New rumor about ATI Southern Islands

[FischOderAal]

to say the truth amd picked nice season for introducing new line. if everything goes right they again will easily take major piece of christmas sales

There was an article on Anandtech a while back and it described how AMD had changed (and what made HD4800 such a success). Instead of thinking of new features their card might have, implement all of them and launch the GPU whenever it's ready, they scrap all the ideas they have that don't fit in the GPU because of the time needed or DIE-space. One of the main goals is to have a new product ready to holiday and when the OEMs plan their new products.

Here's the story. Very good read.
http://www.anandtech.com/show/2679

[mindfury]

http://news.ati-forum.de/index.p ... -der-hd6000er-serie

Juniper(HD5700) rebrand as HD6700
Barts=HD6800
Antilles=HD6990

[madcho]

http://news.ati-forum.de/index.p ... -der-hd6000er-serie

Juniper(HD5700) rebrand as HD6700
Barts=HD6800
Antilles=HD6990

bad link :

http://news.ati-forum.de/index.php/n...mit-rebranding

[Florinmocanu]

The new gpus will be tweaks of the previous generation, not just a re-name. This is not the 8800GTS 512 transforming into 9800GTX, it's a refresh, nothing more.

AMD said it as well, they will refresh their line-up at the end of 2010.

[Pantsu]

Sounds ridiculous to shift all the names if they want to rebrand one mainstream card.

[RaZz!]

Sounds ridiculous to shift all the names if they want to rebrand one mainstream card.

idd. i've always condemned nvidia for their rebranding and if amd starts to do the same they'll lose some plusses in my book as well.
rebranding simply sux, as it's just a cheap attempt to make the average joe think it's some new product.

however, renaming the 6870 to 6970, and 6970 to 6990 doesn't make sense, imo. they'll completely break everyone's association with the naming where x870 = single gpu highend, x970 = dual gpu highend etc... bad idea if you ask me.

i don't think this is going to happen. atleast i hope so

[Sn0wm@n]

loll no concrete info from ati as of now ... so i still hold faith that they will improve the 5K series design and change some things and not do a simple rebrand

[Calmatory]

I'd bet my money on 4+1 -> 4 transition. The rebranding would mean same amount of SPs while actually number of clusters have increased.

[geo]

http://news.ati-forum.de/index.p ... -der-hd6000er-serie

Juniper(HD5700) rebrand as HD6700
Barts=HD6800
Antilles=HD6990

ATI is the new Nvidia

[Sn0wm@n]

I'd bet my money on 4+1 -> 4 transition. The rebranding would mean same amount of SPs while actually number of clusters have increased.

so same numbers of shaders but how they are arranged would be changed ???? and can this improve tesselation and average performance ????

[Florinmocanu]

Well, because a 5 shader array is like a quad-core CPU. It's easier to fill a Dual-core than a quad-core or an octo-core.

So a 4 shader array will be fed easier with data than a 5 shader array. Thus, improving efficiency per array.

And, if you go with 4 shaders array, you have space for more arrays. Keeping the final shader number the same as previous GPUs, but with higher efficiency.

Some people mentioned something about the possibility of having 2 double precision capable shaders + 2 single precision shaders in such a 4 shader array. so a 2:2 ratio instead of the present 4:1. Would improve computational capabilities for sure, but nothing official from AMD.

[Dami3n]

http://news.ati-forum.de/index.p ... -der-hd6000er-serie

Juniper(HD5700) rebrand as HD6700
Barts=HD6800
Antilles=HD6990

bad link :

http://news.ati-forum.de/index.php/n...mit-rebranding

Something is wrong with that info. Where is Cayman?

[Calmatory]

so same numbers of shaders but how they are arranged would be changed ???? and can this improve tesselation and average performance ????

Right now the cores consist of 800/5(4+1) SIMD processors(160). With the 4+1 -> rumour the cores would consist of 800/4 SIMD processors(200). That's a 25 % increase while the SP count remains the same.

The catch here is that with 4+1 configuration there is one SP more or less unused most of the time, 4 SP configuration eliminates this and allows denser packing of the SIMD processors. So they can add more SIMD processors and thus bring the SP count to 800 while working with same die area limitation.

I am very doubtful towards this 4+1->4 theory though, I'd take it as 4+1 -> 3+1. Well, still effectively cutting number of ALUs per SIMD processor from 5 to 4. But the main idea is that there's hardly ever such cases that the instruction level parallelism would be so high that all the ALUs would be busy, so cutting one off isn't a big deal and only gets costy in very rare cases.

[Sn0wm@n]

Well, because a 5 shader array is like a quad-core CPU. It's easier to fill a Dual-core than a quad-core or an octo-core.

So a 4 shader array will be fed easier with data than a 5 shader array. Thus, improving efficiency per array.

And, if you go with 4 shaders array, you have space for more arrays. Keeping the final shader number the same as previous GPUs, but with higher efficiency.

Some people mentioned something about the possibility of having 2 double precision capable shaders + 2 single precision shaders in such a 4 shader array. so a 2:2 ratio instead of the present 4:1. Would improve computational capabilities for sure, but nothing official from AMD.

makes sense ... thanks for the explaination

[mindfury]

Some people mentioned something about the possibility of having 2 double precision capable shaders + 2 single precision shaders in such a 4 shader array. so a 2:2 ratio instead of the present 4:1. Would improve computational capabilities for sure, but nothing official from AMD.

Apparently you dont know how ATI shaders work...

http://www.beyond3d.com/content/reviews/52/9

The structure of a packed VLIW passed for execution by a discrete block is made up of between 1 and 5 64-bit scalar ALU operations and at most 2 64-bit literal constants for a grand total of 7 64-bit words length. Control flow instructions are passed as separate 64-bit words assigned to the branch hardware. The 4 equal ALUs can handle 1 FP MAD/ADD/MUL/dot product per clock, or 1 INT ADD/AND/CMP/LSH*_INT (but not MUL!) per clock, to list just a few instructions.

For single precision floats, MUL and ADD are ½ ulp IEEE using round-to-nearest-even rounding, and MAD is 1 ULP with the same rounding mode. For double precision, as already mentioned, these ALUs are fused in pairs of two to compute 1 DP MAD per cycle across all 4 of them. Only a limited set of instructions are supported in DP (no transcendentals being the obvious omission), and compliance with IEEE754 is relative: denorms are flushed to 0, only round-to-nearest rounding is supported, and a MAD produces different results from MUL+ADD due to rounding. Integer and float instructions can't be processed in parallel.

The transcendental unit (the Rys unit!) is different from its more silhouette conscious brethren: it's (surprisingly!!!) capable of handling transcendentals (cos, sin, log, exp, rcp et al.) at a rate of 1/cycle, INT MUL, due to a slightly higher internal precision (40-bit versus 32-bit, allowing expression of a 32-bit int in the FP exponent) than the other ALUs, and format conversions, all whilst not being able to process dot products or double precision work (so it's idle when double precision processing is happening).

[RaZz!]

Something is wrong with that info. Where is Cayman?

huh? it's there, just open the link


http://news.ati-forum.de/index.php/n...mit-rebranding

[Mechanical Man]

huh? it's there, just open the link


http://news.ati-forum.de/index.php/n...mit-rebranding

I have say, that i cant find any reason for such a rebrand. That would place rebrand 6870 ~ 5870. What would be point in it?

[RPGWiZaRD]

Well perhaps it has to do with surprisingly good performance to come so AMD thought they'd brand them a little higher or so they can more easily price them higher (say $499 for a Cayman XT)? $499 for a HD 6870 might look a lot but if you get a 6970 it looks better, for the knowledgable people it's just a number but for people who don't have a clue about the graphics cards it could have a point when comparing to HD 5870 and 5970.

[Dami3n]

huh? it's there, just open the link


http://news.ati-forum.de/index.php/n...mit-rebranding

laziness effect.
I don´t think that AMD wants to repeat the error of their enemy. Would be ridiculous to see that crappy movements in the red camp too.

[RPGWiZaRD]

laziness effect.
I don´t think that AMD wants to repeat the error of their enemy. Would be ridiculous to see that crappy movements in the red camp too.

Yea I think so too, I hope they don't use this branding, they just start to remind too much of Nviditactics. It really seems to be a bad thing in the long run being at the top and controlling the market which I think they are doing with DX11 generation so far, the companies get only more and more greedy being at the top and start using silly marketing tactics to squeesh even more cash by exploiting people who don't know any better. It happened to Nvidia before, now it could start happening to AMD too in the graphics department with this possible rebranding and rebranding what was meant to be 5870 successor into a higher number etc. But we should take all this with a bit grain of salt as usual. :p

[Mechanical Man]

Still, true or not I hope JF sends certain message to gf part of marketing department. The message being "DONT UP WITH REBRAND". He might not know of their intentions, but that does not mean he could not give feedback from ethusiast part of us.

[Sn0wm@n]

can we stop talking about the rebranding until we have a 100% fact proof of this ..... amd cant afford to make those silly move since they are barely on top on the gpu market ... but still amd as a whole isnt on top of anything ..... so they really need to dominate or at least keep pushing for domination ....

[Calmatory]

I really doubt they are rebranding anything. They're just improving the old design, and the specs are seemingly the same by the numbers, not under the hood.

...so worry not.

[SimBy]

I think AMD is either a) screwing with everyone leaking bogus info b) planting bogus info internally to find out who leaks it

It could also be that leaked information was misunderstood.

I doubt we will see any nVidia style rebranding.

[Sn0wm@n]

id go with calmatory's post .... it might be as simple as a workaround of the current arch so the numbers would most likely look the same on papers thus creating the rebranding rumors ....