AMD Prepares ‘Bristol Ridge’ APU (Carrizo) For Desktop

[kemical]

third quarter of 2016 why are we even talking about this??????????????????//

[patrickjp93]

Regain traction in the server market. Not take over.

Implying perfect core-scaling?

What about the frequencies differences?

As if PCIe lanes matter to this discussion.

Implying imperfect clock scaling.

See first.

Completely wrong. What do you think Ethernet and infinite do run on? It's the PCIE bus.

[LukaP]

I think you confuse AMD's hope for the product(which @Opcode is referring to) with a definitive use/performance of it. That said, In this space where power is the same King as performance, ARM is becoming a threat.

yes, but ARM has been increasingly coming to the similar power segments that x86 has descended to, to keep performance competitive, so i really dont see a rewrite of the backbone of the server industry into ARM code. 

[Opcode]

Where does it show that it is AMD plans/hopes to take over the x86 server world with K12?

It's how you connect the pieces of the puzzle. Look where Seattle went (into the Opteron A1100). K12 will likely be Seattle's successor.

 

yes, but ARM has been increasingly coming to the similar power segments that x86 has descended to, to keep performance competitive, so i really dont see a rewrite of the backbone of the server industry into ARM code. 

ARM has been used among the server ecosystem for quite some time. A lot of your standard Linux distributions have ran on ARM for a while.

[patrickjp93]

yes, but ARM has been increasingly coming to the similar power segments that x86 has descended to, to keep performance competitive, so i really dont see a rewrite of the backbone of the server industry into ARM code.

In the phone/tablet segment you're correct. In the server segment there are 18-core chips barely pulling 40 watts. The only reason IBM's Power 8s draw 80 is out of order processing and that 5GHz clock for a 12-core beast.

[LukaP]

In the phone/tablet segment you're correct. In the server segment there are 18-core chips barely pulling 40 watts. The only reason IBM's Power 8s draw 80 is out of order processing and that 5GHz clock for a 12-core beast.

Ill not go into that, since i dont have the numbers at hand, so i stand corrected

[vm'N]

Completely wrong. What do you think Ethernet and infinite do run on? It's the PCIE bus.

I still don't see how that is relevant.

Are you saying the 5820k doesn't have enough PCIe lanes?

 

It's how you connect the pieces of the puzzle. Look where Seattle went (into the Opteron A1100). K12 will likely be Seattle's successor.

But, the puzzles aren't made from AMD.

[Opcode]

But, the puzzles aren't made from AMD.

Seattle wasn't but K12 will be using AMD's own IP. If you're referring to speculation then yeah, nobody knows, but you can get a good idea from other sources.

 

[Nielips]

What the hell are you on about?

• 28nm process when intel is at 10nm is madness, thats 3x size reduction
• DC shrunk nothing
• TRANSISTOR shrinkage plays a BIGASS fucking difference, thats why AMD have such a high TDP for what their IPC is
• Yes architecture is important, but you cant just stay at a process that is now 4 years old and call it a day

Well even if I am wrong about the devils canyon not beinf a shrink down. If you look at a reliable graph that looks at transistor size comparing power/IPC, we are well past the point were diminishing returns are seen from one die shrink to the next. Also I never stated that they should stay on an old process either, but it's AMD so we already know that all their current CPU's are procuced at this size so why keep going on about it.

[patrickjp93]

I still don't see how that is relevant.

Are you saying the 5820k doesn't have enough PCIe lanes?

 

But, the puzzles aren't made from AMD.

Yes, the 5820k, having only 28 lanes, cannot host both the Fusion I/O disk necessary for big work, a 24-way Infiniband card, and a standard 12-way ethernet card.

[vm'N]

Yes, the 5820k, having only 28 lanes, cannot host both the Fusion I/O disk necessary for big work, a 24-way Infiniband card, and a standard 12-way ethernet card.

Yes, the 5820k cannot do that specifically.

Those who have that kinds of equipment already have a much bigger budget, and are not in concern of small $500...

[patrickjp93]

Yes, the 5820k cannot do that specifically.

Those who have that kinds of equipment already have a much bigger budget, and are not in concern of small $500...

No, but that can translate up to shave $1000 or more in the server chips to compete on price.

[vm'N]

No, but that can translate up to shave $1000 or more in the server chips to compete on price.

Those equipment alone can easily come above $10,000.

Your point is still missing..

[The_Strict_Nein]

Eh, only more performance from upping frequencies not improving technologies.

 

More Cores More Frequencies - AMD 2011-2015

[Khvarrioiren]

Eh, only more performance from upping frequencies not improving technologies.

 

More Cores More Frequencies - AMD 2011-2015

 

The thing is, AMD has been pushing the clocks down for the last couple of years. Irrelevant snarky statement.

 

Edit: Disregard that, it was only on mobile platforms. Hardly relevant in this case.

[patrickjp93]

Those equipment alone can easily come above $10,000.

Your point is still missing..

Are savings savings, or aren't they? 10% is 10%.

[vm'N]

Are savings savings, or aren't they? 10% is 10%.

So, your entire point is the price?

The price is rarely the biggest concern for those who need this kind of equipment.

You save 10%, but lose 10% performance (if we continue with the whole 85% story).

[TheMissxu]

Eh, only more performance from upping frequencies not improving technologies.

 

More Cores More Frequencies - AMD 2011-2015

Intel will have to do it eventually as well... I'm not sure one can say you can do both at the same time.

[patrickjp93]

So, your entire point is the price?

The price is rarely the biggest concern for those who need this kind of equipment.

You save 10%, but lose 10% performance (if we continue with the whole 85% story).

Look again. If the AMD chip has lower power usage, then perf/watt increases which can increase perf/$ in the long run.

[vm'N]

Look again. If the AMD chip has lower power usage, then perf/watt increases which can increase perf/$ in the long run.

I'm looking for what?

You have only pointed out the initialize price as the issue.

You are now including another unknown metric to the speculation..

[patrickjp93]

I'm looking for what?

You have only pointed out the initialize price as the issue.

You are now including another unknown metric to the speculation..

It's not unknown. It's Jim Keller's child after development of the A8X for Apple. 

[vm'N]

It's not unknown. It's Jim Keller's child after development of the A8X for Apple.

It is unknown.