AMD: Volcanic Islands R1100/1200 (8***/9*** series) Speculation/ Rumour Thread

Hi Dave, can you gave us a hint why Tonga is so big? Tonga is just 79mm² smaller than Hawaii, with 1/16 DP and only a 256bit MI, Tonga should easily fit 44CUs/2816SPs like Hawaii.
Full DP is obviously not enabled on the radeon parts. wouldn't be surprised if it was 1/2 DP like hawaii since they are going to make firepros from these.
 
Full DP is obviously not enabled on the radeon parts. wouldn't be surprised if it was 1/2 DP like hawaii since they are going to make firepros from these.

As Dave said, Hawaii is the compute-focused chip, so probably not. There's usually only one AMD GPU with a high DP rate per generation.

44 CUs still sounds a bit too high for a chip of this size but the difference between Tonga and Hawaii (368mm² vs 438mm²) is surprisingly small. I wonder why. I doubt the new color compression technology can account for it, let alone a few extra instructions.

Perhaps the process is simply different? But if so, I'd wonder why the change was made, since it does not seem to yield any power-efficiency benefits. Maybe Tonga XT will clear things up somewhat.
 
As Dave said, Hawaii is the compute-focused chip, so probably not. There's usually only one AMD GPU with a high DP rate per generation.

44 CUs still sounds a bit too high for a chip of this size but the difference between Tonga and Hawaii (368mm² vs 438mm²) is surprisingly small. I wonder why. I doubt the new color compression technology can account for it, let alone a few extra instructions.

Perhaps the process is simply different? But if so, I'd wonder why the change was made, since it does not seem to yield any power-efficiency benefits. Maybe Tonga XT will clear things up somewhat.
It is pretty obvious there is more added to Tonga than there is in Tahiti, They increase the transistor count by 700 million even with the shrink of the DDR5 bus. Why wouldn't AMD make 2 compute capable chips for different segments? It would make sense from the increase that there is at least some increase in compute performance from Tahiti if the die is bigger and more dense.
 
As Dave said, Hawaii is the compute-focused chip, so probably not. There's usually only one AMD GPU with a high DP rate per generation.

44 CUs still sounds a bit too high for a chip of this size but the difference between Tonga and Hawaii (368mm² vs 438mm²) is surprisingly small. I wonder why. I doubt the new color compression technology can account for it, let alone a few extra instructions.

Perhaps the process is simply different? But if so, I'd wonder why the change was made, since it does not seem to yield any power-efficiency benefits. Maybe Tonga XT will clear things up somewhat.
It is pretty obvious there is more added to Tonga than there is in Tahiti, They increase the transistor count by 700 million even with the shrink of the DDR5 bus. Why wouldn't AMD make 2 compute capable chips for different segments? It would make sense from the increase that there is at least some increase in compute performance from Tahiti if the die is bigger and more dense.
 
It is pretty obvious there is more added to Tonga than there is in Tahiti, They increase the transistor count by 700 million even with the shrink of the DDR5 bus. Why wouldn't AMD make 2 compute capable chips for different segments? It would make sense from the increase that there is at least some increase in compute performance from Tahiti if the die is bigger and more dense.

There are more ACEs, but fast DP is very unlikely, because Hawaii adequately addresses the (pretty small) market that actually needs it.

In fact, Hawaii is probably a better point of reference for trying to figure out Tonga, because it's based on more similar IP, probably similar memory PHYs and a closer process. Yet somehow Hawaii is measurably denser, in spite of its 512-bit memory bus. Curious.
 
The bottom line for Tonga: it's a great time to buy a R9280 on the cheap :). Near the end of the 7950's life when stock was still good, I scored one for ~$200 with the Gold bundle. And I'd take a 3GB 280 over 2GB 285 any day.
 
The bottom line for Tonga: it's a great time to buy a R9280 on the cheap :). Near the end of the 7950's life when stock was still good, I scored one for ~$200 with the Gold bundle. And I'd take a 3GB 280 over 2GB 285 any day.

I have buy 2x 7970 when they have been released, and i still use them.. i dont have find yet how to replace them... I use a 1440p monitor and play everything ( i admit specially BF4 this last month ), all maxed.. I dont even worry for the games of the end of 2014...

My gpu's are coupled with a great 6cores Intel CPU SandyBridge-E, watercooled and overclocked, excellent SSD ( Samsung 840Evo 512GB for main system), a big bandwith of ram for the system ( 85000+ Mb/s on a 16GB area ).. it work like a charm for everything, from 3D creation to gaming. Lately i have find that i could still bench them at 1450+ mhz without glitches and problem. I could certainly still keep them 24/24 stable at 1250-1300+mhz who put them even not in shame in Tflops against 2x 780TI or 2x 290x in raw power ( Tflops ) ( including on DP side ) .
 
Last edited by a moderator:
http://videocardz.com/52302/amd-radeon-r9-390x-cooling-pictured The >500mm^2 chip seems to be getting hybrid cooling like R9 295X2

From that slide, it seems that Tonga is HPM as well? Wasn't this supposed to save some power? Or is that really only for leakage? And why are the power savings so minor?

The fact that they need liquid cooling, indicates that AMD still didn't do anything architecturally about the power consumption. This may be survivable for Tonga, but that's really going to bite them in terms for price for this one. Performance better be a big step up from Hawaii and even gm200. Is this the one that's supposed to have HBM? That should help...

Looks like AMD's plan is to combine a tick-tock in one go for 16nm.
 
AMD to release R9 390X to fight GeForce GTX 980 ?

Pictures of a cooler shroud are now showing in Chinese tech forums, it shows a design close to the Radeon R9 295X2, only designed for single-GPU. Specs of the R9 390X are not known to date, the product could even be based on the upcoming "Pirate Islands" family of GPUs. The R9 285 is already part of that family with Tonga. So, is this "Fiji." ??

... Hopefully this does not mean that AMD's Next Generation GPUs will be as famously hot running as the last. But it seems that AMD either were not willing to design a new reference cooler themselves or could not make a cooler powerful enough an quiet enough with an air based cooling solution.

It will be interesting to see what this means for Future aftermarket AMD GPUs from the likes of ASUS, MSI, etc as their custon GPU coolers will have to compete with a reference design which is watercooled. Then again this could means that aftermarket models could be cheaper than reference as their coolers may be cheaper to use than an AIO water cooler.
http://www.guru3d.com/news-story/amd-to-release-r9-390x-to-fight-geforce-gtx-980.html
 
The bad part is that Asetek press release suggests H1/15 release, kinda late to fight 980
 
Meh, until GPUs move to a new process, everything released will be completely uninteresting. It'll be better and faster sure - up to a point. But not very much faster or better. More than three years on the same process = extreme disappointment.
 
There aren't any games that demand new hardware anyway. For those people running surround, 3d or 4k they can SLI and Crossfire their way to victory with existing stuff.
 
There aren't any games that demand new hardware anyway. For those people running surround, 3d or 4k they can SLI and Crossfire their way to victory with existing stuff.

for me its all about the rift. which likely means 90FPS @ ~1440/1600P.


so if this GPU is over 500mmsq what do we recon 64 CU?
if water cooled we could assume 1+ ghz. that would be a pretty mighty chip. I wonder how many CU's tonga currently has disabled?
 
Back
Top