Saturday, July 2nd 2011
AMD Southern Islands GPU Codenames Surface
It was discovered that an early version Catalyst 11.7 contains entries for new GPUs, spelling out the possible codenames of upcoming AMD Southern Islands GPUs. Southern Islands is the next generation of AMD Radeon GPUs, which will possibly see propagation of VLIW4 SIMD design to GPUs in all market segments, and possible adoption of 28 nm manufacturing process. Moving on to codenames, 3D Center observes that the top of the line GPU is codenamed "Tahiti", there is a possibility of two single-GPU variants.
A graphics card that makes use of two Tahiti chips is codenamed "New Zealand". "Thames" is the codenamed of a possible low cost mobile chip, it could also be a performance chip. There's some uncertainty here. Not much is known about another chip codenamed "Lombok", either. Perhaps it's a performance desktop chip.
Source:
3D Center
A graphics card that makes use of two Tahiti chips is codenamed "New Zealand". "Thames" is the codenamed of a possible low cost mobile chip, it could also be a performance chip. There's some uncertainty here. Not much is known about another chip codenamed "Lombok", either. Perhaps it's a performance desktop chip.
71 Comments on AMD Southern Islands GPU Codenames Surface
I hope they are able to squeeze in as many ALU as the famous 5870 (1600 cores) and run it at 1+ghz.
I suspect that the chip will be large and use 28nm well but that its performance isn't going to be that big of a leap. But then we haven't really had big leaps in many years.
I find it strange that they moved to VLIW4 at all with this new design coming to replace it.
also i saws a doubling in perfprmance going from xfired 3870's to my 1x 5870
and 1x4870 equalled my 2x3870's so why your so negative is beyond me in reallity the 6xxx series was i bit weak an upgrade to the 5xxx series (same core similar shaders) imho and hence no card doubled the 5870's power excepct dual gpuagreed afaik it was down to die space on 40nm and prob had some relevance to the lano APU design as 400 4xshaders would take up less space in a APU then 400 5x shaders;) and what with the stars cores taking more room then bulldozer imho it makes some sense as its ALL about size and die area
Also, 5870 is more like only 30-50% faster than 4870. 3870 was a rather poor GPU and 4870 was a big improvement over it however. 3870 occasionally gets slapped around by even 8800GT for some reason. AMD's R600 and RV670 chips had problems with efficiency.
40nm had nothing to do with the switch to VLIW4. Cayman was originally to be 28nm but the delay to 28nm caused them to go back and set it up for 40nm. The chip is the same thing as it would have been on 28nm though. It takes them several years to design GPUs so they can't make major changes in a short period of time.
the 7xxx series is what the 6xxx series would have been had 28nm worked, what they released whas a revision of the core from 5xxx series and some regular un risky shaders and i agree it takes years to dev gpu's but their process fell over late in the day and they backtracked fast simples they needed 28nm to fit southern islands plus the old shader type(what went in anyway) but didnt have the room for the new core on 40 nm given the shader count increase needed to counter nvidia
7xxx new not revised southern islands core and new shader array
all im saying is im optomistic AMD will get it right and imho they playing a wise game due to knights corner + nvdia's project denver APU(what they gona call that APU prob NPU)
and i dont think they will choose to harm graphics performance just to better gpu performance they will still need benchmark scores to sell their gfx cards after all:D
A) The AMD A64 Architecture was the first with IMC ,HYPER TRANSPORT and 64bit !
B) The K10 & K10,5 Architecture waw the first with IMC, NATIVE QUAD CORE and L3 CACHE !!
C) The FIRST REAL CHIP with CPU+IGP in ballance computing is Llano.
--------------------------------------------------------------------------------------------------
Intel imitated all these architectures many years later !!!!
Now AMD want to make the first real cp-gpu = TRINITY !!!
For more information, you can see here:
www.pcper.com/reviews/Graphics-Cards/AMD-Fusion-System-Architecture-Overview-Southern-Isle-GPUs-and-Beyond
techreport.com/discussions.x/21141
i half thought i was in a BD thread there a min sorry
in these slides
wccftech.com/amd-slides-detail-upcoming-radeon-hd-79-series-gpu-architecture/21471/
it looks like their grouping simd units into fours and calling them compute units so i spose my top question is how MANY 320 512 or 640 are my clear faves making for a poss shader count total of 1280 2048 or 2560 but they also hint at multple primitive paths which imho seems like multiple schedulers etc ie more then one gpu (though probably cut down parts thereof) in a modular array making 64 cu per core x 4 cores my prediction making for 1024 shaders, itd make sense, though 8x cores with 64 CU units each makes more with chip binning in mind( ps4, xboxwoteva)
any news on when 7xxx series is due
FYI that dosnt even makes sense
its getting worse this thread, i cane get wot yo be is saying (.)(.)
every company has their own fields, from my point of view AMD is still the best when focus on balancing performance between CPU and GPU, when intel is specialize focusing on CPU, and nvidia are good when focusing on GPU,,
eventhough i had build majority of my cpu's with AMD, i gotta to admit that the fastest CPU is still on Intel's hand..