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 .
In any case, it would be truly impressive if AMD launches a full DX11 lineup from top to bottom. But what does that say about RV740? Scrapped?
 
Laptops?

Since laptop schedule appears to be 3-6 months behind desktop, RV740 presumably has life left there. It is meant to have returned to production.

Jawed
 
Well before the 40nm yield issues were fully exposed I was expecting RV740 to actually do quite well in the notebook market, though it struck me from the outset that AMD had made no significant OEM design win announcements whereas NVDA seems to have racked up a whole bunch. Whatever it is, ATI seems incapable of recovering from that huge market share loss in the mobile space. Maybe the DX11 generation will be a turning point.
 
Sorry, I don't get it: What's your point?
I think the performance hit of using Edge Detect AA compared to MSAA with a tessellated scene will be relatively lower than it has been with traditional fairly low-poly scenes.


ATI Edge Detect AA mode does an edge-detect pass (should scale with resolution not triangle count) to find triangle edges where there is a high contrast change ie where jaggies are most noticeable.
It then focuses high AA application to the detected high-contrast edges.
It applies a lower AA level to the other triangle edges.

MSAA applies the same level of AA on all triangle edges so will scale by triangle count.


Also ATI historically has relatively few ROPs/RBEs compared with NV but AA performance is similar due to higher ATI clocks.
Rumors suggest that the mainstream/high end cards of the new generation will have 32/64 RBEs.
That's more like NV unit numbers but will presumably still be running at ATI's typically higher clock rates.

So if those rumors are true, these cards should be fillrate monsters = Lots of AA performance.
 
Last edited by a moderator:
In any case, it would be truly impressive if AMD launches a full DX11 lineup from top to bottom. But what does that say about RV740? Scrapped?

Well if rumors are true that Rv740 was delayed and was originally meant to launch for last holiday season, then it would be nearing EOL anyways if the R(v)8xx family is on time... I'm afraid the Rv740 will share the same short lifespan as the R520 and R600.

So I'm going to guess that it's pretty much a dead product, killed by TSMC's problems on 40 nm. It'll probably still limp along with what's available but I don't think ATI will be putting anymore orders in for it.

That's of course, assuming recent rumors are true.

And what a bunch of rumors those are.

AMD ready to launch a complete top to bottom lineup featuring up to 5 parts addressing 4 market segments?

And unrelated to AMD, but might help them. A possibility of at least 6 DX11 titles launching this year???

I dunno. It all sounds just a bit too good to be true.

Regards,
SB
 
http://brightsideofnews.com/news/2009/7/22/ati-to-launch-directx-11-gpus-in-seven-weeks.aspx

I know it's theo, one of the most unreliable "sources" for news, but does anyone have any tidbits on if he's in fact correct on this one?

Looking at the article he has used, uncredited, the codenames from previous Inquirer article by Sylvie Barak and attached a bunch of images from a google image search.

Unfortunately with Hemlock appears to have mixed up the Tsuga tree - a medium tallish conifer from North America and East Asia that grows well in the cold with Conium - a small shrub from Europe/North Africa that contains a poison very similar to nicotine.

Could forgive the above if he had added a date and/or event where the launch is to be held. For some reason he did not, as September was also mentioned in the Inquirer article.

Suppose from the information given could search for events in San Francisco around Week 37-38, September 7-18 roughly. The only thing i can think of about then is IDF.....:LOL:
 
Hmm, Sept/Oct in San Francisco, a place that might be forgotten, a full generation of 'earth shattering' chips called Cypress?
How about this ?

Actually they are all streets in San Francisco. Mostly not as famous as that one though.
 
I think the performance hit of using Edge Detect AA compared to MSAA with a tessellated scene will be relatively lower than it has been with traditional fairly low-poly scenes.


ATI Edge Detect AA mode does an edge-detect pass (should scale with resolution not triangle count) to find triangle edges where there is a high contrast change ie where jaggies are most noticeable.
It then focuses high AA application to the detected high-contrast edges.
It applies a lower AA level to the other triangle edges.

MSAA applies the same level of AA on all triangle edges so will scale by triangle count.


Also ATI historically has relatively few ROPs/RBEs compared with NV but AA performance is similar due to higher ATI clocks.
Rumors suggest that the mainstream/high end cards of the new generation will have 32/64 RBEs.
That's more like NV unit numbers but will presumably still be running at ATI's typically higher clock rates.

So if those rumors are true, these cards should be fillrate monsters = Lots of AA performance.

Food for thought: multisampling is at the moment quite a lot faster than supersampling because polygon edge/intersection data is way lower than polygon interior data. Consider what happens if that balance starts to break.

What exactly edge detect has to do with that is beyond my imagination anyway; its an outstanding AA mode frankly but its current performance is all but great at the moment. That shouldn't mean that they can't tune it up in future generations but that's a totall different issue.

NV has released a patent about coalescing for multisampling if you're interested to read into it and not it's not a real solution for the problem at hand either IMHLO.
 
Chiphell's nApoleon confirms some DX11 event on the 24th of September in Beijing.



Justifies Theo's news post?

It should, that czech site Jawed linked to in the GT300 topic said a launch around the first of october, along with the "GTX380."

So, 2 months gentlemen!
 
Maybe he went to Munich, too:

http://translate.google.com/transla...0-jiz-za-par-tydnu&sl=cs&tl=en&hl=en&ie=UTF-8

Amazingly, double the ALUs of GT200, with a texture rate that's basically unchanged. Woah, 6 multiprocessors per cluster.

Jawed

The translation is pretty crappy - maybe because of the typing errors. Here are the key "facts" - I think my translation could be slightly more comprehensible:

Some(?) technical director (no name mentioned) told this to OBR (writer of this news) in Munich after a few beers. According to him, all information about problems with 40nm process and delay of the GeForce is bullshit. There will be enough GPUs for launch, however general availability won't be sufficient (and price will correspond to this fact) and situation will improve in time.

High end model should be based on 480 SPs, 512bit memory bus + 2GB GDDR5. 40nm, DX11, PhysX etc. It's still based on G80/GT200 architecture with some improvements, esp. for GP-GPU aplications. NVIO is separated (OBR was surprised by this fact...)

Gaming performance will be similar to overclocked GTX295, power requirements will be very low in 2D, otherwise similar to todays models. Launch: 1st October.
---

I won't comment it, everybody can make his own conclusion.
 
Last edited by a moderator:
OBR is a huge Nvidia fanboy over at XS...
He rarely has connections/sources but he isn't always completely BS, I would rate him lower than almost any other news site.
 
Isn't he the guy who always claims to have an unreleased card/CPU, promises to benchmark it and then it turns out it's all bs?
 
found at 3dcenter.de (posted from anarchX):

http://gathering.tweakers.net/forum/view_message/32310888

translated with google:

The Source wrote on Wednesday, July 22, 2009 @ 21:05:

Manhattan mobile:

Broadway XT DDR5 -> HD5870
Broadway Pro -> HD5850
Broadway LP -> HD5830
Madison Pro DDR5 -> HD5750
Madison DDR3 XT -> HD5730
Madison LP / Pro -> HD5650
Park XT -> HD5470
Park Pro -> HD5450
Park LP -> HD5430

____

Added some more info:

AMD cliamt Manhattan that better performance per watt and about 135 to 145% faster compared to the M9x series. 40nm, DX11 With Power Xpress, SG Switching, XGP support and pin to pin compatible with the M9x designs.

Broadway -> Ramp E / Aug, Oct MP
Madison -> Ramp E / Aug, Nov MP
Park -> Ramp E / Nov, Nov MP

Now we hope that 40nm can deliver what we expect

Broadway XT -> HD5870
GDDR5
45W-60W
M2

Broadway Pro -> HD5850
GDDR5
30W-40W
M2

Broadway LP -> HD5830
(G) DDR3
29W
128bit
M2

Madison Pro (or XT .. not sure) -> HD5750
GDDR5
20-30W
M2

Madison XT (or Pro .... not sure) -> HD5730
(G) DDR3/GDDR5
20-25W
M2

Madison LP / Pro -> HD5650
(G) DDR3
15-20W
128-bit
M2

Park XT -> HD5470
GDDR5
12-15W
S3/M2

Park Pro -> HD5450
(G) DDR3
10-12W
S3/M2

Park LP -> HD5430
(G) DDR3
<8W
S3
 
Wait, notebook parts in the spotlight that fast?


Unless they're trying to blow the 40nm nVidias away by countering paper with (thicker) paper. :LOL:
 
Now not only rumors of a full top to bottom Dx11 launch by AMD, but also a full top to bottom launch of Dx11 Mobile GPUs?

Yeah, this is starting to make me wonder if someone's been partaking of the crack pipe a little too much.

Regards,
SB
 
Back
Top