The AMD Execution Thread [2007 - 2017]

[Alexko]

"P" vs non "P" are different products with different target markets, and are sold as such - if you want multiple displays, or framelock or to simply put it in a worksation tower you're not going to be buying the passive board with one displayout. A similar analogy would be Radeon 7970 vs 7970M, do we believe these are the same products with the same target markets?

A simple search on the Newegg from the links above vs their respestive non-P branded products should give you some indication on how they are priced differently.

Thanks, I hadn't noticed that. You can still run the same software and get the same performance on both kinds of FirePro, right? I mean can you do CAD and rendering on a FirePro P if you're fine with a single display output and get the same performance?

While I'm at it, I have another question. I know that testing, qualifying and validating pro hardware takes time, and that's why it's usually released some time after consumer variants. But Opterons seem to follow their corresponding Phenoms/FXs quite closely, while FirePros lag behind their Radeon counterparts by several quarters. Why is that?

 

[imaxx]

...probably because for the video cards you have to produce stable, well tested drivers.
For CPU, you do not really need drivers.
Also, GPU needs to design a whole card, whereas for CPU usually a socket.. the motherboard design is not really yours to do, every year or two.

 

[silent_guy]

...probably because for the video cards you have to produce stable, well tested drivers.
For CPU, you do not really need drivers.
Also, GPU needs to design a whole card, whereas for CPU usually a socket.. the motherboard design is not really yours to do, every year or two.

You realize that not having drivers for a CPU makes it harder, not easier, right?

 

[aaronspink]

While I'm at it, I have another question. I know that testing, qualifying and validating pro hardware takes time, and that's why it's usually released some time after consumer variants. But Opterons seem to follow their corresponding Phenoms/FXs quite closely, while FirePros lag behind their Radeon counterparts by several quarters. Why is that?

Server parts tend to require and additional 6-12 months of validation over desktop parts due to the increased complexity(multi-socket, MCM, coherency, memory configuration, etc) of the systems they go in.

In addition, standard GFX parts tend to have a fraction of the validation and rework that a CPU has. Since the drive basically controls everything that "runs" on a GFX part, it is easy to work around a decent number of issues. In addition, you can have a lot of minor errors and such on a GFX part without having significant effect (if your GPU is giving results off by .5-2 ULP, no one really cares for GFX nor will they likely ever know), but on a CPU those same errors would make the part unreleasable.

So if you look at it from a mean time from silicon to release (MTFSTR) (all numbers examples only and no actual relations to reality):
gGPU: 6 months
cCPU: 12-18 months
cGPU: 15-18 months
sCPU: 18-24 months

Diffs:
gGPU->cGPU - 9-12 months
cCPU->sCPU - 6-12 months

The reality is that a gGPU is just quicker to get out as you can more easily work around and ignore bugs that you cannot get away with in a CPU or cGPU.

Another way to look at it, if the GK104 has a crippling bug that makes its compute performance bad, you don't really care, cause you aren't selling it as a compute part.

 

[Alexko]

Server parts tend to require and additional 6-12 months of validation over desktop parts due to the increased complexity(multi-socket, MCM, coherency, memory configuration, etc) of the systems they go in.

In addition, standard GFX parts tend to have a fraction of the validation and rework that a CPU has. Since the drive basically controls everything that "runs" on a GFX part, it is easy to work around a decent number of issues. In addition, you can have a lot of minor errors and such on a GFX part without having significant effect (if your GPU is giving results off by .5-2 ULP, no one really cares for GFX nor will they likely ever know), but on a CPU those same errors would make the part unreleasable.

So if you look at it from a mean time from silicon to release (MTFSTR) (all numbers examples only and no actual relations to reality):
gGPU: 6 months
cCPU: 12-18 months
cGPU: 15-18 months
sCPU: 18-24 months

Diffs:
gGPU->cGPU - 9-12 months
cCPU->sCPU - 6-12 months

The reality is that a gGPU is just quicker to get out as you can more easily work around and ignore bugs that you cannot get away with in a CPU or cGPU.

Another way to look at it, if the GK104 has a crippling bug that makes its compute performance bad, you don't really care, cause you aren't selling it as a compute part.

That makes sense, thank you.

 

[RudeCurve]

Apple

 

[Alexko]

Apple

 

[rpg.314]

Apple

Really, ???????

 

[eastmen]

Really, ???????

Makes little sense to me , if anything apple seems to be moving away from OSX and more and more towards IOS. They already have a strong presence in that world too.


Why would they want a company that can't seem to execute on anything anymore ? Just look at the windows 8 launch coming up. AMD has only 1 new cpu line ready in time for it they have no real desktop presence anymore , brazos 2 years old and has recieved very little updates . It seems like they will hit with new stuff after haswell.

 

[silent_guy]

You can call find many faults at AMD, but when looking at their GPU rollout, lack of execution is the last thing that comes to mind.

 

[Alexko]

You can call find many faults at AMD, but when looking at their GPU rollout, lack of execution is the last thing that comes to mind.

On the technical side of things, I fully agree. Marketing-wise, it's another story.

 

[eastmen]

You can call find many faults at AMD, but when looking at their GPU rollout, lack of execution is the last thing that comes to mind.

Seems like just buying power vr is the easier way to go.

 

[rpg.314]

Seems like just buying power vr is the easier way to go.

No way in hell it's going to happen.

 

[eastmen]

No way in hell it's going to happen.

why not , isn't apple the most valuble company in the world now ? Power vr seems to match their needs more than AMD

 

[rpg.314]

why not , isn't apple the most valuble company in the world now ? Power vr seems to match their needs more than AMD

Apple and Intel own a bit of IMG. They are not going to let anybody else control it.

 

[eastmen]

Apple and Intel own a bit of IMG. They are not going to let anybody else control it.

Does intel own enough of IMG to prevent Apple from buying it ?


Apple getting IMG would disrupt the mobile market.

 

[liolio]

Apple getting IMG would disrupt the mobile market.

How that? I mean the only one that may be bothered would be Intel but for now they are epsilon in this market.

 

[Alexko]

How that? I mean the only one that may be bothered would be Intel but for now they are epsilon in this market.

Lots of chipmakers use IMG's IP: Intel, yes, but also TI, ST-Ericsson, Samsung (though they use Mali too) and many others.

 

[Miksu]

I've been wondering about the AMD's situation regarding Windows 8. All the tablets and laptops released thus far (AFAIK) contain either Intel Atom or Intel Ivy Bridge. I think it's going to be a huge loss to the company if they're going to miss the whole Windows 8 release.

 

[Kaotik]

I've been wondering about the AMD's situation regarding Windows 8. All the tablets and laptops released thus far (AFAIK) contain either Intel Atom or Intel Ivy Bridge. I think it's going to be a huge loss to the company if they're going to miss the whole Windows 8 release.

Microsofts Surface Pro is actually Core i5, not Atom