PCMark2002 and Intel bias

[joe2004]

I noticed that PCMark2002 has been rather bias toward Intel like no other benchmark I know. And unreliable. I have much higher scores on memory when I reboot and run PCMark2002 than when I do that without rebooting. And when I run it 4-5 times every time after the first time I get lower scores. Also hard drive scores are variable as well. Only CPU scores keep consistent. It looks to me that is one fluke of a benchmark. Like Aida32 where all sorts of numbers are popping out, unlike Sandra that is rather consistent.

 

[Shimmishim]

and what's the problem?

 

[Duvie]

So where is the Intel Bias??? Fluctuations in benches except cpu shows some sort of conspiracy???

PCmark2002 performs exactly as one would expect based on sissoft sandra performance...Drop the conspiracy crap and just realize this is a synthetic benchmark and anybody who knows their arse from their head around these parts knows sissoft and synthetic benches don't always mean crap and it is real world performance that counts....

Now lets look at the pc mark numbers shall we???

Cpu is higher then amd equally pr rated systems right??? OK....First we know with the bartons came great skew in the pr rating so don't compare it. Secondly in the cpu test they do some multimedia test which we know favors INtel cause of SSE2 optimization and most all real world multimedia test show this to be true.

Memory....The fact is Intel cpus doe to their quad pumped nature have more theoretical bandwidth...proven fact....It is also proven that dual channel INtel boards have more bandwidth and aremore advantageous to amd dual channel mobos. Amd system the cpu is bandwidth limited whereas until dual channel mobos INtel's chips were memory bandwidth limited.

Amd Barton 3200+ w/ 400fsb has a peak theoretical bandwidth of 3.2gb/s
P4 400fsb chip = 3.2gb/sec
P4 533fsb chip = 4.3gb/sec
P4 800fsb chip = 6.4gb/sec

So where is the shock here???

Repeated running of test showing lower scores, or rebooting showing best scores is no secret!!! well known and most who bench are benching on fresh restarts....

HDD...Now that is a different test altogether but I rarely get much fluctuation here in multiple testing...

 

[joe2004]

So where is the shock here???

Barton 3200 makes about 6,900 on PCMark2002 and Intel 3.2C makes about 8,000. Talking about 16% difference. Now who believes that 3.2C is that much better than Barton 3200?
I don't. Run SuperPi and that is a pure power of CPU, you'll see there is almost no difference or maybe 1 sec on 2 MB score.

Repeated running of test showing lower scores, or rebooting showing best scores is no secret!!! well known and most who bench are benching on fresh restarts....

That is false. Variety of benchmarks are consistent and do not need "fresh" start, for example Sandra, for example 3DMark, for example SuperPI etc ect. Do you run your programs by first rebooting PC? That is silly, to say at least.
If you are only benching to make fake claims then OK but I rather know what is the power of my system than fake with my scores. If a benchmark runs so much of a stddev then you need to run it 10 times and take average else your scores are fake.

 

[Duvie]

Originally posted by: joe2004

So where is the shock here???

Barton 3200 makes about 6,900 on PCMark2002 and Intel 3.2C makes about 8,000. Talking about 16% difference. Now who believes that 3.2C is that much better than Barton 3200?
I don't.

Repeated running of test showing lower scores, or rebooting showing best scores is no secret!!! well known and most who bench are benching on fresh restarts....

That is false. Variety of benchmarks are consistent and do not need "fresh" start, for example Sandra, for example 3DMark, for example SuperPI etc ect. Do you run your programs by first rebooting PC? That is silly, to say at least.

As for the barton 3200+ why don't you go look at some multimedia testing like encoding and such where a 2.4c can take out a Barton 3200+, so don't give me that crap..Also remeber how much better in multimedia scoring the p4 3.2c does versus the 3200+. That synthetic bench about confirms this synthetic bench...Does this sound silly to you??? You are putting to much value in these benches anyways.

I didn't say a restart is needed in every app...I agree I get fluctuation up or down in 3dmark...however in sissoft benches and pcmark2002 many of us get best marks after a fresh restart....Why??? Cause we often run programs that when they are close may not release all the memoryor leave an app running in the background. A fresh restart for many of us who have real limited number of things at startup allows the system to refresh and have the maximum amount of memory defraged ready to go...Again I see this mainly in the synthetic test and ot real world benches which you should concern yourself with instead of trying to use these things to compare cross-platform. Again like I stated most around these parts don't put a lot of stock in thease numbers anyways....

SuperPi I think doesn't even test the system memory at all, and that is where I think the system restart helps anyways not in CPU performance.

 

[joe2004]

SuperPi I think doesn't even test the system memory at all

That is false.
I suggest you use 2-2-2-5 timmings and then 3-4-4-8 timmings and see how much difference you'll get.

Cause we often run programs that when they are close may not release all the memoryor leave an app running in the background.

Aha? So ... is that what benchmark isn't capable of? So which score is then the one that is to be counted on, with or without memory defragmentation?

As for the barton 3200+ why don't you go look at some multimedia testing like encoding and such where a 2.4c can take out a Barton 3200+,

There are also benches like Winstone where Barton 3200 will take on 3.2 GHz and kill it outright. One way or another Barton 3200 is not that much worse than Pentium as PCMark2002 programmers would like you to believe.

 

[Duvie]

I still like to run it multiple times anyways and then average the score....Ie like 5 times, and I don't restart for each test....I usually only run this synthetic crap when I am first ocing or setting up my system and then never run them again. I run all the sissoft programs (cpu/MM/mem) then go into 3dmark2k1se and then Prime95 bench and then onto prime95 testing and memtesting....That is it. Either way a minor fluctuation most see anyways from fresh to not so fresh is neglible when we are talking about 50-100 pts in a test that tops out at 9000-14000, right???

I wasn't sure about the superpi but then again I don't dwell on those things...My memory timings are already tweaked to more important real world apps I run by the time I run superpi so I don't chaneg the timings to get a better score while sacrificing something else...


as for your 16% barton 3200+ to P4 quandry...I suggest you look at these numbers again...hell reread the whoel article cause it shows the 16% to be quite true in real world apps if not conservative!!!!

Divx 5.05 testing ( 3.2c vs 3200+ )
well over 16% and the 2.4c even best it!!!



3d rendering ( 3.2c vs 3200+ )
Again the p4 3.2c wins by over 16%

Through in the UT flyby-botmach test and again the 16% doesn't look like a stretch!!!

MP3 encoding ( 3.2c vs 3200+ )
Again over 16% and the 2.4c wins another!!!

AV mpeg encoding ( 3.2c vs 3200+ )
Agian one over 16% and one at 10%...2.4c does well on both!!!

Winrar archiving and Seti@Home ( 3.2c vs 3200+ )
And once again..OUCH


I think you ned to realise SSE2 optimizatios in the cpu help and if the pcmark2002 is figuring for those then the spread is quite real....

I think you need to take off the AMD blinders!!!! The Barton was not that great of a processor. They gave it cache but each time they neutered it by reducing clock speed yet raising pr rating. Noticein 3200+ and some 3000+ reviews the number of times the xp 2800+ beat the the bartons with less clock speed....

 

[joe2004]

Tom's hardware?
BWAHAHAHAHAHAHAHA!!!!

This guy came recently as the only web site that made how Athlon FX51 got beaten by Pentium 3.2EE.
Plus in the past he has shown his bias many times, you have whole thread in overclockers.com with examples of his bias.

http://forum.oc-forums.com/vb/showthread.php?s=da335f94428c0db48a6830092fac41a3&threadid=235524

If THG is known for anything besides his video card comparisons, which surprisingly show Athlon XP 2700 pulling a lead over Intel(!), is for his Intel bias.

3.2C is faster than Barton 3200 but not even close Tomshardware would tell you. read Anandt for more balanced comparison, I am sure you can find it.

 

[Shimmishim]

the point is...

you're using a synthetic benchmark...

woop dee doo that intel beats amd at 2k2 benchies but is within 1-2 seconds on superpi...

and you can't compare a 3.2 ghz p4 to a 3200+ barton...

amd even claims its not to compare to p4's but for some reason people do anyway...


this is argument / statement / whatever has been discussed in this forum one too many times...

just accept the fact that intel is better than amd at encoding/decoding whatever..

and that amd is better than intel at business aps...

and that they are almost equal in gaming performance...


(i have no bias towards amd or intel)

 

[joe2004]

Yup shimmishim, that is my opinion too. I have both Intel and Barton machines and I don't see any dramatic differences.

 

[Duvie]

Originally posted by: joe2004
Tom's hardware?
BWAHAHAHAHAHAHAHA!!!!

This guy came recently as the only web site that made how Athlon FX51 got beaten by Pentium 3.2EE.
Plus in the past he has shown his bias many times, you have whole thread in overclockers.com with examples of his bias.

http://forum.oc-forums.com/vb/showthread.php?s=da335f94428c0db48a6830092fac41a3&threadid=235524

If THG is known for anything besides his video card comparisons, which surprisingly show Athlon XP 2700 pulling a lead over Intel(!), is for his Intel bias.

3.2C is faster than Barton 3200 but not even close Tomshardware would tell you. read Anandt for more balanced comparison, I am sure you can find it.

I did you frekkin moron that is where the first bunch of links are from..LOL HAHAHA!!!! Tomshardware is not that far off and every reputable site declared the barton 3200+ a terrible comparison....

You new members who start threads with no evidence to back up your FUD and then wont listen to ppl who have evidence..I am tired of you punks!@!! GO live in your synthetic test world and hand-jack yourself to your sandra, Pcmark and 3dmark benches....

 

[Duvie]

Originally posted by: joe2004
Yup shimmishim, that is my opinion too. I have both Intel and Barton machines and I don't see any dramatic differences.

Where did he say that??? He said they were comparable in gaming....NOw figure in the ass-kicking they take in the multimedia and rendering and we have a different ballgame...

I own both AMD and INtel right now, do you???

 

[Duvie]

Originally posted by: Shimmishim
the point is...

you're using a synthetic benchmark...

woop dee doo that intel beats amd at 2k2 benchies but is within 1-2 seconds on superpi...

and you can't compare a 3.2 ghz p4 to a 3200+ barton...

amd even claims its not to compare to p4's but for some reason people do anyway...


this is argument / statement / whatever has been discussed in this forum one too many times...

just accept the fact that intel is better than amd at encoding/decoding whatever..

and that amd is better than intel at business aps...

and that they are almost equal in gaming performance...


(i have no bias towards amd or intel)

Not much of a discussion here...he throws out a bunch of wild claims he doesn't back and then can't see the point that pcmark is obviously in the cpu testing multimedia performance and that is likley where the p4 gets its pcmark numbers. the real world multimedia test show this 16%+ lead to somewhat verify these synthetic test he should not be worrying about anyways....

I think he is a lame-ass!!!

 

[joe2004]

Duvie, a little civility would not hurt you. And a little less punctuation marks as well. Take it easy.

 

[Duvie]

Go troll elsewhere....Seems to be a rash of you ppl lately!!! Got some wild conspiracies on the prescott and cpubench2003 as well???

I like punctuations!!!

 

[Shimmishim]

Originally posted by: joe2004
Duvie, a little civility would not hurt you. And a little less punctuation marks as well. Take it easy.

not to rub salt in the wound, but HE is right...

and i value his opinion a little more than yours...

 

[Duvie]

Originally posted by: Shimmishim

Originally posted by: joe2004
Duvie, a little civility would not hurt you. And a little less punctuation marks as well. Take it easy.

not to rub salt in the wound, but HE is right...

and i value his opinion a little more than yours...

Who are you referring to Shim???

 

[Shimmishim]

Originally posted by: Duvie

Originally posted by: Shimmishim

Originally posted by: joe2004
Duvie, a little civility would not hurt you. And a little less punctuation marks as well. Take it easy.

not to rub salt in the wound, but HE is right...

and i value his opinion a little more than yours...

Who are you referring to Shim???

YOU YOU IDIOT!

j/k j/k

i value your opinion a little more than his.. SHEESH...

you need to stop smokin' that crack dude!

 

[joe2004]

Lucky him, and both of you two.

 

[Jeff7181]

What the hell Duvie... don't think I've ever seen you resort to name calling before.

I'm wondering where Tom came up with 1 hour 22 minutes for the P4's SETI@Home test. He must have run two threads and then averaged the total time. Cause there's no way in hell a P4 finished a WU in 1 hour 22 minutes... go check out the DC forums if you don't believe me.

 

[Duvie]

I am just tired of these new members floating around here....


I am not sure about Seti test specifically but I gave both AT and Tomshardware...I can go get pctechreport, Aceshardware, and others but you know all about the 3200+ barton...The fact is the 16% in Pcmark2002 can be justified in real world benches so JOe shouldn't be so surprised.....The point his where was any proof of his but making a statement. I gave links to support my statement....


Name calling is happening more and more as I get tired of these new guys...I fell like Wingznut when he took a brief vacation for posting in the AT threads....

 

[andreasl]

Anyone who is interested in knowing what PCmark2002 REALLY tests, you should read this article:

PCMark2002 - A First Look

 

[Slammy1]

I'm a bit afraid to post in this thread now. I just want to make the point that we are all enthusiasts, and our goal is to improve the computing experience for the members (ourselves included). Everyone's opinion is meaningful. And, of course any benchmark that tests CPU and memory speed will be biased towards Intel over AMD.


j/k Pleas don't flame me : Puppydogeyes :

 

[Duvie]

Originally posted by: andreasl
Anyone who is interested in knowing what PCmark2002 REALLY tests, you should read this article:

PCMark2002 - A First Look

And there you go!!! Just as the test I have shown above...The cpu test has quite a few multimedia test and therefore is not really that much of a shock when you see all the real-world benches that show INtel leading in those test...

Yes it is a synthetic test. I have stated this numerous times don't put that much stock in it, but don't claim it is biased unless you have proof. It is good for what that article tested...same architecture and same platform comparison...

I think the article was ather wish-washy on some of its conclusions throughtout the article....

I don't use this test anyways. I usually go past all the sissoft, pcmark, sysmark, and 3dmark test when I do read reviews...

 

[joe2004]

Bias is the claim based on the selection of tests.
That is an integrated bench, or synthetic if you like, where the tasks were chosen. These chosen tasks are the way that Intel gets advantage whether intended or unintended, the bias is there. We know full well that many science tasks will show that Intel Pentium IV HT has little or no advantage over Barton 3200.

As far as THG goes I think there is a concensus in OC community that that site is obviously bias toward Intel, I gave you the link to read the discussion.

And for one I don't know what are you foaming about. I claimed bias and you yourself are telling that the choice of tasks (multimedia) favors Intel and then you resort to name calling like I stole something from you. By the way what is a newbie to you? I might well be older than you and with more computing experience than you, how would you know?
Take it easy.