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 .
bandwith makes dramatic differences in performance, so that's fine, they can do whatever. I appreciate that the model numbers don't lie.

How many people know that a geforce 9500GT gddr3 is much faster than a GT220 ddr2, and that a GT220 gddr3 is much faster than a 9500GT ddr2?
 
Here's a review of a Juniper CE HD5670: http://www.inpai.com.cn/doc/hard/124320.htm

Based on these numbers (except for Vantage), I'd say a HD5670 name is justified :(
If I'd have to guess I'd suspect there are only 8 ROPs but there are no fillrate tests (the one from Vantage would be insufficient anyway) to say anything about rops unfortunately (nor is there anything to confirm 640SP as far as I can tell, Vantage Perlin Noise would do but the feature tests weren't run).
 
Here's a review of a Juniper CE HD5670: http://www.inpai.com.cn/doc/hard/124320.htm

Based on these numbers (except for Vantage), I'd say a HD5670 name is justified :(
If I'd have to guess I'd suspect there are only 8 ROPs but there are no fillrate tests (the one from Vantage would be insufficient anyway) to say anything about rops unfortunately (nor is there anything to confirm 640SP as far as I can tell, Vantage Perlin Noise would do but the feature tests weren't run).

Even so, I still don't understand why except perhaps its an extreme salvage part? I.E. A way to use ~100k or so Juniper dies which would have otherwise been scrapped.
 
Here's a review of a Juniper CE HD5670: http://www.inpai.com.cn/doc/hard/124320.htm

Based on these numbers (except for Vantage), I'd say a HD5670 name is justified :(
If I'd have to guess I'd suspect there are only 8 ROPs but there are no fillrate tests (the one from Vantage would be insufficient anyway) to say anything about rops unfortunately (nor is there anything to confirm 640SP as far as I can tell, Vantage Perlin Noise would do but the feature tests weren't run).
I think 8 ROPs would explain these results. It could be the same BW situation as with HD5830...
 
Not necessarily, that kind of volume is only possible on the lower end side of business and for those people it wouldn't have made sense to wait for a 500mm²+ chip's reviews.
 
Better photo-equipment could help much...

but it seems there's a central section (red - crossbar?), 4 columns / 10 rows of square blocks (green - SPs?) x2 and some slighly different square block at the end of each row (blue - TMUs?).

 
I'd say the cores should be interpreted at 90 degrees rotation.

The block of squares on each side of the red area is 5x8. That looks to me like 5x4 x2, with the cores back-to-back. So above the red area there are two sets of 5 cores, making one shader engine of 10 cores. Below the red is the same again.

So that would put the TMUs at the far left and far right.

The red needs to be extended leftwards, I reckon, to match how it is on the right.
 
So, about 37% of the die is the shader cores, I reckon (122mm²). Admittedly that's pretty woolly and seems a bit low (e.g. I was thinking it might be 150mm²).
 
Back
Top