PDA

View Full Version : [H] Cpu Scaling Review


Pages : [1] 2

Gar
06-28-04, 08:58 PM
Found this review very interesting.

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

I started a thread at [H] about this and it was requested by a member there to ask some of you who have a 6800U to post your scores using same settings, game, speed (core @ 400mhz), and drivers (61.34) using FRAPS. Could you please post a pic with your results. He feels that something is considerably wrong with Brent's review and its results. I appreciate it.

Edit: Oops, this should be in Benchmark central. Sorry :).

Vagrant Zero
06-28-04, 09:07 PM
Yes please, HOCPs figures are absolutely ridiculous and they need to be corrected. Preferabley by some of you guys with the actual cards. :ORDER:

MUYA
06-28-04, 09:08 PM
Gar - Tis ok...it's also about NV cards performances.


Vagrant - they also have cards. What exactly do u see wrong with them dude?

Waffles
06-28-04, 09:15 PM
I'd like to see other reviewers try to do a comparison that is similiar to the [H] reviews. [H] is the only site I've seen that makes the NV40 look soo terrible. a FULL resolution LOWER to MATCH the x800 XT PE speed?

Is this really the case?

MUYA
06-28-04, 09:23 PM
Clay will have a one coming shortly. So be patient ;)

burningrave101
06-28-04, 09:26 PM
Yea i've looked over that review and it nowhere near compares to the performance results shown in half a dozen or more other new reviews using the 61.34 drivers or higher.

The new review at Anandtech shows completely the opposite for results.

http://www.anandtech.com/video/showdoc.html?i=2096

I dont know what the deal is but somehow HardOCP manage to always make the nVidia cards perform alot worse then anyone else that reviews them lol.

What about all this talk about FRAPS vs Time Demos? Do most sites like Anandtech use Time Demo's? Is FRAPS a better indicator of real world performance?

We've got a rather intensive thread going on right here:

http://www.hardforum.com/showthread.php?t=769566&page=10

combat_wombat
06-28-04, 09:32 PM
that websites reviews totally suck.

Gar
06-28-04, 09:33 PM
We've got a rather intensive thread going on right here:

http://www.hardforum.com/showthread.php?t=769566&page=10



Yeah I know, I started it :headbang:

Brent
06-28-04, 09:38 PM
FRAPS gives you a realworld view of how performance is in a game since it captures the framerate every second (if you select that option) to gather data. How can it be anything but accurate of game performance?

FYI Recorded timedemos in FarCry do NOT playback the AI, there is no shooting, no damage being taken, nothing, it is like a flyby, which is not showing 100% gameplay.

jimmyjames123
06-28-04, 09:40 PM
[H]OCP reviews are just fine. They present the data that they see. Essentially all reviews are imperfect to a certain degree.

No reviewing style is perfect. [H]OCP's reviewing style has somes ups and downs. They want to show us what they think are the best settings to use so that the game is still playable on the card. The potential pitfalls of this method include:

1) limited amount of games tested

The [H]OCP graphs are very in-depth, and their relatively complex tests and graphical presentations probably mean that there is less time available to test a larger suite of games.

2) the complexity of the graphs are daunting to many new users

The graphs are very nice to see for those who know how to interpret them. However, the reality is that most people, especially newbies, have much more trouble interpreting and reading these more complex graphical displays. Also, the lines from one video card tends to smear or overlap data from another video card. Maybe a smoothing of these graphs would make things easier to follow for some.

3) there is an added element of subjectivity that is inherent in each review

The reviewer(s) need to make a subjective decision on what is playable, in a review that is supposed to be as objective as possible. In reality, there is no perfect agreement on what is "playable". [H]OCP also tends to choose higher resolution over higher AA/AF settings, something that not everyone tends to do, especially those who use displays with native resolutions.

If you see some other reviews, you will easily notice how reviewing style can differ. I like reading all of them to get a sense of the big picture. I tend to prefer reviews that show performance as resolution increases AND as AA/AF increases. These reviews cater to various types of individuals, depending on what resolution and AA/AF settings they tend to prefer in their games, and they make it easier for each reader to come to their own conclusion regarding performance at various settings. I am especially fond of the Tech Report reviews, as one can easily see how performance changes with resolution and AA/AF by looking at easy to read graphs.

Waffles
06-28-04, 09:41 PM
Man, I love how peaceful the NVNEWS forums can be. For the most part (and its natural given the name NVNEWS) people seem to be favorable towards NV cards. You try to say ANYTHING positive about nvidia in the [H] forums and you get skewered.

Brent
06-28-04, 09:44 PM
3) there is an added element of subjectivity that is inherent in each review

The reviewer(s) need to make a subjective decision on what is playable, in a review that is supposed to be as objective as possible.

Yes, it is subjective, we are not hidding that fact. We evaluate the gameplay experience, and that is very much subjective.

Waffles
06-28-04, 09:44 PM
I just can't believe that the X800 PRO is the same speed as the 6800 Ultra in those reviews.

CAn anyone look me in the eye and honestly tell me the 6800 Ultra is the same speed as an x800 pro?

MUYA
06-28-04, 09:44 PM
Can we please leave reference to other forums and websites out of this forum please. Can u discuss, CPU scaling of the new g en cards as opposed to other websites etc. THANK YOU very much.

Edit: The reason is, every website has a feedback section, we would like for u to exercise those if u have a problem. No need to come in here and "complain" about other websites pls.

jimmyjames123
06-28-04, 09:46 PM
Brent, I hope you soak in some of my sincere and constructive feedback. I respect the fact that you guys are trying something different, and I can see you really believe strongly in what you do.

Brent
06-28-04, 09:49 PM
Brent, I hope you soak in some of my sincere and constructive feedback. I respect the fact that you guys are trying something different, and I can see you really believe strongly in what you do.

I listen to and soak in all feedback :)

jimmyjames123
06-28-04, 09:55 PM
Reviewing video cards is becoming more and more complex with each new generation of cards. These graphics companies are giving more and more options all the time.

With NV for instance, the reviewer has the option to: enable tri optimizations, enable aniso optimizations; enable tri, disable aniso optimizations; disable tri, enable aniso optimizations; disable tri, disable aniso optimizations, etc.

With ATI, the reviewer can enable temporal AA, etc.

And then, within each game, there are various settings that the reviewer can toggle on or off that will effect performance.

When you factor in all the various resolutions, AA settings, and AF settings that each card can use, there are an amazingly large number of variables to consider.

Certainly not an easy task to comprehensively review a video card.

grimreefer
06-28-04, 10:04 PM
a little off topic:
r ati cards using Optimized LOD-Fraction?

BIGeMpTy
06-28-04, 10:09 PM
FRAPS gives you a realworld view of how performance is in a game since it captures the framerate every second (if you select that option) to gather data. How can it be anything but accurate of game performance?

FYI Recorded timedemos in FarCry do NOT playback the AI, there is no shooting, no damage being taken, nothing, it is like a flyby, which is not showing 100% gameplay.

Brent.

You guys explain your methodologies in painful detail and it all sounds just ducky. But then your results seem to directly contradict every other review I have seen out there. And not just on mickeymouse websites. Every other major hardware site that has reviewed these cards has shown the 6800GT kicking the crap out of a X800Pro...but then you guys come out saying that the Ultra is barely keeping up with the Pro.

I'm not saying your benchmarks should fall in line perfectly with everyone elses, but I am saying that if you're going to use a new methodology and get wildly different results from every other site out there, you should probably have a sentence or two stating why your results are right and everyone else is horribly wrong. Seriously, this latest review has me seriously questioning the validity of your reviews since no one can seem to back up the results you are getting.

On a lighter note, I think CPU scaling benchies are awesome. It seemed to me though that both cards scale about the same. Has anyone ever considered creating a database of all the different benchmark numbers obtained on all these different systems with all these different drivers?

-eMpTy

shinrai
06-28-04, 10:11 PM
I have no issue with the way Hardocp reviews. As far as the 6800U goes, i think they just need a refresh of their benchmarks but maybe they are waiting for worldwide available drivers and worldwide available retail cards. I personally am sick of reading reviews when so few cards are available. If Hardocp are waiting for a full release, then i can respect that.

And as far as these forums go, nvnews is far more nvidia fanatical than the guys are ATI fanatics at Hardocp. I've seen heeps of comments criticising 3Dc for unclear reasons but very few giving it to the severe and disgraceful lack of support for PS3.0 at the moment.

Trouble is, nvidia have marketed us to death. And i know because i want to buy a 6800GT.

Waffles
06-28-04, 10:14 PM
Muya is gonna get pissed but I gotta ask it. I apologize in advance :)

When using fraps how do you make sure you are doing the EXACT same thing? With AI getting so advanced it should react different each time you play through a situation. At least SOMEWHAT different. Doesn't this affect the outcome of the benchmarks? Even if just a little? (Granted not making the x800 pro as fast as the 6800 Ultra)

Brent
06-28-04, 10:15 PM
Brent.

You guys explain your methodologies in painful detail and it all sounds just ducky. But then your results seem to directly contradict every other review I have seen out there. And not just on mickeymouse websites. Every other major hardware site that has reviewed these cards has shown the 6800GT kicking the crap out of a X800Pro...but then you guys come out saying that the Ultra is barely keeping up with the Pro.

I'm not saying your benchmarks should fall in line perfectly with everyone elses, but I am saying that if you're going to use a new methodology and get wildly different results from every other site out there, you should probably have a sentence or two stating why your results are right and everyone else is horribly wrong. Seriously, this latest review has me seriously questioning the validity of your reviews since no one can seem to back up the results you are getting.

On a lighter note, I think CPU scaling benchies are awesome. It seemed to me though that both cards scale about the same. Has anyone ever considered creating a database of all the different benchmark numbers obtained on all these different systems with all these different drivers?

-eMpTy

I can't speak for other websites and their methods.

All I know is that our method accuretly reflects the gameplay performance these cards provide in the games tested. It is as real world as it gets.

If only I could have y'all right here beside me as I'm playing the games, you'd see it it is accurately depecting real world game performance.

Brent
06-28-04, 10:18 PM
Muya is gonna get pissed but I gotta ask it. I apologize in advance :)

When using fraps how do you make sure you are doing the EXACT same thing? With AI getting so advanced it should react different each time you play through a situation. At least SOMEWHAT different. Doesn't this affect the outcome of the benchmarks? Even if just a little? (Granted not making the x800 pro as fast as the 6800 Ultra)

You are correct. We run through at least 3 times on each card in the same place to make sure the performance is very close each run. The results graphed per second is real world game performance exhibited by each card.

BIGeMpTy
06-28-04, 10:23 PM
I can't speak for other websites and their methods.

All I know is that our method accuretly reflects the gameplay performance these cards provide in the games tested. It is as real world as it gets.

If only I could have y'all right here beside me as I'm playing the games, you'd see it it is accurately depecting real world game performance.

See this is my point. I don't think you should just throw the numbers out and when people ask why they look wierd just shrug your shoulders and pretend nothing is out of the ordinary. I'm sure you read reviews from other sites just like we do. So what do you think accounts for this disparity? Do you think Nvidia has optimized for timedemos? When you got your results did you get a little curious and try to duplicate the results from other sites and then try to find what was making the difference?

This disparity needs to be explained, otherwise your review is just tossed out as the lone irregularity in an otherwise normal grouping of reviews.

For the sake of your own reputation, I humbly recommend that you dig a little deeper.

-eMpTy

Waffles
06-28-04, 10:27 PM
So when you are playing the games, you just get the FEEL that the x800 Pro is doing the same thing as the 6800 Ultra?

Man...this is heavy...I had my $500 in-hand just ready to get the 6800u. Hrmph. THANKS a ton Brent :) dammit...