AMD: R7xx Speculation
- Thread starter Unknown Soldier
- Start date
- Status
OCW reports there will no demo-showcase of RV770 at Computex.
So it seems that RV770 may move in Q3.
So it seems that RV770 may move in Q3.
Interesting way to read it you have there, I mean, they say there might not be showcase of the cards there, not to mention that not showing it there (in "public" doesn't mean anything about not being able to bring the card this quarter.
Not very strong language in that post. "might not" "we suspect"
I'm not saying it can't be correct, just that I'll believe it when Computex has come and gone and there was no public showing of RV770.
Sound_Card
Regular
When is Cebit exactly?
I take it you mean Computex, since Cebit already took place this last March
Computex runs June 3-7 in Taipei.
Sound_Card
Regular
I take it you mean Computex, since Cebit already took place this last March
Computex runs June 3-7 in Taipei.
oh my bad.
I hope they show it. By all means, the last thing they need is delays.
I think it's rather silly to believe they won't be showing RV770 at one of the biggest trade shows in the industry a month+ after shipping chips to AIB partners.
Just remember, showcasing doesn't mean it won't be shown.
After all, last computex, Intel didn't showcase their X38 boards (technically they were under NDA still) but every manufacturer/partner had their X38 boards in full display in all their variants and specs, so it doesn't mean anything.
What they likely mean is that ATI themselves will not have a showcase/presser for the RV770 but will AIB/partners have them on display? In all likelihood... yes
After all, last computex, Intel didn't showcase their X38 boards (technically they were under NDA still) but every manufacturer/partner had their X38 boards in full display in all their variants and specs, so it doesn't mean anything.
What they likely mean is that ATI themselves will not have a showcase/presser for the RV770 but will AIB/partners have them on display? In all likelihood... yes
In his latest musing, http://www.fudzilla.com/index.php?option=com_content&task=view&id=7240&Itemid=34, Fuad seems more confused than ever. I don't believe that the RV770XT would have a 512-bit memory interface but not the RV770Pro. Is there an agreement (between rumours at least) on those names btw?
The width itself doesn't make much sense, combined with GDDR5, unless it's some cunning design for dual chips to access memory each with its 256-bit path. (Do I know what I am talking about? Absolutely not, but this thread has been kinda dead lately, and I find it depressing, so maybe offending those who know will tempt them to reveal something, anything)
The width itself doesn't make much sense, combined with GDDR5, unless it's some cunning design for dual chips to access memory each with its 256-bit path. (Do I know what I am talking about? Absolutely not, but this thread has been kinda dead lately, and I find it depressing, so maybe offending those who know will tempt them to reveal something, anything)
It wouldn't be a new ATI architecture if it didn't increase the ALU-Tex ratio.
The general consensus is that applications will become more and more shader intensive (both in complexity and length), especially with GPU's becoming more and more programmable. A further ALU endowed architecture would make sense, wouldn't it?
About the multi-die issue: So long as the card doesn't rely on software performance profiles, I'll be happy. Cards like the 7950 GX2, HD3870 X2 and 9800 GX2 do very little to interest me. Performance is just way too shaky, and hiccups always occur in games (I've got a HD3870 CrossFire setup--everything isn't smooth sailing).
Let's try being a bit more pragmatic here: even if the ALU-TEX ratio was unchanged, it would effectively go down (i.e. more 'effective' TMUs) because a higher-end chip is naturally used at higher-end resolutions which require less trilinear/AF and more simple bilinear.
I doubt that the RV770 will run games at a higher resolution than the r600 did when it was released - games today require a bit more horsepower than they did a year ago. (it however depends on how big a performance improvement the RV770 offers).
Jawed
Legend
I wonder how much value there is in comparing the 2xMSAA performance of RV670 against the 4xMSAA performance of G92 as a potential indicator of RV770's 4xMSAA performance.
There's a high expectation that RV770 will double Z fillrate (though ATI has a sickening habit of not delivering in this regard) and there's no way in hell that it should be more bandwidth bound running 4xMSAA than G92 is. So I'm wondering if 2xMSAA performance on RV670 gives a reasonable indication of RV770's 4xMSAA performance.
Jawed
There's a high expectation that RV770 will double Z fillrate (though ATI has a sickening habit of not delivering in this regard) and there's no way in hell that it should be more bandwidth bound running 4xMSAA than G92 is. So I'm wondering if 2xMSAA performance on RV670 gives a reasonable indication of RV770's 4xMSAA performance.
Jawed
Here's a question that's been nagging me for a while, because I don't have the fundamental understandings to answer it myself:
I can see that a lot of shading algorithms result in the shaders producing textures (for instance for shadows). Wouldn't those textures require any filtering in order to use them, and in that case wouldn't any increase in ALU power still require an increase in filtering power?
I'm asking here because that question pops up in my mind everytime the ALU-tex ratio is mentioned in this thread. And thanks to whoever can take the time to answer it.
I can see that a lot of shading algorithms result in the shaders producing textures (for instance for shadows). Wouldn't those textures require any filtering in order to use them, and in that case wouldn't any increase in ALU power still require an increase in filtering power?
I'm asking here because that question pops up in my mind everytime the ALU-tex ratio is mentioned in this thread. And thanks to whoever can take the time to answer it.
Unknown Soldier
Veteran
So much for R7xx not getting GDDR5 memory.
GDDR5 in Production, New Round of Graphics Cards War Imminent
Qimonda, a leading manufacturer of advanced dynamic random access memory (DRAM), said that it could deliver next-generation GDDR5 memory for graphics cards and other applications that require high memory bandwidth in volume. At this point Qimonda can supply makers of graphics boards GDDR5 memory with up to 4.50GHz clock-speed.
“Qimonda was the first to announce samples of GDDR5 back in November 2007. We have proven the technology and we can deliver in volume production to the market today,” said Glen Haley, communications director of Qimonda in North America, in an interview with X-bit labs web-site.
Presently Qimonda has 512Mb (16Mx32) GDDR5 chips at 3.60GHz, 4.0GHz and 4.50GHz clock-speeds in PG-TFBGA-170 packages in production. It is interesting to note that current-generation GDDR3 chips from Qimonda use PG-TFBGA-136 packing, which means that GDDR5 has more pins and requires more complex print-circuit boards of graphics cards.
It is projected that GDDR5 will play a substantial role in the next round of war between the leading designers of graphics processing units (GPUs) because it can double bandwidth available for graphics chips. Unfortunately, the price of GDDR5 memory from Qimonda is unclear. But while Qimonda does not unveil the exact pricing, it is likely that GDDR5 will be more expensive than GDDR3 and GDDR4, at least initially.
Read More: http://www.xbitlabs.com/news/video/...New_Round_of_Graphics_Cards_War_Imminent.html
GDDR5 in Production, New Round of Graphics Cards War Imminent
Qimonda, a leading manufacturer of advanced dynamic random access memory (DRAM), said that it could deliver next-generation GDDR5 memory for graphics cards and other applications that require high memory bandwidth in volume. At this point Qimonda can supply makers of graphics boards GDDR5 memory with up to 4.50GHz clock-speed.
“Qimonda was the first to announce samples of GDDR5 back in November 2007. We have proven the technology and we can deliver in volume production to the market today,” said Glen Haley, communications director of Qimonda in North America, in an interview with X-bit labs web-site.
Presently Qimonda has 512Mb (16Mx32) GDDR5 chips at 3.60GHz, 4.0GHz and 4.50GHz clock-speeds in PG-TFBGA-170 packages in production. It is interesting to note that current-generation GDDR3 chips from Qimonda use PG-TFBGA-136 packing, which means that GDDR5 has more pins and requires more complex print-circuit boards of graphics cards.
It is projected that GDDR5 will play a substantial role in the next round of war between the leading designers of graphics processing units (GPUs) because it can double bandwidth available for graphics chips. Unfortunately, the price of GDDR5 memory from Qimonda is unclear. But while Qimonda does not unveil the exact pricing, it is likely that GDDR5 will be more expensive than GDDR3 and GDDR4, at least initially.
Read More: http://www.xbitlabs.com/news/video/...New_Round_of_Graphics_Cards_War_Imminent.html
Freak'n Big Panda
Regular
Not every shader needs texture data, the trend in the industry as of late as been an increase in the amount of math needed per pixel hence ATI's rather aggressive ALU:Tex ratio. You can write shaders that don't need any texture data whatsoever so in short an increase in ALU power does not necessitate an increase in texure filtering ability.
Arun, could you elaborate here please? Are you saying as a result of this that enabling AF should cause less of a performance hit as resolution scales, or are you saying that end-users shouldn't be applying AF @ high resolution because it causes the GPU to do unnecessary work?
- Status
