A comparison between a few of my boxes

Duvie

Elite Member
Feb 5, 2001
16,215
0
71
**** System #1,#2 and #4 are now sold so no more test will be done****

Replacing it with an E6300 system I am hoping to get to 3.4ghz range like the one I built for a buddy and the one MarkFW900 built. Components are ordered and are proven winners. They will go into my 265's case with ample case cooling, 550watt Antec 2 true power 2 PSU and 300GB SATA2 HDD.


The fact is these test as well as my Folding results for the last month or so of my E6600 and my buddies E6300 show me get these things over 3ghz and they will outproduce the quad boxes, In actually things I do like below the results are very clear. NOT MUCH TAKES ADVANTAGE OF THE 4 CORES...Therefore I get better use out of these systems.

The quad 270 box is going next with the 6800GT agp card for either another E6300 or perhaps a jump at a woodcrest system, if they will be compatabile with adding a 4 core chip later like on my Giga 965p chipsets...


Systems:

#1: (2) opteron 270 (2ghz) w/ 2gb Ram DC (per cpu) cas 2-3-3-6
#2: (2) opteron 265 (2.5ghz (9x277)) w/ 1gb Ram SC (per cpu) cas 2.5-3-3-8 133divder
#3: (1) E6600 (3.26ghz (7x466)) w/ 2gb Ram DC cas 5-5-5-15 (933ddr2)
#4: (1) E6300 (3.40ghz (7x486)) w/ 1gb Ram DC cas 4-4-4-12 (972ddr2)
#5: (1) QX6700 (3.2ghz (8x400)) w/ 512gb Ram SC cas 5-5-5-15 (1000ddr2)
#6: (1) QX6700 (3.3ghz (8x412)) w/ 1gb Ram DC cas 5-5-5-15 (1030ddr2)
#7: (1) QX6700 (2.66ghx (10x266)) w/ 1gb Ram DC cas 5-5-5-15 (800ddr2)
#8: (1) Q6600 (2.4ghz (9x266)) w/ 1gb Ram DC cas 5-5-5-15 (800ddr2)


Cinebench 9.5 (multi-cpu render)

#1--------------553 (dual)
#1--------------979 (quad)
#3--------------996 (dual)
#4--------------1074 (dual)
#2--------------1176 (quad)
#8--------------1281 (quad) **NEW**
#7--------------1422 (quad) *NEW*
#5--------------1704 (quad)
#6--------------1756 (quad)


*NEW* = #6 is 23% higher clock speed and the score scales 23% higher. Cinebench scores linear improvement..Likely not realistic...

**NEW** = #8 again scales linear to both #7 and #6 systems...basically the faster you can go and the score will climb....

Dvdshrink

I used Gone in 60 Seconds DVD. I ripped to HDD with DVD-decryptor to isolate cpus. I did a full back up less foreign languages...It is a 65% compression to keep it to DVD 4.7gb standard. I use sharp adaptive.

w/ deep analysis

#1------------------------21:01 (quad)
#2------------------------N/A
#3------------------------15:06 (dual)

w/o deep analysis:

#1------------------------24:54 (dual)
#1------------------------15:40 (quad)
#2------------------------N/A
#3------------------------11:02 (dual)
#4------------------------10:59 (dual)
#8-------------------------9:14 (quad) **NEW**
#7-------------------------8:28 (quad) *NEW*
#5-------------------------7:58 (quad) **max cpu usage was 85-88%**
#6-------------------------7:12 (quad)
#6-------------------------6:59 (quad) Rev w/ multiple HDDs

*NEW* = #6 is 23% faster in clock but score was only 17.5% faster. Max cpu usage was similar as seen in system #5 though it may have held a higher average sustainable usage. I can only submit this may be the sign of the program being IO limited. Perhaps I will test this again once I have my Raid setup back up.

**NEW** = #6 is 38% faster in clock speed over #8 but only represents a 28% gain in performance...#7 is 11% faster then #8 but only represents 9% gain in performance. NOt quite as bad but it is likely we start seeing quads below 2.4ghz actually being slow enough the HDD can keep up...




TMPGenc 2.52

First off I decided to convert a Hidef AVI file to MPEG2 DVD standard....Clip is a Trailer from Xmen 3...

First test was to use Automatic Variable Bitrate (setting 80 hi6000/lo2000) 8 bit and normal motion precision.

#1-----------------------2:00 (dual)
#1-----------------------1:44 (quad)
#2-----------------------1:42 (dual)
#2-----------------------1:29 (quad)
#3-----------------------0:57 (dual)
#4-----------------------0:55 (dual)
#5-----------------------0:56 (quad)

* note: as mentioned in earlier review 2 cores stays pegged around 48-50%, but 4 cores only shows around 58-60% cpu usage average with an occasional spike to 75%. It didn't use 3 cores effectively let alone 4. So the real comparison comes against system #1 using just 2 cores. It is 2x as fast

Second test was to use 2 pass Variable Bitrate (setting 80 hi6000/lo2000) 8 bit and normal motion compensation.


#1-----------------------3:56 (dual)
#1-----------------------3:28 (quad)
#2-----------------------N/A
#3-----------------------1:54 (dual)
#4-----------------------N/A
#5-----------------------1:53 (quad) ** cpu usage was around 54% max**

* note: Again same as above and same as noted previously...the real comparison is the system #1 with 2 cores. Again 2x as fast.

Third test was to use Automatic Variable Bitrate (setting 80 hi6000/lo2000) 10 bit component precision and High Quality motion precision. Merely higher quality finished product.


#1-----------------------3:00 (dual)
#1-----------------------1:53 (quad)
#2-----------------------N/A
#2-----------------------1:27 (quad)
#3-----------------------1:25 (dual)
#4-----------------------1:19 (dual)
#8-----------------------1:11 (quad) **NEW**
#7-----------------------1:05 (quad) *NEW*
#5-----------------------0:58 (quad)
#6-----------------------0:54 (quad)
#6-----------------------0:52 (quad) Rev w/ multiple HDDs

*NEW* = #6 is 23% faster in clock speed and is 20% faster in the test versus test system #7. It scales close to linear. In both case cpu usage was very consistent. May be hints of the program becoming more IO limited as the cpu speed increases.

**NEW** = #8 scales to #7 near linear..similar to above results


* note: as mentioned in previous thread with the addition of higherprecision and higher motion compensation the 4th core was used more with usages spike in the 80-88% range.
Here we see once again 2 cores of #3 is about 2x as fast as 2 cores of #1. Almost a wash with quad 2.5ghz versus dual 3.26ghz of the C2D.


I think what we see here is same as in my previous thread...

http://forums.anandtech.com/messageview...hreadid=1830732&enterthread=y&arctab=y

and in Anandtech's recent QX6700 review. Many programs like these are well suited for dual core but do not effectively use 4 cores at this time. More test showing HD content will show more usage of the 3rd and 4th core and the 2 cores of the E6600 will start to fall behind the quad 265 box. However it is amazing a 300 dollar chip is laying waste to 600-800 dallars worth of opterons.

may have to sell some of these opteron boxes and get more C2D's....
 

myocardia

Diamond Member
Jun 21, 2003
9,291
30
91
Hey, Duvie, since system #2 is so much faster than system #1, why is it left out of nearly every benchmark?
 

Duvie

Elite Member
Feb 5, 2001
16,215
0
71
Originally posted by: myocardia
Hey, Duvie, since system #2 is so much faster than system #1, why is it left out of nearly every benchmark?


I only left it out of DVDshrink and 2nd test of the TMPGenc....I coudln't do Dvdshrink cause since that was just afolding box I never installed a DVD-rom to rip the movie...

The 2nd test I just didn't do because I felt the 1st and 3rd were sufficient to get the point across....



 

Duvie

Elite Member
Feb 5, 2001
16,215
0
71
Originally posted by: avi85
Duvie, maybe you could do some Photoshop CS2 benchmarks?



Send me a link to the app...I originally had it installed and now it wont run without purchasing it I guess...
 

Duvie

Elite Member
Feb 5, 2001
16,215
0
71
Added my latest system...Though higher clock speed and ram timings the L2 cache is less and hence perhpas noticeable...

Take a look! Either way it still beats my quad 2.5ghz box in everything except cinebench.
 

Yellowbeard

Golden Member
Sep 9, 2003
1,542
2
0
Good work Duvie. This should even appease those who always demand real world results over pure benchmarks. I started building computers for my wife's video editing business and I always tested builds/OCs/tuning by doing encoding tests and by testing the machine using what it was built for as a guage.

IMO, you might even want to see if some people around here want to agree on a standardized benchmark setup like you have here. It would be a good way to get apples v apples cpmparisons.
 

Duvie

Elite Member
Feb 5, 2001
16,215
0
71
updated QX6700 to 3.3ghz...thinking about doing scok 2.667ghz speed as well as perhaps stock 2.4ghz Q6600 speeds in there
 

Duvie

Elite Member
Feb 5, 2001
16,215
0
71
updated QX6700 at stock settings of 2.66ghz and 800ddr2...probably what most common non ocers may run....


I need to try to run DVDshrink with some different adaptive correction to see if I can force a higher cpu usage and those get closer to the scaling of the speeds I have tested...Above stock speeds is only 14% slower then the chip at 3.3ghz eventhough it represents a 23% increase in clock. It suggest that perhaps I am being IO limited and thus average cpu usage may be less as the pcu is constantly waiting for the HDD to catch up in the writing process...

Darn it we need faster drives...Cpus are far outpacing the HDDs..i will retest this on my Raid setup when I get it back up...

 

Duvie

Elite Member
Feb 5, 2001
16,215
0
71
updated at Q6600 speeds and a 800ddr2 system....

As I expected....DVDshrink actually hit higher cpu utilization by a few percentage points but actually also held steady in the 90-84 range with only a spike now and then. this is a clear sign it is being limited by IO performance....So really cpu speed is becoming moot until we see faster drives...

All you have to do is watch the buffer count in dvdshrink and she how it correlates to sharp down spikes in cpu usage on task manager..the biggest thing I see is sharper and more frequent spikes with the faster speed as the cpu is really waiting for the HDD...
 

avi85

Senior member
Apr 24, 2006
988
0
0
Originally posted by: Duvie
updated at Q6600 speeds and a 800ddr2 system....

As I expected....DVDshrink actually hit higher cpu utilization by a few percentage points but actually also held steady in the 90-84 range with only a spike now and then. this is a clear sign it is being limited by IO performance....So really cpu speed is becoming moot until we see faster drives...

All you have to do is watch the buffer count in dvdshrink and she how it correlates to sharp down spikes in cpu usage on task manager..the biggest thing I see is sharper and more frequent spikes with the faster speed as the cpu is really waiting for the HDD...

Cheetah 15k.5's in RAID-0 FTW!
 

Duvie

Elite Member
Feb 5, 2001
16,215
0
71
Tried to test IO limitation theory by using 2 HDDs...source on one drive writing to a 2nd drive...speed up was 3-4% in both Dvdshrink and TMPGenc test....

During test max cpu usage increased 2-3% and average cpu usage stayed between 87-83 for most of the duration with little or no spikes downward as seen in previous testing of 3.3ghz....Seems as if the cpu wasn't waiting as much...