AMD: Southern Islands (7*** series) Speculation/ Rumour Thread

Does STREAM SDK ring a bell?
Interesting you should mention this. When you google for it and end up on their clusterfsck of a website, you'd initially be mistaken to think that it still exists. Only after a click-through, you end up on a different site, designed by what must be a descendent from Stalin, where you can find out that it's now called 'AMD APP SDK'. So that's CTM, Stream and APP, but maybe I've lost count? So much for a consistent marketing message. Way smart, BTW, to use the super generic term 'app' to make your product stand out in search engines. CUDA may sound dumb, but there's absolutely no ambiguity where it will lead you searching for something.

Let's search for some kind of information. Say, gpu to gpu communication. For CUDA, you google "cuda gpu to gpu communication" and get a ton of useful links. Try 'ati stream gpu to gpu communication' or 'amd app gpu to gpu communication' and you basically come back empty handed.

After quite a bit of digging, you discover that they have a profile and a kernel analyzer and even a couple of libraries. But it pales compared to the competition. There's minimal 3rd party support. No advanced deployment support. The highest rated book on Amazon is sold out. The one section in the AMD forums on OpenCL is anemic.

One can go on and on. But unless your app is perfectly suited to unlock the higher number of flops of AMD GPUs or you're mostly interested in theoretical benchmarks, you must be a very bold and courageous person ( (c) Sir Humprey Appleby) to use AMD APP.
 
So that's CTM, Stream and APP, but maybe I've lost count? So much for a consistent marketing message
CTM was never a brand, it was the first vestages of the machanism for getting to the hardware. The only brands have been Stream and APP. Given the point at which it Stream started and the AMD/Radeon/Fusion strategy is evolving over the interviening period a brand change is not really to be unexpected.

Way smart, BTW, to use the super generic term 'app' to make your product stand out in search engines. CUDA may sound dumb, but there's absolutely no ambiguity where it will lead you searching for something.
Depends on what you are tyring to do. CUDA/Stream may mean something the the techies and the developers, but very little to the common user. APP/Apps mean more to the common user.
 
Dave Baumann said:
APP/Apps mean more to the common user.
If the goal was to come up with a name that was as bland and generic as possible, why not call it 'thingie' ?

But it is true that the naming part should be the least of your worries...
 
Depends on what you are tyring to do. CUDA/Stream may mean something the the techies and the developers, but very little to the common user. APP/Apps mean more to the common user.
You can see CUDA logo badge sticker on almost every retail box with GeForce card, witch is a consumer grade product. ;)
 
APP/Apps mean more to the common user.
If it means anything to common users it's total confusion, since AMD APP and what common folks know as an app aren't even remotely the same thing.

It's a terrible name from start to finish, whomever came up with it needs a whuppin', and the person who cleared it as a product name needs firing.

...Just my personal opionion, of course. :)
 
Indeed, but ask your average Best Buy customer what it means...

Surely at some point they'll need to be taught what it means. How else are you going to show people who've probably never bought an AMD CPU in their life that their programs are going to run better on a Llano chip vs something from Intel?
 
CTM was never a brand, it was the first vestages of the machanism for getting to the hardware. The only brands have been Stream and APP. Given the point at which it Stream started and the AMD/Radeon/Fusion strategy is evolving over the interviening period a brand change is not really to be unexpected.


Depends on what you are tyring to do. CUDA/Stream may mean something the the techies and the developers, but very little to the common user. APP/Apps mean more to the common user.

I don't know. hardly anyone has heard of Stream (except maybe the Firestream cards) ; and APP is a random three letter acronym.
OpenCL has more recognition, even though it's not much used.
though I agree that a naming such as "accelerated applications" is useful to the user.

trouble is, the common user thinks their computer is a "Vista" one or a "windows 7" one, and has no clue about what an athlon II or an E-350 is and what's the difference with a core2duo, etc.
 
APP/Apps mean more to the common user.

Actually it means very little to the regular consumer and is at best misleading. AMD still has work to do in the marketing dept. "App" is a pretty generic term and communicates nothing about the product. Maybe there's a marketing campaign out there but I've never seen or heard of "APP" before this thread.

Is there anything you can tell us about where AMD thinks the GPU compute market is heading and what things require more attention in the future?
 
trouble is, the common user thinks their computer is a "Vista" one or a "windows 7" one, and has no clue about what an athlon II or an E-350 is and what's the difference with a core2duo, etc.
Absolutely - most users only care about what a computer can do for the money that they have available to spend, so spending time educating them what "Stream" or "Compute Unified Device Architecture" is probably somewhat wasted. It’s more important to give them an understanding on the capabilities and experiences a platform can bring.

And if you ask that same guy what "app" means, he'll either say he doesn't know, or that it means "application". But APP is not really an application now, is it?
At the most basic level APP accelerates applications.

Dave, can you tell us something about the plans regarding gdebugger?
I am somewhat removed from this; there is a rather large team under Manju that has ownership of all aspects of this.

Is there anything you can tell us about where AMD thinks the GPU compute market is heading and what things require more attention in the future?
Heterogeneous compute is a fundamental component to making the fusion strategy work from a consumer/client perspective. Other than that, I would suggest following the Fusion Developer Conference very closely.
 
According to Charlie, SI taped out in February and more details will be unveiled very soon, at AMD's Fusion Developer Summit.

I sure hope all that is true, because I may have been a little bit out of touch lately, but I've heard exactly nothing about Southern Islands, be it release date, foundry choice, or anything architecture-related.

Plus, I'm wondering whether I should wait for SI or just get a HD 6950.
 
According to Charlie, SI taped out in February and more details will be unveiled very soon, at AMD's Fusion Developer Summit.

I sure hope all that is true, because I may have been a little bit out of touch lately, but I've heard exactly nothing about Southern Islands, be it release date, foundry choice, or anything architecture-related.

Plus, I'm wondering whether I should wait for SI or just get a HD 6950.

Im in the same boat as you, almost pulled the trigger on a 6950 but decided to wait instead :???:

From that link you posted, Charlie also says its on TSMC. Second gen may go to GF and i think thats sounds reasonable. In the long run i see them going exclusively with GF. Since they're going to be manufacturing CPUs/APUs at GF, it makes sense to go with the same process for the GPUs as they'll already have experience with the process.
 
Im in the same boat as you, almost pulled the trigger on a 6950 but decided to wait instead :???:

From that link you posted, Charlie also says its on TSMC. Second gen may go to GF and i think thats sounds reasonable. In the long run i see them going exclusively with GF. Since they're going to be manufacturing CPUs/APUs at GF, it makes sense to go with the same process for the GPUs as they'll already have experience with the process.

He also mentioned recently (can't remember where, sorry) that he thought GloFo would be ready a little earlier for volume 28nm production. So at this point, who knows?
 
Back
Top