Sneak Peek at ATi's CrossFire Graphics System

Kez writes "While at Computex in Taipei HEXUS.net grabbed some benchmarks of an ATi CrossFire powered system. They have since had the chance to reconstruct a similar system and perform the same benchmarks with other cards and configurations to give us an idea of how CrossFire will perform. Obviously, CrossFire's performance will almost certainly change before release time, but in the very least the article provides an idea of what to expect. Interestingly, from these tests it looks like Nvidia's SLI may remain top-dog for graphics performance."

Comments

What about the price?

[Saiyine]

Will it cost more than the computer alone?

Re:What about the price?

[rpozz]

Given that a single Radeon X850 XT as used in the article retails at around $450, I'd say so. $900 would go quite a long way for the rest of the computer.

architecure

[imboboage0]

It would be interesting to compare diagrams of the architectures that SLI/Crossfire use to see why one would be better than the other.

Wait and see

[Radicode]

That motherboard they used for testing looks like a monster! 8 sata connectors... I don't want to think about the noise produced by 8 HDs spinning.

Anyway, as with any ATI products... it's better to wait for the final before declaring it a winner or a loser. I tested many beta revisions of their TV wonder USB2 and I saw the performances change with every release, sometimes good, sometimes bad.

-Radicode

Available in SGI Prism systems?

[CyricZ]

Are these cards compatible with SGI Prism systems? The current SGI Prism systems appear to include a ATI FireGL card.

http://www.sgi.com/products/visualization/prism/

Bias?

[LTC_Kilgore]

In the last test (3DMark05 - 1280x1024 4xAA 16xAF), they are running the Nvidia cards at 4x Anti-Aliasing, while the ATI cards are running at 6x.

Re:Bias?

[MightyPez]

At first glance that is always the conclusuon people draw. But Nvidia and ATI do different methods of anti-aliasing depending on the the level that is chosen. In Nvidia's case they use super-sampling, multi-sampling, and sometimes both. ATI uses multi-sampling only. The result is the same level of AA on each card will produce different visual results.

This article goes into depth about the FSAA issue between ATI and Nvidia. Look at page 12 and beyond for the full poop.

Re:Bias?

[LTC_Kilgore]

I don't think Nvidia cards CAN run that 6x mode, it's ATI only. Perhaps these were driver defaults? As in, the driver defaults one is supposed to have set for "official" 3dMark tests?

Then why not run the ATI cards at 4x like the Nvidia ones?

Re:Bias?

[LTC_Kilgore]

I believe that has changed with the modern line up of ATI cards. I could be wrong however, but I think that the differences in AA quality were more prominent with the older generation cards from both manufacturers.

Re:Bias?

[DeathByDuke]

Why not? Cos ATi cards are more efficient at FSAA than nvidias cards. If you see all the benchmarks, ATi gets less of a hit in 4x AA compared to nvidia. In fact in some cases, you get same hit with 6x on ATi as with 4x on nvidia... they're just balancing it out.

Re:Bias?

[LTC_Kilgore]

The point of a benchmark is not to 'balance things out'. The ATI card is pushing more pixels in 6x mode. Period. Whether it takes a bigger performance hit or not isn't the issue, it's doing calculations on more pixels, which will affect it's score in an adverse way.

Re:CrossFire

[DreadCthulhu]

Trademarks only generally only valid for one sort of product, and exist to prevent consumer confusion between the makers of similar products. If two companies/people make different products, that are not likely to be confused. they can have the same name for their trademarks - for example, Apple Records and Apple Computers. But if you tried to start a computer company called "Appletastic Computers" Apple Computer could sue you and probably win. Or for a real life example, think of the Lindows case, and how that could get confused the MS Windows OS.

3Dmark

[taskforce]

They tested it on 3Dmark... that's totally irrelevant to anyone looking to buy the card; Nvidia are notorious for optimising their drivers for synthetic benchmarks, meaning Nvidia cards almost always perform much better in tests like 3D mark, but when you get the cards into a game anything can happen.

Re:3Dmark

[softends]

Both sides are notorious for cheating, but 3DMark05 benchmark is generally a good indicator of gaming performance.

Re:3Dmark

[Anonymous Coward]

The 3DMark05 benchmark is generally a good indicator of 3DMark05 performance. Any similarity to real gaming is coincidental.

Re:3Dmark

[Guy LeDouche]

Why was this modded funny? It is completely true. 3DMark scores have almost no bearing on real-world game performance.

Re:3Dmark

[ionpro]

That's what people bitched out when 3dMark03 came out and gave advantages to ATI's cards. Guess what? Those 9700 Pros really did blow away the 5800 Ultras in next-gen games. 3DMark is much maligned, but I think they've done an excellent job of trying to predict the future 2 years in advance and figure out what games will be using then. Can it be perfect? No. Will the performance deltas differ between 3dmark and current-gen games? Absolutely. But when Unreal Engine 3 comes out, go back and look at those 3dmark05 scores and say, "Hmm. They were pretty much spot on."

Re:CrossFire

[bumptehjambox]

I thought it was a trademark of Milton Bradley.

ATi should borrow their briliiant advert's song:
Crossfire, you'll get caught up in the... Crossfire!

The R520 might make this comparison meaningless

[LTC_Kilgore]

If the predictions are accurate, these tests will be meaningless when the R520 based card from ATI is released. The comparison that matters in the uber-high end will be the 7800GTX in SLI vs. R520 in Crossfire.

Re:The R520 might make this comparison meaningless

[Thomas DM]

Yes, but it may still take two or more months before the R520 will be available.

And I guess that NVIDIA will start shipping a new faster G70 chip by the time that ATI will launch its R520.

Klunky AND slow?

[TheSkepticalOptimist]

When ATI first announced their CrossFire solution, hyping about the fact you could use older cards with newer generations for improved performance, I thought this was a great idea. Spending $500+ for a video card today, only to have it replaced a year or two later is kind of a waste, but if it still could be used to contribute to improved gaming performance, then I could see spending the money.

Then details about CrossFire came out. It requires using only CURRENT generation ATI cards, the X850 and X800, a very expensive CrossFire generation video card AND the fact you need a CrossFire compatible motherboard, of which, currently only ATI makes a chipset for. All this adds up to an expensive system, and not very practical.

If the benchmarks and real-time performance of a CrossFire platform shows significant gains in performance, then it may be worth it to get a system that meets ALL these conditions, but as of yet, nothing suggests that this kind of system offers anything better then what is available today.

With nVidia's SLI, sure you need 2 expensive and matching cards to work, but that is it, you don't need any specialized motherboards. I think this will be CrossFire's major downfall, the requirement for specialized hardware, especially if VIA decides not to make their own CrossFire compatible chipset.

Time will tell if CrossFire lives up to the hype, but I think that ANY dual card configuration is only a gimmick that won't last, like 3DFX original SLI hardware. It seems like next-generation video cards are already boasting the capability to out perform current generation dual card systems, with only ONE GPU. Wasting $1000+ to get a dual system today to find out a $500 video card 6 months from now outperforms it would be quite dissapointing.

Do my eyes deceive me?

[Travoltus]

It looks like those video cards overlap not just one, but TWO empty expansion slots that I could use for other cards!

http://img.hexus.net/v2/features/dfi_crossfire_com putex/images/crossfire_big.jpg

This is why I have avoided upgrading to these new generation of cards... I have the lowly 6600 now and that's going to be it, perhaps. I don't like onboard sound (I prefer my Audigy 2, especially for Linux), thank God for the on board USB, FireWire and NIC though; I have a video capture card and a SCSI card for legacy stuff, and there'd be no room for these two cards in any PCI-E system I'd upgrade to... they all come with fewer slots now.

Re:The important thing to remember is that...

[william_w_bush]

not true, nvidia's solution has them creating profiles for popular games for them to function, though i believe you can sli awareness to a game to increase it's support.

my current release has at least a hundred games, and there aren't that many popular games that need this kind of graphics firepower out there.

Re:But will it still be ATI crap.

[Jthon]

I've seen this problem before and it turned out a bad stick of ram was causing the random videocard lockups. Took me forever and a couple different cards to figure this out. Try running a memory test or pulling out and swapping DIMs to see if this improves the problem.

Bad RAM seems to cause lots of computer problems in various other components.