AMD: R9xx Speculation

If AMD are going to drop the price to match the performance levels, then the name doesn't really matter. It just means that people who would have bought 6800 now have to buy 6900 as the range gets shuffled downwards through the price ranges.

As long as they are not doing the opposite and shuffling lower products upwards to sell us lower performing parts at higher prices. The prices of the 5800s are already over the top.
 
gongo said:
Twice as much pressure meaning....fan has to spin twice as fast... or half the volume of air possible at twice the fan rpm...i think it makes a different..

Lol @ the 3DVantage score...its been 2 years and AMD still cannot deliver a gpu twice as fast as 4870 1GB ...in perf/$...the days of rapid gpu evolution are gone...i loved the times when it went from ps/vs to unified and then hundreds of sp..it made the just launched consoles look dated already....
Click to expand...

Hmm.. if you look at the Extreme score, the stock 5870 is almost twice as fast than the stock 4890. And this 6870, while not being the fastest GPU of the new ATI family, is more than 50% faster in this regard (and a little faster than the 5850, too). But more interesting than the 3Dmark scores would have been some performance in real games, as 3Dmark rarely represent a valid point of reference for other applications.
 
gongo said:
Twice as much pressure meaning....fan has to spin twice as fast... or half the volume of air possible at twice the fan rpm...i think it makes a different..
Click to expand...

If fluids were so simple, there would be so CFD simulations.

I suggest you inform yourself a little before taking such a negative stance on something you don't understand. Just look up what the Reynolds number means and what the value is for air.
 
UniversalTruth said:
Yep, only one, better than nothing. :LOL: I think I am missing this generation.
Click to expand...
Actualy, that one positive thing, if true, might be enough for me to get it. :)

While the next generation on 28nm should deliver, seeing the problems of 40nm, expecting them to be magnified for 28nm, I wouldn't count on next gen arriving that soon. If they can get it for Xmas next year, it would be a success in my book. Therefore I think the current generation should be a safe bet for most. :)
 
I'm hoping Vantage Extreme scores are less CPU-bound (though apparently newer drivers help a lot):

Code: 
       [URL="http://www.overclockersclub.com/reviews/sapphire_hd5850/16.htm"]OCClub[/url]   [URL="http://techgage.com/article/sapphire_radeon_hd_5850_toxic/10"]TechGage1[/url]  [URL="http://techgage.com/article/nvidia_geforce_gtx_480_-_gf100_has_landed/12"]"2[/URL]     [URL="http://www.xbitlabs.com/articles/video/display/asus-matrix-5870_11.html#sect1"]Xbit[/URL]    PConline
       i7-920   i7-975     "      "       i7-940
        @3.00+   @3.33      "      "       @2.93
        9.9      9.10       10.3   10.7    ???
5870:  8246     8213       [b]9041   8870[/b]    
5850:  6590     6566
5770:           4311
480:                       9237   9828
470:                              7844
6870:                                     [b]7538[/b]
6850:                                     [b]6549[/b]

6870 seems ballpark 5850+.
 
Bouncing Zabaglione Bros. said:
If AMD are going to drop the price to match the performance levels, then the name doesn't really matter. It just means that people who would have bought 6800 now have to buy 6900 as the range gets shuffled downwards through the price ranges.

As long as they are not doing the opposite and shuffling lower products upwards to sell us lower performing parts at higher prices. The prices of the 5800s are already over the top.
Click to expand...

Still much less that what was in the past. I remember times were bad-performing mid-range card were sold for tons of money.. for example 8600 GTS, 32sp and sold for 200$+
 
Code: 
card       X-Score        GPU-clock    SPs    GFLOPs    GB/s

6850        X6549        775MHz        800    1208    128

6870         X7538         900MHz        960    1728    134

difference    [B]+15,1%[/B]        +16,1%        +20%    [B]+43%[/B]    +4%

So there is something wrong with specs/benchmark-scores or Barts is still front-end-limited.
 
AnarchX said:
Code: 
card       X-Score        GPU-clock    SPs    GFLOPs    GB/s

6850        X6549        775MHz        800    1208    128

6870         X7538         900MHz        960    1728    134

difference    [B]+15,1%[/B]        +16,1%        +20%    [B]+43%[/B]    +4%

So there is something wrong with specs/benchmark-scores or Barts is still front-end-limited.
Click to expand...


I noticed that. It's quite a big difference between the real difference and the expected difference And it cannot be bandwidth limited, i mean fasters ram chips are around, why don't use them?
Maybe they are both 800sp.
 
I'm dubious about 800/960. While it's compatible with 80 ALU lanes per SIMD, it's not compatible with 64 ALU lanes per SIMD (which is one configuration for VLIW-4).

Also, some guy with a pair of cards in his hands (the card shown doesn't match any existing HD5xxx card I can find) and nothing else has no way to tell the count of lanes. He needs documentation, and he hasn't shown any.

But of course VLIW-4 isn't confirmed.
 
The original poster for that thread says that the Barts Pro card has a single PCI Express power socket on the back.

The card is quite a bit longer though. Some of that will be due to the doubling of memory chips, I presume. But I imagine the power regulation is beefier too. Not sure what effect on the length there'd be due to moving the power stuff to the other end of the board (between GPU and backplane). Actually, I suspect the power stuff is split in two: one section for the GPU and another section (remaining at the rear of the card) for memory.
 
Well I actually thought VLIW-4 was an even less reliable rumour than 960. /shrug

Seeing these numbers I wonder what AMD has planned for $100-$150. Will 256-bit chips/Barts now fall into this category or will they be able to improve on Juniper while staying at 128? If these rumoured numbers are close to the truth it would seem the performance gain from Juniper->Barts is all bandwidth based. (at least in 3dmark).
 
gongo said:
Twice as much pressure meaning....fan has to spin twice as fast... or half the volume of air possible at twice the fan rpm...i think it makes a different..
Click to expand...
The vents do not alter the pressure, what is important is the effectivness of the thermals inside at getting the heat off the ASIC. Cypress has lower operating temps and a less agressive fan table than either RV770 XT or RV790 yet it has a higher TDP and the stacked DVI. The slots at the back don't make much difference with regards to the "pressure".
 
nyt said:
90% of buyers don't know/don't care. For those that do, they buy perf/watt or $, and couldn't care less which name the card has. I know I won't loose sleep over what 5770 replacement (read more perf/$) will be named, I just care about the end result (fps at 8xAA on my 22"). I really don't get all the fuss over all this naming thing. FX5200 sold well although it was worthless at DX9. G92 got rebranded many times and was a cash cow for NV. History teaches us one thing, marketers from both IHVs know better than B3D forum fillers :) Can we now wait for NDA expiry or some hard facts? Please?
Click to expand...

I think you missed the part of my post where I made it clear that I specifically targeted the naming of mobile solutions.

And that there was a bit of logical reasoning as well: Given that parts are renamed when mobile to masquerade as higher end desktop parts, this will serve to drive unaware customers to buy "down" as well as "up", since not everyone buys the top part - indeed, by all accounts the volume is further down. So it is an open question whether the naming practise actually drives revenues at all. What is definitely clear though is that anyone who finds out they got less than they thought is also going to be less than happy once they find out. And that a fair part of the people who are informed enough to be aware what is going on find it less than impressive.

So - is there a gain in revenue at all, and if so, is it sufficient to offset the loss of goodwill?
That was what I questioned. And as far as I can see, it's a valid question. I don't pretend to know the answer.
 
no-X said:
I'm talking about ATi's margins. I mean, that with ATi's market share and mentioned results, they can't have high margins.
Click to expand...

You mean AMD. There hasn't been an ATI company for years.
 
You must log in or register to reply here.

Similar threads

F
RDNA5 Speculation
Replies
2
Views
1K
Erinyes
E
N
AMD: Volcanic Islands R1100/1200 (8***/9*** series) Speculation/ Rumour Thread
141 142 143
Replies
3K
Views
675K
pMax
pMax
S
AMD: Sea Islands R1100 (8*** series) Speculation/ Rumour Thread
69 70 71
Replies
1K
Views
389K
AnarchX
A
A
  • Locked
News & Rumors: Xbox One (codename Durango)
613 614 615
Replies
12K
Views
2M
BRiT
BRiT
Unknown Soldier
AMD Talks ATI Radeon 3000-Series
Replies
6
Views
6K
Silent_Buddha
S
Back
Top