AMD: R8xx Speculation

How soon will Nvidia respond with GT300 to upcoming ATI-RV870 lineup GPUs

  • Within 1 or 2 weeks

    Votes: 1 0.6%
  • Within a month

    Votes: 5 3.2%
  • Within couple months

    Votes: 28 18.1%
  • Very late this year

    Votes: 52 33.5%
  • Not until next year

    Votes: 69 44.5%

  • Total voters
    155
  • Poll closed .
fp16 might have been ATI's original solution to the question of increasing texture-filtering quality.

Indeed the high quality filtering in G80 etc. might be a hang-over from an expectation that strict quality was going to be defined for D3D10. But it was dropped from the standard for some reason?

Jawed
 
Theo just got an epic smackdown, its on the same level if not worse than Sander. :LOL: He should have know better for guessing around too long, just surprised why CJ didnt oust him earlier?

Hahahaha I had no idea that there was more to that article than the snippet Jawed linked. Just read the whole thing. Such intrigue!! Drama!! But yeah Theo got called out like a punk and although it's only Charlie's take on the story it sounds like he deserves it.
 
Compiling the previous mobile information into a table with the TDP:

Broadway HD5870M 45-60W
Broadway HD5850M 30-40W
Broadway HD5830M 29W
Madison HD5750M 20-30W
Madison HD5730M 20-25W
Madison HD5650M 15-20W
Park HD5470M 12-15W
Park HD5450M 10-12W
Park HD5450M < 8W

Now trying to compare to the current M9x TDP Figures(From here and here - obviously take from a german site i cant find):

RV770 HD4870M 47-55W
RV740 Hd4860M ~44W ??
RV770 HD4850M 40-48W
RV740 HD4830M ~30W ??
RV730 HD4670M 28-30W
RV730 HD4650M 12-25W
RV710 HD4570M 12-15W
RV710 HD4530M 8-12W
RV710 HD4330M 7W

Hopefully can see in the above roughly Broadway is set to replace RV740 and RV770, Madison the RV730 and Park the RV710. They are stated as pin compatible so that implies Park is a 64bit part. Looks like AMD is trying to make these drop in replacements so wont need to go through as full ordeal verfication to get them in current designs.

Now when TSMC announced 40nm G they stated: The 40G process is claimed to be up to 30 percent faster than TSMC’s 65nm GP process at the same leakage, or up to 70 percent lower leakage at the same speed.

As the power figures are roughly equal it looks as if AMD has taken nearly all the 40nm gains in terms of increased performance. If the improvement they got was really 35%-45% than current parts, then say roughly 20% is due to the shrink(as comparing 40nm to 55nm not 65nm) and at least 15-25% is due to design improvements and or/more units.

As it sounds like alot of gain, i think some of that 15-25% could also be due in part to better memory ie low voltage DDR3 or GDDR5

Finally particularly note the 3 40nm 128bit chips: RV740M, Madison and Broadway. Madison is clearly below the RV740 and Broadway just above the RV740 in performance.
 
Mr. Hands

http://www.fudzilla.com/content/view/14815/1/

We can confirm that today at 1 AMD place in Sunnyvale we had a chance to see DirectX 11 card, up and running. We had one in our hands on one of the cards and we’ve seen one running a few DirectX 11 technology demos. So DirectX 11 hardware from ATI is real and not a power point presentation.

Confidence radiates.

Hemlock vs. D12U, which codename do you prefer?
 
Last edited by a moderator:
R600 had additional point samplers in there, not just the wider bilerp lanes. As for the possible native FP16 tex impl in R800, it would be justified, if game dev's are to be lured by the new FP compression formats and start using more extensively HDR texturing all around.
Hmm, I've got no idea if the point samplers were necessitated by the inclusion of fp16 filtering in R600, so no idea whether they'd return if Evergreen returns to fp16. Maybe that was more an artefact of the ring-bus based "split" texturing architecture.

But, yeah, the "first class citizen" status of fp16 in D3D11 might be relevant to the design.

Jawed
 
We saw Evergreen, DirectX 11 card (Fuad)
We can confirm that today at an AMD place in Sunnyvale we had a chance to see DirectX 11 card, up and running. We had one in our hands on one of the cards and we’ve seen one running a few DirectX 11 technology demos. So DirectX 11 hardware from ATI is real and not a Power Point presentation.



ChinaJoy 2009 (Shanghai, 2009-07-26) - AMD DX11 demonstration pictures

6e9dc849-8d6f-4db1-86e5-f91d0bf6e1bd.png


High resolution:
· Image 1
· Image 2
 
2 pin PCIE. That's not exactly what we saw at previous showcases, and I doubt that was running at real low clocks (probably "840" instead)

Couldn't they have made their watermark even more opaque?

The size approach is better if you have a noisy image that is intolerant of touchups. :LOL:


The official Chiphell source has been having a run- 770/30/10 coolers way earlier than launch, 770 SP numbers, and now the flagship single GPU card.
 
Last edited by a moderator:
2 pin PCIE. That's not exactly what we saw at previous showcases, and I doubt that was running at real low clocks (probably "840" instead)
I think the use of RV740 cooler was a dead giveaway on the fact that it was a middlerange card tops
 
Back
Top