AMD: R9xx Speculation

It's quite elegant considering slide below:
Well I think it's both quite elegant and brute force...
Elegant because the vertex data doesn't have to move on/off the SM while being processed...
Brute force because simply there are lots of these units...

I'm wondering if AMD couldn't try to double some more logic instead of only the rasterizer without too much effort - that might still not be as elegant as a fully distributed setup, but probably a lot cheaper...
 
Elegant because the vertex data doesn't have to move on/off the SM while being processed...
From earlier presentations I seem to remember sentiments like "When no amplification takes place, the data stays inside the SM" implying also that it has to be re scheduled when there's additional geometry involved.
 
From earlier presentations I seem to remember sentiments like "When no amplification takes place, the data stays inside the SM" implying also that it has to be re scheduled when there's additional geometry involved.

When no amplification takes place, you get jack on a Geforce(not for any technical reason, mind you). And they don't exactly stay on SM, but rather take a round-trip via the L2 most of the time.
 
191031052025.jpg


After several weeks of speculation and false rumor spreads, we now have at NordicHardware to define the most important details about AMD's new graphics card series, Radeon 6800th Towards mid-range market, AMD to launch Radeon HD 6850 and HD 6870 October 18
AMD Radeon HD 6850 (Bart Pro) and HD 6870 (Barts XT) will be launched on 18 October (19 th of certain countries) and the new Northern Islands architecture is refined and optimized to provide even better performance per watt. We should not expect any breakthrough performance increase over today's model range, but more on that soon

 
Everything suggests this to be a fake. It seems to me that is leaked in order to build a negative attitude towards the products. Just look at these slides. I have the impression that NVidia leaked them. :LOL:
 
If NordicHardware's info is true (they seem confident about it ) , then I have to say , these power consumption figures looks cool .
How so? Especially Barts Pro would be very disappointing imho. Put another way, the performance estimate "worse than HD5830" is "hardly better than HD5770". So something like 10% faster than HD5770 with 10% higher power draw wouldn't quite seem like progress to me. I sure hope though at least the performance estimate for this is a bit off...
 
How so? Especially Barts Pro would be very disappointing imho. Put another way, the performance estimate "worse than HD5830" is "hardly better than HD5770". So something like 10% faster than HD5770 with 10% higher power draw wouldn't quite seem like progress to me. I sure hope though at least the performance estimate for this is a bit off...

How so? Well, lets see here...they can't get the information about the 5770 correct for starters :p
But a 108 watt 256 bit memory bus actually would be pretty good though.
 
But a 108 watt 256 bit memory bus actually would be pretty good though.
:LOL:

How so? Especially Barts Pro would be very disappointing imho. Put another way, the performance estimate "worse than HD5830" is "hardly better than HD5770". So something like 10% faster than HD5770 with 10% higher power draw wouldn't quite seem like progress to me. I sure hope though at least the performance estimate for this is a bit off...
HD 6870 (Barts XT) looks good though !
 
It seems that when you look at the slides some websites are leaking as prolly accurate it tells more about the bias toward AMD/Nvidia than any post they could make :D
 
HD 6870 (Barts XT) looks good though !
"Slightly better than HD5850" - again with what looks like slightly higher power draw. Granted they are using same manufacturing tech, so miracles can't be expected but I was hoping for a bit more for the supposed efficiency improvements...
 
If NordicHardware's info is true (they seem confident about it ) , then I have to say , these power consumption figures looks cool .
Check the table, their info contradicts it. They say even better performance per watt, but according to the table HD6870 is slightly faster than HD5850 and consumes slightly more power, too... that means the same performance per watt...

/edit: mczak was faster...
 
Back
Top