Any near-term plans for a client taking advantage of GCN?

Moderators: Site Moderators, PandeGroup

Any near-term plans for a client taking advantage of GCN?

Postby j3pflynn » Thu Dec 22, 2011 12:03 pm

Now that it's near "volume" release time?
j3pflynn
 
Posts: 5
Joined: Sun Mar 28, 2010 9:53 pm

Re: Any near-term plans for a client taking advantage of GCN

Postby Zagen30 » Thu Dec 22, 2011 6:08 pm

Southern Islands won't require a new client; all that should be required is an update to add them to the whitelist of supported GPUs.
Image
Zagen30
 
Posts: 1589
Joined: Tue Mar 25, 2008 12:45 am

Re: Any near-term plans for a client taking advantage of GCN

Postby Nathan_P » Thu Dec 22, 2011 6:24 pm

Zagen30 wrote:Southern Islands won't require a new client; all that should be required is an update to add them to the whitelist of supported GPUs.

No, but it may well need a new core, arch is a lot closer to Fermi than previous AMD/ATI products
Image
Nathan_P
 
Posts: 1405
Joined: Wed Apr 01, 2009 9:22 pm
Location: Jersey, Channel islands

Re: Any near-term plans for a client taking advantage of GCN

Postby Zagen30 » Thu Dec 22, 2011 6:30 pm

I haven't been paying much attention to Southern Islands since I'm not really a fan of AMD, but since Core 16 is written in OpenCL, shouldn't it work (albeit not at max efficiency) on the 7000 series once they're on the whitelist?
Zagen30
 
Posts: 1589
Joined: Tue Mar 25, 2008 12:45 am

Re: Any near-term plans for a client taking advantage of GCN

Postby 7im » Thu Dec 22, 2011 6:36 pm

Fermi was a new architecture, and needed a new fahcore. GCN is likely the same, based on past experience. PG is unlikely to have anything in the works until the hardware is available publicly, which is weeks if not months away. So plenty of time to open a ticket to request the whitelisting, and a much lesser concern.

Also, I'm not sure the Gay Chistian Network will like that AMD has jacked their acronym (no affiliation). Graphics Core Next isn't even on the first page of results from Google. ;)
How to provide enough information to get helpful support
Tell me and I forget. Teach me and I remember. Involve me and I learn.
User avatar
7im
 
Posts: 14648
Joined: Thu Nov 29, 2007 4:30 pm
Location: Arizona

Re: Any near-term plans for a client taking advantage of GCN

Postby mdk777 » Thu Dec 22, 2011 7:26 pm

Fermi was a new architecture, and needed a new fahcore. GCN is likely the same, based on past experience.


Yeah, that was when the core ran on CUDA.

Since the new hardware is completely Open CL compliant, and the core is written for Open CL and not any AMD hardware at all(past or future)....no such new core will be required.

A simple white listing is all that is required.

PS. Public availability will be in January. Not some months. :mrgreen:
Transparency and Accountability, the necessary foundation of any great endeavor!
mdk777
 
Posts: 813
Joined: Fri Dec 21, 2007 4:12 am

Re: Any near-term plans for a client taking advantage of GCN

Postby 7im » Thu Dec 22, 2011 7:41 pm

Yes, I read the same early Jan release date articles. Duh. Didn't all those same articles question whether it would be available in any kind of volume that normal people could actually obtain one? ;)

Available is not the same as obtainable. :roll:
User avatar
7im
 
Posts: 14648
Joined: Thu Nov 29, 2007 4:30 pm
Location: Arizona

Re: Any near-term plans for a client taking advantage of GCN

Postby mdk777 » Thu Dec 22, 2011 8:05 pm

Could have shipped earlier, but not early enough to catch the holiday sales.

Hence, pushed out to avoid conflict.

What is considered volume for the $500 + graphics card market is open to wide interpretation.

Consumer availability of the 7970 will be January....complete line(7950 and bellow) to follow over the next few months. :wink:
mdk777
 
Posts: 813
Joined: Fri Dec 21, 2007 4:12 am

Re: Any near-term plans for a client taking advantage of GCN

Postby codysluder » Thu Dec 22, 2011 11:28 pm

Zagen30 wrote:I haven't been paying much attention to Southern Islands since I'm not really a fan of AMD, but since Core 16 is written in OpenCL, shouldn't it work (albeit not at max efficiency) on the 7000 series once they're on the whitelist?


As OpenCL advances from 1.0 to 1.1 to 1.2 some additional capabilities will be added. FAH may use some of those capabilities and not use others.

As ATI GPUs progress from one series to the next, additional hardware capabilities are added. FAH may take advantage of some of those capabilities and not use others.

If FAH is using capabilities that are supported in OpenCL 1.1 and in hardware in my HD5000 series GPU, that code will still work fine if the drivers are 1.2 and your hardware is GCN. If FAH needs something that's only in the HD7000, or only in OpenCL 1.2, either the driver or the FahCore must figure out how to compute the same result in spite of the limitations of my hardware, even if it has to compile some slower code for my HD5000 than 1.2 would be able to compile for the HD7000. The same logic applies to Cayman's DPFP, but FAH doesn't seem to ever use that capability.

What I do not know is whether FahCore16 needs to be revised to use OpenCL 1.2, and if it does, what implications that might have for performance for any GPU. I suppose that would also demand an upgrade of the OpenCL drivers if they're 1.0 or 1.1.

What I also doe not know is whether OpenCL 1.2+ will eventually work better on nvidia than CUDA (probably not) and if so, whether FAH will use it.
codysluder
 
Posts: 2128
Joined: Sun Dec 02, 2007 12:43 pm


Return to V7.1.52 Windows/Linux

Who is online

Users browsing this forum: No registered users and 1 guest

cron