More Mainstream DX10: AMD's 2400 and 2600 Series
by Derek Wilson on June 28, 2007 8:35 AM EST- Posted in
- GPUs
A Closer Look at RV610 and RV630
The RV6xx parts are similar to the R600 hardware we've already covered in detail. There are a few major differences between the two classes of hardware. First and foremost, the RV6xx GPUs include full video decode acceleration for MPEG-2, VC-1, and H.264 encoded content through AMD's UVD hardware. There was some confusion over this when R600 first launched, but AMD has since confirmed that UVD hardware is not at all present in their high end part.
We also have a difference in manufacturing process. R600 uses an 80nm TSMC process aimed at high speed transistors, while their RV610 and RV630 GPU based cards are fabbed on a 65nm TSMC process aimed at lower power consumption. The end result is that these GPUs will run much cooler and require much less power than their big brother the R600.
Transistor speed between these two processes ends up being similar in spite of the focus on power over performance at 65nm. RV610 is built with 180M transistors, while RV630 contains 390M. This is certainly down from the huge transistor count of R600, but nearly 400M is nothing to sneeze at.
Aside from the obvious differences of transistor count and the number of different units (shaders, texture unit, etc.), the only other major difference is in memory bus width. All RV610 GPU based hardware will have a 64-bit memory bus, while RV630 based parts will feature a 128-bit connection to memory. Here's the layout of each GPU:
RV630 Block Diagram
RV610 Block Diagram
One of the first things that jump out is that both RV6xx based designs feature only one render back end block. This part of the chip is responsible for alpha (transparency) and fog, dealing with final z/stencil buffer operations, sending MSAA samples back up to the shader to be resolved, and ultimately blending fragments and writing out final pixel color. Maximum pixel fill rate is limited by the number of render back ends.
In the case of both current RV6xx GPUs, we can only draw out a maximum of 4 pixels per clock (or we can do 8 z/stencil-only ops per clock). While we don't expect extreme resolutions to be run on these parts (at least not in games), we could run into issues with effects that make heavy use of MRTs (multiple render targets), z/stencil buffers, and antialiasing. With the move to DX10, we expect developers to make use of the additional MRTs they have available, and lower resolutions benefit from AA more than high resolutions as well. We would really like to see higher pixel draw power here. Our performance tests will reflect the fact that AA is not kind to AMD's new parts, because of the lack of hardware resolve as well as the use of only one render back end.
Among the notable features that we will see here are tessellation, which could have an even larger impact on low end hardware for enabling detailed and realistic geometry, and CFAA filtering options. Unfortunately, we might not see that much initial use made of the tessellation hardware, and with the reduced pixel draw and shading power of the RVxx series, we are a little skeptical of the benefits of CFAA.
From here, lets move on and take a look at what we actually get in retail products.
The RV6xx parts are similar to the R600 hardware we've already covered in detail. There are a few major differences between the two classes of hardware. First and foremost, the RV6xx GPUs include full video decode acceleration for MPEG-2, VC-1, and H.264 encoded content through AMD's UVD hardware. There was some confusion over this when R600 first launched, but AMD has since confirmed that UVD hardware is not at all present in their high end part.
We also have a difference in manufacturing process. R600 uses an 80nm TSMC process aimed at high speed transistors, while their RV610 and RV630 GPU based cards are fabbed on a 65nm TSMC process aimed at lower power consumption. The end result is that these GPUs will run much cooler and require much less power than their big brother the R600.
Transistor speed between these two processes ends up being similar in spite of the focus on power over performance at 65nm. RV610 is built with 180M transistors, while RV630 contains 390M. This is certainly down from the huge transistor count of R600, but nearly 400M is nothing to sneeze at.
Aside from the obvious differences of transistor count and the number of different units (shaders, texture unit, etc.), the only other major difference is in memory bus width. All RV610 GPU based hardware will have a 64-bit memory bus, while RV630 based parts will feature a 128-bit connection to memory. Here's the layout of each GPU:
One of the first things that jump out is that both RV6xx based designs feature only one render back end block. This part of the chip is responsible for alpha (transparency) and fog, dealing with final z/stencil buffer operations, sending MSAA samples back up to the shader to be resolved, and ultimately blending fragments and writing out final pixel color. Maximum pixel fill rate is limited by the number of render back ends.
In the case of both current RV6xx GPUs, we can only draw out a maximum of 4 pixels per clock (or we can do 8 z/stencil-only ops per clock). While we don't expect extreme resolutions to be run on these parts (at least not in games), we could run into issues with effects that make heavy use of MRTs (multiple render targets), z/stencil buffers, and antialiasing. With the move to DX10, we expect developers to make use of the additional MRTs they have available, and lower resolutions benefit from AA more than high resolutions as well. We would really like to see higher pixel draw power here. Our performance tests will reflect the fact that AA is not kind to AMD's new parts, because of the lack of hardware resolve as well as the use of only one render back end.
Among the notable features that we will see here are tessellation, which could have an even larger impact on low end hardware for enabling detailed and realistic geometry, and CFAA filtering options. Unfortunately, we might not see that much initial use made of the tessellation hardware, and with the reduced pixel draw and shading power of the RVxx series, we are a little skeptical of the benefits of CFAA.
From here, lets move on and take a look at what we actually get in retail products.
Facebook
Twitter
RSS
96 Comments
View All Comments
DerekWilson - Thursday, June 28, 2007 - link
I agree that we need to know dx10 performance, which is why we're doing a followup.I would think it would be clear that, if I were buying a card now, I'd buy a card that performed well under dx9.
All the games I play are dx9, all the games I'll play over the next 6 months will have a dx9 codepath, and we don't have dx10 tests that really help indicate what performance will be like on games designed strictly around dx10.
We always recommend people buy hardware to suit their current needs, because these are the needs we can talk about through our testing better.
TA152H - Thursday, June 28, 2007 - link
OK, that recommendation part is a little scary. You should be balancing the two, because as you know, the future does come. DX9 will exist for the next six months, but there are already games using DX10 that look better than DX9. Plus, Vista surely loves DX10.But, we can agree to disagree on what's more important. I think this site's backward looking style is obvious, and while I fundamentally disagree with it, at least you guys are consistent in your love for dying technology. Then again, I still prefer Win 2K over XP, so I guess I'm guilty of it too, but in this case my primary concern would be DX10. It's better, noticeably so. But, the main thing is, you're judging something for what it's not made for. AMD's announcement made it very clear that DX10 was the main point, and HD visual effects. Yet you chose to test neither and condemn the hardware for legacy code. Read the announcement, and judge it on what's it's supposed to be for. Would you condemn a Toyota Celica because it's not as fast as a Porsche? Or a Corvette because it's got bad fuel economy? I doubt it, because that's not why they were made. Why condemn this part without testing it for what it was for? I didn't see DX9 mentioned anywhere in their announcement. Maybe that was a hint?
Chaotic42 - Thursday, June 28, 2007 - link
Yes, but how many people are going to purchase low-to-mid range cards to play games that aren't coming out for several months?poohbear - Thursday, June 28, 2007 - link
celica compared to a porsche?!?! dude that analogy is waayyyyyy off. How the hell is a toyata celica supposed to represent DX9 & a porsche DX10?!?! considering a porsche u can see instant results and enjoy it instantly, there's nothing out right now on a DX10 and i dont think even in 3 years the DX10 AP would ever encompass the differences between a celica and a porsche. get over yourself.KhoiFather - Thursday, June 28, 2007 - link
Wow, what worthless cards! Like does ATI really think people are going to buy this crap? Maybe for a media box and that's about it but for us mid-range gamers, it's worthless! All this hype and wait for nothing I tell ya!Chadder007 - Thursday, June 28, 2007 - link
Yeah, WTF?? They are all sometimes WORSE than the X1650XT!!! What is going on? According to the specs it should be better, could it be driver issues still??tungtung - Thursday, June 28, 2007 - link
I don't think driver alone will help much ... beside ATI has never really known to be able to magically put strong numbers out through driver updates.Personally I'd say the 2xxx line that AMD/ATI has just sunk to the deep abyss. First it was months late, and the performance was light years behind ... all the while the price is just well not right.
As much as I hate saying this ... it seems that we'll have to wait till Intel dips their giant feet into the graphic industry before nVidia and (especially) AMD/ATI woke up and think carefully about their next products (that is if they can bring a competitive product) ... especially in the mainstream and value market.
OrSin - Thursday, June 28, 2007 - link
Very easy to guess what is happening here. Both camps are targeting the the high engame that switch to vista and cna afford the high end cards. And the OEM cards for deal so they push Vista again on people. Niether company want to lower their high sells by releasing a mid-level part. I just wonder if the cards are just more expensive to make for D10. I don't see a reason it would be, maybe I'm wrong.Until Vista is used by more gamer my guess is they will not release a mid range card.
Early adaptors of software is getting screwed.
DigitalFreak - Thursday, June 28, 2007 - link
Agreed. I couldn't help laughing when I read the Final Words section. Kinda like "The Nvidia 86xx/85xx cards suck, and the ATI 26xx/24xx suck worse!"WTF happened this generation? The only cards worth their salt are the 88xx series. Nvidia dropped the ball with their low end stuff, and AMD.... well, AMD never really showed up for the game.
smitty3268 - Thursday, June 28, 2007 - link
I think it's clear that with these low end cards, ATI and NVIDIA both came to the conclusion that they could either spend their transistor budget implementing the DX10 spec or adding performance, and they both went with DX10. Probably so they could be marketed as Vista compatible, or whatever. It's still a mystery why they didn't choose to make any midrange cards, as they tend to sell fairly well AFAIK. Perhaps these were meant to be midrange cards and ATI/NVIDIA were just shocked by how badly performance scaled downwards in their current designs, and were forced to reposition them as cheaper cards.