Interview with Rick Bergman of AMD

[thilanliyan]

http://www.insidehw.com/Editorials/...t-Mean-Death-To-The-Graphics-Card-Market.html

 

[VirtualLarry]

What I can say is that GPU product cycles tend to run faster than on the CPU side. AMD is working to incorporate the GPU pace of innovation and execution across all of our products, we call this AMD Velocity.

They're attempting to speed up CPU development cycles, to be more like GPU development cycles? Are they crazy? They can't even match Intel's R&D cycles for CPU development, nevermind surpass them.

 

[alyarb]

that's just his way of saying that he hopes they are toe-to-toe by the time we get to 22 nm.

 

[Cogman]

Q:will Fusion have anything to offer in the mid, perhaps even high class as far as GPU performance goes?

A: There are all sorts of synthetic tests, but we are interested in the performance gains we are able to achieve in actual applications. In fact, our goal is to achieve high performance in that segment exclusively.

Translation: No.

 

[alyarb]

i dont know why people are expecting these APUs to be super fast. it's just the next evolutionary step from a radeon 4200, geforce 9300, etc. it's the same integrated segment.

 

[BD231]

i dont know why people are expecting these APUs to be super fast. it's just the next evolutionary step from a radeon 4200, geforce 9300, etc. it's the same integrated segment.

Cheaper motherboards for SFF computers and maybe some more income for AMD? Kinna seems like one of those " nothing to see here " situaitions.

I'm not a fan of added components on the CPU, Intel has already shown us what limitations can come of adding even something as simple as PCI-E controller to a CPU. A GPU is going to add a lot of heat, RIGHT next to your CPU core which is not attractive. I'm sure many were just hoping there might of been something to gain with such an integration, when in fact its more limiting (dare I say worthless?) to a power user.

 

[VirtualLarry]

when in fact its more limiting (dare I say worthless?) to a power user.

I disagree. With DirectCompute (and rivals) taking off, and things like GPU transcoding coming to the desktop, I would rather have 480SPs under the hood than not.

 

[cbn]

i dont know why people are expecting these APUs to be super fast. it's just the next evolutionary step from a radeon 4200, geforce 9300, etc. it's the same integrated segment.

They are still small GPUs, but seem proportionally larger to IGPs in the past.

According to articles on review sites from November 2009 the Llano GPU is supposed to be 480 stream processors. A lot of folks still using older 12x10 LCDs (built circa 2004) will probably find it quite useful for playing slightly older games.

 

[cbn]

Cheaper motherboards for SFF computers and maybe some more income for AMD? Kinna seems like one of those " nothing to see here " situaitions.

But wouldn't cheaper SFF mainboards make a heck of a lot of sense in this global economy?

In fact, if these little computers could become as cheap or cheaper than a gaming console it might help inspire more young kids (from other countries) to enter IT as a career. I don't mean to knock on gaming consoles, but they are quite limited as educational tools (at this moment).

I'm not a fan of added components on the CPU, Intel has already shown us what limitations can come of adding even something as simple as PCI-E controller to a CPU. A GPU is going to add a lot of heat, RIGHT next to your CPU core which is not attractive. I'm sure many were just hoping there might of been something to gain with such an integration, when in fact its more limiting (dare I say worthless?) to a power user.

Well power users will probably want Discrete cards, that is true.

However, consider something different:

Could APUs find a niche with a larger audience?

Make a little case with a "built-in" Tower cooler. Then let the young kids try to overclock it? Maybe they will get fascinated enough to want to become an engineer?

 

[SmCaudata]

Having that extra core on the CPU package could lead to interesting things in the near future. With proper coding and architecture you can easily offload physics or some other specific task to that die. Obviously software and hardware would have to work together for this, but there is potential for these to be useful even for gamers.

 

[nyker96]

Having that extra core on the CPU package could lead to interesting things in the near future. With proper coding and architecture you can easily offload physics or some other specific task to that die. Obviously software and hardware would have to work together for this, but there is potential for these to be useful even for gamers.

I think they already doing this with OpenCL which offloads floating point calculations into the gpu been an open standard, I'm sure it can be implemented for either dedicated pci card or on board gpu package. shouldn't matter.

 

[cbn]

I think they already doing this with OpenCL which offloads floating point calculations into the gpu been an open standard, I'm sure it can be implemented for either dedicated pci card or on board gpu package. shouldn't matter.

I think I have asked this question before, but does anyone think the era of heterogeneous computing could lead to the practical elimination of the multi-core era?

If x86 code can be handled a GPU, why continue to multiply CPU cores? Wouldn't making CPU cores wider lead to larger gains per dollar?

 

[StrangerGuy]

I think I have asked this question before, but does anyone think the era of heterogeneous computing could lead to the practical elimination of the multi-core era?

If x86 code can be handled a GPU, why continue to multiply CPU cores? Wouldn't making CPU cores wider lead to larger gains per dollar?

From what I see so far x86 code is full of branches so it doesn't execute well on GPUs despite them being massively parallel.

 

[BD231]

LMAO yes bottleneck lets not forget the children, haha. Doesn't ATI make the xbox GFX sub system? They're not going to undermine themselves like that of course, there's to much money in console gaming and kids aren't the only ones playing these days. SFF market should bode well no doubt, but its still the same side port memory/more flops situation. I think a lot more could have been done with the integration and others probably did as well.

Maybe later revisions will have more to offer.

How many years have quad core processors been around SM? and, how many gaming apps are actually capable of putting all those cores to use? Can't imagine it would be any easier for on chip GPU's to get the attention of programmers who don't even use what they have at their complete disposal. Just a thought though, I certainly won't have use for fusion myself.

 

[cbn]

LMAO yes bottleneck lets not forget the children, haha. Doesn't ATI make the xbox GFX sub system? They're not going to undermine themselves like that of course, there's to much money in console gaming and kids aren't the only ones playing these days.

Sorry I didn't mean for this to sound like "ethical dilema" for a large corporation like AMD/ATI.

I was thinking more on the consumer level.

In some of the poorer countries parents might be much more inclined to buy their kids a small SFF desktop rather than a specialized gaming console. So for this reason there might not be any "overlap" in sales to reduce AMD profits.

 

[BD231]

No apology necessary, really. I could never fault a soul for caring and you're clearly a positive person which is admirable in anyone. Poor countries are historically quite a few product cycles behind us that's all, it got me laughing.

 

[cbn]

Poor countries are historically quite a few product cycles behind us that's all, it got me laughing.

Well that and maybe "gaming" wouldn't really seem like the best use of a child's time (from a strict parent's perspective).