While AMD will tell us that R600 is not late and hasn't been delayed, this is simply because they never actually set a public date from which to be delayed. We all know that AMD would rather have seen their hardware hit the streets at or around the time Vista launched, or better yet, alongside G80. But the fact is that AMD had quite a few problems in getting R600 out the door.
While we couldn't really get the whole story from anyone, we heard bits and pieces here and there during our three day briefing event in Tunis, Tunisia. These conversations were short and scattered and not the kind of thing that it's easy to get a straight answer about when asking direct questions. Keeping that in mind, we do have some information and speculation about a few of the road bumps AMD faced with R600.
Apparently, the first spin of R600 silicon could only communicate over the debugging interface. While the upside is that the chip wasn't totally dead, this is not a good problem to have. We also overheard that a later revision of the hardware suffered from fragments getting stuck in pixel shaders. We even overheard one conversation where someone jokingly remarked that AMD should design hardware but leave the execution to NVIDIA.
In a wild bout of pure speculation on our part, we would have to guess about one other problem that popped up during R600's creation. It seems to us that AMD was unable to get their MSAA hardware to work properly and was forced to use shader hardware to handle MSAA rather than go back for yet another silicon revision. Please know that this is not a confirmed fact, but just an educated guess.
In another unique move, there is no high end part in AMD's R600 lineup. The Radeon HD 2900 XT is the highest end graphics card in the lineup and it's priced at $399. While we appreciate AMD's intent to keep prices in check, the justification is what we have an issue with. According to AMD, it loses money on high end parts which is why we won't see anything more expensive than the 2900 XT this time around. The real story is that AMD would lose money on a high end part if it wasn't competitive, which is why we feel that there's nothing more expensive than the 2900 XT. It's not a huge deal because the number of people buying > $399 graphics cards is limited, but before we've started the review AMD is already giving up ground to NVIDIA, which isn't a good sign.
More than anything, we'd guess that the lack of a high end part has a lot to do with the delays and struggles AMD saw this time around in bringing R600 to market. We expect to see the return of a very high end part by the time R700 comes around, assuming that there aren't similarly debilitating delays.
The delays and lack of a high end would be beyond perfect if the Radeon HD 2900 XT could do to NVIDIA what the G80 launch did to ATI, unfortunately the picture just isn't that rosy. ATI's latest and greatest doesn't exactly deliver the best performance per watt, so while it doesn't compete performance-wise with the GeForce 8800 GTX it requires more power. An ultra high end power requirement in a sub-$400 graphics card isn't exactly ideal.
Despite all of this, there's a great deal of cool technology in the R600, and as ATI is now a part of a CPU company, we received more detail on the GPU than we've gotten during any other GPU launch. AMD takes graphics very seriously, and it recently reaffirmed its commitment to continue to deliver high end discrete graphics cards, so amidst countless delays and rumors of strange problems, the R600 architecture is quite possibly more important to AMD than the graphics cards themselves. An eventual derivative of this architecture will be used in AMD's Fusion processors, eventually making their way into a heterogeneous multi-core AMD microprocessor.
With AMD's disappointing Q1, it can't rest too much on the hope of Fusion changing the market, so we'll have to start by looking at where R600 is today and how it stacks up to NVIDIA's latest and almost greatest.
The new ATI Radeon HD 2900 XT runs at 740 MHz and accesses its 512 MB GDDR3 memory at 1.65 GHz (825 MHz x 2), using a new 512-bit memory interface, with boosts the memory maximum theoretical transfer rate to 105.6 GB/s – Radeon X1950 XTX has a memory maximum transfer rate of 64 GB/s and GeForce 8800 GTX, of 86.4 GB/s, but the new GeForce 8800 Ultra reaches 103.6 GB/s.
Since it is based on Shader 4.0 (DirectX 10), this video card uses a unified shader processing architecture, i.e. instead of having separated units for processing pixel shader, vertex shader, physics and geometry, it has several generic units, called “stream processors”. This model has 320 processing units, against 128 on GeForce 8800 GTX and Ultra.
There are several other new features on this new video card, like the new generation of Avivo – 2D-enhancement technology –, dubbed Avivo HD, now capable of decoding HD-DVD and Blu-Ray discs directly on the GPU and thus freeing up the CPU from this task, and the use of 128-bit High Dynamic Range (HDR), against 64-bit on the previous ATI generation. We’ve written a full article explaining the architecture of this new video card and we recommend you to read this article to complement the present review.
For a full comparison between ATI Radeon HD 2900 XT and competing chips from nVidia, please read our tutorial nVidia Chips Comparison Table.
On Figures 2 and 3 you can see the reviewed card from AMD, which is a reference model. When a high-end video card is released, AMD (or nVidia) manufactures the cards (actually the cards are manufactured by an Asian manufacturer contracted by them) and then sells them to their partners, the video card manufacturers, which add their sticker, changed the BIOS with their settings, add their cables and CDs and put everything into their box. In this case, the cards provided by different manufacturers are exactly the same. This is the case with Radeon HD 2900 XT, at least right now
Since it is based on Shader 4.0 (DirectX 10), this video card uses a unified shader processing architecture, i.e. instead of having separated units for processing pixel shader, vertex shader, physics and geometry, it has several generic units, called “stream processors”. This model has 320 processing units, against 128 on GeForce 8800 GTX and Ultra.
There are several other new features on this new video card, like the new generation of Avivo – 2D-enhancement technology –, dubbed Avivo HD, now capable of decoding HD-DVD and Blu-Ray discs directly on the GPU and thus freeing up the CPU from this task, and the use of 128-bit High Dynamic Range (HDR), against 64-bit on the previous ATI generation. We’ve written a full article explaining the architecture of this new video card and we recommend you to read this article to complement the present review.
For a full comparison between ATI Radeon HD 2900 XT and competing chips from nVidia, please read our tutorial nVidia Chips Comparison Table.
On Figures 2 and 3 you can see the reviewed card from AMD, which is a reference model. When a high-end video card is released, AMD (or nVidia) manufactures the cards (actually the cards are manufactured by an Asian manufacturer contracted by them) and then sells them to their partners, the video card manufacturers, which add their sticker, changed the BIOS with their settings, add their cables and CDs and put everything into their box. In this case, the cards provided by different manufacturers are exactly the same. This is the case with Radeon HD 2900 XT, at least right now
ไม่มีความคิดเห็น:
แสดงความคิดเห็น