DH Review: XFX Geforce 8800 GT 512MB XXX / Sapphire HD 3870 512MB / Asus EAH 3870 512

ET3D

Interesting review. I understand that the different resolutions were forced by the games, but what about the AA level? It seemed to vary a lot between games, and I don't know why.

In terms of benchmarks, I think I've seen more useful reviews. I liked the rest of the review, though.

In particular, I really appreciated the discussion of support. I'd like to see more of this in the future, or a special review of support provided by various companies. I guess you compared to BFG because that's an experience you had, but it still felt strange considering that the review was for XFX, Asus and Sapphire cards. Or maybe that was the 2900XT card? It wasn't listed on the Test System page, for some reason.

The Crossfire 2900XT + 3870 test was very cool. Another part of what made the article worth reading.

All in all, though I've read quite a few reviews of the new ATI cards, there were still interesting things to learn from this review, so well done.

Veridian3

The AA levels varied basically to cover as many settings as possible. We try to do the same with resolutions... some 1280x720 and a few using 12x8 etc ... this is something we try to do in all reviews as not everyone uses the same resolutions.

Quite interested in your comment about benchmarks not being useful, can you give me an example of what you would have liked to see instead?

BFG were chosen because they are a direct competitior to XFX, the reason we concentrated on XFX on the support page was that they launched this new support technique so we felt it fair to compare to a direct competitior. If you are interested I finally got a response to my support ticket on Thursday...took 4 days.

[hobo]eclipse

Veridian3, any comment on the IQ differences, im gettn jumpy for an 8800gt but the screen caps from crysis that i linked to in post 13 show quite a difference, in my opinion, between the mountains on the right hand side, everything exceptt he mountain looks pretty much identical to me. can you comment on this?

ET3D

The reason I felt that they were not useful is partly because of the variations. When each game is tested at just one resolution and AA level, it's hard (or impossible) to get a good idea of how the cards perform. It's often clear, seeing benchmarks on sites that do test multiple resolutions for each game, that some cards are better than others with AA on, or at certain resolutions.

People who probably read only the review at DH will find any benchmarks necessary, but anyone who's read other reviews (and I'm sure most hardware enthusiasts have) will already know the performance figures. Which is why I thought that the parts of the reviews which aren't standard were the most interesting and/or useful to know.

Skull

It an interesting point however I think the days of testing a card at 14 resolutions and eye candy combinations is a waste of time. Its only useful for a few people, the majority of people who buy a midrange card like these aren't partnering it with a 30inch screen. I think the metholodogy of testing the game for the public with eye candy at the resolution end users will be running it at is the most useful way. I have read reviews of 8800 ultras and there are benchmarks in the mix at 1024x768 which is a waste of everyone's time.

Just my opinion.

ET3D

There's truth to this, but there's at least a need, IMO, to choose a line and walk it. If reviewing for normal end users, use a resolution of 1280x1024, for example, as it's a very common monitor resolution these days. As you say, few people buying a mid-range card will run at 1920x1200, for example, and few if any will run their mid range card at 8x AA.

That said, I read reviews not only to be able to choose a card better, but also to learn more about them. I'd never have a 3870+2900XT combination and most likely won't run anything at 8x AA, but I'd be interested in these tests. It's just that having a single 8x AA test doesn't tell me much.

As for testing multiple games at multiple resolutions, it's still done and hopefully will continue to be done. Three resolutions at two different AA/AF combinations (none, 4x/16x) can be pretty informative. I like Digit-Life's 3Digests for this reason.

Skull

1280x1024 however is a 4:3 resolution and I think that the majority of tech people who come to this forum are using a widescreen monitor?

Again I see your points and im certainly not dismissing them, the 3870+2900XT was included for a specific reason after speaking with Veridian3, that being a scenario of a user already having a 2900XT from the last hardware refresh and a crossfire capable motherboard. It seemed a nice novelty factor to include it. We already had numerous emails about that thanking us for the testing as no one else covered it. (I think there was even a guy in this thread who mentioned it).

Sure, its a point you make no one could invalidate as it is something you want in a review. I know DH reviews used to have a plethora of resolutions and settings and it was a informed decision to narrow these down. Obviously its something I can't comment on but ill pass it onto Z. Thanks for the constructive feedback ET3D.

ChaosMinionX

Yep, I will agree 100% that seeing the cross series crossfire between the 2900XT and the 3870 was quite nice to see...Might be a redeeming quality for those that spend alot on the 1GB 2900XT and want to get a little extra life outta it, maybe toss in a 3870/3850.

[hobo]eclipse

so am i the only one that sees such a drastic difference on the mountain on the right in the 2 crysis IQ comparison shots from the article i linked above?

Asmoday

Nope I see the difference too...the ATI card seem to render a more detailed image, there is also small differences in the water and vegetation although not as drastic as on the mountain.

[hobo]eclipse

cool gald im not nutz, cant believe this isnt more of an issue to others, maybe its the fact the game is newer and both drivers and game needs to mature a bit. any others care to chime in with thoughts about why and how?

tc17

Either something is wrong, or ATI improved their drivers since that review. For example... my computer runs off a AMD 5000+ x2 black edition (OC: 3.0ghz), and an ATI 3870 card. (I paid $100 for the cpu on Amazon, vs. the $1000 cpu in the review)

Using the same settings for World Of Conflict, I get the exact same fps, maybe a little higher as you did for the 3870.

In Crysis, I'm not sure what you used to determine your fps, but I used the external benchmark program. I get the same 15fps as you, and if I raise my resolution to 1680x1050, I get even more fps at 20fps using the same settings, which is odd in itself.

I'm using the Catalyst 7.12 drivers under Vista x64.