Intel Kaby Lake + AMD Radeon product *spin-off*

  • Thread starter Deleted member 13524
  • Start date
Bondrewd said:
Yet not fully connected, 2 hop diameter inside the package (well, at least to my understading).
Click to expand...

That would be a question of physical layout. You could place four dies close together, having their corners overlap an embedded silicon interposer, allowing all-to-all connections.

Also, two hops might not matter. The topology is effectively a ring bus, which works fine for a low number of drops as long as you have ample bandwidth.

Cheers
 
Bondrewd said:
Goddammit, this is a semi-custom chip, RTG only did the design.

They don't need to, GPU users will be begging for EPYC themselves anyway.
Click to expand...
Sigh.
It IS the Radeon GPU division that has the deal at the expense of the CPU division (because it solidifies Intel CPUs in a high margin segment), just because it is a semi-custom design and contract still has relevance to a particular division and influence on corporate and division strategy..
Or do you now say the VP of Radeon Gaming is not actually VP of Radeon Gaming (he is the one that announced this from an AMD side)?
Anyway as I said The Register article sums it up nicely that you ignore, which they differentiate as well between the two divisions and fits my experience working with various VPs and engineering divisions and sales teams.
But feel free to ignore how hostile Intel is to the CPU division of AMD while working with the Radeon GPU Division, there is some disconnect going on there as that aggressive approach by Intel to AMD CPUs was at the time they were negotiating with another side of AMD (Radeon),
Raja as the overall VP for the Radeon group has been looking at ways to leverage the division's IP,technology,HW, including semi-custom projects - this is something he has touched upon many times.

Edit:
In fact looking back Intel made it clear in their announcement:
Helping to deliver on our vision for this new class of product, we worked with the team at AMD’s Radeon Technologies Group. In close collaboration, we designed a new semi-custom graphics chip, which means this is also a great example of how we can compete and work together, ultimately delivering innovation that is good for consumers.
Click to expand...
 
Last edited:
Bondrewd said:
You do understand that this is only happening because AMD is not planning to adress this segment in 2018? At all.
Or you don't?
Click to expand...
As I explained earlier with a context; that argument means it is OK for AMD CPU division to team up with Nvidia in the HPC segment (especially scientific/analytics) because that would be perfect way to combat a hostile and very dominant Intel, and AMD Radeon are not going to compete there until late 2018 at best (putting aside their ecosystem is still not fully developed).
Like I said there is balance between what is good for the division and the company as a whole, but we only know some aspects of the deal with Intel albeit from AMD side the announcement came from Radeon gaming VP who talks about AAA gaming, content creators, enthusiasts all of which are high margin segments if one provides the complete solution.
 
CSI PC said:
team up with Nvidia in the HPC segment
Click to expand...
Why would they bother with that?
Whoever wants GPU compute racks will beg for EPYC anyway without any partnerships with asshats from nVidia.

I think you need to calm down, take your pills and rethink your life, while possibly selling your $AMD.
 
CSI PC said:
Sigh.
It IS the Radeon GPU division that has the deal at the expense of the CPU division (because it solidifies Intel CPUs in a high margin segment)
Click to expand...

It's the high end, no doubt. But I doubt it is higher margin than integrated solutions; Costs are going to be a lot higher with this solution.

Crystall Well was a lot more expensive than similar APUs without the eDRAM L4, but it still had lower margins

Cheers
 
Bondrewd said:
Why would they bother with that?
Whoever wants GPU compute racks will beg for EPYC anyway without any partnerships with asshats from nVidia.

I think you need to calm down, take your pills and rethink your life, while possibly selling your $AMD.
Click to expand...
You are the only one doing insults...
Can you really not see the benefit for AMD CPU division working closely with Nvidia in the HPC/AI/scientific/analytics market?
I guess it was pure luck that IBM with Nvidia has some of the largest supercomputer contracts for 2017/2018.
 
CSI PC said:
Can you really not see the benefit for AMD CPU division working closely with Nvidia in the HPC/AI/scientific/analytics market?
Click to expand...
There's absolutely no benefits in this over selling EPYCs to hyperscalers.
And you don't even need to deal with nVidia!

*[mod edit] we can do without the personal attacks
 
Last edited by a moderator:
Gubbi said:
It's the high end, no doubt. But I doubt it is higher margin than integrated solutions; Costs are going to be a lot higher with this solution.

Crystall Well was a lot more expensive than similar APUs without the eDRAM L4, but it still had lower margins

Cheers
Click to expand...
Remember it is the CPU division that has the best profits, AMD was slowly sinking until Zen launched.
And the deal with Intel will be pretty low margins.
 
would it be safe to assume the HBM is 2048bits?

also when they mention AMD sells the chip to Intel and provides driver support like they do for consoles, does that mean AMD supports drivers with their driver's package like on laptop GPUs or something else? like Intel handles the package?
.
 
Bondrewd said:
.Your hateboner for RTG is too stronk.
Click to expand...
Can please tone down the language?

I have no need for teenage conversations about hateboners and hairy assholes. CSI PC and others are trying to have a rational discussion and you’re welcome to be part of it and disagree. But you’re making it very hard to not be seen come as somebody who needs to grow first before being promoted to the adult section. Thanks.
 
Gubbi said:
Why do you think that ?

Cheers
Click to expand...
Your right in that we do not know any details of the deal itself so it can only be opinion, and there are still Ifs regarding what exactly is the scope in terms of its use when it launches and importantly what happens 6-12 months down the line.
Couple of general reasons, Intel will want to protect their margins as much as possible and even more so if this product is going into any Apple devices; Intel are not known for being generous to weaker companies (context relative to their own perception) and would look to cut AMD's margins more so than theirs.
So far Radeon group product margins has been pretty low-to-average, same could be said for other semi-custom projects albeit consoles has both soft and hard benefits while providing revenue without cutting off one division over the other and being a long term consistent mass sales while ending up also controlling that market.

Time will tell either way once we get the financial analyst conference a quarter later once this product starts shipping, although even then it may be hidden within the general semi-custom figures.
However it is fair to say AMD would have greater margins if they actually competed in this high margin space; AMD has destroyed the MSRP of Ryzen (went too far IMO as it is probably their most viable product) and impacted margins and yet still are pretty profitable from it.
Anyway with little known about the deal, it does present potential risk in some areas for AMD, while also guaranteeing it means no high performance solution will now be completely viable as it would potentially be competiting against this, which is a shame considering how competitive the latest generation of AMD CPUs are in terms of power demand and performance in context of gaming and certain other prosumer workloads against Intel.
 
Last edited:
What a weird product. It's like a failed console solution sent over to the PC market. I'm not sure I grasp the appeal of this considering it's going to be expensive and not clearly superior to just getting CPU + other discrete GPU.
 
swaaye said:
What a weird product. It's like a failed console solution sent over to the PC market. I'm not sure I grasp the appeal of this considering it's going to be expensive and not clearly superior to just getting CPU + other discrete GPU.
Click to expand...
High end portable Intel CPU + AMD GPU + focus on thin design. Are you sure you can't see who this would appeal to? That isn't super concerned with saving every last cent when it comes to price/performance?
(Which also addresses one of the questions regarding if it makes sense for AMD - it is by no means certain that their CPUs would have a chance of selling in that arena anyway.)
 
CSI PC said:
Your right in that we do not know any details of the deal itself so it can only be opinion, and there are still Ifs regarding what exactly is the scope in terms of its use when it launches and importantly what happens 6-12 months down the line.

Couple of general reasons, Intel will want to protect their margins as much as possible and even more so if this product is going into any Apple devices; Intel are not known for being generous to weaker companies (context relative to their own perception) and would look to cut AMD's margins more so than theirs.
So far Radeon group product margins has been pretty low-to-average, same could be said for other semi-custom projects albeit consoles has both soft and hard benefits while providing revenue without cutting off one division over the other and being a long term consistent mass sales while ending up also controlling that market.

Time will tell either way once we get the financial analyst conference a quarter later once this product starts shipping, although even then it may be hidden within the general semi-custom figures.

However it is fair to say AMD would have greater margins if they actually competed in this high margin space; AMD has destroyed the MSRP of Ryzen (went too far IMO as it is probably their most viable product) and impacted margins and yet still are pretty profitable from it.
Anyway with little known about the deal, it does present potential risk in some areas for AMD, while also guaranteeing it means no high performance solution will now be completely viable as it would potentially be competiting against this, which is a shame considering how competitive the latest generation of AMD CPUs are in terms of power demand and performance in context of gaming and certain other prosumer workloads against Intel.
Click to expand...


I really am not sure what you point really is.
To me, it seems you are fully unaware of AMD heterogeneous computing and their pursuance of their APU. You don't seem to understand that a GPU + CPU using the same resources is more efficient than a separate CPU & GPU using their individual resources.

Intel, nor Nvidia are as advanced in heterogeneous computing, as AMD.



AMD has a 15w APU chip that competes very nicely. The mere fact you can not conceive of a future 45w APU from AMD, seems like your are omitting this on purpose. Nvidia has NOTHING to compete with in this market. So AMD needs nothing from NVidia.

Intel is already seen the light and are hiring AMD to make their GPU's for them, because AMD has better technology. Nvidia doesn't fit into the mobile market, because it is not deep learning. Pascal offers nothing new and Volta is 8 month away. Nvidia also doesn't have right to x86, so no APU. AMD is win/win and you do not have to even break it down into the Radeon Group.



Lastly, as other have mentioned, why are you so overly concerned with the "deal" that AMD and Intel struck? Intel's cpu+gpu MCM will not be as good as AMD's APU (zen + vega). There is zero risk for AMD in selling Intel graphics chips/design. The only entity harmed, is Nvidia in a loss of sales.
 
swaaye said:
I'm not sure I grasp the appeal of this considering it's going to be expensive and not clearly superior to just getting CPU + other discrete GPU.
Click to expand...
Why would it be all that expensive? One single stack of HBM without the interposer, is that really going to be so costly? And Intel seems to think their silicon bridge tech is pretty cost-effective as well, as they're essentially promoting it as the way of the future (after slagging off epyc as a "glued-together" processor... :LOL:)

The appeal would be power efficient, space efficient applications, so thin performance-class notebooks, where a traditional solution would occupy significantly more board space and slurp more juice as well.
 
You must log in or register to reply here.
Back
Top