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 .
No idea about DDR3 but supposedly GDDR5 vs GDDR3 is only ~10% price increase.

Edit- elsence, I like your third option the best for Redwood. It was what I was speculating it to be quite awhile ago.

The problem i have with this option (8ROPs/480SPs) is that with many Q4 2009-Q1 2010 games the 8ROPs will be a problem. (even at 1280X1024)

Also it will have better SP/ROP ratio than even high-end 5870 which is kinda strange (but not impossible, HD2000 & HD3000 era...) What performance in the majority of games can the design extract from 480SP with only 8 ROPs?

Also option 2 (16ROPs DX11 RV730) with same die size will be faster overall imo.

In Redwood option 1 (DX11 RV730) the ROPs are also 8, but it will also be cheaper than option 3.

Like i said i like option 1 probability because its easier for ATI take the RV730 and make it DX11 (easier on the hardware side & on the software side)

That's very nice of you :) But I know you were moving the planchette...

Ahh, you got me.

That's very nice of you :)
I agree. I think this is the best way to get optimal performance/mm² while keeping board costs at a reasonable level, though it would help to know if a 64-bit board with GDDR5 costs more or less than a 128-bit board with DDR3.



Regarding 128bit DDR3 vs 64bit GDDR5 options i am split.
Regarding GDDR5, i have the following issues:

1.If you see the 5750 512MB (109$) PCB you will see a lot of staff related to GDDR
5 power states etc. so i don't know what cost GDDR5 will add to the overall cost.

2.I think that the 128bit vs 64bit PCB related cost is not that great (eg. low cost Nvidia 9400GT based cards are 128bit)

3.DDR3 prices finally is cheaper than DDR2 and probably in Q1 2010 the trend will, be the same.

4.DDR3 power consumption is excellent and this will help some HTPC designs.

5.with 64bit bus and with current GDDR5 ICs the maximum memory will be 512MB, while DDR3 can easily be 1GB (mostly for marketing reasons...)
 
Last edited by a moderator:
After I saw the slide that compares 'core' count I laughed hysterically for 30 seconds.

Well, the last page does say "The information presented in this document is for informational purposes only and may contain technical inaccuracies, omissions and typographical errors."
 
http://www.semiaccurate.com/2009/10/06/nvidia-will-crater-gtx260-and-gtx275-prices-soon/

The numbers in this (about 2/3rds down) are *SLIGHTLY* blurred from real volume prices. Getting hard number on this stuff is hard, but you can if you try. Basically 10% more from GDDR3 @~1GHz to GDDR5@1GHz. Going to GDDR5@1.25 adds about another 10%.

The question to ask is how much PCB cost, and pin/package costs does it save by going from 3 to 5? Low bin GDDR5 has more or less the same bandwidth (1) as GDDR3 at the top bin with double the bit width. That is significant, especially on a small die.

-Charlie

(1) Unless your GDDR5 controller is totally hosed. No names. In totally unrelated news, did you see Fudo posted a story ~2 weeks ago about the cool new feature of Fermi? It supports GDDR3! I wonder why? 384 bit GDDR3 feeding 512 shaders is how much more than 512 bit GDDR3 feeding 240 shaders? My calculator must be broken, it keeps putting a dash in front of the results.

I love how every other post you make at B3D starts or ends with a link to your website and to your articles. You really think highly of yourself, don't you :D

Yeah, NVIDIA is supporting GDDR3 on Fermi because GDDR5 controller is hosed. That's logical. ;)
 
I didn't notice that, where does it say that? That means that T cannot issue FMA, only MAD, I guess.

Jawed

It's on slide 22 - pick your poison: [H] or [pcgh].

So, that'd also mean, that they can optionally issue a traditional MAD over all 5 ALUs with no penalty or alternatively a concurrent 320 MADs on the T-units - otherwise all the glory of 2,720 GFLOPS of glory would be bust and making 'em sueable. And that will not be the case.
 
Last edited by a moderator:
...
So, that'd also mean, that they can optionally issue a traditional MAD over all 5 ALUs with no penalty or alternatively a concurrent 320 MADs on the T-units - otherwise all the glory of 2,720 GFLOPS of glory would be bust and making 'em sueable. And that will not be the case.

Interesting enough, all the numbers are in favour of ATI. Is it correct?
 
It's selective counting, but yes, essentially if you agree on this kind of counting and terminology (which Nvidias uses in their own kind also), then it's correct, AFAIK.
 
It's selective counting, but yes, essentially if you agree on this kind of counting and terminology (which Nvidias uses in their own kind also), then it's correct, AFAIK.
Well, without the clock rate, that sure is selective. At least the DP numbers (except ADD) are not in AMD's favor if you factor that in :). Even AMD acknowledges Fermi should be quite a beast for HPC apparently...
One thing I noticed, there are lots of ??? when some information isn't known yet, however AMD seems to be sure Fermi can't drive 3 displays. I don't think I've seen any information about display controller capability yet of this chip, so I wonder what makes AMD so sure about it...
 
I don't think I've seen any information about display controller capability yet of this chip, so I wonder what makes AMD so sure about it...

You only need one leaky board partner. but hey.. you're talking about a card that nvidia says is going to launch in six weeks.. they should have samples out by the bucket-loads already ;)
 
Not too R800 related, but anyway...

http://news.mydrivers.com/1/146/146585.htm

Chinese, selftranslate.

Says that Apple bought 2 months' volume of RV770s, majority of them being specced as Radeon 4850s. Not a design win decided mostly by Apple but rather through AMD's initiatives (low price spamming?)

iMac refresh err whut


Clearing out inventory in advance of EOLing the RV770. AMD probably want to do that ASAP if they get better margins on 58xx.
 
Funny though- Apple considers RV770s to be adequate enough for a long period of baseline OpenCL. So much for commitment for Mac users touting SL's OpenCL loveyness here and there.

(Just wondering if the LDS will be another barrier for devs in general)
 
5890: A real video card or not?

Is the 5890 a real video card or was this an attempt at fud to keep people from buying 5870 (or something else)?
 
Is the 5890 a real video card or was this an attempt at fud to keep people from buying 5870 (or something else)?

Has AMD announced it?
Is that a good enough answer?

On the other hand, there is more than enough potential to release an overclocked part (with more memory?) that caters the gap between Cypress and Hemlock and would be a direct competitor to Fermi.. but hey.. that's my $0.02
 
Has AMD announced it?
Is that a good enough answer?

On the other hand, there is more than enough potential to release an overclocked part (with more memory?) that caters the gap between Cypress and Hemlock and would be a direct competitor to Fermi.. but hey.. that's my $0.02

A question can never be interpreted as an answer.
I am curious if the rumors are true or not on this issue.
 
A question can never be interpreted as an answer.
I am curious if the rumors are true or not on this issue.

His question was the answer. AMD hasn't made an annoucement. Therefore, it's up to you to determine whether you think the rumor is true or not.

All you're going to find out there are various rumors and speculations on rumors.

Regards,
SB
 
Back
Top