AMD: R8xx Speculation

[digitalwanderer]

Luckily, LG is bringing borderless TVs already, so I guess there will be similar meant as monitors soon too
http://www.gadgetfolder.com/lg-sl8000-and-sl9000-borderless-lcd-tvs-ifa-2009.html

However, the customary frame has not changed much, and the display does not occupy the entire front of the TV, just outfitted with new items special laminated film, which hides the frame and simultaneously reduces glare.

Still has a frame.

 

[Kaotik]

Well, if you're a DIY guy, you can always disassemble few monitors and make a new, borderless version out of 'em

 

[Silent_Buddha]

I can already imagine a DIY market for that similar to the console modding market.

Regards,
SB

 

[MfA]

Well, if you're a DIY guy, you can always disassemble few monitors and make a new, borderless version out of 'em

The physical panels have borders, can't DIY that away.

 

[neliz]

http://www.fudzilla.com/content/view/15471/1/
. Poor Fuad.


5770 @ $150-200, 285+- performance with half the power draw?

You can see which side influences Fuad's "news"

 

[brain_stew]

"Because we support GPU-accelerated physics, our $129 card that’s shipping today is faster than their new RV870 (code name for new AMD chips) that sells for $399." - nVidia




"President and CEO of NVIDIA Corp. (NVDA) Jen Hsun Huang sells 260,475 shares of NVDA on 09/11/2009 at an average price of $16.07 a share." (link)

Well I didn't think it'd be long before the FUD campaign got going but that's just embarassing. Really now, this is Sony level spin.

 

[Jawed]

I think Juniper is 256bit and Cypress 512bit. I hope im right

Hmm, with Cypress coming in 1GB and 2GB configurations...

But Juniper being 256-bit is hardly economical and wouldn't be a replacement for RV740.

On the other hand, if there is a chip between Juniper (256-bit) and Cedar (64-bit), then it (Redwood) sorta makes sense as the 128-bit replacement for RV740.

This patent, in theory, would allow for a radical increase in I/O density:

http://v3.espacenet.com/publication...T=D&date=20090205&CC=US&NR=2009032941A1&KC=A1

since power can be routed across the outside of the GPU chip. This means that power for the peripheral areas of the chip can be routed from central areas of the chip. This then means that the peripheral areas of the chip, where the I/O lives, don't lose balls to power. All the balls can be I/O.

Jawed

 

[brain_stew]

By the way... There are a few more videos linked from where the new ruby vid arrived.


Some kind of deferred lighting demo.
http://pc.watch.impress.co.jp/video/pcw/docs/315/056/html/19.flv.html

QUOTE]

That's Frostbite 2 (from Dice), the demo is suppose to show off 1000 point lights and is using compute shaders.

 

[no-X]

Hmm, with Cypress coming in 1GB and 2GB configurations...

But Juniper being 256-bit is hardly economical and wouldn't be a replacement for RV740.

On the other hand, if there is a chip between Juniper (256-bit) and Cedar (64-bit), then it (Redwood) sorta makes sense as the 128-bit replacement for RV740

HD5870 2GB has 16 memory modules: 16x Samsung GDDR5 K4G10325FE:

http://www.wsgfmedia.com/generaladmission/pics/3x2setup-card.jpg
http://www.samsung.com/global/business/semiconductor/productList.do?fmly_id=675

16pcs of 32bit modules could be used for 512bit configuration, but the GPU is only < 340 mm² large, so 256bit clamshell configuration (like FS9270) seems to be more likely.

Anyway, official slide says nothing about memory bus:
http://www.abload.de/img/2ym9i.jpg

 

[Jawed]

Yes, 256-bit clamshell 2GB makes much more sense. But we can dream...

Jawed

 

[Raqia]

Nice report by the enthusiasts.

http://www.widescreengamingforum.com/forum/viewtopic.php?t=16780

Thanks, quite informative.

 

[Wirmish]

Anyway, official slide says nothing about memory bus:
http://www.abload.de/img/2ym9i.jpg

LINK -> "Grafikspeicher-Anbindung : 256 Bit" = "memory bus width : 256 bit"

 

[Nebuchadnezzar]

LINK -> "Grafikspeicher-Anbindung : 256 Bit" = "memory bus width : 256 bit"

That site has just bogus specs before launches to create publicity, the same with GT200 and 770 if I remember right.

 

[ZerazaX]

LINK -> "Grafikspeicher-Anbindung : 256 Bit" = "memory bus width : 256 bit"

Still, ATI hasn't released any official slides/info on the bus width yet

 

[Alexko]

That's true, but if the bus were 512 bit they'd probably be advertising their crazy 250+GB/s bandwidth on the official slides.

 

[no-X]

I'm very skeptical to 512bit theory, but just imagine a hypotetical situation: ATi wants to release a 512bit part, but they want surprise us (not to shout about 512bit bus and cause exaggerated expectation as in R600 times) - what would be the best strategy? To avoid any mention of bus width, any mention of bandwidth and speak only about capacity and clock speed. And use a configuration, which could be explained alternatively: as a 256bit clamshell.

 

[Jaaanosik]

That's true, but if the bus were 512 bit they'd probably be advertising their crazy 250+GB/s bandwidth on the official slides.

Why would they. It's supposed to be a secret weapon.
Remember, 'dominate' is going to have a new meaning!

Where does the Eyefinity power come from?

 

[Arty]

You can see which side influences Fuad's "news"

Fuad is saying AMD is lucky if there is one DX11 title this year, some spin that is.

 

[Ninjagnu]

'By a rough specification-based estimate, the Radeon HD 5870 could end up two times faster than the Radeon HD 4870 but realistically, you should expect the new card to be faster by about 60 percent across the board.'

http://www.fudzilla.com/content/view/15436/34/

That is a quality humor site, or maybe my Saturdays or just really boring.

 

[nAo]

That is a quality humor site, or maybe my Saturdays or just really boring.

What part does it make you laugh the most?