every process and service I could think of turned off, leaving 177MB used by Windows when idle)

256MB memory
bench: utime=93.953s *2 = 187.906s

512MB memory
bench: utime=32.828s *2 = 65.656s

960MB memory
bench: utime=34.234s *2 = 68.468s




Chipset
------------------------------------------------------------------------------

Northbridge Intel i975X rev. C0
Southbridge Intel 82801GB (ICH7) rev. 01
Graphic Interface PCI-Express
PCI-E Link Width x16
PCI-E Max Link Width x16
Memory Type DDR2
Memory Size 1024 MBytes
Memory Frequency 267.0 MHz (1:1)
CAS# 4.0
RAS# to CAS# 4
RAS# Precharge 4
Cycle Time (tRAS) 12
Bank Cycle Time (tRC) 16

Memory SPD
------------------------------------------------------------------------------

DIMM #1

General
Memory type DDR2
Manufacturer (ID) Corsair (7F7F9E0000000000)
Size 512 MBytes
Max bandwidth PC2-4300 (266 MHz)
Part number VS512MB533D2

Attributes
Number of banks 1
Data width 64 bits
Correction None
EPP no

Timings table
Frequency (MHz) 200 266
CAS# 3.0 4.0
RAS# to CAS# delay 3 4
RAS# Precharge 3 4
TRAS 9 12
TRC 11 15

- DIMM 2 is identical - they're both Corsair Value Select PC2-4200 512MB.
- MSI 975X Platinum PowerUp Edition motherboard - I wish I bought one with a less convoluted name :)
- AviSynth 2.5.7
- WinXP SP2 32-bit
- WD Caviar SE16 7200RPM 320GB hard disk.


[edit]Forgot to multiply them! Results are a lot less impressive now compared to Moitah's :(

@Moitah: Could you try underclocking your RAM to 266MHz?

[BigDid]

... I don't know how I'm beating Moitah's times with a slower CPU and cheapskate Corsair Value Select PC2-4200 RAM (512MB pair) O_o...
Hi,

Maybe the timings? Would you mind to share your cpu-z memory tab screenshot?

Did

[Jeremy Duncan]

Your supposed to multiply the results by 2.

[zilog jones]

Oh :(
I was wondering what was going on with people doing that above...