backtop


Print E-mail del.icio.us 154 comment(s) - last by Lakku.. on Apr 25 at 11:08 PM


"R600" OEM image courtesy of PCinlife
320-stream processors, named ATI Radeon HD 2900

AMD has named the rest of its upcoming ATI Radeon DirectX 10 product lineup. The new DirectX 10 product family received the ATI Radeon HD 2000-series moniker. For the new product generation, AMD has tagged HD to the product name to designate the entire lineup’s Avivo HD technology. AMD has also removed the X-prefix on its product models.

At the top of the DirectX 10 chain, is the ATI Radeon HD 2900 XT. The AMD ATI Radeon HD 2900-series features 320 stream processors, over twice as many as NVIDIA’s GeForce 8800 GTX. AMD couples the 320 stream processors with a 512-bit memory interface with eight channels. CrossFire support is now natively supported by the AMD ATI Radeon HD 2900-series; the external CrossFire dongle is a thing of the past.

The R600-based ATI Radeon HD 2900-series products also support 128-bit HDR rendering. AMD has also upped the ante on anti-aliasing support. The ATI Radeon HD 2900-series supports up to 24x anti-aliasing. NVIDIA’s GeForce 8800-series only supports up to 16x anti-aliasing. AMD’s ATI Radeon HD 2900-series also possesses physics processing.

New to the ATI Radeon HD 2900-series are integrated HDMI output capabilities with 5.1 surround sound. However, early images of AMD’s OEM R600 reveal dual dual-link DVI outputs, rendering the audio functions useless.

AMD’s RV630-based products will carry the ATI Radeon HD 2600 moniker with Pro and XT models. The value-targeted RV610-based products will carry the ATI Radeon HD 2400 name with Pro and XT models as well.  

The entire AMD ATI Radeon HD 2000-family features the latest Avivo HD technology. AMD’s upgraded Avivo with a new Universal Video Decoder, also known as UVD, and the new Advanced Video Processor, or AVP. UVD previously made its debut in the OEM-exclusive RV550 GPU core. UVD provides hardware acceleration of H.264 and VC-1 high definition video formats used by Blu-ray and HD DVD. The AVP allows the GPU to apply hardware acceleration and video processing functions while keeping power consumption low.

Expect AMD to launch the ATI Radeon HD 2000-family in the upcoming weeks, if AMD doesn’t push back the launch dates further.


Comments     Threshold


This article is over a month old, voting and posting comments is disabled

Awesome
By tuteja1986 on 4/12/2007 10:33:13 PM , Rating: 2
I so can't wait. Looks like a total monster but i do hope it consume less than 250W.




RE: Awesome
By Lakku on 4/12/2007 10:41:16 PM , Rating: 1
All indications point to otherwise, unless they were able to get the process down to 65nm like they were trying. At least, no information exists from leaks that suggests anything less then 240w, with some suggesting up to 280 if used with the 8-pin connector. Who knows, but unless AMD decides to take a big price hit, as in not get as much profit per card, I can't see this not being one helluva expensive card, due to the complex PCB for the 512-bit memory interface and the fact the chip will have horrible yields.


RE: Awesome
By AntDX316 on 4/12/2007 11:49:16 PM , Rating: 1
i was reading this and i was like yea what yea whatever yea whatever then i saw 320 Stream processors? wtf! that should have insane performance


RE: Awesome
By mezman on 4/13/2007 3:51:00 PM , Rating: 2
I don't know. How can ATI possibly keep all those stream processors fed all the time? Is that memory bus wide enough or the mem fast enough? I guess we'll have to wait for some benchies to be sure.


RE: Awesome
By InsaneScientist on 4/13/2007 4:28:10 PM , Rating: 4
Between the 512-bit bus, and the rather high Memclocks, it's runored that the high end version will have roughly 135GB/s memory bandwidth.

Keep in mind, that's just a rumor.

It does make sense, though... 1050MHz DDR (probably GDDR4), or an effective 2100MHz memclock (not unreasonable at all, considering the X1950's 1GHz, 2GHz effective memclock) * a 512 bit bus = approx 134.4GB/s


RE: Awesome
By aftlizard01 on 4/13/2007 12:40:33 AM , Rating: 5
I can't remember where I read it but part of the delay is that AMD wanted to move it all to 65nm and just have a small offering of 80nm ones. Also supposably the power savings from 80-65nm is something like 30%, dropping the wattage requirements below 200watts. Of course it is all conjecture on my part with only hearsay for me to go by.


RE: Awesome
By Dactyl on 4/13/2007 3:47:09 AM , Rating: 2
The Inq is pushing that story hard and I believe their reporting.

AMD of course doesn't want to admit it because then people won't want to buy the 80nm cards.

AMD should simply be honest and open about it: mark the 80nm cards differently, let people know they're hotter/slower, and charge less money for 'em.


RE: Awesome
By kalak on 4/13/07, Rating: -1
RE: Awesome
By Spoelie on 4/15/2007 8:44:52 AM , Rating: 2
3*75 or 225 is the absolute maximum power draw the boards have, assuming 100% draw from every power connector, more is physically impossible. Your numbers are way off.

The actual power draw was more along the 200w. The cooling device was designed for a maximum of 250w. And these numbers are from the early boards, one would hope that AMD has lowered these numbers over time.


RE: Awesome
By Lakku on 4/25/2007 11:08:51 PM , Rating: 2
Except the R600 uses an 8-pin connector and a 6 pin, allowing for a maximum of 300 watts. 8-pin PCIe 2.0 offers 125+ watts I believe. I know you can probably use two six pin connectors, but whose to say to get maximum benefit (i.e. for the XTX version) you don't need the 8-pin connector?


RE: Awesome
By hrah20 on 4/12/2007 10:55:54 PM , Rating: 2
Me too, can't wait for this card, just hope they really come out this time.


RE: Awesome
By aurareturn on 4/12/2007 10:59:12 PM , Rating: 1
Don't get too excited. It all sounds good but I doubt it will live up to its hype.

They always say that "ooh, our new card can do 1 teraflop, our new card has 200gbs of bandwidth, 5x the shading power, blah blah blah". In the end, the card will probably end up around 1.8x the speed of X1950XTX.


RE: Awesome
By WayneG on 4/12/2007 11:23:31 PM , Rating: 2
quote:
by aurareturn on April 12, 2007 at 10:59 PM
Don't get too excited. It all sounds good but I doubt it will live up to its hype.

Erm what are you on about... 320 stream processors are 320 stream processors. I'll be honest I saw this and thought it was a late april fools. These specs look way too good to be true. Fact is though that we won't see the performance from this card really shine until a DX10 title comes along. The CPU that would be needed to feed this beast would be rediculous in DX9. I am soo buying this when the XTX is released. Moving from an oldie X1900XTX!


RE: Awesome
By Lakku on 4/12/2007 11:59:36 PM , Rating: 2
The CPU needed will matter in DX10 titles as well. With my current 8800gtx setup, I find situations more often then not where my quad core at 3ghz holds back the GPU, though this is less common in games like Oblivion and STALKER. With DX10 titles, many of them are going to take advantage of multi-core CPUs. Alan Wake, Bioshock, UT 3, SupCom (DX10 patch coming supposedly, but in its current form, a quad core really helps), Crysis... they ALL take advantage of dual and quad core CPUs. Hell, Alan Wake requires at least a dual core, with quad core preferred. The system is a bottle neck now a days, so 320 stream processors may not mean much, especially if it uses 240 watts and puts out 150 degrees+ of heat. As a side note, we don't know what kind of stream processors it will be using, Beyond3d speculates about it if you are technically minded. It could be that one of ATi's shader processors isn't as efficient or powerful as one of nVidia's, but time will tell.


RE: Awesome
By DingieM on 4/13/2007 4:52:27 AM , Rating: 2
Err AMD (or ATI) is at its second generation of its unified shader architecture so it has the experience and the technology to be more efficient than nVidia.
Multiple sources within AMD has confirmed that the price-tag of the 65nm parts, which are all of them (!) will drop down considerable meaning the consumer will have to pay $100 less for the same performance. Just an example.
I suspect that the 320 stream processors do have calculations with only 1 parameter instead of 4 or 5 parameters within the 48 (3 groups of 16) shader array of the Xenos inside the Xbox360. It may be true that the R600 has 64 physical shader arrays that can calculate 4 or 5 parameters at the same time, if I can describe it that way. Well if you multiply 64 by 5 you get 320! nVidia 8800GTX has 128 physical shader arrays albeit it can only calculate 1 parameter at a time. So one can say 320 stream processors is marketing talk to compare it to nVidia but still has only 64 physical shader arrays (I think).

So the mass production cards shall have 65nm low-power, low-price and very high performance. Next to this built-in crossfire, HDMI and other goodies makes this a true winner.

The major disadvantage is that the main CPU can't keep up.
That's why the games consoles are so much more efficient with main memory, hence the consoles can push their graphics accelerators much more to the limit (which especially counts for the Xenos inside the Xbox360).


RE: Awesome
By Lakku on 4/13/07, Rating: 0
RE: Awesome
By Pirks on 4/13/2007 2:41:49 PM , Rating: 1
quote:
AMD can't make a 200 to 500 million transistor CPU at 65nm
bullshit, AMD manufactures 65nm brisbanes for several months now :P OTOH I agree AMD won't make R600 65nm for a while, just because capacity is not there yet - all AMD 65nm fabs are busy with cranking out brisbanes, I heard (correct me if I'm wrong) so any new 65nm production lines AMD gets will immedietaly go brisbane way, not R600 way, so forget about 65nm R600 for a while, my guess it's a year from now


RE: Awesome
By Goty on 4/13/2007 4:50:37 PM , Rating: 4
AMD isn't going to handle the manufacturing of R600 right now, it's still being outsourced to TSMC.


RE: Awesome
By SilverMirage on 4/14/2007 3:07:37 PM , Rating: 3
Correct me if I'm wrong, but the manufacturing process for Brisbanes (SOI) and R600 (something else...I'm pretty sure) are substantially different to the point that even if they were both 65nm, the Brisbane fabs would have to completely change the equiptment they have.

This is the reason it's taking so long to make the AMD Fusion. It is hard to fuse AMD CPUs with ATI GPUs because the companies process of manufacturing are so different.


RE: Awesome
By Hydrofirex on 4/14/2007 3:08:32 PM , Rating: 3
Keep in mind the two companies aren't using unified manufacturing! ATi presumably still only has the manufacturing capability it had before AMD ever came along. Now, as that situation changes we just might see that AMD and ATi knew exactly what they were doing. As AMD moves out ahead in processors and retires fab space for new generations the Ati segment can come in behind and utilize the advances for much cheaper than having to outsource manufacturing like Ati has been.

HfX