Brent's latest review........ the way it should be done?

[madshi]

1) Find the "playable sweet spot" for each card. (Highest image quality where you deem it "playable.")
2) Post benchmarks for BOTH cards at the settings for its own sweet spot AND the other cards sweet spot. Do not mix benchmark results from two different settings in one graph. Too confusing.

This might be interesting for technical interested people like us. But for pure gamers I think the current way how Brent does the reviews is just fine. There's only one graph per game, which keeps the review fairly short, although much information is there and lots of games are tested.

Well, that's my opinion only, of course.

3) Post image quality comparisons / shots for each card at its sweet spot.

Agreed!

Especially if you consider things like that NVidia's 4xAA doesn't look much better than 2xAA. This is the one big thing which I'm missing in Brent's review. In that one graph the 5900XT looks much superior with 4xAA compared to the 9600XT's 2xAA. But in screenshots the difference will probably barely noticable... And when both cards do 4xAA, the 9600XT should be quite superior in IQ.

Would I would greatly welcome is if Brent would do a 5 second video of each card in each game at their sweet spots. But I guess that won't come anytime soon...

 

[Nazgul]

Hey Joe, that's plagiarism
I will grant you this though, you said it better.

and it's an excellent idea. the "experience" is important, but benchmark scores at equal settings would still be a useful tool. Having each card tested at its opponent's sweet spot seems like a good way to do it without having to run a whole slew of tests.

 

[Gambler FEX online]

Hi!

are these tests testing any geometry performance or vertex shader performance? Most is AA and AF on/off and no low resolution tests.

i ask because i want more polygons in games, and polygonised blood and holes in bodies and better dismembered bodies etc.

UT 2004 is game with most polygons lately but it is slow even at lowest detail.

nvidia have always showed us high poly tech demos.

 

[Joe DeFuria]

1) Find the "playable sweet spot" for each card. (Highest image quality where you deem it "playable.")
2) Post benchmarks for BOTH cards at the settings for its own sweet spot AND the other cards sweet spot. Do not mix benchmark results from two different settings in one graph. Too confusing.

This might be interesting for technical interested people like us. But for pure gamers I think the current way how Brent does the reviews is just fine. There's only one graph per game, which keeps the review fairly short, although much information is there and lots of games are tested.

The only problem is you don't really get a good idea of what the graphs are trying to show. Hell, I was skimming through the review, and I was reading blurbs about how one card was faster than the other, but the "graphs showed the opposite."

IMO, it's not too much to ask. We're (tehcnial people or pure gamers) used to get all kinds of graphs at several settings anyway. It's just that now, the "settings" that are tested are determined by "playability", instead of just doing "one at 640x480, one with AA, one with Aniso...." etc.

Would I would greatly welcome is if Brent would do a 5 second video of each card in each game at their sweet spots. But I guess that won't come anytime soon...

That would be cool, but I think that's not practical...mostly because compressing video (to make size practical) would probably destroy any chance at having a fair image quality comparison.

 

[mczak]

I still don't get it why in the blue moon does Brent insits Prince of Persia is a DX9 title. Again, AFAIK it's using DX8.1 vertex shader effects, no more no less.

Well, at least Brent has now corrected Max Payne 2 to be DX8.1. Though I'm sure he now gets feedback "but it needs DX9 installed, it's a DX9 game!"

 

[cthellis42]

Also, some readers who just skim the article and look at the numbers may miss that some of the benches were run at different settings. This could lead to erroneous conclusions like -- the 5700U was faster than the 9600XT in many of the games when this was not really the case.

Anyone who looks only at charts and reads absolutely NO text, nor processes any of the information presented (even simply looking at resolution and quality), has no hope of ever understanding anyway, IMHO. <shrugs> Not useful to worry about types like that.

On the whole, I just wouldn't have minded the numbers matching for direct comparison's sake. I can certainly make value judgements without them, however.

 

[dan2097]

I really like Brents review including that one. I cant remember whether he deos at the moment, but a link to comparisons of the different cards af/fsaa settings probably in an older hardocp article could be useful as a refresher as to what the settings mean when comparing.

I personally would also prefer if maybe one game which traditionally runs better on Nvidia hardware is included i.e. NWN, X2-the threat, KOTOR. JK2 academy and RTCW ET are also possible but I dont think they are taxing enough.

I mean you might say they have Nvidia specific coding (bar X2-the threat) but it is something that should be taken into account, that many older games are Nvidia optimized.

 

[Pete]

It was a solid review, but I would have liked to have seen IQ comparison screenshots for each game, and more discussion as to how the 5900XT's AA compared to the 9600XT's (not in screenshots, but in-game, as that's the theme of the review). Some the graphs didn't seem to correspond to the scores, too. Specifically, the Halo graph didn't correlate to the benchmark table at all. The 5700U looks at least 10fps faster than the 9600XT throughout, yet is only listed as 4fps faster.

And I'm still not a fan of starting the Y-axis at anything but 0.

Brent seems to be the only reviewer that consistently rates the 9600XT above the 5700U. Most reviews show the 5700U benching faster than the 960XT, particularly with AA, so I'm somewhat confused why it does so poorly with Brent? Is it an issue of game settings? Of minimum fps? Of everyone being basically benchmark bots, and Brent actually FRAPSing his way through these games?

 

[StealthHawk]

I mean you might say they have Nvidia specific coding (bar X2-the threat) but it is something that should be taken into account, that many older games are Nvidia optimized.

People generally don't buy new cards to play old games- or at least, not for performance reasons. Because old games typically play well on all new hardware. Who really cares if card A is faster than card B in three year old game X when both are more than adequate.

If you're going to do something to test old games, I'd rather reviews mention that NVIDIA cards have hybrid SS/MS FSAA modes that can give higher IQ in said old games compared to ATI cards, or something of that nature. IQ is what's the primary concern when playing old games anyway.

 

[dan2097]

Brent seems to be the only reviewer that consistently rates the 9600XT above the 5700U. Most reviews show the 5700U benching faster than the 960XT, particularly with AA, so I'm somewhat confused why it does so poorly with Brent? Is it an issue of game settings? Of minimum fps? Of everyone being basically benchmark bots, and Brent actually FRAPSing his way through these games?

I think it comes down to:
(positives)
a) not using cheated synthetics
b) not using old quake 3 based games where the frame rates tend to be high anyway
c) not using manufacture time demos or non gameplay i.e. flybys
d) takes account of image quality differences when cards appear to be at comparable settings

(negatives IMO)
e) not using any games that have pro Nvidia tendancies i.e gun metal, X2-the threat, jedi academy, KOTOR, NWN, RTCW

 

[dan2097]

I mean you might say they have Nvidia specific coding (bar X2-the threat) but it is something that should be taken into account, that many older games are Nvidia optimized.

People generally don't buy new cards to play old games- or at least, not for performance reasons. Because old games typically play well on all new hardware. Who really cares if card A is faster than card B in three year old game X when both are more than adequate.

If you're going to do something to test old games, I'd rather reviews mention that NVIDIA cards have hybrid SS/MS FSAA modes that can give higher IQ in said old games compared to ATI cards, or something of that nature. IQ is what's the primary concern when playing old games anyway.

X-2 the threat and gunmetal (I dont support benching this game) are not old games and performance does matter in them. Same with KOTOR.

EDIT: well maybe old games should really read, opengl based (bar COD)

I think Nvidia's hybrid modes should be mentioned although I personally find ATI's Fsaa worked on the older games I tried bar UT (D3d) although thats ATI specific, I probably just dont have enough old games.
Anwyay yes anything like that should be mentioned if it could be important in a purchasing decision

 

[martrox]

d) takes account of image quality differences when cards appear to be at comparable settings

Not really...... he seems to test at what gives the best overall I.Q that is still speed usable. While many here seem to also want direct comparisons (which I understand), those can be had at almost all test sites, both speed at differant resolutions & AF/FSAA. Brent's differance is he gives you an exact idea of just what is playable with each card. And while some have stated that Brent is the only reviewer that shows the 9800XT being superior to the 5700Ultra, remember that only Brent seems to take the route of game playability. That's why his reviews give the end user the kind of information that is really usable. Far more real life information than any other reviewer.

 

[digitalwanderer]

I still think Brent's reviews are a bit lacking, he really needs to include some photos of himself posing topless with the cards too. 8)

 

[Rugor]

I liked the review too.

I figure the B3D reviews are great for showing the technical aspects of a card but Brent has no peer for the comparative "what it's like to play a real game" reviews.

However, by avoiding the NV-optimised games it does two things. First it shows the real performance of NV3x cards on a level playing field, which is good. Second, which is in some ways less good, it doesn't show that there are games their hardware is very strong for.

Still for overall reviews they're very good.

 

[dan2097]

d) takes account of image quality differences when cards appear to be at comparable settings

Not really...... he seems to test at what gives the best overall I.Q that is still speed usable. While many here seem to also want direct comparisons (which I understand), those can be had at almost all test sites, both speed at differant resolutions & AF/FSAA. Brent's differance is he gives you an exact idea of just what is playable with each card. And while some have stated that Brent is the only reviewer that shows the 9800XT being superior to the 5700Ultra, remember that only Brent seems to take the route of game playability. That's why his reviews give the end user the kind of information that is really usable. Far more real life information than any other reviewer.

Well I mean on say flight simulator 2004 most/all sites bar hardocp havnt commented that Nvidias af is producing very poor quality (texture stage optimizations in control panel af Id expect)

 

[dan2097]

I liked the review too.

I figure the B3D reviews are great for showing the technical aspects of a card but Brent has no peer for the comparative "what it's like to play a real game" reviews.

Nordichardware also does good reviews like that. They bench a wide variety of games Last time I read a review there they had counterstrike in the lineup, all the cards getting 160fps and the gaming comment being it doesnt matter which one you use for this game they're all good or something to that effect

 

[crusher_pt]

"Brent seems to be the only reviewer that consistently rates the 9600XT above the 5700U. Most reviews show the 5700U benching faster than the 960XT, particularly with AA, so I'm somewhat confused why it does so poorly with Brent? Is it an issue of game settings? Of minimum fps? Of everyone being basically benchmark bots, and Brent actually FRAPSing his way through these games?"

http://www.oc-zone.com/modules.php?name=Reviews&rop=conteudo&id=61

Use a translator... sorry.

 

[StealthHawk]

X-2 the threat and gunmetal (I dont support benching this game) are not old games and performance does matter in them. Same with KOTOR.

EDIT: well maybe old games should really read, opengl based (bar COD)

I was commenting on your sentiments that some "NVIDIA biased" games should be tested because a lot of old games are also NVIDIA biased. This doesn't make much sense to me for the reasons I outlined. However, I do agree that it may be good to benchmark games like KOTOR and X-2.

I think Nvidia's hybrid modes should be mentioned although I personally find ATI's Fsaa worked on the older games I tried bar UT (D3d) although thats ATI specific, I probably just dont have enough old games.
Anwyay yes anything like that should be mentioned if it could be important in a purchasing decision

Old games tend to use a lot of alpha tests for which multisample AA does nothing. Much like some members of this forum have expressed their distaste for games that have gimmicky features because one small part of the game looks good while the rest looks like crap, there's just something about seeing alphas with aliasing when the rest of the scene has AA that bugs me. I want consistency whenever possible.

 

[DemoCoder]

Older games used alot more alphatextures to simulate holes in geometry (fences, stairs, trees, etc) for which using actual vertices would be too expensive.

 

[StealthHawk]

Older games used alot more alphatextures to simulate holes in geometry (fences, stairs, trees, etc) for which using actual vertices would be too expensive.

Yes, and if you have an ATI card you're SOL since they won't implement Supersampling for Windows drivers.