So, where is AMD Seattle?

[IntelUser2000]

Edit: Though it certaintly does not look good for seattle to have such low spec rate scores.

Shouldn't SpecRate be the worst-case for Atom C because Seattle has a much better memory subsystem and SpecRate is memory/cache sensitive benchmark?

http://www.anandtech.com/show/8718/t...xynos-review/6

Not really optimized and preliminary but A57 is pretty much on par with A15 on the same process (20nm).

Is this opposite day?

Because A57 according to there is 18% faster in Spec clock normalized and both are 20nm.

 

[Enigmoid]

Shouldn't SpecRate be the worst-case for Atom C because Seattle has a much better memory subsystem and SpecRate is memory/cache sensitive benchmark?

Is this opposite day?

Because A57 according to there is 18% faster in Spec clock normalized and both are 20nm.

I'm talking about efficiency. A57 is faster than A15 but it looks like ARM went for at least a 1:1 power/performance design. It uses at least more than 18% more power for that 18% more performance.

As far as Seattle vs. Avaton. I'm not quite sure. May be preliminary tests are what we are seeing or AMD is having trouble with cache or the memory controller.

 

[IntelUser2000]

I'm talking about efficiency. A57 is faster than A15 but it looks like ARM went for at least a 1:1 power/performance design. It uses at least more than 18% more power for that 18% more performance.

I see, that's right.

(BTW, its Avoton, not Avaton: http://ark.intel.com/products/77987/Intel-Atom-Processor-C2750-4M-Cache-2_40-GHz

Avaton is S|A's pre-release code-name guesses)

As far as Seattle vs. Avaton. I'm not quite sure. May be preliminary tests are what we are seeing or AMD is having trouble with cache or the memory controller.

That's what happens with a product that's delayed a year. It's never a good sign when its delayed. Broadwell sucked because its delayed. 14nm too. Bulldozer was delayed. First Pentium 4, Netburst, Nvidia NV30. Conroe/Merom beat the forecast by 3 months, and came out to be an outstanding product.

(BTW, to using GeekBench for server comparisons. It might be relevant for a mobile benchmark, but not for server. GeekBench scales linearly with clock, meaning there's a problem with the benchmark - its not realistic. It might be that Avoton has much better scaling than Seattle, which is why it loses in ST but wins in MT. Also, server scenarios are sensitive to other things, like interconnects)

 

[JDG1980]

Frankly, AMD should can Seattle. It's not going to get many sales and they can't afford the distraction. Why are they dabbling in ARM chips at all? The only reason they should even try is if they have a specific customer who wants a semi-custom solution with an AMD GPU. AMD's two advantages over other companies are some excellent GPU IP, and the x86 license. Playing around with ARM servers takes advantage of neither.

 

[ShintaiDK]

(BTW, to using GeekBench for server comparisons. It might be relevant for a mobile benchmark, but not for server.

Its not even good for mobile

 

[ShintaiDK]

Frankly, AMD should can Seattle. It's not going to get many sales and they can't afford the distraction. Why are they dabbling in ARM chips at all? The only reason they should even try is if they have a specific customer who wants a semi-custom solution with an AMD GPU. AMD's two advantages over other companies are some excellent GPU IP, and the x86 license. Playing around with ARM servers takes advantage of neither.

This is really the core question. AMD must somehow have determined inside the company that the future is ARM for them. In other words they will abandon x86 on the long run.

So unless AMD plans to sell vanilla cores and save a truckload of R&D. Then its simply a losers path. While competiting with Intel in x86, its also their guardian against all the other ARM companies. While competiting in the ARM space its just a pure wild wild west game.

Short conclusion seems that AMD have shoot themselves in the foot...again.

 

[Fjodor2001]

AMD server solutions:

ARM: For those that require best perf/watt and price, and does not have to run x86 SW.
x86: For those that must run x86 SW.

Seems quite logical to provide both solutions. Especially since designing an ARM based chip based on an ARM core already designed by ARM is not that expensive to do.

Also, it's an entry to what will come, AMD's K12 (ARM) and Zen (x86) for future generations.

 

[Idontcare]

This is really the core question. AMD must somehow have determined inside the company that the future is ARM for them. In other words they will abandon x86 on the long run.

So unless AMD plans to sell vanilla cores and save a truckload of R&D. Then its simply a losers path. While competiting with Intel in x86, its also their guardian against all the other ARM companies. While competiting in the ARM space its just a pure wild wild west game.

Short conclusion seems that AMD have shoot themselves in the foot...again.

I think they thought they were second only to Intel in the MPU realm, and beating their head against that rock wasn't giving them the profit opportunities they desired, so they figured competing against the ARM companies would be like shooting fish in a barrel for their design teams.

Now you fast forward 4 yrs, those ARM teams are quite flush with talent themselves (a surprising amount being ex-AMDers too) and the ARM chips AMD thought would be gravy to produce and sell are turning out to perhaps not be all that.

But, just like the case with Fusion and APU, you can't spend four years selling your BoD and shareholders on the vision of an ambidextrous AMD portfolio only to then come back and "can" the whole thing just because the ARM competition showed up and trumped you.

You can only "can" the effort if you have a much better (i.e. sellable) strategy and vision to pitch to the BoD and shareholders in the very same meeting/presentation wherein you are killing off the vision of the past many years.

AMD needs something to pivot into, they pivoted from a bad Phenom situation to "fusion is the future"...and then they pivoted from fusion to "ambidextrous computing, ARM & x86, plus semi-custom, and a kitchen sink as well!" under Rory, now that Lisa is at the helm she will determine what is the catalyst for the next pivot point that is sellable to AMD's ever patient shareholders and ever blind BoD.

 

[ShintaiDK]

Talk about mistakes in a row. I wonder if Lisa will create a new thing of her own, while she cans yet another hopeless AMD project that a previous CEO thought of. She already had to put SeaMicro to the grave as a complete loss.

The ambidextrous platform already seems dead. And it will drag everything down with it from the looks of it. I would guess AMD ends up going from "shooting fish in a barrel" to playing the anonymous fiddle in the vanilla ARM core band. While the x86 parts fade into nowhere. Simply to keep cost under control while they desperately try to get the last money out of the former ATI that is also starting to be a ghost of its former self.

And lets be honest, when AMD will finally field its A57 Seattle fiasko. You can buy A72 cores all over the place.

 

[jpiniero]

This is really the core question. AMD must somehow have determined inside the company that the future is ARM for them. In other words they will abandon x86 on the long run.

They have no choice really because they are going to lose the license eventually. Either AMD is going to get bought or go bankrupt. The goal is go completely semi-custom after the reorg, and for that there's not much reason to do x86 anyway.

 

[Arachnotronic]

I think they thought they were second only to Intel in the MPU realm, and beating their head against that rock wasn't giving them the profit opportunities they desired, so they figured competing against the ARM companies would be like shooting fish in a barrel for their design teams.

Now you fast forward 4 yrs, those ARM teams are quite flush with talent themselves (a surprising amount being ex-AMDers too) and the ARM chips AMD thought would be gravy to produce and sell are turning out to perhaps not be all that.

But, just like the case with Fusion and APU, you can't spend four years selling your BoD and shareholders on the vision of an ambidextrous AMD portfolio only to then come back and "can" the whole thing just because the ARM competition showed up and trumped you.

You can only "can" the effort if you have a much better (i.e. sellable) strategy and vision to pitch to the BoD and shareholders in the very same meeting/presentation wherein you are killing off the vision of the past many years.

AMD needs something to pivot into, they pivoted from a bad Phenom situation to "fusion is the future"...and then they pivoted from fusion to "ambidextrous computing, ARM & x86, plus semi-custom, and a kitchen sink as well!" under Rory, now that Lisa is at the helm she will determine what is the catalyst for the next pivot point that is sellable to AMD's ever patient shareholders and ever blind BoD.

I know, know! AMD's ambidextrous strategy needs moar architectures. ARM and x86 aren't enough, so what AMD needs to do to complete this strategy is to build SPARC, POWER, and MIPS64 cores.

One performance target, 5 architectures. It's brilliant!

 

[ShintaiDK]

They have no choice really because they are going to lose the license eventually. Either AMD is going to get bought or go bankrupt. The goal is go completely semi-custom after the reorg, and for that there's not much reason to do x86 anyway.

Dropping x86 is good on the perspective of selling the company. But as time goes on, there is less and less to sell. And who would want to buy? Its a company filled with skeletons in the closets while everything is run into the ground. Just covered by a thin layer of shiny paint.

 

[Arachnotronic]

Seems quite logical to provide both solutions. Especially since designing an ARM based chip based on an ARM core already designed by ARM is not that expensive to do.

Yeah, but designing a good one is.

 

[mrmt]

AMD needs something to pivot into, they pivoted from a bad Phenom situation to "fusion is the future"...and then they pivoted from fusion to "ambidextrous computing, ARM & x86, plus semi-custom, and a kitchen sink as well!" under Rory, now that Lisa is at the helm she will determine what is the catalyst for the next pivot point that is sellable to AMD's ever patient shareholders and ever blind BoD.

I always saw Rory strategy as "we have to get out of Intel's way ASAP and find some place with lower margins to duck", and I must say that the first part was executed flawlessly. He cut down or axed all the projects for the x86 market that he could, and wrecked their consumer CPU business with a 70%+ drop in revenues.

The problem I see is that he failed miserably on the second part. Their ARM efforts seems to be fizzling and their ARM server strategy was thrown in disarray by Intel's latest moves. He also failed to hit a much easier and softer target than all these ventures he proposed: Nvidia.

But I think AMD's next pivot won't take long to arrive The company is bleeding money and they certainly cannot afford to burn more cash in dead end projects like Seattle.

 

[BigDaveX]

Cancelling the FX versions of Steamroller and Excavator was an understandable move, since Steamroller definitely wouldn't have been able to compete with Haswell, and likely Excavator wouldn't have done either (to say nothing of the fact that it would most likely have been going up against Skylake). But it seems like they took much too long to get any sort of new direction in place.

Really, what AMD could do with is something akin to their first major success, the K6-2; a chip that can't compete against Intel's best, but is still a very reasonable performer for most users, and cheap to produce.

 

[JDG1980]

You can only "can" the effort if you have a much better (i.e. sellable) strategy and vision to pitch to the BoD and shareholders in the very same meeting/presentation wherein you are killing off the vision of the past many years.

The best strategy would be to focus on the few areas where AMD has scored major design wins: most notably, the consoles.

AMD is the only company that can offer an all-in-one x86 gaming solution. Nvidia doesn't have access to the x86 architecture, and Intel is way too far behind in GPU technology. The consoles are a proof of concept of this. If AMD can get to the point where a single chip can provide competitive 1080p gaming performance on PC titles, in addition to doing all the usual productivity stuff people want, this would open up new possibilities for all-in-one OEM systems. "Gaming Ready". Such systems might well be a better overall deal for many buyers than Intel systems.

To do this, AMD needs to focus on the following:

• Zen. The Bulldozer architecture is a dead end; the cat cores are nice, but need a sizable IPC boost to get single-threaded performance up for gaming. An IPC on par with Sandy Bridge would be just about perfect.
• GCN improvement. The architecture is better than people think (AMD insists on shipping their products with higher clocks and voltages than they really should, hurting efficiency), but it still needs some improvements to be competitive with Maxwell.
• HBM. This is the key to a good all-in-one gaming chip, because GDDR5 is expensive, power-hungry, and complicates motherboard design, while DDR3 (and even DDR4) is way too slow for decent gaming. This is what hamstrings Kaveri more than anything else. On the other hand, give a gaming APU 8GB of shared HBM2 memory, and you're really cooking.
• Node improvements. This isn't totally within AMD's control, but they will need 16nm/14nm FinFET+ to get this working at acceptable power levels.
• Better marketing and coordination with OEMs.

Admittedly, this is something of a niche market, but it's a substantial niche and one that AMD could completely own. They can still sell stand-alone versions of Zen and discrete graphics cards to whoever wants them, while focusing on the core "single-chip x86 gaming" vision. I think they make some money on HPC, so for the biggest discrete GPU, they'll want to maintain current levels of Double Precision performance. The gaming focused chips don't need this (and indeed, current chips below Hawaii and Tahiti don't have it).

 

[tweakboy]

No not in seattle but in North Dakota

 

[AtenRa]

Cancelling the FX versions of Steamroller and Excavator was an understandable move, since Steamroller definitely wouldn't have been able to compete with Haswell,

An 8-core Steamroller with L3 would be able to compete very easily with 4C/8T Haswell in MT loads. Take Kaveri vs Core i3 Haswell and extrapolate.

 

[NTMBK]

AMD is the only company that can offer an all-in-one x86 gaming solution. Nvidia doesn't have access to the x86 architecture, and Intel is way too far behind in GPU technology.

I doubt that the next console generation will be x86, even if they are made by AMD. 64-bit ARM will be available, offering improved efficiency over x86 along with a nice, clean instruction set- and one which is familiar to thousands of mobile developers.

 

[mrmt]

AMD is the only company that can offer an all-in-one x86 gaming solution. Nvidia doesn't have access to the x86 architecture, and Intel is way too far behind in GPU technology. The consoles are a proof of concept of this. If AMD can get to the point where a single chip can provide competitive 1080p gaming performance on PC titles, in addition to doing all the usual productivity stuff people want, this would open up new possibilities for all-in-one OEM systems.

This would be the entire "the future is fusion" rehashed, the same vision Rory killed the day he assumed his CEO position at AMD.

 

[AtenRa]

This would be the entire "the future is fusion" rehashed, the same vision Rory killed the day he assumed his CEO position at AMD.

where did you see RR killing the Fusion business ??? They went full throttle with APUs, OpenCL and HSA.

 

[mrmt]

I doubt that the next console generation will be x86, even if they are made by AMD. 64-bit ARM will be available, offering improved efficiency over x86 along with a nice, clean instruction set- and one which is familiar to thousands of mobile developers.

There aren't as many companies that could provide the GPU for the console chips, and only one of them willing to swallow sub-20% margins, so yes, I think x86 has a chance, especially if AMD's ARM adventure continue going the way of the dodo like it is going now.

 

[Arachnotronic]

I doubt that the next console generation will be x86, even if they are made by AMD. 64-bit ARM will be available, offering improved efficiency over x86 along with a nice, clean instruction set- and one which is familiar to thousands of mobile developers.

I don't think the benefits of the ARM ISA outweigh the backwards compatibility that future consoles could possibly have with previous generation consoles that sticking with x86 would bring.

 

[JDG1980]

This would be the entire "the future is fusion" rehashed, the same vision Rory killed the day he assumed his CEO position at AMD.

The problem with AMD's existing attempts at "fusion" APUs are failures of implementation, not concept. The CPU side has mediocre single-threaded performance, and the GPU side doesn't perform all that much better than Intel's best iGPUs because DDR3 creates a hard bottleneck. The result is that Trinity, Richland, and Kaveri can barely play AAA titles at 720p30 low settings; a noticeably degraded experience. You lose power efficiency and CPU performance, and don't get much in return.

But the Playstation 4 design already manages to solve some of these issues. The Jaguar cores still have lower IPC than a gaming console would like, but at least they're power-efficient. The GPU gets the fast RAM (GDDR5) that it needs to run many AAA titles at 1080p. And this is all done on 28nm. Add a better CPU architecture (doesn't need to catch up to Intel's newest, just narrow the gap), the new GPU technology in Tonga and beyond, HBM in place of GDDR5, all on 16nm/14nm FinFET+ for more CPU/GPU oomph at lower power usage... and suddenly you have a very competitive product indeed, for both the PC market and the next generation of consoles.

Add into that the fact that the gaming consoles are about the only thing keeping AMD alive right now, and doubling down on this strategy starts to look a lot better.

 

[NTMBK]

I don't think the benefits of the ARM ISA outweigh the backwards compatibility that future consoles could possibly have with previous generation consoles that sticking with x86 would bring.

They seemed happy enough to ditch it for the latest generation. And "backwards compatible" means it is harder to sell HD remakes. (Or 4k remakes, I guess!)