The AMD Execution Thread [2007 - 2017]

Status
Not open for further replies.
AMD Betting Everything on OpenCL

BT: So you really think CUDA's not going to be around in a few years?

TM: CUDA's going to have a niche place in the industry – I think it'll exist for very specific high-performance computing applications - stuff like genetic mapping and specific things that universities are researching. I think that a lot of those places like that be happy with the infrastructure and support, and they'll stick around.
Is it me, or AMD is just cautiously backing off from the GPU HPC competition with Nvidia?
It looks like they hope for some miracle in the mass market to drive the open standard by itself to profiit levels -- a very lazy attitude in my opinion.
 
AMD Betting Everything on OpenCL


Is it me, or AMD is just cautiously backing off from the GPU HPC competition with Nvidia?
It looks like they hope for some miracle in the mass market to drive the open standard by itself to profiit levels -- a very lazy attitude in my opinion.

Alternatively, if nVidia goes to far catering to niche HPC markets, they risk becoming the next Matrox, so specialized they're unable to compete in the mainstream. Not getting caught up in a race to the high end has been one of ATI/AMD's best business decisions of the last decade.
 
AMD Betting Everything on OpenCL


Is it me, or AMD is just cautiously backing off from the GPU HPC competition with Nvidia?
It looks like they hope for some miracle in the mass market to drive the open standard by itself to profiit levels -- a very lazy attitude in my opinion.
AMD as always are going for mainstream first. Nothing wrong to push OpenCL (also heavily pushed by Apple, used in consoles, etc) and DirectCompute (MS will make sure its a standard on PC platform anyway).

AMD isnt strong or rich enough to create such market on its own, therefore they're betting on the direction market is going, and aiding to make transition faster. Fusion are outstripping supply, ~4 mln. APUs per Q1, and only will increase as new models are launched and capacity increases. AMD is getting there, from market penetration (APU+GPU), as well as more and more apps are supporting OpenCL/DC.

CUDA will never reach such mass adoption as OpenCL/DC, more like it will shrink to specific niche market, as mentioned in the article.
 
Actually, there's one niche GPGPU market where AMD is thriving and even dominating -- the bitcoin currency mining network. It is an example of virtually accidental success in the field, with almost zero input/effort by AMD's side. But I would be very cautious of such company policy, relying on "accidents" for its market development... it's a very steep betting curve.
 
But I wonder, as PC users progressively use their browsers for almost everything they do (web-browsing, music listening, video watching, document viewing and editing, etc), I wonder what's all that software that justifies using OpenCL in it, besides games, image and video editing software.
 
But I wonder, as PC users progressively use their browsers for almost everything they do (web-browsing, music listening, video watching, document viewing and editing, etc), I wonder what's all that software that justifies using OpenCL in it, besides games, image and video editing software.

With bandwidth caps hitting Canada and now the united states I think as soon as the next compression tech is done h.265 i think its called , many will switch over to it and todays cpus and gpus will have problems dealing with its acceleration which will cause many to want ot upgrade again.

Monitors are also still getting bigger and cheaper and image quality is going up . My gfs father just picked up a 23inch led monitor for $140 . I believe that soon monitor makers will want to move up to the next resolution size to gain more sales. So even there we will need more power for video.
 
Anand is saying B0 and B1 were no good so Bulldozer is pushed to late July with a B2 stepping.

Elsewhere september is being touted. Not sure it really matters whether its late Q2 or late Q3.

AMD Betting Everything on OpenCL


Is it me, or AMD is just cautiously backing off from the GPU HPC competition with Nvidia?
It looks like they hope for some miracle in the mass market to drive the open standard by itself to profiit levels -- a very lazy attitude in my opinion.

Alternatively, AMD are looking for the biggest markets to get share and traction in.

From that same link, Terry says Llano is AMD's biggest product at the moment. Presumably with several million shipped for revenue in this quarter, it'll be in shipping products this quarter i.e. launched. Last month of the quarter is june, launch both the enthusiast consumer processor and the mainstream consumer processor in the same month? Need a big budget for PR, lots of product managers and OEM relations peeps to do that. Get all the design wins, get the coverage, get the samples out there, have everybody enough time to test it before the next one... lots to do coming together in one month. Doable with Intel resources. Practical for AMD?

Not saying there's no performance issue or bug fix that needs a stepping revision to fix, just saying it might be gamesmanship with Intel vs. IB for BD release. The big BD release has always been Q3, for server and workstation products.
 
No BD in June is fail, no BD in July is tragic, no BD in August is catastrophic,No BD in September is game over. End of times for AMD in that segment. My bet is they already know this, and hope to get a few years foot hold in with Fusion, till they bleed red again... slow death. Its a WTF moment here. Hope i see some not so fast BD's anyways in early July...
 
from what people are saying of the live stream ( it kept dieing ) they said bulldozer next month , which is july (it was the first in tawiian ) .

If thats true then hopefully we will see it in systems by august
 
I wouldn't worry much, AMD has seen far worse before and has survived.

Well, they're out of fabs to sell, and the Intel suing business has probably run dry for a while.

Maybe Abu Dhabi wants to buy some more shares?
 
Maybe Abu Dhabi wants to buy some more shares?

Absolutely. But i wouldn't bet on that being too soon.

Unrelated, there was a discussion here some pretty long time ago (pre-fermi speculations) where we established the fact that metal spins will get nowhere in terms of clock speed increase or power consumption. Charlie (perhaps others?) was hinting on C0 for BD instead of B3, So either it's a bug in the chip or we will have a full respin on our hands?
 
No BD in June is fail, no BD in July is tragic, no BD in August is catastrophic,No BD in September is game over. End of times for AMD in that segment. My bet is they already know this, and hope to get a few years foot hold in with Fusion, till they bleed red again... slow death. Its a WTF moment here. Hope i see some not so fast BD's anyways in early July...



Bulldozer (consumer version) is now a secondary product for AMD, aimed at a niche market.
Llano has a much broader audience and has features that Intel isn't yet able to compete with. APUs are AMD's focus right now, as they've stated many times.


In fact, I'd go as far as stating that CPU-only chips are a dying breed that will serve the sole purpose of trying out new tech to embed into future APUs.
 
Bulldozer (consumer version) is now a secondary product for AMD, aimed at a niche market.
Llano has a much broader audience and has features that Intel isn't yet able to compete with. APUs are AMD's focus right now, as they've stated many times.
I'm not sure it's a good thing, BD is aimed at a market with way higher margins. May be AMD knows already that the chip doesn't stand a chance vs Ivy bridge.
In fact, I'd go as far as stating that CPU-only chips are a dying breed that will serve the sole purpose of trying out new tech to embed into future APUs.
Well I would like to see X86 Soc sooner than later. ARM CPU are nice but even Nvidia doesn't expect them to match the power of netbook in the next five years.
 
How atractive is a new FM1 mainboard for desktop users ? When u buy it, than u are stuck with Lliano cpu-s only and cant upgrade to bulldozer.
Phenon users can buy now FX990 mainboards and upgrade later to the 8 core bulldozer.
 
No BD in June is fail, no BD in July is tragic, no BD in August is catastrophic,No BD in September is game over. End of times for AMD in that segment.

Why? Why is it all of a sudden fail/tragic/catastrophic? I mean, wasn't BD first supposed to appear in 2009? And now it's gone from Q2 11 to Q3 11? What's with the sky is falling act? If June 11 was ok, then so is Sept. 11. SB appeared in January, SB-EX hasn't yet, IB is due Q2 12... seems like there is lots of room.
 
Status
Not open for further replies.
Back
Top