VideoHelp.com Video Benchmarking

rallynavvie · Posted: Jan 09, 2005 23:13

That one P4 result seems a bit fast compared to the other P4s. Which chipset are you running that P4 on Delerious?
_________________
NE CEDE MALIS SED CONTRA AUDENTIOR ITO

edDV · Posted: Jan 10, 2005 01:53

Oh fun thread !
I'll toss my favorite bomb into the room re: Intel vs. AMD

http://www.digit-life.com/articles2/intelamdcpuroundupvideo/

From the mainconcept MPeg2 summary:

"For MPEG2 encoding the Hyper-Threading works correctly, and Intel's processors that support it get a good gain. Just compare Pentium 4 2.4C and its sibling without Hyper-Threading. By the way, if the bus bandwidth is not indicated, it means that the RAM speed has a weak effect on performance, and the comparison of the Pentium 4 2.4C and Pentium 4 2.4 comes to comparison of performance of the processors with HT and without it. Performance doesn't depend much on the cache size, though there's some difference between the processors with the large and small L2 caches. But the worst disadvantage of the Celeron is not the cache but the lack of the Hyper-Threading. The gap between Celeron 2.0 and Pentium 4 2.0 is not that great, it's the same for Celeron 2.4 GHz and Pentium 4 of the same clock speeds without Hyper-Threading. And the breakaway of Pentium 4 2.4C is caused not by the large cache... AMD's processors look flabby in this test - AMD's flagships Athlon XP 3200+ and Athlon 64 FX-51 go on a par with Pentium 4 2.4C and fall behind Pentium 4 3.2 GHz. But the worst thing for AMD is not the defeat of its top CPUs but the fact that Athlon XP 2500+ looks equal to Celeron 2.4 GHz! "

Delerious · Member Joined: 22 Aug 2003

offline · Mondo Cane Joined: 05 Jun 2002 Location: 7 Craven Road

Delerious,

I have the same board. Where can I get this hacked
bios?

BJ_M · Patron Joined: 30 Jul 2002 Location: Canada

BJ_M · Patron Joined: 30 Jul 2002 Location: Canada

note: Full PAT is eanabled with 533 cpu's by default.
_________________
"Each problem that I solved became a rule which served afterwards to solve other problems." - Rene Descartes (1596-1650)

Delerious · Member Joined: 22 Aug 2003

BJ_M pointed you in the right direction. Follow the steps precisely, and
remember to heed any warnings. Here's a screen shot of another
program showing my memory bandwidth around 6000 MBytes/sec.

J

Twizzle · Member Joined: 09 Jun 2002

WooHoo 44 seconds... And my details are as follows:

Number of CPUs 1
APIC ID 0
Name AMD Athlon 64 3200+
Code name NewCastle
Specification AMD Athlon(tm) 64 Processor 3200+
Family/Model/Stepping FF0
Extended Family/Model F/F
Brand ID 1
Package Socket 939
Core Stepping DH7-CG
Technology 0.13µ
Instructions Sets MMX, Extended MMX, 3DNow!, Extended 3DNow!, SSE, SSE2, x86-64
Clock Speed 2202.7 MHz
Clock multiplier x10.0
HTT Bus Frequency 220.3 MHz
L1 Data Cache 64 KBytes, 2-way set associative, 64 Bytes line size
L1 Instruction Cache 64 KBytes, 2-way set associative, 64 Bytes line size
L2 Cache 512 KBytes, 16-way set associative, 64 Bytes line size
L2 Speed 2202.7 MHz (Full)
L2 Location On Chip
L2 Data Prefetch Logic yes
L2 Bus Width 128 bits

rallynavvie · Posted: Jan 15, 2005 09:33

Stuff is all over the place now. How is it that 3200 A64 is markedly faster than Cobra's OC'd 3500? And how is BJ_M's 2.66 P4 markedly faster than offline's OC'd P4? The results here indicate that overclocking slows your system down, at least for this real-world application benchmark.

Twizzle I need RAM specs and chipset specs. See the first post for specifics on submitting results.
_________________
NE CEDE MALIS SED CONTRA AUDENTIOR ITO

Cobra · Moderator Joined: 06 May 2002 Location: UK

rallynavvie · Posted: Jan 15, 2005 11:08

Cobra · Moderator Joined: 06 May 2002 Location: UK

I am wondering how we could have a system to actually prove results. I am more than happy to participate in any such scheme - my system scored 59 seconds and I could prove it.
_________________

Forum Rules :: PM me with any questions/problems/suggestions!

rallynavvie · Posted: Jan 15, 2005 13:47

Already thought of it, but the only thing I could think of is to have the person PM or email the TMPGEnc logfile to me to and a checksum of the final MPEG file. That way it would be harder to spoof results if your logfile and checksum don't reflect what they should be.

I've got a pretty large LAN party coming up in a couple weeks that I'm going to use to test some systems in person. I'll just put everything I need on a CD and take it from person to person. Since there's nothing to install it should go quick and nobody should worry about about trying it out. Hopefully there will be a good mix of systems so I can get a barometer of what to gauge other results with. I wasn't looking for massive differences between like systems. If two people had an A64 3000 and one had a couple seconds' faster time we could figure out where the other system is being slowed and thus help members optimize their systems for video. Having results scattershot like they are now doesn't really help anyone.
_________________
NE CEDE MALIS SED CONTRA AUDENTIOR ITO

Twizzle · Member Joined: 09 Jun 2002

Sorry guys. my full specs are as follows:

As before its an A64 3200 running and 2200Mhz on an Abit AV8. I have 1Gig of DDR 333 and there is one HDD which is partitioned but working on the same partition. I am using WinXP home and the system is a clean install (3 days old)

The rest is from CPUz

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

Northbridge VIA K8T800 Pro rev. 0
Southbridge VIA VT8237 rev. 0
Memory Type DDR
Memory Size 1024 MBytes
Memory Frequency 183.6 MHz (CPU/12)
CAS# 2.5
RAS# to CAS# 3
RAS# Precharge 3
Cycle Time (tRAS) 7
Bank Cycle Time (tRC) 10
DRAM Idle Timer 16

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

Module #1

General
Memory type DDR-SDRAM
Manufacturer (ID) Micron Technology (2CFFFFFFFFFFFFFF)
Size 512 MBytes
Max bandwidth PC2700 (166 MHz)
Part number 8VDDT6464AG-335D1

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

Timings table
Frequency (MHz) 133 166
CAS# 2.0 2.5
RAS# to CAS# delay 3 3
RAS# Precharge 3 3
TRAS# 6 7

Module #2

General
Memory type DDR-SDRAM
Manufacturer (ID) Micron Technology (2CFFFFFFFFFFFFFF)
Size 512 MBytes
Max bandwidth PC2700 (166 MHz)
Part number 16VDDT6464AG-335C2

Attributes
Number of banks 2
Data width 64 bits
Correction None
Registered no
Buffered no

Timings table
Frequency (MHz) 133 166
CAS# 2.0 2.5
RAS# to CAS# delay 3 3
RAS# Precharge 3 3
TRAS# 6 7

Cobra · Moderator Joined: 06 May 2002 Location: UK

I tested my system again now that it has bedded in and I've installed all my crap on it. My first set of results were on this setup:

http://www.videohelp.com/~cobra/cpuz.htm
(Athlon64 XP3500 @ 2.42GHz (so just over an XP3800), FSB255, RAM DDR510 dual-channel mode)

With a fresh copy of XP - 0:59 minutes
With a now well-used copy of XP - 1:00

I also clocked it back to standard settings (so the CPU is on-par with Twizzle's, just the RAM is quicker) - details from CPU-Z here:

http://www.videohelp.com/~cobra/non_overclock.htm
(Athlon64 XP3500 @2.2GHz, FSB200, RAM DDR400 dual-channel mode)

Scored - 1:09 minutes

This makes Twizzle's machine somehow 44% faster. That's an awful big difference - can anyone explain it?

As for overclocking in theoretical terms and actual terms I have:

Theoretical gain on CPU: 10%
Theoretical gain on HTT/FSB: 27.5%
Actual gain (in video encoding) : 15%

I believed that the speed of the CPU was really all that accounted for the rate you could encode video, but evidently the FSB also plays a big role.

Well, I hope these results are useful and interesting to you guys. If anyone has a tamper-proof way I can substantiate these results I am more than happy to do so.

Cobra
_________________

Forum Rules :: PM me with any questions/problems/suggestions!

Twizzle · Member Joined: 09 Jun 2002

Hmmm, Im a little concerned now that I must have done something wrong... so just to confirm:
1. I am using TMPGEnc Plus 2.59.47.155
2. I downloaded both parts of the video file and they extracted to one file that is 3.14mb in size
3. I used the mcf setting file from this thread
4. I ticked all of the boxes under the CPU setting (on a side note, when I left the Cache Setting one blank it slowed things down to about 1.05)
5. I ran it.... this time 45 seconds but I do have a program running in the background.

Im more than happy to prove the settings / results if some one tells me how they want to see it.

offline · Mondo Cane Joined: 05 Jun 2002 Location: 7 Craven Road

This test should be restarted and restructured in a fashion
that prevents data access. So far the data seems
corrupted. I'm not talking about the AMD results.

Twizzle · Member Joined: 09 Jun 2002

I have just downloaded the latest version of cpuz as I noticed mine was a little bit out of date. My actual CPU results are as follows:

Number of CPUs 1
APIC ID 0
Name AMD Athlon 64 3200+
Code name Winchester
Specification AMD Athlon(tm) 64 Processor 3200+
Family/Model/Stepping FF0
Extended Family/Model F/1F
Brand ID 4
Package Socket 939
Core Stepping DH8-D0
Technology 0.09µ
Instructions Sets MMX, Extended MMX, 3DNow!, Extended 3DNow!, SSE, SSE2, x86-64
Clock Speed 2202.8 MHz
Clock multiplier x10.0
HTT Bus Frequency 220.3 MHz
P-Rating 3200+
L1 Data Cache 64 KBytes, 2-way set associative, 64 Bytes line size
L1 Instruction Cache 64 KBytes, 2-way set associative, 64 Bytes line size
L2 Cache 512 KBytes, 16-way set associative, 64 Bytes line size
L2 Speed 2202.8 MHz (Full)
L2 Location On Chip
L2 Data Prefetch Logic yes
L2 Bus Width 128 bits

I think that the only difference that it has picked up is that it is an Winchester core at 0.09u and not the Northwood core at 0.13u. (Would explain why I had a big argument with the salesman about what he had sold me when I got home! OOps

)

Still, this shouldn't make that much of a difference shoukd it?

Faustus · insane. Joined: 11 Apr 2002 Location: Dallas, TX

Hrm... the Winchester core is very over clockable.

Cobra · Moderator Joined: 06 May 2002 Location: UK

If you're talking about your core and my core, mine is a Newcastle. I thought I was getting a Winchester, but I didn't... Shame - the Winchester is a nice core.

The actual difference between the two cores is power consumption and heat production - no speed difference. The move to 0.09u technology allows more transistors to be put on a chip and therefore more complex CPUs to be made (I think).

offline - I'm not sure I understand what you mean by "data access". Do you have something in mind to help our results?
_________________

Forum Rules :: PM me with any questions/problems/suggestions!

Faustus · insane. Joined: 11 Apr 2002 Location: Dallas, TX

I think he means the results collected in a way that everyone cannot see them until he has a good ammount of samples.

offline · Mondo Cane Joined: 05 Jun 2002 Location: 7 Craven Road

-edit-

no. @ cobra. I can't think of a simple solution.

We need a foolproof way of measuring that
makes each computer start as equal as possible and
that has a non modifiable output & also what flaystus says.

Cobra · Moderator Joined: 06 May 2002 Location: UK

Agreed. Basically, exclude the user from having to make any settings or tweaks, automatically capture their system settings (CPU-Z dump) and report the time with some sort of checksum on it so you can't fake it. Is this what you are proposing?

I don't know enough about this to offer help with constructing a system like this, or to even know if it is possible.
_________________

Forum Rules :: PM me with any questions/problems/suggestions!