|
||
|
|
|
              |
|
 |
 |
||||
 |
|
|||
By William Van Winkle |
||||
And I thought the seagate/Maxtor deal was big news. The AMD/ATI merger announced late in July clearly takes the headline cake for the year so far. The interesting thing is that few questioned whether the hard drive vendor melding made sense. Seagate was the stronger business with better drives, Maxtor had the superior creativity and retail presence. It all worked. But with AMD and ATI...man, it's anybody's guess. The growing consensus is that this merger makes sense. Intel has long proven that the way to control the industry is through the chipset. For whatever reasons, AMD abandoned doing chipsets many years ago, and ATI, after three or four attempts, at last has a strong, AMD-compatible chipset portfolio. The Xpress 3200 overclocks like mad, and ATI finally has a southbridge that works. I have always and continue to be befuddled by the relative lack of creativity ATI shows in its core logic versus NVIDIA. The nForce line, by any standards, is on a level far above anyone else, and just because ATI now overclocks like NVIDIA doesn't mean that the Xpress line can hold a candle to nForce for overall value in the mainstream and business segments. No doubt, this will change as we head into 2007. The opportunities AMD and ATI will have when developing next-gen platforms in the same lab should prove substantial. ATI core logic stability issues should finally be a thing of the past (on the AMD side at least), and it's a fair bet that AMD will benefit by being able to promote graphics as part of its corporate stable image platform rather than having to push people uphill to NVIDIA for graphics assurance. Expanding from CPUs to CPUs with chipsets as one cohesive package has nothing but upside for AMD. "This merger will provide system builders with another level of choice," says AMD's Patrick Moorhead, vice president of global channel marketing. "Their options on integrated solutions—chipsets with graphics plus CPUs—has been one, and that's been Intel. Now, there's a second choice with AMD and ATI. But the big differentiator here is this: We will also be providing an open ecosystem to all partners. Our competitor sets up pretty high hurtles for royalty fees and royalty structures to operate in their ecosystem. So if you're an NVIDIA or an ATI today, you're paying these massive royalty fees to Intel to be able to participate. We're not going to do that. That's not what brought us to the party and made us successful." OK, two things here. One, this stance is a negation of the dogma AMD has preached for years about not tying anyone down to anything. You've never heard AMD complain about Intel bundling its CPUs and IGPs together because AMD, just like Intel, had the ability to sit alongside IGPs from ATI, NVIDIA, VIA, and SiS. There is no "integration" between the chipset and CPU. If you want ATI integrated graphics with AMD, you can go out today and plant an Athlon 64/FX on a board using the Radeon Xpress series chipset. The difference is that the two are tied together by market choice, not through any sort of manufacturer-level bonding. Picking AMD was all about the freedom to select whatever components you like. As a case in point, here is a statement made to me in March by AMD product manager Matt Mantini when we were discussing whitebooks and the Turion line: "We bring to the table the ability for manufacturers and system builders to leverage best-of-breed technologies from the likes of ATI, NVIDIA, SiS, and then a host of wireless solutions from Airgo, Broadcom, etc. We give the channel the flexibility to pick and choose the types of solutions that meet the types of customers they're going after. With a Centrino platform, it's the same in many cases regardless of which manufacturer you purchase. I think our flexibility brings a lot of value to the reseller community." I can't believe that AMD will continue this brand of agnosticism when it comes to graphics. In fact, speaking of whitebooks, given that ATI controls the majority of mobile GPU sockets now in production, this merger may be AMD's saving grace as a path into the whitebook space, because the company obviously hasn't fared well there to date. Extrapolating further, perhaps this is the opportunity NVIDIA has been needing to grow within whitebooks. Centrino models with beefed-up graphics have generally gone for ATI chips. If Intel has any pull with the ODMs, perhaps NVIDIA will now get the nod as the preferred graphics choice for higher-end Centrino designs, a possibility well worth keeping in mind as the VBI initiative cranks into full swing during the second half and we prepare for fourth-gen Centrino (Santa Rosa) early next year. But my second thought about Moorhead's statement regards royalties. Regardless of whether the fees that Intel charges chipset manufacturers to design for its platforms are "massive" or not, we have to question the logic of not charging those royalties. It's a given that ATI won't be paying AMD such royalties because ATI is going to be AMD. To then turn around and charge NVIDIA royalties would jack up the costs of any AMD-compatible nForce product and leave it less price competitive versus the equivalent Intel-based offering. So if AMD wants NVIDIA to continue making platforms on which it can continue to move AMD processors, it has to keep the fee field level. There's nothing high-minded or altruistic about it. Moreover, business is business. If you lose revenue from royalties, those dollars have to get made up somewhere else. Sure seems like a good time to debut the 4x4 platform and start having people plant two CPUs on each motherboard instead of one, eh? The same strategy worked like magic for NVIDIA with SLI, after all. Now, speaking of giving away the farm, a look at some financial numbers might shed more insight on why this deal came to pass and why you're less likely to see an overly tight bond result between NVIDIA and Intel. ATI and NVIDIA have similar technologies as well as similar revenue—roughly $2.5 billion. However, ATI's operating margin as listed on Yahoo! Finance is 1.1% with 3,469 employees. NVIDIA's operating margin is 15.34% with 2,737 employees. The picture grows bleaker for ATI when you look at a merger slide deck posted on AMD.com showing that gross margin for ATI's chipset business is 22% while the discrete graphics business hits only 24%—and chipset margins are notoriously low in this industry. AMD predicts that it can lift the discrete graphics numbers to 32% to 37 percent. Meanwhile, according to NVIDA's 10-Q statement, the nForce chipset business is hitting about 35% margins and the discrete graphics line is pulling in over 42 percent. No wonder NVIDIA is loathe to plant SLI on Intel chipset motherboards. It doesn't need to! I wanted to see if NVIDIA's stance on this had altered of late and was told this morning by NVIDIA's Drew Henry, senior director of platform products, "No, we have no plans to offer SLI on Intel core logic." Moreover, ATI and NVIDIA spend roughly the same amount on R&D, but ATI seems to have less to show for it in the desktop and workstation spaces. NVIDIA continues to lead in multi-GPU innovation, physics processing, professional multi-monitor support, and HDV and HDTV handling. On the other hand, Microsoft spent two years subsidizing ATI's work on a new graphics engine for the Xbox 360, none of which has shown up in desktop products yet, and beyond that comes speculation about blending CPUs with GPUs on one chip. While I tend to roll my eyes at a lot of Inquirer stories, one in particular (www.theinquirer.net/default.aspx?article=33219) makes an interesting case for how future CPUs following the Niagara "mini-core" model from Sun, in which many small, dumbed-down cores far exceed the work of one large one, will soon envelope and leave in the dust the processing load handled by the GPU. Intel is already well along in its mini-core research, and AMD without ATI doesn't have a prayer of catching the mini-core with graphics boat. Intel, of course, has a mounting bed of experience with embedded graphics, and you can bet this will become a hot focus for the company going forward. Meanwhile, we have NVIDIA preparing for this scenario...how? Good question. Fortunately, it's one the channel won't have to worry about for two or three years. Meanwhile, with AMD's present sway in the workstation world, what better time to update that market with multi-GPU graphics? You rarely if ever see workstation boards with more than one graphics slot. NVIDIA on its own will have a far harder time moving the same mountain. NVIDIA working in tandem with Intel on the same problem? That's a more intriguing proposition...provided the two companies are willing to stand side-by-side rather than one behind the other, something neither is particularly good at. And in my own heart of hearts, I'd somewhat written the AMD LIVE! platform off as an abortive distraction against Intel's Viiv. But the merger now changes things. ATI's Xilleon line owns 80% of the digital TV chip market. Could there be ways to integrate future Xilleon chips to add greater platform functionality with AMD CPUs and/or ATI chipsets? You bet. "We definitely see ways to enhance the AMD LIVE! platform and even substantially impact the user experience," notes AMD's Moorhead. "Now, the deal isn't closed until fourth quarter, and there are some regulations on how close we can get with ATI before that. But [Xilleon] is absolutely a way for us to bolster the AMD LIVE! solution." Does all of this leave NVIDIA as the odd man out in tomorrow's graphics scene? Probably. But that may not be a bad place to be for a while. "We now become the only company in the world that can supply GPUs and platform products to either CPU," says NVIDIA's Drew Henry. "To us, the opportunity with AMD is as exciting as it was before. The Intel marketplace is now perhaps even more exciting. So I think being in this position is right where we want to be. For our corporate customers, we will be the only company offering a stable image platform for both processor families. This means a company will be able to standardize on our driver set and ship whichever CPU they please." On top of that, NVIDIA has a lot of momentum with some of the highest margin centers in the market. Henry notes that NVIDIA has shipped over 10 million SLI GPUs and over 4 million SLI chipsets. In contrast, he says he would be surprised if 100,000 CrossFire motherboards have been sold to date. CrossFire support is bound to vanish from Intel core logic, which won't help matters for ATI in the short-term, and while the major OEMs may be embracing CrossFire more openly now, sales through distribution to resellers have been relatively slim. In the end, I think everybody wins on this deal. AMD will have far more value to offer, ATI will shed many of its past inefficiencies, NVIDIA may get to play both sides of the fence for a bit but will ultimately end up being Intel's buddy (at least for a while), and Intel will keep on dominating the market with integrated graphics, not really caring much what GPUs go on its boards. Everyone will have to become more competitive and find new values to add, and that spells nothing but good news for system builders. |
||||
 Back to top |
||||
Copyright © 2007 RAM Magazine. All rights reserved.
Do not duplicate or redistribute in any form. |
||||
