AMD CPU speculation... and expert conjecture

[bmacsys]

Juan, ARM will win in small devices...this is clear. Which is what was said to you many moons ago when you started all this mess.

x86 can do FAR too many things that would make life far more complicated for ARM to even bother with. Could ARM try? Sure...would they unseat x86 for this? Not very likely...perfect example is Windows RT. People will want windows on x86, why? Because it does all the things they want. Additionally...you will never get consumer Joe Q Public to switch his perfectly fine PC at home for some "new fangled" device just because a promo slide you showed by some speculator says ARM will win. He does not want to buy a new x86 rig to do what he wants better, so why would he abandon all that familiarity to go to something "untried" in the public PC sector? Are they fine with it in a tablet? Sure...would they trust their jobs and lives to it? Well, let us just say, FAR too many people are still on XP. The proof is in the pudding...ARM will not win anything that it is not already winning. I might add that x86 is already trying very flagrantly to steal market share as well.

I would argue, if x86 hardware competes at PPW for ULP devices...ARM dies a quick death. Look at the number of new tablets now that are running x86 compared to what was offered a year ago even. Then, consider many x86 devices can run android because it is linux, essentially, which means that the software would not have any issues adapting. That is all people really care about...will this play angry birds, and will it run android? If the answer is yes, then there is no concern from the public.

I think you are wrong. Little by little ARM devices are eating away at x86. PC sales continue to fall. There is real momentum being built. ARM device sales dwarf x86 device sales right now. A few years ago people would have laughed at this ever happening. I think eventually Apple will build an ARM powered MacBook. It will sell. Then you will see all the big Asian company's follow suit by building cheap Android desktops. Microsoft knows this. That is why they built the Surface RT. Intel will be a bit player in 10 years. ARM is winning the battle right now. Intel gets huge margins with x86. Those kinds of margins don't exist in the ARM world. Processors are $20 not $200. Intel can't survive selling $20 processors.

Because ARM cores go into disposable devices like tablets and phones.

The fact that x86 is stagnating is not because ARM is "winning", it is because the PC market is saturated. No amount of ARM sales that have been made to date are erroding any part of any segment of x86 market share, the 2 do not compete except in very small portions of niche segments (i.e. convertibles).

x86 is honestly gaining ground in those markets where ARM is traditionally dominant as well. Granted, it is not gaining market share at 2 to 1, but when you are talking about 100s of millions of devices even 2-3% gain is a substantial amount of devices that are now x86 that were not 12 months ago.

That just isn't true. There is only so much disposable income. People are using it to buy tablets and smartphones over new pc's.

 

[juanrga]

Spoiler

Another thing to note is that Carrizo is nowhere to be seen on AMD's updated roadmap. It goes from Kaveri, Beema, Mullins and Seattle in 2014 to Project Skybridge in 2015. No Excavator, no Carrizo anywhere.

With limited resources, it wouldn't surprise me if AMD has thrown all it's chips on the table (no pun intended) for Skybridge, canned Excavator, and at most we'll see a Kaveri refresh at some point to plug any gaps until Skybridge arrives, but not a full blown Excavator/Carrizo.

Are you sure?

http://benchmarkhardware.com/images/stories/2014/Mayo/semana2/ADM-Roadmap-SkyBridge-BH.JPG

Well, I took it straight from the horse's mouth from AMD's press release page and if you follow the link at the bottom of the page where it says "The core roadmap and presentation can be viewed at ir.amd.com," you get the official roadmap for graphics and cpu/apu in pdf form.

It doesn't say anything about Excavator canceled. In fact, it only show the 2013-2014 roadmaps. In any case I asked AMD if Carrizo/Toronto (Excavator) had been canceled. And the answer is that new plans don't affect previous x86 plans. Thus I will continue expecting Excavator x86 cores for Carrizo in 2015 and then K12 cores in 2016.

 

[juanrga]

First, it is you who are confused, you need more than 12 ARM processors from Apple or NVidia to compete with 1 single i7-4930k. Show me ONE x86 metric where ARM even COMPETES 1 to 1.

Also, you continue trying to shift the conversation off topic.

ARM will not scale well upward. Period.

I care not what Papermaster said...you take what Keller said out of context, and Rory Read knows less about APU/CPU architecture than he knows about Tarot Card readings. He came from Lenovo, he knows how to run a business, which is his job...he is not an Engineer, nor was he ever.

You continue very confused.

As shown by Anand review of cyclone architecture, this is a "desktop class" design, which competes in complexity and IPC with a Ivy Bridge or Haswell big core.

Apple will be using its SoC on phones/tablets and thus has to limit its SoC at ~2W TDP

Although any industry expert claims that Apple core design is the best in its class, Apple cannot do miracles and cannot pack i7-4930k performance in ~2W TDP. The i7-4930k is rated at 130 W TDP.

Intel cannot offer anything that competes with ARM for the same TDP, because Intel needs four x86 cores @1.6GHz to match the performance of two apple cores @ 1.3GHz.

There is nothing magical in x86 that prohibits ARM scale up to 100W or even 200W.

I see you start insulting AMD president, despite he is doing an excellent work and re-floating the company. I see that you admit that you don't care about that AMD says, but I already knew since the past year!

A small resume of how you are living in a different universe than rest of us:

AMD: the new K12 core is a high-frequency high performance core that will span the range from laptops to desktops to servers—and not just microservers.
8350rocks: the new K12 core is only for ULP and microservers.

AMD: The TAM for x86 is declining. The TAM for ARM is increasing.
8350rocks: x86 is gaining ground and replacing ARM.

AMD: ARM and x86 are sister cores and treated as first-class citizen
8350rocks: ARM is only an "experiment" (my friend-at-AMD said so, trust me).

AMD: Our ARM core will be a bigger-engine that our x86 core
8350rocks: ARM will not scale well upward. Period.

AMD: We are abandoning the CMT design for our new cores
8350rocks: No. AMD new cores will be based in a redesigned CMT architecture

AMD: The Bulldozer architecture was as failure. The chief architect and the rest of the team were fired.
8350rocks: No. It rocks.

AMD: AMD now takes a bold step forward and has become the only company that can provide high-performance 64-bit ARM and x86 CPU cores paired with world-class graphics
8350rocks: Rory Read knows less about APU/CPU architecture than he knows about Tarot Card readings.

AMD: All our 28nm products will be 28nm bulk
8350rocks: Kaveri is delayed because is made on 28nm FD-SOI, trust me

AMD: We are migrating to 20nm bulk and then FinFET on bulk
8350rocks: AMD is returning to FD-SOI for 20nm and then FinFET on SOI, trust me

AMD: Steamroller has same ALUs per core than Piledriver
8350rocks: I know for a "fact" that Steamroller has extra ALU per core

AMD: We are transforming into an APU/Soc company
8350rocks: No. AMD will continue designing dCPUs forever.

AMD: We selected a 10TFLOP APU for our exascale supercomputer
8350rocks: No, AMD will use dCPU and dGPU

AMD: only six cores of our PS4 SoC are available to games
8350rocks: No. Eight cores are fully available to games because the OS is run in a separate chip.

AMD: The FX-9000 series is the last.
8350rocks: No. AMD is will be releasing FX-Steamroller CPU for AM3+.

And so on. And so on.

 

[colinp]

It doesn't say anything about Excavator canceled. In fact, it only show the 2013-2014 roadmaps. In any case I asked AMD if Carrizo/Toronto (Excavator) had been canceled. And the answer is that new plans don't affect previous x86 plans. Thus I will continue expecting Excavator x86 cores for Carrizo in 2015 and then K12 cores in 2016.

I'm not sure I ever saw a roadmap making a reference to a cancelled product.

What I do know is that there is a slide that very clearly shows Kaveri, Seattle, Beema and Mullins going directly to Skybridge with no intermediate step. What reason would they have to leave out Carrizo from that picture?

 

[de5_Roy]

Adaptive Clocking in AMD’s Steamroller
http://www.realworldtech.com/steamroller-clocking/

 

[Cazalan]

Adaptive Clocking in AMD’s Steamroller
http://www.realworldtech.com/steamroller-clocking/

Seems like that combined with Intel's proposal for a better CPU load metric for task scheduling would be a superb combination.

Intel Developer Proposes A New Linux CPU Load Metric
http://www.phoronix.com/scan.php?page=news_item&px=MTY3MjU

 

[de5_Roy]

Seems like that combined with Intel's proposal for a better CPU load metric for task scheduling would be a superb combination.

Intel Developer Proposes A New Linux CPU Load Metric
http://www.phoronix.com/scan.php?page=news_item&px=MTY3MjU

i wonder if either company dabbled in increasing power efficiency in multicore (>2C or 2C/4T) cpus. may be the reason we're seeing these now is because the ulp sector is moving to multicore cpus.

 

[8350rocks]

First, it is you who are confused, you need more than 12 ARM processors from Apple or NVidia to compete with 1 single i7-4930k. Show me ONE x86 metric where ARM even COMPETES 1 to 1.

Also, you continue trying to shift the conversation off topic.

ARM will not scale well upward. Period.

I care not what Papermaster said...you take what Keller said out of context, and Rory Read knows less about APU/CPU architecture than he knows about Tarot Card readings. He came from Lenovo, he knows how to run a business, which is his job...he is not an Engineer, nor was he ever.

You continue very confused.

As shown by Anand review of cyclone architecture, this is a "desktop class" design, which competes in complexity and IPC with a Ivy Bridge or Haswell big core.

Apple will be using its SoC on phones/tablets and thus has to limit its SoC at ~2W TDP

Although any industry expert claims that Apple core design is the best in its class, Apple cannot do miracles and cannot pack i7-4930k performance in ~2W TDP. The i7-4930k is rated at 130 W TDP.

Intel cannot offer anything that competes with ARM for the same TDP, because Intel needs four x86 cores @1.6GHz to match the performance of two apple cores @ 1.3GHz.

There is nothing magical in x86 that prohibits ARM scale up to 100W or even 200W.

I see you start insulting AMD president, despite he is doing an excellent work and re-floating the company. I see that you admit that you don't care about that AMD says, but I already knew since the past year!

A small resume of how you are living in a different universe than rest of us:

AMD: the new K12 core is a high-frequency high performance core that will span the range from laptops to desktops to servers—and not just microservers.
8350rocks: the new K12 core is only for ULP and microservers.

AMD: The TAM for x86 is declining. The TAM for ARM is increasing.
8350rocks: x86 is gaining ground and replacing ARM.

AMD: ARM and x86 are sister cores and treated as first-class citizen
8350rocks: ARM is only an "experiment" (my friend-at-AMD said so, trust me).

AMD: Our ARM core will be a bigger-engine that our x86 core
8350rocks: ARM will not scale well upward. Period.

AMD: We are abandoning the CMT design for our new cores
8350rocks: No. AMD new cores will be based in a redesigned CMT architecture

AMD: The Bulldozer architecture was as failure. The chief architect and the rest of the team were fired.
8350rocks: No. It rocks.

AMD: AMD now takes a bold step forward and has become the only company that can provide high-performance 64-bit ARM and x86 CPU cores paired with world-class graphics
8350rocks: Rory Read knows less about APU/CPU architecture than he knows about Tarot Card readings.

AMD: All our 28nm products will be 28nm bulk
8350rocks: Kaveri is delayed because is made on 28nm FD-SOI, trust me

AMD: We are migrating to 20nm bulk and then FinFET on bulk
8350rocks: AMD is returning to FD-SOI for 20nm and then FinFET on SOI, trust me

AMD: Steamroller has same ALUs per core than Piledriver
8350rocks: I know for a "fact" that Steamroller has extra ALU per core

AMD: We are transforming into an APU/Soc company
8350rocks: No. AMD will continue designing dCPUs forever.

AMD: We selected a 10TFLOP APU for our exascale supercomputer
8350rocks: No, AMD will use dCPU and dGPU

AMD: only six cores of our PS4 SoC are available to games
8350rocks: No. Eight cores are fully available to games because the OS is run in a separate chip.

AMD: The FX-9000 series is the last.
8350rocks: No. AMD is will be releasing FX-Steamroller CPU for AM3+.

And so on. And so on.

1. Please quote where I said the new core is only ULP, I was not referring to K12, and FYI, K12 != ARM.

2. The TAM for x86 is increasing and it is gaining ground against ARM, did I say it would overtake it? No...in fact I said specifically it was gaining low single digits in market share in the tablet/convertible market...and may make it into phones in the near future.

3. AMD's venture into ARM is NOT going to replace their x86 business anytime soon. It will be a niche market they try to gain foothold in.

4. Their ARM core was never stated, at any point, by anyone you referenced to be more powerful than their x86 parts.

5. I said AMD may go SMT in some form or fashion, (note: CMT is a version of SMT), though they would not go with HTT.

6. I never said BD/PD/SR was the greatest thing since sliced bread, but it is not terrible either. It does some things well, but others not as well.

7. Rory Read is a good BUSINESSMAN AMD is expanding into other markets in an attempt to boost profit and gain market share in developing markets. Of course that is the course he is going to take, he would be a fool to do so otherwise. Though, you should note that Rory Read does not give tech speeches, and his answers about new technologies should be considered as coming from a second hand source. Notice they are all very vague? If you were getting answers from Keller or Kudori you would have more specific examples.

8. I said, and I quote, "If AMD really does go 28nm bulk, it will cost them clockspeed, and any performance gains will essentially be negated due to high leakage, lower clocked parts."

9. Please quote anyone from AMD saying that specifically.

10. Negative, I never said SR will have extra ALUs, in fact, I provided some commentary on the new architecture about better pipelines, and greater efficiency of code execution to improve single threaded performance in addition to greater parallelism.

11. They cannot produce a 10 TFLOP APU for doing anything right now...that pipedream is a long way off, and please quote the reference.

12. This is true...do you have a PS4 SDK? No? Ok, I think I have proved my point...show me where anyone from Sony has stated your "fact"

13. That reference is AGES old. I have since said on numerous occasions that AMD would like to do a FX steamroller part, but the fabs are the issue, and doing it on bulk would be suicidal at best, ludicrous at worst. In addition to the fact it would eat into APU production schedules.

Now that I have corrected all your misinformation, I request any further commentary you provide come directly from AMD, and AMD only. That means no speculation sites.

Thanks for playing.

EDIT: Next time you try to put words in my mouth, make sure you at least get what I said right.

 

[blackkstar]

http://semiaccurate.com/2014/05/07/intel-touts-dominance-chromebooks/

Intel saying that they are winning in Chromebooks against ARM. What is going on Juan? Can you explain this to me? I thought that ARM was superior and everyone would be flocking to it?

It looks like ARM is only doing well in very low power situations and in highly disposable devices like cell phones and smart phones. It would be odd if that would skew sales numbers and cause people to think that one is overtaking the other when one gets purchased more.

It would be kind of like saying that every high end restaurant is going out of business because they don't sell as much food as McDonalds.

 

[jimmysmitty]

http://semiaccurate.com/2014/05/07/intel-touts-dominance-chromebooks/

Intel saying that they are winning in Chromebooks against ARM. What is going on Juan? Can you explain this to me? I thought that ARM was superior and everyone would be flocking to it?

It looks like ARM is only doing well in very low power situations and in highly disposable devices like cell phones and smart phones. It would be odd if that would skew sales numbers and cause people to think that one is overtaking the other when one gets purchased more.

It would be kind of like saying that every high end restaurant is going out of business because they don't sell as much food as McDonalds.

The biggest problem with ARM is that it was designed with low power in mind and it has been mentioned before that it is not nearly as complex as anything Intel has.

ARM just added OoOE to some of its newer designs but both Intel and AMD have way more experience in that and Intel has much better and more advanced predictors than ARM does.

I don't think ARM will ever take over as some people think it will. I think it will still do very well in low power situations and that Intel will push there but I doubt ARM will ever take over servers and desktops as people think they will.

 

[8350rocks]

http://semiaccurate.com/2014/05/07/intel-touts-dominance-chromebooks/

Intel saying that they are winning in Chromebooks against ARM. What is going on Juan? Can you explain this to me? I thought that ARM was superior and everyone would be flocking to it?

It looks like ARM is only doing well in very low power situations and in highly disposable devices like cell phones and smart phones. It would be odd if that would skew sales numbers and cause people to think that one is overtaking the other when one gets purchased more.

It would be kind of like saying that every high end restaurant is going out of business because they don't sell as much food as McDonalds.

The biggest problem with ARM is that it was designed with low power in mind and it has been mentioned before that it is not nearly as complex as anything Intel has.

ARM just added OoOE to some of its newer designs but both Intel and AMD have way more experience in that and Intel has much better and more advanced predictors than ARM does.

I don't think ARM will ever take over as some people think it will. I think it will still do very well in low power situations and that Intel will push there but I doubt ARM will ever take over servers and desktops as people think they will.

You 2 get it...

Maybe a higher power will hit Juan with a bolt of lightning and he will finally get it too...

 

[Cazalan]

i wonder if either company dabbled in increasing power efficiency in multicore (>2C or 2C/4T) cpus. may be the reason we're seeing these now is because the ulp sector is moving to multicore cpus.

They keep adding more P states and the turbo ranges are widening. I'd say it's a progression of the efficiency in overclocking 1 or 2 cores higher and doing the work quicker, vs turning on more cores and running a slower frequency. And less register copying when processes are normally jumping cores.

 

[jdwii]

http://linustechtips.com/main/topic/148246-update-2-official-excavator-to-be-amds-last-modular-core-new-high-performance-core-coming/

Just in case some of you didn't watch this video.

 

[jimmysmitty]

http://linustechtips.com/main/topic/148246-update-2-official-excavator-to-be-amds-last-modular-core-new-high-performance-core-coming/

Just in case some of you didn't watch this video.

I had a feeling that AMD would not give up x86 server designs. That would be the dumbest move they could make. The server market is worth more than the DT market is. Look at CPU prices there.

I think what AMD is using the ARM for is low power servers, much like the 1U Atom blade servers. And while they have a market, there is no way a single ARM based CPU will be able to do what a current high end Opteron or Xeon can do.

As I said, I am interested to see a full server workload on AMDs ARM design and eve other ARM designs like say Appls A7 SoC. Mainly because I think under high workload situations that Opteron and Xeons currently deal with (i.e. a hyper-V with 8 virtual servers including a EMC, AD and multiple other heavily used servr roles) they will choke in comparison.

If we actually look at it as well, currently Windows Server only runs of x64 and I don't see ARM support yet.

Still should be interesting but I still don't see ARM just swooping in and taking over like people think it will.

As well it looks like AMD has seen that the CMT design is not that great. Server or DT, it really didn't take off for either market. They need to get back to K8 class CPUs where they focused on making stronger cores. If you take 4 strong cores it will do better than 8 weak cores. That shows in the i7 4770K vs the FX 8350. Lower clock speed and less of a virtual core yet in most tasks the i7 keeps up with the FX.

I hope the new designers help AMD become more competitive. I want to see Pentium 4 to Core 2 jumps again. Those are way more fun than the miniscule jumps we are seeing right now.

 

[-Fran-]

I'll just leave this here for you guys to read before adding ANYTHING further to the ARMvX vs x86-64 (or ARM vs Intel/AMD) argument.

http://research.cs.wisc.edu/vertical/papers/2013/hpca13-isa-power-struggles.pdf

Enjoy.

Cheers!

EDIT: This bit is part of the conclusion (which you should all read fully)

"Our study suggests that at performance levels in the range of A8 and higher, RISC/CISC is irrelevant for performance, power, and energy. Determining the lowest performance level at which the RISC/CISC ISA effects are irrelevant for all metrics is interesting future work"

 

[Cazalan]

I posted that one before. Prepare for some copypasta from you know who.

 

[juanrga]

Spoiler

A small resume of how you are living in a different universe than rest of us:

AMD: the new K12 core is a high-frequency high performance core that will span the range from laptops to desktops to servers—and not just microservers.
8350rocks: the new K12 core is only for ULP and microservers.

AMD: The TAM for x86 is declining. The TAM for ARM is increasing.
8350rocks: x86 is gaining ground and replacing ARM.

AMD: ARM and x86 are sister cores and treated as first-class citizen
8350rocks: ARM is only an "experiment" (my friend-at-AMD said so, trust me).

AMD: Our ARM core will be a bigger-engine that our x86 core
8350rocks: ARM will not scale well upward. Period.

AMD: We are abandoning the CMT design for our new cores
8350rocks: No. AMD new cores will be based in a redesigned CMT architecture

AMD: The Bulldozer architecture was as failure. The chief architect and the rest of the team were fired.
8350rocks: No. It rocks.

AMD: AMD now takes a bold step forward and has become the only company that can provide high-performance 64-bit ARM and x86 CPU cores paired with world-class graphics
8350rocks: Rory Read knows less about APU/CPU architecture than he knows about Tarot Card readings.

AMD: All our 28nm products will be 28nm bulk
8350rocks: Kaveri is delayed because is made on 28nm FD-SOI, trust me

AMD: We are migrating to 20nm bulk and then FinFET on bulk
8350rocks: AMD is returning to FD-SOI for 20nm and then FinFET on SOI, trust me

AMD: Steamroller has same ALUs per core than Piledriver
8350rocks: I know for a "fact" that Steamroller has extra ALU per core

AMD: We are transforming into an APU/Soc company
8350rocks: No. AMD will continue designing dCPUs forever.

AMD: We selected a 10TFLOP APU for our exascale supercomputer
8350rocks: No, AMD will use dCPU and dGPU

AMD: only six cores of our PS4 SoC are available to games
8350rocks: No. Eight cores are fully available to games because the OS is run in a separate chip.

AMD: The FX-9000 series is the last.
8350rocks: No. AMD is will be releasing FX-Steamroller CPU for AM3+.

And so on. And so on.

1. Please quote where I said the new core is only ULP, I was not referring to K12, and FYI, K12 != ARM.

2. The TAM for x86 is increasing and it is gaining ground against ARM, did I say it would overtake it? No...in fact I said specifically it was gaining low single digits in market share in the tablet/convertible market...and may make it into phones in the near future.

3. AMD's venture into ARM is NOT going to replace their x86 business anytime soon. It will be a niche market they try to gain foothold in.

4. Their ARM core was never stated, at any point, by anyone you referenced to be more powerful than their x86 parts.

5. I said AMD may go SMT in some form or fashion, (note: CMT is a version of SMT), though they would not go with HTT.

6. I never said BD/PD/SR was the greatest thing since sliced bread, but it is not terrible either. It does some things well, but others not as well.

7. Rory Read is a good BUSINESSMAN AMD is expanding into other markets in an attempt to boost profit and gain market share in developing markets. Of course that is the course he is going to take, he would be a fool to do so otherwise. Though, you should note that Rory Read does not give tech speeches, and his answers about new technologies should be considered as coming from a second hand source. Notice they are all very vague? If you were getting answers from Keller or Kudori you would have more specific examples.

8. I said, and I quote, "If AMD really does go 28nm bulk, it will cost them clockspeed, and any performance gains will essentially be negated due to high leakage, lower clocked parts."

9. Please quote anyone from AMD saying that specifically.

10. Negative, I never said SR will have extra ALUs, in fact, I provided some commentary on the new architecture about better pipelines, and greater efficiency of code execution to improve single threaded performance in addition to greater parallelism.

11. They cannot produce a 10 TFLOP APU for doing anything right now...that pipedream is a long way off, and please quote the reference.

12. This is true...do you have a PS4 SDK? No? Ok, I think I have proved my point...show me where anyone from Sony has stated your "fact"

13. That reference is AGES old. I have since said on numerous occasions that AMD would like to do a FX steamroller part, but the fabs are the issue, and doing it on bulk would be suicidal at best, ludicrous at worst. In addition to the fact it would eat into APU production schedules.

Now that I have corrected all your misinformation, I request any further commentary you provide come directly from AMD, and AMD only. That means no speculation sites.

Thanks for playing.

EDIT: Next time you try to put words in my mouth, make sure you at least get what I said right.

I see that you continue seriously confused:

AMD: K12 is a new high-performance ARM-based core.
8350rocks: No. K12 != ARM.

AMD: The TAM for x86 is decreasing, while it's increasing for ARM.
8350rocks: No. The TAM for x86 is increasing and it is gaining ground against ARM.

AMD: ARM and x86 cores will be treated as first class citizen. ARM will win over x86 in the long run.
8350rocks: No, AMD considers ARM will be a niche market.

AMD: Bulldozer family was a failure.
8350rocks: No. Bulldozer family is not terrible. It does some things well.

AMD: We will abandon CMT and will return to classic SMT design.
8350rocks: No. AMD new cores will be based in a redesigned CMT architecture (note: CMT is a version of SMT).

AMD: We promise 10TFLOP APU by 2020
8350rocks: No. AMD cannot produce a 10 TFLOP APU for doing anything right now

Sony: only six cores are available to games two cores are reserved by OS and for background/dev tasks
8350rocks: No. Eight cores are fully available to games because the OS is run in a separate chip.

Sony: Jaguar cores are clocked at 1.6GHz
8350rocks: No. Jaguar cores are clocked at 2.75GHz.

 

[juanrga]

It doesn't say anything about Excavator canceled. In fact, it only show the 2013-2014 roadmaps. In any case I asked AMD if Carrizo/Toronto (Excavator) had been canceled. And the answer is that new plans don't affect previous x86 plans. Thus I will continue expecting Excavator x86 cores for Carrizo in 2015 and then K12 cores in 2016.

I'm not sure I ever saw a roadmap making a reference to a cancelled product.

What I do know is that there is a slide that very clearly shows Kaveri, Seattle, Beema and Mullins going directly to Skybridge with no intermediate step. What reason would they have to leave out Carrizo from that picture?

You aren't sure, but you used that roadmap to sustain your speculation about AMD canceling Excavator and releasing a Kaveri refresh.

I asked to someone from AMD and didn't confirm your speculation.

 

[-Fran-]

I posted that one before. Prepare for some copypasta from you know who.

Really? I didn't read it nor notice it... Sorry for having missed it.

It's a great read to dismiss most of the speculative talk and get some hard evidence on things.

Cheers!

 

[juanrga]

http://semiaccurate.com/2014/05/07/intel-touts-dominance-chromebooks/

Intel saying that they are winning in Chromebooks against ARM. What is going on Juan? Can you explain this to me? I thought that ARM was superior and everyone would be flocking to it?

It looks like ARM is only doing well in very low power situations and in highly disposable devices like cell phones and smart phones. It would be odd if that would skew sales numbers and cause people to think that one is overtaking the other when one gets purchased more.

It would be kind of like saying that every high end restaurant is going out of business because they don't sell as much food as McDonalds.

Years ago, before Bulldozer family, AMD had better CPU designs than Intel, but Intel bribed OEMs to not use AMD processors.

This is happening again. Intel is bribing OEMs to get its mobile chips inside instead of ARM chips. Analysts estimate that Intel is paying OEMs $50 per chip they use. There are at least two reasons why this tactic will not work as it did against AMD: (i) Intel is loosing about one billion per quarter those days and (ii) the competitors aren't small defenceless companies as AMD; e.g., Samsung alone spend about $14 billions annual only on marketing. They could start counter-bribing OEMs if needed.

 

[Cazalan]

And China bribes their exporters with compensation packages funded by their currency manipulation. Call it state funded contra-revenue. All is fair in love an war right?

 

[juanrga]

The biggest problem with ARM is that it was designed with low power in mind and it has been mentioned before that it is not nearly as complex as anything Intel has.

ARM just added OoOE to some of its newer designs but both Intel and AMD have way more experience in that and Intel has much better and more advanced predictors than ARM does.

I don't think ARM will ever take over as some people think it will. I think it will still do very well in low power situations and that Intel will push there but I doubt ARM will ever take over servers and desktops as people think they will.

ARM has been designed for efficiency: i.e., to provide the maximum performance within a given power budget. Traditionally ARM has focused its designs to low power devices such as phones. But the last designs are focused to servers, workstation, and supercomputers as well.

As shown by Anand analysis of cyclone architecture this is a complex design which is at level of complexity of a big x86 core like Haswell. I mentioned this before, but the myth of ARM is simpler and low performance persists. I repeat the analysis:

To begin with, Cyclone is very wide. It can decode, issue, and retire up to six instructions per clock cycle. [...] There is also a massive 192-entry re-order buffer (ROB) — the same size as Haswell’s ROB (which makes sense, given they both make heavy use of OoOE (out-of-order execution).

The brand mispredict penalty goes up slightly, but interestingly there’s a range of penalties from 14 to 19 cycles — the same range as Intel’s newer CPU cores (Sandy Bridge and later [including Haswell]).

[...]

On the actual number crunching side of things, Cyclone is seriously beefy: It has four FPUs (up from Swift’s two), two load/store units (up from one), two branch units (up from one), and there are three FP/NEON units. Working together with the six decoder units and 192-entry ROB, Cyclone can sustain three FP/NEON adds in parallel per clock. To accommodate all of this beastliness, Cyclone doubles the instruction and data caches to 64KB each (per core).

In short, Cyclone is a serious CPU. In the words of Anandtech, “With six decoders and nine ports to execution units, Cyclone is big… bigger than anything else that goes in a phone.” When Apple announced the A7 SoC, one of the slides said it had a “64-bit desktop-class architecture” — and now we know that wasn’t just marketing hyperbole. Where Swift was very similar to Krait and other mobile ARM cores, Cyclone is a big departure from the usual thin-and-light approach of building mobile CPUs.

Again as Anand mentioned:

Apple didn't build a Krait/Silvermont competitor, it built something much closer to Intel's big cores.

It would be a good thing if the old myths are abandoned. AMD's Keller team is developing something more bigger and complex than the cyclone core. During last conference Keller mentioned how he can extract more performance from ARMv8 than from x86-64. Keller also mentioned that the ARM core that his team is designing will have a "bigger engine" than the x86 sister.

From the covering of the conference:

If it enables the folks building data centers to get the same sort of per-thread performance from ARM-based servers that they can from x86 processors, the K12 could obviate the need for x86 CPUs almost entirely. It could help key a dramatic transition from a single dominant ISA to two competing options—or it might even spark a longer transition away from the Intel-dominated x86 world to one ruled by the more open and expansive ARM ecosystem.

And Nvidia has exactly the same plan for servers

http://www.techpowerup.com/172683/nvidia-to-take-on-xeon-and-opteron-with-a-boulder.html

 

[Cazalan]

Really? I didn't read it nor notice it... Sorry for having missed it.

It's a great read to dismiss most of the speculative talk and get some hard evidence on things.

Cheers!

I agree it's a good read. I hope someone does a similar paper next year after more ISA updates are on the market.

 

[juanrga]

I'll just leave this here for you guys to read before adding ANYTHING further to the ARMvX vs x86-64 (or ARM vs Intel/AMD) argument.

http://research.cs.wisc.edu/vertical/papers/2013/hpca13-isa-power-struggles.pdf

Enjoy.

Cheers!

EDIT: This bit is part of the conclusion (which you should all read fully)

"Our study suggests that at performance levels in the range of A8 and higher, RISC/CISC is irrelevant for performance, power, and energy. Determining the lowest performance level at which the RISC/CISC ISA effects are irrelevant for all metrics is interesting future work"

As Caza mentions, that paper was cited before in this thread. It was then analyzed and explained why the conclusions doesn't stand up on close inspection. It was also mentioned then that one of the authors is closely bounded to Intel (what includes receiving grants). I will leave you to guess who

 

[palladin9479]

Samsung alone spend about $14 billions annual only on marketing.

Samsung Telecommunications is not a singular company, they are multiple companies that all fall under the umbrella of the Samsung Group. It's a big corporation that's run by a single Chaebol family here in SK. They do construction, life insurance, financial investments, medical research, defense research and have their hands in dozens of industries. Heck they used to be my ISP until I switched to LG Powercomm (which is part of the LG group). They have an annual revenue 5x greater then Intel and could buy Intel if they really wanted to.

https://en.wikipedia.org/wiki/Samsung
https://en.wikipedia.org/wiki/Intel
https://en.wikipedia.org/wiki/Apple_Inc.

Nobody should be underestimating the behemoth that is Samsung.