After a slightly longer than planned break, we jump straight back into the heat of things and see whether or not 4850 twosomes are recommendable.  As has become customary, we have some slightly boring stuff that no one reads placed at the beginning of the article, which gives an idea of what we've done.

• We've moved from having a large heterogenous mix of in-game settings to testing according to 3 exact presets:

  • BASELINE: this is the game set to its maximum quality settings, but without any AA or AF enabled

  • HIGH QUALITY: same as above, but 4X AA gets enabled alongside 16X AF

  • EXTREME QUALITY: game settings remain at their maximum respective value, but AA gets bumped to 8X, whilst AF remains pegged at 16X

  • CROSSFIRE EXTREME QUALITY: still maximum in-game settings and 16X AF, but we're using the Crossfire exclusive 16X AA mode- this mode can only be forced through the CCC and thus will only work in games that support such forcing; we're investigating it since it probably is of interest to quite a few people out there

  • For games that have no support for AA- Stalker, Timeshift, Bioshock DX10, Gothic 3, the baseline setting has 16X AF enabled

• We're using the presets in an effort to improve overall readability and usability of the data we provide.

• Most in-built benchmarking utilities have been relinquished in favor of FRAPS runs - while we most certainly don't consider this to be the be-all end-all of testing, it is probably a better way of showing how the games tested will actually perform in practice

• We've also moved from averaging three 3-minute runs with FRAPS to averaging six of them.  This should help remove some of the inherent variability associated with FRAPS testing.

• All tests are run at 1920x1200 unless otherwise specified.

• For each game, a graph will be presented to you showing the percentile increase of card A vs card B- the percentile increase is calculated as (A-B)/B  and expressed in percentage points

Since we're looking at Crossfire (CF), we'll have to talk about the most recently discovered doomsday device, one which sends shudders down multi-GPU enthusiasts' spines - micro-stuttering. A fairly extensive bit of the 4850 closing thoughts (yep, those are almost here) will be dedicated to figuring out what this horror story is all about and whether or not the world of multi-GPU users is ending.

We won't be recycling general comments about each game (it's a poor way of giving the illusion of extra content), but rather go ahead and show results directly, with what we hope will be relevant comments made where necessary.

Here are the specs for the system we used during testing:

As you can see we're using a slightly overclocked 3870X2 card from Gecube for illustrating CF performances of the old lineup, and we've also achieved driver parity for this round of tests, as opposed to the single card ones where we had to use the slightly older Catalyst 8.5s for the 3870.

Version used:  1.2 X64

Testing method:  FRAPS run through the first part of the Assault level, the results are the average of 6 three minute runs

Crysis Settings (click for full image)

We're controlling anisotropic filtering in-game with the r_TexMaxAnisotropy console command.  Be advised that surfaces to which POM (Parralax Occlusion Mapping) is applied won't receive any AF, and since those form the majority, AF won't be very noticeable in-game. AA is controlled through the game's menu.

If one would see only the average framerate, one might think that the game was actually nearly playable without AA at these elevated settings.  However the minimum framerate achieved takes care of that misconception.  $200 further along and we're still not in Crysis Nirvana...and looking around it seems that we could as well be $1800 above the single 4850 and still not achieve aforementioned enlightenment.  It's the "Crysis-complex" that all GPUs and combinations of GPUs seem to have.

What is interesting to notice here is that the 4850 crossfire configuration gets considerably better scaling than the X2 does compared to the single 3870. Both cards are running out of VRAM, especially once AA gets enabled, but the 4850 seems to handle it better.  It also is seemingly immune to a certain level of FPS jumpiness that was apparent with the X2.  X2 quirk or improvements to CF in the new lineup?  We'll have to draw that conclusion once we've been through all of our gaming tests.

Be advised that the FRAPS run we're using is slightly more demanding than the demo we were using before, so results aren't directly comparable with the ones we got on other occasions in Crysis (not to mention the fact that we have to correct those results anyhow since an error made its way amongst them).

Version used: 1.1.1.0

Testing method: FRAPS run through first part of the second sub-level of the first episode, the results are the average of 6 three minute runs

Game Settings

Game Settings Cont.

AA and AF levels are controlled through the game's menu.

Both cards scale well in COJ, but just like in the single-card bout, something within the game is preventing the 4850s from gaining a decisive lead.

Version used: 2.300.0.24

Testing method: FRAPS run through the entire Wolfheze mission (the first mission you play with the Germans), the results are the average of 3 playthroughs

Game Settings

AA was controlled through the game's menu.  AF didn't to impact either visual quality or performance in our experience so we've considered it non-working for CoH. Graph bath:

Lovely- 8X AA becomes a realistic option for CoH when using the 4850 CF system. The dominance that the single 4850 experienced over the 3870 is repeated here. It seems that most of the improvements implemented in the RV770 chip are taken advantage of in CoH.

It is worth mentioning (we'll get back to this in the closing section of the CF series) that load times seemed to be reduced on the 4850 CF rig when compared to the X2.  This has been an area of complaint for CF users, and it seems that some work went into alleviating the issue.

Version used: 1.0.0.8

Testing method: In-built performance test, average of 6 runs

Game Settings

Game Setings Cont.

We're controlling AA and AF through the game's menu.  Since WiC doesn't support more than 4X AA natively and AA forcing through the CCC doesn't work there are no Extreme Quality numbers to be seen here:

Nothing to note here. Same story as usual, when you add the eye-candy the RV770 starts to distance itself with ease. The numbers we're seeing from the 4850 CF configuration are near to the ones a 3870 Quad configuration produced in our older tests. Not too shabby.

Version used: 1.18704.70.4256 X64 (Latest single-player patch)

Testing method: FRAPS run through the "Tottenham Court Road" level, results are the average of 6 three minute runs

Game Settings

We're controlling AA and AF through the game's menu. In demon dominated London, Medium=4X AA.  That is as high as one can go with an ATi card, with 8X AA being a seemingly divine artefact that would hamper their domination and thus must not be included as an option - which means no Extreme Quality numbers (again):

Once more, the 3870X2 is smacked around quite badly...which is impressive given the fact that this card was introduced just a few months ago at  a price point slightly higher than what you'd pay for a pair of 4850s today.

This is another title where CF seemed to be more well-behaved in its 4850 incarnation, with slightly shorter loading times and less jumpy framerates.  This  last aspect can be attributed to the fact that the RV770 is considerably more powerful than the RV670, and thus handles worst-case scenarios better). Again we see better scaling from the 4850s, so perhaps some CF magic was implemented in the new cards.

It should be mentioned that the drivers we used introduced a bug related to the DX10 Dynamic Smoke option that only manifested itself on the 4850 CF configuration. We're showing you a picture of it below, but the driver team is on top of things and it's probable the final 8.7 Catalysts won't suffer this issue:

Version used: 1.1

Testing method: FRAPS run through the "Welcome to Rapture" level, results are the average of 6 three minute runs

Game Settings

Bioshock is, as we've shown before, not quite a challenge for modern GPUs. The drivers we're using include a tweaked CF profile for it though which improved performance when compared to older ones;  we're skimming the edges of being CPU limited here in spite of the rather generous overclocking of our processor.

Version used: 1.4

Testing method: FRAPS run through the first level of the game, results are the average of 6 three minute runs

Game Settings

and more...

and more...

AA and AF are controlled through the game's menu. Now, does the prettiness run fast?

Still not the level of performance we'd like to see but that can be in part attributed to the game itself.  We're looking at exchanging Lost Planet with Devil May Cry 4 which is based on a newer version of the same engine and seems to perform considerably better;  that's something we'll explore in the future. The take-home note here is that the game is playable (with indulgence and taking into account its profile) with 8X AA and 16X AF on the 4850 CF configuration which is something that can't be said about the 3870X2

Version used: 1.0.3340.131

Testing method: FRAPS run through the "Impasse" level, results are the average of 6 three minute runs

Game Settings

AA only goes up to 4 samples and in spite of our repeated attempts being cheeky and working around that by adjusting the configuration file, Gears seemed intent on keeping itself within its "factory" set parameters. This also means that 16X AF didn't want to be disabled-even when changing the .ini file and setting its properties to read-only, so AF is enabled even in the Baseline profile:

Gears is capped at 64 FPS so that makes it less than ideal for evaluating the more powerful GPU configurations available - at least if you only care about absolute number comparisons.  It is this very limitation that makes the game interesting because it makes one look in other places for notable differences.  In this case we find them in the minimum framerates area. The 4850 CF system has a 13 FPS advantage here. Correlating that with the average framerates and the aforementioned framerate cap it's fairly easy to conclude that the CF 4850s provide a better game play experience.

The drivers we're using seem to have taken care of the bug that caused hitches in 2 precise points of our run-through, which in previous tests generated artificially low minimum framerates.

Version used: 1.2

Testing method: FRAPS run through the "Masyaf" level, results are the average of 6 three minute runs

Game Settings

AA only goes up to 4 samples and we're controlling both it and AF through the game's configuration file.

Assassin's Creed doesn't show us much, just that we're running into CPU limitations with both cards and that we need to setup another test run since this one is no longer adequate, which has the unequivocal effect of filling us with joy. Due to this limitation, we can't judge scaling all that well in this particular title.

As stated in the introduction, with this article we are nearing the end of our in-depth exploration of the HD 4850 cards.  Stay tuned for our final tests and complete conclusions covering these cards as we look towards the higher-end HD4000 series cards and additonal articles on Image Quality and other interesting aspects we've discovered in our ongoing testing.

Ready to enjoy the RV770 experience yourself? Don't miss on your chance to win a Radeon HD4850 Crossfire combo, sponsored by ATi!