which processor for avisynth encoding

[Pookie]

Oi, your system is fast enough to do a full test. Use the Test.avs on the first page. Thanks for testing, I was interested in particular at the E6300/E6400/E6600.

BTW, Intel is going to add 4MB L2 cache on all of their E series procs starting in March.

[zilog jones]

Ah, it all makes sense now :D

Here's the full test:
256MB memory
bench: utime=303.313s

512MB memory
bench: utime=156.204s

960MB memory
bench: utime=103.546s

<2nd try>

256MB memory
bench: utime=303.812s

512MB memory
bench: utime=156.860s

960MB memory
bench: utime=79.391s

...OK, it's significantly worse :confused:

[alfixdvd]

Specification Intel(R) Pentium(R) 4 CPU 2.80GHz
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=2048 MB

512MB memory
bench: utime=317.172s

1024MB memory
bench: utime=86.078s

1984MB memory
bench: utime=75.843s


With SetMTMode(2)

Specification Intel(R) Pentium(R) 4 CPU 2.80GHz
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=2048 MB

512MB memory
bench: utime=225.907s

1024MB memory
bench: utime=69.609s

1984MB memory

ffmpeg crashes : Runtime Error at frame21

[Pookie]

You can't MT with MVtools. This test will probably run with MT 3, but in a real world application it won't. Maybe that should be a sticky somewhere.


Ah, it all makes sense now :D


...OK, it's significantly worse :confused:

Sigh... too many variables are getting changed. I think the 2 things we've learned from this benchmark are - 1) The E6000 series is the proc of choice if you're planning on taking advantage of the latest advancements in denoising and deinterlacing - namely motion compensated processing. 2) 50% of total RAM seems to be the optimal value to setmemorymax()

[Jeremy Duncan]

I ran the test on the first page.

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=809.188s

252MB memory
bench: utime=432.422s

440MB memory
bench: utime=621.157s

[Taurus]

Just for grins:
I used the "stripped off" BigDid Version of the script on an old
AMD 2200XP out of curiosity.
What makes me wonder, after fiddling around with the avs plugins, there's quite a difference in speed by changing plugins around. Even holes in the encodes happen as you can see in the first lineup.
So I think everybody should use the same plugins and versions to make a fair test. Maybe somebody can throw a bundle of the "right" ones into this thread.
What amazes me the most, how memory allocation and bus speed pushes the numbers.
Multiply numbers by x2 please.

Specification AMD Athlon(tm) XP 2200+
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 256 KBytes, 16-way set associative, 64-byte line size

Total Memory Installed=512 MB 133Mhz:D
-------------------------------------------------------
with new removegrain.dll:

128MB memory
bench: utime=626.971s

256MB memory
bench: utime=0.000s

448MB memory
bench: utime=241.718s
--------------------------------------------------------
with older removegrain.dll

128MB memory
bench: utime=644.537s

256MB memory
bench: utime=261.616s

448MB memory
bench: utime=252.723s

[Pookie]

Just for grins:
Maybe somebody can throw a bundle of the "right" ones into this thread.


That would be considered not adhering to the GNU General Public License. You can't just throw the needed ones into a zip file. And most of all, I have too much respect for the plugin developers to even seem like I'm dissing them. It really ruffles their feathers.

Then there's the version of Windows, Avisynth, etc...

[ThEdEaDLiVe]

Using modified test.avs from trim(100,129) to trim(100,114) and tested after fresh reboot.

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=2048 MB

512MB memory
bench: utime=48.875s *2 = 97.75s

1024MB memory
bench: utime=28.297s *2 = 56.594s

1984MB memory
bench: utime=27.859s *2 = 55.718s

[Moitah]

@Moitah: Could you try underclocking your RAM to 266MHz?
1024 MB Test

266 MHz:
1st: 37.735s
2nd: 37.797s

400 MHz:
1st: 37.375s
2nd: 37.032s

I am using MVTools 1.4.13. I noticed when I tried MVTools 1.6.2 it took more like 50 seconds. I'll list the versions of everything:

Windows Server 2003 Standard (32 bit)
Tritical's Modified AviSynth 2.5.7 CVS 2006-Nov-25 - I switched back to this after upgrading to 2.5.7 final official because it was crashing not only during the 1984 MB test but also during my normal encodes.
MVTools 1.4.13
MedianBlur 0.8.4
MaskTools 2.0a30
TTempSmooth 0.9.4
FluxSmooth 1.1b
RemoveGrain/Repair 0.9 SSE3 - I tried it with the regular (non-SSE3) versions, didn't seem to make a noticeable speed difference on this script.

[Pookie]

Now THAT is good information. Perhaps it'll save someone a few $ when they are configuring RAM for a new Avisynth machina...

[*.mp4 guy]

I said before that it was the timings, not the bandwidth of the ram that mattered for stuff like this, could you please test the ram with different timings and see how that effects the speed.

[Adub]

Have we established completely why Moitah is faster than Squid_80 and my test results, even thought the ram speed doesn't really matter. Also, It seems that my over clocked processor was performing about the same as stock, what the heck is going on?

Note: I do have Folding@Home running in the back ground. Maybe if I kill it it will improve my numbers, but I doubt it. It is a idle priority program.

[Pookie]

You're going to have to look at other variables, Merlin. Moitah's running Windows 2003; Squid_80 is running XP64. That alone is a huge difference. In your case, perhaps the L2 cache size is more important than raw clock speed when you hit these amazing numbers. Another possibility is that you've hit a ceiling of performance and the bottleneck is software related. Does the E6000 series throttle back during periods of high heat ? ffmpeg can turn a proc into a furnace.

[Moitah]

I said before that it was the timings, not the bandwidth of the ram that mattered for stuff like this, could you please test the ram with different timings and see how that effects the speed.
333 MHz, 4-4-4-12: 37.406s, 37.062s, 37.078s
333 MHz, 5-5-5-15: 37.531s, 37.297s, 37.375s

I tried to test 6-6-6-18 but it was too unstable. Shouldn't it have been more stable, if anything? I'm not really into overclocking, so I'm not sure.

[*.mp4 guy]

Some ram can be very particular about the timings, most ram modules work best with certain timings, but usually using more lax timings then the optimal ones doesn't result in instability. Anyway All I can surmise from your results is that c2ds don't benefit from tight timings like amd64's do, It looks like your fsb might be saturated, but at this point I'm just guessing.

[Taurus]

That would be considered not adhering to the GNU General Public License. You can't just throw the needed ones into a zip file. And most of all, I have too much respect for the plugin developers to even seem like I'm dissing them. It really ruffles their feathers.

Then there's the version of Windows, Avisynth, etc...

You're right, but for the sake of compatibility it would be nice to know the versions and/or date of release.
Never mind, it's a nice benchmark anyway. Thanks for your work and intelligence.

[Audionut]

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=1024 MB

256MB memory
bench: utime=181.703s

512MB memory
bench: utime=98.719s

960MB memory
bench: utime=89.906s


OC to 3.15GHz, with crap ram running at 702mhz 5-6-6-17 timings.
Using sse3 versions of removegrain and repair.

edit: using all the scripts etc from the first page. It looks as if you guys might be using different ones now.

[Pookie]

Those are very different results than the other 6600s. That's more than software version differences affecting your performance. Open the "Cpuinfo.txt" file that gets created and have a look at the first section for any anomalies. If something looks really off, like bus or core speeds, your bios settings might need adjustment.

[*.mp4 guy]

I think its becuase he didn't use multithreading.

[Audionut]

Here it is. minus a few lines to bring the post under 16000 characters long.

-------------------------
CPU-Z version 1.38
-------------------------

Processors Map
------------------------------------------------------------------------------------

Number of processors 1
Number of threads 2

Processor 0
-- Core 0
-- Thread 0
-- Core 1
-- Thread 0


Processors Information
------------------------------------------------------------------------------------

Processor 1 (ID = 0)
Number of cores 2
Number of threads 2 (max 2)
Name Intel Core 2 Duo E6600
Codename Conroe
Specification Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz
Package Socket 775 LGA (platform ID = 0h)
CPUID 6.F.6
Extended CPUID 6.F
Core Stepping B2
Technology 65 nm
Core Speed 3150.1 MHz (9.0 x 350.0 MHz)
Rated Bus speed 1400.0 MHz
Stock frequency 2400 MHz
Instructions sets MMX, SSE, SSE2, SSE3, SSSE3, EM64T
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
FID/VID Control yes
FID range 6.0x - 9.0x
max VID 1.213V


Thread dumps
------------------------------------------------------------------------------------

CPU Thread 0
APIC ID 0
Topology Processor ID 0, Core ID 0, Thread ID 0
Type 01008001h
Max CPUID level 0000000Ah
Max CPUID ext. level 80000008h


CPU Thread 1
APIC ID 1
Topology Processor ID 0, Core ID 1, Thread ID 0
Type 01008001h
Max CPUID level 0000000Ah
Max CPUID ext. level 80000008h


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

Northbridge Intel P965/G965 rev. C1
Southbridge Intel 82801HB (ICH8) rev. 02
Graphic Interface PCI-Express
PCI-E Link Width x16
PCI-E Max Link Width x16
Memory Type DDR2
Memory Size 1024 MBytes
Memory Frequency 350.0 MHz (1:1)
CAS# 5.0
RAS# to CAS# 6
RAS# Precharge 6
Cycle Time (tRAS) 17


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

DIMM #1

General
Memory type DDR2
Manufacturer (ID) (0000000000000000)
Size 512 MBytes
Max bandwidth PC2-5300 (333 MHz)
Part number

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

Timings table
Frequency (MHz) 200 266 333
CAS# 3.0 4.0 5.0
RAS# to CAS# delay 3 4 5
RAS# Precharge 3 4 5
TRAS 9 12 15
TRC 12 16 20


DIMM #2

General
Memory type DDR2
Manufacturer (ID) (0000000000000000)
Size 512 MBytes
Max bandwidth PC2-5300 (333 MHz)
Part number

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

Timings table
Frequency (MHz) 200 266 333
CAS# 3.0 4.0 5.0
RAS# to CAS# delay 3 4 5
RAS# Precharge 3 4 5
TRAS 9 12 15
TRC 12 16 20


Dump Module #1
0 1 2 3 4 5 6 7 8 9 A B C D E F
00 80 08 08 0E 0A 60 40 00 05 30 45 00 82 08 00 00
10 0C 04 38 01 02 00 03 3D 50 50 60 3C 1E 3C 2D 80
20 20 28 10 18 3C 1E 1E 00 00 3C 69 80 18 22 00 00
30 00 00 00 00 00 00 00 00 00 00 00 00 00 00 12 73
40 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
50 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
60 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
70 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00


Dump Module #2
0 1 2 3 4 5 6 7 8 9 A B C D E F
00 80 08 08 0E 0A 60 40 00 05 30 45 00 82 08 00 00
10 0C 04 38 01 02 00 03 3D 50 50 60 3C 1E 3C 2D 80
20 20 28 10 18 3C 1E 1E 00 00 3C 69 80 18 22 00 00
30 00 00 00 00 00 00 00 00 00 00 00 00 00 00 12 73
40 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
50 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
60 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
70 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00


DMI
------------------------------------------------------------------------------

DMI BIOS
--------
vendor Award Software International, Inc.
version F9
date 01/11/2007


DMI System Information
----------------------
manufacturer Gigabyte Technology Co., Ltd.
product 965P-DQ6
version unknown
serial unknown
UUID FFFFFFFF-FFFFFFFF-FFFFFFFF-FFFFFFFF


DMI Baseboard
-------------
vendor Gigabyte Technology Co., Ltd.
model 965P-DQ6
revision unknown
serial unknown


DMI System Enclosure
--------------------
manufacturer Gigabyte Technology Co., Ltd.
chassis type Desktop
chassis serial unknown


DMI Processor
-------------
manufacturer Intel
model Intel(R) Core(TM)2 CPU 660
clock speed 3150.0MHz
FSB speed 350.0MHz
multiplier 9.0x


DMI Memory Controller
---------------------
correction 8-bit parity
Max module size 1024MBytes


DMI Memory Module
-----------------
designation A0


DMI Memory Module
-----------------
designation A1
size 512MBytes (single bank)


DMI Memory Module
-----------------
designation A2


DMI Memory Module
-----------------
designation A3
size 512MBytes (single bank)


DMI Port Connector
------------------
designation PRIMARY IDE (internal)
connector On Board IDE


DMI Port Connector
------------------
designation SECONDARY IDE (internal)
connector On Board IDE


DMI Port Connector
------------------
designation FDD (internal)
port type 8251 FIFO Compatible
connector On Board Floppy


DMI Port Connector
------------------
designation COM1 (internal)
port type Serial Port 16450
connector 9 Pin Dual Inline (pin 10 cut)
connector DB-9 male


DMI Port Connector
------------------
designation COM2 (internal)
port type Serial Port 16450
connector 9 Pin Dual Inline (pin 10 cut)
connector DB-9 male


DMI Port Connector
------------------
designation LPT1 (internal)
port type Parallel Port ECP/EPP
connector DB-25 female
connector DB-25 female


DMI Port Connector
------------------
designation Keyboard (internal)
port type Keyboard Port
connector PS/2


DMI Port Connector
------------------
designation PS/2 Mouse (internal)
designation Detected (external)
port type Mouse Port
connector PS/2
connector PS/2


DMI Port Connector
------------------
designation USB (internal)
port type USB
connector Access Bus (USB)


DMI Port Connector
------------------
designation USB (internal)
port type USB
connector Access Bus (USB)


DMI Extension Slot
------------------
designation PCI
type PCI
width 32 bits
populated no


DMI Extension Slot
------------------
designation PCI
type PCI
width 32 bits
populated yes


DMI Extension Slot
------------------
designation PCI
type PCI
width 32 bits
populated no


DMI Extension Slot
------------------
designation PCI
type PCI
width 32 bits
populated no


DMI Physical Memory Array
-------------------------
location Motherboard
usage System Memory
correction None
max capacity 4096MBytes
max# of devices 4


DMI Memory Device
-----------------
designation A0
format DIMM
type unknown


DMI Memory Device
-----------------
designation A1
format DIMM
type unknown
total width 64bits
data width 64bits
size 512MBytes


DMI Memory Device
-----------------
designation A2
format DIMM
type unknown


DMI Memory Device
-----------------
designation A3
format DIMM
type unknown
total width 64bits
data width 64bits
size 512MBytes


Software
------------------------------------------------------------------------------

Windows Version Microsoft Windows XP Professional Service Pack 2 (Build 2600)
DirectX Version 9.0c


Resources
------------------------------------------------------------------------------

Port I/O Space, BA=0xCF8, size=0x8
Port I/O Space, BA=0xD000
Port I/O Space, BA=0xC000
Port I/O Space, BA=0xC400
Port I/O Space, BA=0xC800
Port I/O Space, BA=0xCC00
Port I/O Space, BA=0xF000
Port I/O Space, BA=0xFC00
Port I/O Space, BA=0x500
Port I/O Space, BA=0xD800
Port I/O Space, BA=0xDC00
Port I/O Space, BA=0xE000
Port I/O Space, BA=0xE400
Port I/O Space, BA=0xE800
Port I/O Space, BA=0xEC00
Port I/O Space, BA=0x7000
Port I/O Space, BA=0x9000
Port I/O Space, BA=0x9400
Port I/O Space, BA=0x9800
Port I/O Space, BA=0x9C00
Port I/O Space, BA=0xA000
Port I/O Space, BA=0xB000
Port I/O Space, BA=0x408, size=0x4
Port I/O Space, BA=0x480, size=0x40
Memory I/O Space, BA=0xFB105000
Memory I/O Space, BA=0xFB100000
Memory I/O Space, BA=0xFB104000
Memory I/O Space, BA=0xFB106000
Memory I/O Space, BA=0xF4000000
Memory I/O Space, BA=0xE0000000
Memory I/O Space, BA=0xF5000000
Memory I/O Space, BA=0xFB000000
Memory I/O Space, BA=0xFA000000
Memory I/O Space, BA=0xF0000000, size=0x1000
Memory I/O Space, BA=0xFEE00000, size=0x1000
SMBus I/O Space, BA=0x500
PCI Express I/O Space, BA=0xF0000000, size=0x1000
PCI I/O space, BA=0xCF8, size=0x8


It was only using one cpu/thread while running the test.

[Audionut]

Here's the results using setmtmode(2,0)

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=1024 MB

256MB memory
bench: utime=129.687s

512MB memory
bench: utime=67.109s

960MB memory
bench: utime=55.937s

[squid_80]

I only used multithreading in the posts where I said so. I made this post (http://forum.doom9.org/showthread.php?p=933579#post933579) to spell that out pretty clearly.
Try winding the FSB back to stock speed and set the memory timings back to auto and see what happens.

[Audionut]

Using 256MB of memory, my overclocked cpu is quite a bit faster.
However when using more memory, my cheap shit house memory shows just how cheap and crap it is.

edit: I can't reboot yet, but will try a few things on the weekend.

[BigDid]

Hi, some updates

I re-arrangered my previous post: http://forum.doom9.org/showthread.php?p=937879#post937879 to have all the infos grouped, borrowed the good idea from Moitah

@Pookie, do you have the version number for the fluxsmooth.rar you have uploaded?

Now that it is working and to avoid the the scattered infos I will try to update my previous post only if some changes.
One could be using the 1.4.13 mvtools instead of the 1.3.00 if quicker.
Another would be finding why I (and a few others) are having better results with more ram, even more than half of the total installed; atm I think about the old memory manager I use (optimize memory) but I could be wrong.

If/when I change my gear to go on the darkside twice (Intel C2d) :D I will try to make another summary.

Did

[Pookie]

It has no version info associated with it.

Created:Tuesday, August 23, 2005, 9:44:24 PM
72.0 KB (73,728 bytes)

I think your performance results are consistent with your processor and ram. I've got a similar machine at work and I'll do a comparison.

[Taurus]

I think it's useless comparing speed as long as not everybodies using the same setup. Look at my post one page back.
By just changin around some plugins I've got a performance boost:
-------------------------------------------------------
Before optimizing:
-------------------------------------------------------
Specification AMD Athlon(tm) XP 2200+
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 256 KBytes, 16-way set associative, 64-byte line size

Total Memory Installed=512 MB 133Mhz
-------------------------------------------------------
with new removegrain.dll:
-------------------------------------------------------
128MB memory
bench: utime=626.971s

256MB memory
bench: utime=0.000s

448MB memory
bench: utime=241.718s
--------------------------------------------------------
with older removegrain.dll
--------------------------------------------------------
128MB memory
bench: utime=644.537s

256MB memory
bench: utime=261.616s

448MB memory
bench: utime=252.723s
--------------------------------------------------------
after optimizing:
--------------------------------------------------------
128MB memory
bench: utime=425.972s

256MB memory
bench: utime=230.962s

448MB memory
bench: utime=125.360s
--------------------------------------------------------
Of course: Same script/same environment.
Don't forget to multiply the numbers x2 :D

[Pookie]

What does "after optimizing" mean ?

[Taurus]

What does "after optimizing" mean ?
just fiddling around with the avisynth plugins......
Trial and error principle.

[BigDid]

just fiddling around with the avisynth plugins......
Trial and error principle.
Hi Taurus and happy new year,

Would you mind giving your plugins versions and/or your other optimizations?
I changed the mvtools from 1.3.0 to 1.4.13 and got around 10% less time, will update my summary post with complete data later.

Did

[Pookie]

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= 3072MB

768MB memory
bench: utime=107.595s

1536MB memory
bench: utime=77.892s

3008MB memory
bench: utime=91.692s

---------------------------
Using Bigdid's memory amounts:

256test
bench: utime=389.350s

512test
bench: utime=309.475s

960test
bench: utime=80.266s

----------------------------------------------------
medianblur.dll Version: 0,8,4,1

mt_masktools.dll Version: 2,0,30,0

TTempSmooth.dll Version: 0,9,4,0

mvtools.dll Version: 1,6,2,0


The rest of the Dlls won't return version info

If you want to check dll versions, download

http://helpdesk.kixtart.org/Download/Utils/GetVers.exe

At the command prompt, type in the following

for %a in (*.dll) do getvers %a >>dllvers.txt

[Adub]

medianblur=0.8.4.1
mt_masktools=2.30
ttempsmooth=0.9.4
mvtools=1.6.2.0


No multithreading.

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=2048 MB

512MB memory
bench: utime=158.766s

1024MB memory
bench: utime=43.750s

1984MB memory
bench: utime=43.532s


Multithreading:
setMTmode(2)

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=2048 MB

512MB memory
bench: utime=69.328s

1024MB memory
bench: utime=21.500s

1984MB memory
bench: utime=21.750s

[Moitah]

mvtools=1.6.2.0
That explains it... your result is quite good then.

[Pookie]

Amazing performance, but remember, it wouldn't work on a "real" video file because of the MVtools/MT incompatibility issue. Doesn't change the results - if you got the dough, the E6600 is the way to go....

Also mentioned in another thread - the E6300 and E6400 are going to get 4MB L2 cache as a standard beginning in Q2 2007

[BigDid]

Regarding MVT 1.6.2, comparison updated from here: http://forum.doom9.org/showthread.php?p=937879#post937879

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)

For an Atlon monocore, MVT 1.4.13 is always better than MVT 1.3.0.0; and is better than MVT 1.6.2 excepted for 960Mb

Anybody could confirm for another config and/or more RAM?

Did

[Taurus]

Regarding MVT 1.6.2, comparison updated from here: http://forum.doom9.org/showthread.php?p=937879#post937879

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)

For an Atlon monocore, MVT 1.4.13 is always better than MVT 1.3.0.0; and is better than MVT 1.6.2 excepted for 960Mb

Anybody could confirm for another config and/or more RAM?

Did

That's exactly what I'v got too:
----------------------------------------
128MB memory
bench: utime=425.552s

256MB memory
bench: utime=229.490s

448MB memory
bench: utime=191.626s Mvtools 1.4.13.0
----------------------------------------
128MB memory
bench: utime=464.598s

256MB memory
bench: utime=254.927s

448MB memory
bench: utime=137.037s MVtools 1.6.2.0
----------------------------------------
I guess we all should stay at this plugins for benchmark purposes:

Fluxsmooth.dll Vers.: unknown version build date: Thursday, 9. November 2006 Size: 72,0 KB (73.728 Bytes)

Mt_masktools.dll Vers.: 2.0.30.0

Medianblur.dll Vers.: 0.8.4.1

Mvtools.dll Vers.: 1.6.2.0

Removegrain.dll Vers.: unknown version build date: Sunday, 1. Mai 2005 Size: 23,0 KB (23.552 Bytes) from authors Website (0.9)

Repair.dll Vers.: unknown version build date: Sunday, 1. Mai 2005 Size: 21,5 KB (22.016 Bytes) from authors Website (0.9)

TTempSmooth.dll Vers.: 0.9.4.0

[zilog jones]

Jeremy Duncan already tried to start off a more controlled approach to this test using specific plugin versions in this thread (http://forum.doom9.org/showthread.php?t=120819).

[BigDid]

Jeremy Duncan already tried to start off a more controlled approach...
Hi,

I will let the "controlled approach" to it's author.

...
I guess we all should stay at this plugins for benchmark purposes:
...
Mvtools.dll Vers.: 1.6.2.0

Precisely this is the point aka in real time what will be the quickest settings/plugins? I'm not sure 1.6.2 is the answer.

Until now I had a setmemorymax.avsi set to 192 but that was for avisynth 2.56; what for 2.57? no setmemorymax needed cause no more memory leak?
If this is true maybe Pookie could give a modified runme.bat (adding a no fixed mem size, and/or returning optimal mem size used internally) to verify this particular point?

I still have a long way to go to master avisynth so that's why this thread is interesting :)

Did

[zilog jones]

I will let the "controlled approach" to it's author.
I was just saying, as a response to Taurus suggesting we use specific versions of the plugins that Jeremy already suggested the exact same thing (and put it in a less cluttered new thread). If we did this along with not changing the order of loading the plugins and whatever else we'd have a chance of getting more comparable results.

[Pookie]

It's all good... I sorta knew this thread was going to splinter into a bunch of different directions once we uncapped the 128MB memory limits. As far as "Dunc's" branch is concerned - I think the train has already left the station on this issue. Besides, if you like Peter Sellers movies, you can't help but be entertained.

Remember, the original goal was to determine the fastest processor if one was going to buy/build a new machine today. Done.

Changing file versions of the Avisynth plugin dlls is interesting, but again, there's no way to standardize the other O/S based dlls. We're all running different versions of Windows, including different service packs. Then there's the issue of Mvtools 1.62 vs 1.4x.
You're trading speed for quality on that one, especially during fade-ins/outs. Is it worth the 10% speedup/penalty? Depends on you and your source file.

Regarding 2.56 vs 2.57 - I don't know. Notice Moitah's posting regarding having to switch to Tritical's 2.57 build from the official because of a crash he was experiencing at total RAM-64MB (the third test). Then again, he's running Windows 2003 server, so who knows how that is affecting RAM allocation with Avisynth. I used those percentages for the test due to an old posting from Ian B stating that the largest setmemorymax statement is Total Ram -16MB, and that using a number higher than total RAM would result in pagefile swapping. Perhaps that forumula changed in the latest version. Still, 50% of total RAM seems like a stable and performance optimized value to use, and that works for both 2.56 and 2.57 builds.

Pretty head-spinning - too many variables all around :D

[zilog jones]

Does SetMemoryMax take into account memory already used by other running (non-AviSynth) processes? If it doesn't (which I assume not), this could vary what the optimal settings will be greatly for people depending on what else they're running. For example, right now I'm just running Opera, iTunes and some other junk, and about 500MB is already being used :(

[Pookie]

You can download http://www.stereopsis.com/eatmem/eatmem.zip

Run it in another Command Prompt Window

Syntax : Eatmem 256 (or whatever amount you want to allocate). Wait until it has allocated the amount you have specified on the command line (it'll tell you), then run the Avisynth test. See if the results differ from your original test.

[Taurus]

Pretty head-spinning - too many variables all around :D
You never will get exact reasults as Pookie pointed out in his post.
There are to many variables to take into account.
But for a rough guideline on processor performance it's almost clear to see which ones on top.
So this is all information really needed.

Still it is interesting to see different plugins behavior on memory management.
My own tests clearly shows the script was getting slower as memory gets saturated. Page file swapping occurs at the -64 entry. Part of the OS and background tasks got swapped.
Whole system became sluggish.
So the only logical conclusion is to make sure you never allocate to much Setmemorymax=? to your scrypt.
Just watch taskmanager how much memory is already in use.
Maybe kill some unneccesary background tasks.
Kill networking, because if there's heavy disc traffic, no benchmark will show correct numbers.
Examples:
With almost all Background tasks and some services disabled:
------------------------------------
128MB memory
bench: utime=420.174s

256MB memory
bench: utime=227.607s

448MB memory
bench: utime=117.639s
------------------------------------
Normal windows environment:
------------------------------------
128MB memory
bench: utime=425.552s

256MB memory
bench: utime=229.490s

448MB memory
bench: utime=191.626s
-----------------------------------
Draw your own conclusions.
-----------------------------------
I've got the information I wanted from this thread.
Now I know which processor to buy if I can afford it:p

Thanks again Pookie for starting this.

[mooms]

wohoo i have the fastest processor in history !

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

Total Memory Installed=1024 MB

256MB memory
bench: utime=0.828s

512MB memory
bench: utime=1.062s

960MB memory
bench: utime=0.625s

what am i doing wrong ?

[Audionut]

what am i doing wrong ?

Have you got all the required plugins?
Ensure that you are using the correct file path in the test.avs.

[Taurus]

what am i doing wrong ?
Is your test.avs playing in any mediaplayer? If not: look @ Audionut's post.

[Audionut]

A little update. Bus speed 400mhz. Cpu 8x multi 3.2Ghz. Mem at 400mhz 5-6-6-18.

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=1024 MB

256MB memory
bench: utime=131.250s

512MB memory
bench: utime=65.453s

960MB memory
bench: utime=48.937s

A fraction slower at 256mb, but about 7 secs faster at 960mb.

[BigDid]

Same config as my summary post: http://forum.doom9.org/showthread.php?p=937879#post937879

Was:
with 15 frames (trim 100,114) and MVTools 1.4.13.0 :
256MB memory
bench: utime=85.641s
512MB memory
bench: utime=41.656s
960MB memory
bench: utime=49.875s

Same config + thunderbird+Firefox+trays programs+eatmem 256 (539Mb free):
256MB memory
bench: utime=129.594s
512MB memory
bench: utime=102.812s
960MB memory
bench: utime=235.047s

Same config + thunderbird+Firefox+trays programs+eatmem 256 + mem manager tuned up:
256MB memory
bench: utime=143.562s
512MB memory
bench: utime=59.000s
960MB memory
bench: utime=190.875s

Did

[Pookie]

Wow, interesting info, BigDid. There are a few freeware apps out there which will allow you auto close a list of running apps and services which you specify. They are mostly intended to keep CPU utilization down for DVD burning, but they can certainly work as part of the transcoding procedure for Avisynth. FSautostart is a good one. You can tell it to stop all unecessary apps and svcs, run an avisynth script, then restart the same apps and svcs after the script is complete.

http://www.wspilots.com/index.php?page=4&action=file&file_id=944
http://www.softpedia.com/get/Tweak/System-Tweak/FSAutoStart.shtml

[BigDid]

Wow, interesting info, BigDid. There are a few freeware apps out there which will allow you auto close a list of running apps and services which you specify.
Hi Pookie,

Thanks. I'll redo the test with an eatmem 384 or 512 to get less than 512 free mem.

... FSautostart is a good one. You can tell it to stop all unecessary apps and svcs, run an avisynth script, then restart the same apps and svcs after the script is complete.

http://www.wspilots.com/index.php?page=4&action=file&file_id=944
http://www.softpedia.com/get/Tweak/System-Tweak/FSAutoStart.shtml
Until now I was more concerned by free mem or allocated mem between avisynth and the encoder+codec; now I will also take tray programs into account :thanks:

Did

Edit: Have tuned the mem manager and updated the above results.

[mooms]

Is your test.avs playing in any mediaplayer? If not: look @ Audionut's post.


http://img297.imageshack.us/img297/774/sanstitrepm6.png

which plugin is causing that message ?