|
ATI Crossfire Xpress 3200 (RD580) Technology Preview
Inside RD580
"True" Dual x16 PCIe Crossfire Xpress 3200 is known by many enthusiasts as RD580, but the engineers who poured their heart and soul into it endearingly named it Skeletor because, as Joe Macri said, "who's bigger and badder than Skeletor?" As I mentioned before, RD580 was designed from scratch, and it was about 8 months before it went into production. Just 24 hours after receiving the first silicon, ATI was able to get Crossfire up and running stable. After the now infamous bug with R580, it would look as though ATI's luck is about to change. The chip is manufactured on TSMC's 0.11um process, and it sports the smallest northbridge die on the market at only 39mm2, containing 22 million transistors and with a Total Designed Power (TDP) of 8W. The chip will be passively cooled by a fanless heatsink, which means less noise and more stability, as it is common for heatsink fans to die. As you will see, ATI has made great claims about real performance benefits which can be seen thanks to their "true dual x16 PCIe." Before I go into detail about ATI's solution, let's examine NVIDIA's method of passing data from card-to-card in an AMD-based nForce4 dual x16 SLI setup. For this example, we'll use ASUS' A8N32-SLI motherboard, which uses a C51 Northbridge chip for the North Bridge and an nForce4 SLI chip for the South Bridge. Each of these two chips drives a single x16 slot. What's interesting, however, is the fact that there is a bandwidth bottleneck between these two chips, where the data is NOT carried on the PCIe bus. Although both cards communicate with a chip via the x16 PCIe bus, the communication between the two chips, which is vital for SLI, negates the benefit of having dual x16 in the first place, or so says ATI at least.  ATI's Crossfire Xpress 3200 chipset communicates with both video cards, each across x16 PCIe, on what ATI has labeled the "Xpress Route." The main difference is that communication is more efficient and each card will be communicating with the other through one central chip - no bottlenecks. ATI claims up to 50% performance gains with mid-range GPUs and up to 100% performance gains with low-end GPUs, in comparison with NVIDIA's solution. This is what got me to thinking and set off some bells in my head.   Sure, the NVIDIA dual x16 solution suffers from a bottleneck, but that is assuming that the cards are communicating directly between the two chips. With a high-end SLI setup NVIDIA says that you must use the SLI bridge for optimal performance, this is precisely because of the bottleneck pointed out by ATI. With the SLI bridge connection, you eradicate the bottleneck, and therefore you do get dual x16 PCIe. Of course, if you are using some mid-range or low-end parts which don't require a bridge connection, then you will suffer from the bottleneck ATI has pointed out. I don't really doubt ATI's claims of boosted performance in these areas, but I do not think there will be any difference in speed attributable to ATI's "true" dual x16 implementation in the high-end segment.    Overclocking One of the things ATI really tried to get across to us is that RD580 is a superb overclocker. All the logic is massively overspecced, so you should be able to achieve incredible overclocking results on stock cooling and without upping voltage. More specifically, the HT link interface has been optimized so that it can run at over 1.5GHz (50%+ overclock) and the PCIe interface can run up to 3.5GHz (40% overclock). Another plus is that there is no internal division when raising the HTT reference clock, translating to a 1:1 internal clock increase. South Bridge The information I was given about RD580 shows that motherboard manufacturers can choose to use either the ULi 1575 or ATI SB450 South Bridge. I'm not positive on this, but based on the fact that the ASUS A8R32-MVP specifications say nothing about SATA II, I believe it will use ATI's SB450 South Bridge. The SB450 has received harsh criticism for not only a lack of SATA II (and NCQ) support, but also for poor USB performance. I don't know at this time what the other manufacturers will decide to use for the South Bridge on their boards.
Related Articles Editorials August 27, 2009
Editorials August 18, 2009
Editorials March 3, 2009
Editorials February 12, 2009
Add Comment
2 User Comments 1 - Posted by
Kurtis
on March 8, 2006 - 6:39 am I just noticed this article over at PCPerspective. Ryan investigates ATI's claims about bandwidth bottlenecks on NVIDIA's SLIx16 boards and comes to the conclusion that these claims were not based in fact. Have a look: http://www.pcper.com/article.php?aid=212&type=expe... 2 - Posted by
Frosty
on March 8, 2006 - 5:24 pm i knew it those bastards lied!
Add Comment To add a comment without being a member, you may omit the password field, but you must enter your name (or nickname) along with your comment. * Denotes required fields. |
Recent Reviews Video Cards September 23, 2009
Gaming Devices July 17, 2009
Portable Media Players July 16, 2009
Thumb Drives July 14, 2009
Recent News Tom's Hardware Nov. 6, 2009 - 4:45 pm
Boing Boing Nov. 3, 2009 - 8:22 pm
Destructoid Nov. 3, 2009 - 7:14 pm
Fidgit Nov. 3, 2009 - 6:57 pm
Gizmodo Nov. 2, 2009 - 11:58 pm
Rock Paper Shotgun Oct. 30, 2009 - 7:12 pm
Fidgit Oct. 27, 2009 - 11:10 pm
|
|||||||||||||||
