3.06 performance at [H]: do we miss something?

chainbolt · 15th November 2002, 12:27 AM

Well, below is a selection of 3.08 reviews. This is without doubt a mighty gun. And H/T works obviously well in certain situations. But all reviews show that with or without HT you get the same score in 3DMark 2001, as expected one might say. There is only one exception: {H} is not only showing a much higher score in 3DMark 3001 than anybody else, but also a substantial gain with H/T. Everybody is getting with a 3.06@3.06 and a Radeon 9700 around 15000~15500 in 3Dmark 2001 at the default test settings. Even tech-report gets only 15400 with a RDRAM system, but H gets 16900 with a DDR (!) system. And on top of the miraculously high score they get 17,850 with HT enabled:

"A large gain from Hyper-Threading technology here. The gain was so significant that I reran the test several times to ensure that the results we got were indeed accurate."

Nobody else is getting any gain with H/T in 3DMark 2001.

Do I miss something?

Anand
http://www.anandtech.com/cpu/showdoc.html?i=1746&p=14

gamePC
http://www.gamepc.com/labs/view_cont...Ftest=1&bhcp=1

Hardocp
http://www.hardocp.com/article.html?art=Mzg4

hardtecs
http://www.hardtecs4u.com/reviews/20...hz/index19.php

Suhian
http://www.sudhian.com/news.cfm#2

Techreport
http://tech-report.com/reviews/2002q...6/index.x?pg=9

Tech report: 3.06@3.06 / Radeon 9700 / 512 MB RDRAM

This is the "changed chart. It showed initially a difference of almost 1000 points. After they realized that all other sites don't get any difference they "reduced" their "test" by around 650 points

{H} 3.06@3.06 / Radeon 9700 / 512 MB DDR XMS 3500

hardtecs: 3.06@3.06 /Radeon 9700 Pro

Lombers · 15th November 2002, 12:44 AM

I just noticed that too Chainbolt. It seems the HardOCP hyper threading enabled tests all show a much more significant increase than any other review I have read. I don't understand how this is possible, as they all seem to be using pretty much the same hardware. Unfortunately, it seems that HardOCP is the odd one out here, hence they either a) know something we don't or have a specific setting in the bios or in windows or b) their benchmarks are innacurate.

The only thing I can assume is that Windows would require a fresh installation for a 3Ghz P4 to ensure that hyper threading is working correctly, as I'm assuming a different kernel is used. If the other reviews are just ghosting the system, this could lead to innacurate results. This is only a guess, but I do find it very strange indeed.

chainbolt · 15th November 2002, 12:48 AM

It's getting worse! First the difference was 1000 points as noticed by some OCAU members, and now it's suddenly "only" 450 points. I guess tomorrow morning the difference is zero! This is no good. If they made a mistake they should say so. But secretly changing the values looks very "fishy".

MnM · 15th November 2002, 12:56 AM

Hmmm I dont like it at all ...yes IT WAS 1K difference when I red the review ... too bad that [H] will go to that extent to get a free cpu

...can anybody say tomshardware II?

Lombers · 15th November 2002, 12:57 AM

Quote:

Originally posted by chainbolt
It's getting worse! First the difference was 1000 points as noticed by some OCAU members, and now it's suddenly "only" 450 points. I guess tomorrow morning the difference is zero! This is no good. If they made a mistake they should say so. But secretly changing the values looks very "fishy".

OMG!!! I just noticed that too. The hyper threading enabled score was definetely 1000 points more than the normal score with hyper threading disabled. Will the other benchmarks numbers change now as well?? It does seem perhaps HardOCP have some explaining to do. I'll be keeping an eye on this one. So far, I have found The Tech-Report's review to be one of the best.

pilot6 · 15th November 2002, 12:59 AM

Yeah I knew all along those figures couldnt be right, people with with high Ati9700 Vid O/C's on an 3.2Gig + O/C P4 processor are just managing to hit those scores, and even the 3.06 HT disabled score is ridiculous, yep someone is full of it over there.

chainbolt · 15th November 2002, 1:05 AM

it's really funny, maybe you remember our discussions about "sites" and how much we can trust them? A sites is only as good as the author of the respective article. And this guy is in shit now. I guess Mr. Kyle will be a little angry with him

Lombers · 15th November 2002, 1:14 AM

Thank god for IE6 cache

This is what the graph posted above looked like:

As you can see, quite a difference. Haven't noticed any other changes in benchmarks. Results look a little too skewed for my liking here. I still tend to think Anandtech and Techreport's numbers are most accurate.

P.S. If the above picture is down, it's due to optusnet problems. I'm sure it will be all fixed soon

nutty · 15th November 2002, 1:24 AM

hmm this is susprising
the fact they have changed it has made it look even more sus
lets hope they havnt done a Tom

SuperBeast · 15th November 2002, 1:30 AM

they even left each of them with 29 on the end..

really makes u wonder which ones u can trust

chainbolt · 15th November 2002, 1:36 AM

Quote:

Originally posted by chainbolt
it's really funny, maybe you remember our discussions about "sites" and how much we can trust them? A sites is only as good as the author of the respective article. And this guy is in shit now. I guess Mr. Kyle will be a little angry with him

wow, I am shocked, I just realized that Kyle himself is the author.

By the way, I also think that the Anand article is well balanced one.

azaroth · 15th November 2002, 1:40 AM

I just checked the review out at [H]ardocp and it appears to be different yet again, its 879 points difference at this present hour.

chainbolt · 15th November 2002, 1:47 AM

Quote:

Originally posted by azaroth
I just checked the review out at [H]ardocp and it appears to be different yet again, its 879 points difference at this present hour.

this is too funny! it's changing with the hour

MnM · 15th November 2002, 1:50 AM

Hehehe Intel must have a "morphing" CPU ...is fast when they advertise it ...but then when you get home with it, it "morphs" into a slower CPU

thst why [H] cant get the real speed

looktall · 15th November 2002, 2:09 AM

Quote:

Originally posted by chainbolt

wow, I am shocked, I just realized that Kyle himself is the author.

By the way, I also think that the Anand article is well balanced one.

kyle wasn't the author, he was the editor. steve lynch is the author.

15th November 2002, 12:27 AM	#1
chainbolt Little member Join Date: Jun 2001 Location: Tokyo Posts: 76,465	3.06 performance at [H]: do we miss something? Well, below is a selection of 3.08 reviews. This is without doubt a mighty gun. And H/T works obviously well in certain situations. But all reviews show that with or without HT you get the same score in 3DMark 2001, as expected one might say. There is only one exception: {H} is not only showing a much higher score in 3DMark 3001 than anybody else, but also a substantial gain with H/T. Everybody is getting with a 3.06@3.06 and a Radeon 9700 around 15000~15500 in 3Dmark 2001 at the default test settings. Even tech-report gets only 15400 with a RDRAM system, but H gets 16900 with a DDR (!) system. And on top of the miraculously high score they get 17,850 with HT enabled: "A large gain from Hyper-Threading technology here. The gain was so significant that I reran the test several times to ensure that the results we got were indeed accurate." Nobody else is getting any gain with H/T in 3DMark 2001. Do I miss something? Anand http://www.anandtech.com/cpu/showdoc.html?i=1746&p=14 gamePC http://www.gamepc.com/labs/view_cont...Ftest=1&bhcp=1 Hardocp http://www.hardocp.com/article.html?art=Mzg4 hardtecs http://www.hardtecs4u.com/reviews/20...hz/index19.php Suhian http://www.sudhian.com/news.cfm#2 Techreport http://tech-report.com/reviews/2002q...6/index.x?pg=9 Tech report: 3.06@3.06 / Radeon 9700 / 512 MB RDRAM This is the "changed chart. It showed initially a difference of almost 1000 points. After they realized that all other sites don't get any difference they "reduced" their "test" by around 650 points {H} 3.06@3.06 / Radeon 9700 / 512 MB DDR XMS 3500 hardtecs: 3.06@3.06 /Radeon 9700 Pro Last edited by chainbolt; 15th November 2002 at 2:58 AM.

15th November 2002, 12:48 AM	#3
chainbolt Little member Join Date: Jun 2001 Location: Tokyo Posts: 76,465	It's getting worse! First the difference was 1000 points as noticed by some OCAU members, and now it's suddenly "only" 450 points. I guess tomorrow morning the difference is zero! This is no good. If they made a mistake they should say so. But secretly changing the values looks very "fishy". Last edited by chainbolt; 15th November 2002 at 12:51 AM.

15th November 2002, 12:56 AM	#4
MnM Member Join Date: Jun 2001 Posts: 490	Hmmm I dont like it at all ...yes IT WAS 1K difference when I red the review ... too bad that [H] will go to that extent to get a free cpu ...can anybody say tomshardware II? __________________ -==MnM==- Last edited by MnM; 15th November 2002 at 12:56 AM.

15th November 2002, 1:14 AM	#8
Lombers Feed Me! Feed Me! Join Date: Jun 2001 Location: Melbourne Posts: 4,026	Thank god for IE6 cache This is what the graph posted above looked like: As you can see, quite a difference. Haven't noticed any other changes in benchmarks. Results look a little too skewed for my liking here. I still tend to think Anandtech and Techreport's numbers are most accurate. P.S. If the above picture is down, it's due to optusnet problems. I'm sure it will be all fixed soon __________________ You must always push the limits, because if you never fail, you will never succeed. Overclockers Australia Folding@Home Team Founder Overclockers Australia Folding@Home Team Dedication Award Last edited by Lombers; 15th November 2002 at 1:15 AM.

15th November 2002, 1:24 AM	#9
nutty Member Join Date: Feb 2002 Location: brisbane.queensland.au Posts: 1,687	Re: 3.06 performance at [H]: do we miss something? hmm this is susprising the fact they have changed it has made it look even more sus lets hope they havnt done a Tom __________________ My s15 200sx![/size]

15th November 2002, 12:44 AM	#2
Lombers Feed Me! Feed Me! Join Date: Jun 2001 Location: Melbourne Posts: 4,026	I just noticed that too Chainbolt. It seems the HardOCP hyper threading enabled tests all show a much more significant increase than any other review I have read. I don't understand how this is possible, as they all seem to be using pretty much the same hardware. Unfortunately, it seems that HardOCP is the odd one out here, hence they either a) know something we don't or have a specific setting in the bios or in windows or b) their benchmarks are innacurate. The only thing I can assume is that Windows would require a fresh installation for a 3Ghz P4 to ensure that hyper threading is working correctly, as I'm assuming a different kernel is used. If the other reviews are just ghosting the system, this could lead to innacurate results. This is only a guess, but I do find it very strange indeed. __________________ You must always push the limits, because if you never fail, you will never succeed. Overclockers Australia Folding@Home Team Founder Overclockers Australia Folding@Home Team Dedication Award

15th November 2002, 12:59 AM	#6
pilot6 Member Join Date: Aug 2001 Location: Melbourne Posts: 558	Yeah I knew all along those figures couldnt be right, people with with high Ati9700 Vid O/C's on an 3.2Gig + O/C P4 processor are just managing to hit those scores, and even the 3.06 HT disabled score is ridiculous, yep someone is full of it over there.

15th November 2002, 1:05 AM	#7
chainbolt Little member Join Date: Jun 2001 Location: Tokyo Posts: 76,465	it's really funny, maybe you remember our discussions about "sites" and how much we can trust them? A sites is only as good as the author of the respective article. And this guy is in shit now. I guess Mr. Kyle will be a little angry with him

15th November 2002, 1:30 AM	#10
SuperBeast Member Join Date: Mar 2002 Location: Melbourne Posts: 353	they even left each of them with 29 on the end.. really makes u wonder which ones u can trust __________________ Successful trades: Lots intel 3570k @ 4.2 \| 2x8gb G.Skill Ares 1866 \| 2x Gigabyte GTX 670 OC \| ASRock Z77 Extreme6 \| 2x 256gb Samsung 830 Series \| Corsair H100 \| NZXT Phantom \| Corsair HX-1050 M17X-R3 Nebula Red \| i7-2720QM \| 8GB \| Radeon HD 6970m \| 120GB OCZ Vertex 3 \| 500GB Hybrid HP ProLiant MicroServer \| AMD N36L \| 3GB \| 2x2TB \| 1x1TB

15th November 2002, 1:40 AM	#12
azaroth Member Join Date: Aug 2001 Location: Sydney Posts: 477	I just checked the review out at [H]ardocp and it appears to be different yet again, its 879 points difference at this present hour. __________________ i7-2600K - 8GB Corsair Vengeance - Asrock Z68-Ext3-Gen3 - Sapphire 6970 2GB - 120gb Corsair Force 3 SSD - Antec Tri-power Trio 650W - Acer X243 HQ 24" & LG W2252TQ 22" G15/Razer Naga Molten Ed If your not with us, then your with the trolls Bargain downloadable games thread

15th November 2002, 1:50 AM	#14
MnM Member Join Date: Jun 2001 Posts: 490	Hehehe Intel must have a "morphing" CPU ...is fast when they advertise it ...but then when you get home with it, it "morphs" into a slower CPU thst why [H] cant get the real speed __________________ -==MnM==- Last edited by MnM; 15th November 2002 at 1:52 AM.