AMD CPU speculation... and expert conjecture

[juanrga]

I do agree GIMP is not bad but it's not going to be easy to get people to switch.

The thing is that ARM will never be a majority on desktop, workstation, and mobile workstation and laptops.

Your comparison between the shift from ARM on tablets/phones are not the same as the markets I'm mentioning. There were no mobile phones which offered x86 compatibility before the advent of ARM.

The dominant OS on mobile before 2010 was Symbian. How many people are going to miss all those Symbian applications when they move to Android, iOS, or WP8?

And how many people would have applications they would miss giving up Windows and any OS that has Windows compatibility (WINE, DARWINE, etc)? My guess is a lot.

I don't even agree with you that the Wintel monopoly is crumbling.
https://en.wikipedia.org/wiki/Usage_share_of_operating_systems

If anything these ARM chips are being built to serve a different purpose than x86 chips, and they're finding success by doing what x86 doesn't. It's not meant to replace it and it never will on DT and laptop.

But as I've said it will have a home on tablets and phones forever and x86 won't go anywhere there. Classic x86 applications are useless on that form factor, and x86 has the same problem ARM has on DT where there's not as much software (and yes I'm pretty sure Android VM works on x86 just fine).

ARM will do fine in HPC because those types of folks are running fully custom software stacks that they can do what they please with. Recompiling when you have source code makes it possible, but it's not something people want to do.

For a real world example of people not wanting to go through hoops to recompile things, compare the popularity of distributions like Linux From Scratch and Gentoo to Ubuntu and variants of it (or even Debian and all variants, which all variants of Ubuntu are).

AMD would have a chance of pushing ARM on DT and laptop if they could do like Loongson and implement some x86 instructions for backwards compatibility but they aren't, they're just licensing cores.

ARM's problem when competing with x86 in markets where x86 still dominates is an extremely common problem for hardware that doesn't have software to back it up and the most current example I can think of off the top of my head is Wii U.

It is a good innovative system and people would argue that it's better than xbone and PS4, but it doesn't have that many games. Even with Nintendo releasing a few killer pieces of software (like you're implying ARM will somehow do when they make their giant push for ARM DT) it's still not going to be enough to drag people away from CoD/Halo/Killzone and the vast libraries on Xbox 360 and PS3. Nintendo is only holding on by the fact that it basically has fans and ARM has no consumer level fans for that to work.

Curiously enough ARM DT would suffer the same problem that Ouya console suffers. People look at it and think it's a great idea until they actually go to buy one and run around going "where's COD/Mario/Killzone/Battlefield/etc"

And the same situation will happen with ARM on DT and laptop no matter how much software ARM somehow manages to bring with it.

Of course I think a simpler example of why ARM won't go anywhere on DT and Laptop is comparing Windows RT to Windows 8. Windows 8 is not doing that well but Windows RT is abysmal. Care to guess why?

ARM will stay in mobile like Phones/Tablets, HPC where clients are using open source and can recompile, and in other closed ecosystems (Apple, Smartwhatevers like SmartTVs, SmartFridges, etc).

By suggesting that ARM will win in DT and laptops for actually doing things is suggesting that ARM will replace those products with x86 in them by not being able to do the things that people who buy x86 products buy them for.

GIMP is good and can replace Photoshop, but can you replace all the 3d modelling software? Software like FL Studio, Pro Tools, etc?

Adobe Creative Suite alone is an impossible task for open source to tackle in the next few years and that's not including Autodesk and tons of other companies that make industry standard software.

And lets not even touch on the subject of games again. Would you trade all the games you can play on PC for all the games in Tegra Zone and Google Play + a few titles that get ARM ports?

Once again, the point is not which is better, GIMP or Photoshop, but that GIMP on ARM shows that Adobe could port Photoshop to ARM if there is market for it. The point is that there is not anything magical in x86.

I continue holding my position that ARM will take the desktop.

I was not referring to people shifting from a x86 phone/tablet to a ARM phone/tablet, but shifting from a x86 desktop to an ARM phone/tablet.

That Wintel is over cannot be observed in a shot of the current market share of OS. It can be observed in Microsoft partnership with ARM and development of an OS exclusively optimized for that architecture. It can be observed in Microsoft plans to merge the three Windows in one

http://www.fudzilla.com/home/item/33219-microsoft-to-merge-windows-operating-systems

It can be observed in Intel partnership with Google. It can be observed in Intel plans to scale up Android to get a Windows replacement

http://timesofindia.indiatimes.com/tech/tech-news/software-services/Intel-to-turn-Android-into-Windows-like-OS/articleshow/26349633.cms

Who even mentioned that people will be recompiling its applications at home? Nobody. People will download/purchase applications for ARM, just as today purchase/download applications for x86.

When apple was migrating to an all ARM product line, customers will purchase software for the new architecture and probably will run older software using some emulator.

ARM64 is a clean and modern ISA. It makes no sense to dirt it with a x86 mesh for unneeded backward compatibility.

The Windows RT fiasco has many sources: bad marketing, slow performance compared to alternatives (e.g. IPad/iOS), heavy restrictions to developers (e.g. there is no Firefox version for RT machines),...

And of course no problem with games:

My sources have confirmed for me that both Sony and Microsoft felt that MIPS didn’t have the right size developer ecosystem or the horsepower to power the new consoles. Then it came down to ARM versus X86 architecture. I am told there was a technical “bake-off”, where prototype silicon was tested against each other across a myriad of application-based and synthetic benchmarks. At the end of the bake-off, ARM was deemed as not having the right kind of horsepower and that its 64-bit architecture wasn’t ready soon enough. 64-bit was important as it maximized memory addressability, and the next gen console needed to run multiple apps, operating systems and hypervisors. ARM-based architectures will soon get as powerful as AMD’s Jaguar cores, but not when Sony or Microsoft needed them for their new consoles.

This was years ago when ARM 64-bit was not ready. Next year, AMD presents Seattle and Hierofalcon with the first ARM 64-bit cores and AMD already admits that are better than jaguar cores used in both consoles. By 2015 Nvidia comes with its custom ARM64 core, which will be faster. If all that "bake-off" were made today, probably the new PS4 and XboxOne would be ARM-based.

I am convinced that next gen consoles (PS5...) will be build over the ARM64 architecture.

Then we will have ARM in phones, tablets, servers, supercomputers, desktops, laptops, workstations, and consoles.

 

[esrever]

ARM has as much a chance to take the desktop as MIPS has at taking over HPC.
Next to nobodys going to buy a desktop in 5 years and the people who do will want legacy software support. GAMES AND WORKSTATIONS. Nobody wants a stationary computer if its not able to run high performance applications. Nobody wants to run high performance applications on ARM if they don't need to. Nobody needs to do it because desktop is a dying market segment and there is no need for new players.

 

[Ags1]

ARM has as much a chance to take the desktop as MIPS has at taking over HPC.
Next to nobodys going to buy a desktop in 5 years and the people who do will want legacy software support. GAMES AND WORKSTATIONS. Nobody wants a stationary computer if its not able to run high performance applications. Nobody wants to run high performance applications on ARM if they don't need to. Nobody needs to do it because desktop is a dying market segment and there is no need for new players.

The desktop will still be here in 5 years, in a big way. In my development organization, it would be senseless to give all the devs laptops, when they currently have nicely maintainable desktops with multiple monitors. Laptops and tablets solve a problem we don't have. I think this goes for any IT-intensive organization.

It makes sense that the desktop platform with a consumer spin for high end software (like ever more immersive VR - and not just for games) will still be around for a long time to come. Especially with VR, the idea of a lightweight carry-able computer is pretty meaningles when your interface is a headset and you want to keep your hands free. It makes more sense to have a box in the corner of the the room with hefty processing power and a strong signal to connect wirelessly to the headset.

 

[Cazalan]

This was years ago when ARM 64-bit was not ready. Next year, AMD presents Seattle and Hierofalcon with the first ARM 64-bit cores and AMD already admits that are better than jaguar cores used in both consoles. By 2015 Nvidia comes with its custom ARM64 core, which will be faster. If all that "bake-off" were made today, probably the new PS4 and XboxOne would be ARM-based.

I am convinced that next gen consoles (PS5...) will be build over the ARM64 architecture.

ARM lost the round. Seatle/Hiero being faster than a part on the market for 1 year already doesn't say a whole lot. NVidia having a part faster than a 2 year old product doesn't say a whole lot. If AMD sits on their thumbs and lets Jaguar/Beema tank then that's just self prophecy.

5 years is a long time in CPU world. The performance/watt disparity gap has already collapsed. If ARM doesn't create a significant advantage they won't switch.

This has gotten way off topic and needs to be moved to a different thread.

 

[blackkstar]

AMD ARM64 parts will not be competing with Temash, they will be competing with Beema and Mullins.

You know, the one in the AMD marketing slides that say it's double the performance/watt.

The problem where you seem to be getting confused is that Beema, Mullins, and Temash are in both consumer and embedded space and the ARM parts are only for embedded (as per the roadmaps). ARM looks to be replacing x86 but only for embedded markets. You know, where Geode was completely useless anyways. The biggest threat this poses is to Via's market and it has nothing to do with x86 dying overall.

http://techreport.com/r.x/2013_11_13_AMD_reveals_2014_APU_roadmap_for_tablets_convertibles/roadmap.png

That is the mobile roadmap. There are no ARM CPUs on it. The embedded one is the one with ARM. The embedded market AMD is talking about moving ARM into is the part of AMD's product stack that involves Geode, G-Series, and R-Series. Opteron is also on the list but AMD doesn't even have ARM anywhere on their roadmaps.
http://www.amd.com/us/products/embedded/processors/Pages/embedded-processors.aspx

Hopefully this helps you understand where ARM fits in with AMD's embedded product roadmap and why AMD ARM isn't going to replace x86 on desktop.
https://en.wikipedia.org/wiki/Embedded_system

I used to be like you and hate x86 until I reasonably looked at the alternatives and realized it's completely not feasible to have a computer you can actually do work on that's not x86.

But the whole thing is like I said. ARM will do well where x86 can't (like mobile toys, embedded, hpc which is dominated by GPGPU now anyways, etc) and x86 will dominate everywhere else.

 

[esrever]

ARM has as much a chance to take the desktop as MIPS has at taking over HPC.
Next to nobodys going to buy a desktop in 5 years and the people who do will want legacy software support. GAMES AND WORKSTATIONS. Nobody wants a stationary computer if its not able to run high performance applications. Nobody wants to run high performance applications on ARM if they don't need to. Nobody needs to do it because desktop is a dying market segment and there is no need for new players.

The desktop will still be here in 5 years, in a big way. In my development organization, it would be senseless to give all the devs laptops, when they currently have nicely maintainable desktops with multiple monitors. Laptops and tablets solve a problem we don't have. I think this goes for any IT-intensive organization.

It makes sense that the desktop platform with a consumer spin for high end software (like ever more immersive VR - and not just for games) will still be around for a long time to come. Especially with VR, the idea of a lightweight carry-able computer is pretty meaningles when your interface is a headset and you want to keep your hands free. It makes more sense to have a box in the corner of the the room with hefty processing power and a strong signal to connect wirelessly to the headset.

I was more taking about consumer space, which most people seem to be content with laptops and prefer them to desktops. And with the introduction of smartphones and tablets, even more people have moved away from desktop computing. Desktops will definitely be here for a long time for business and work environments because most people don't need to carry their work around with them. It is also cheaper for the business to supply desktops than laptops. But in the end, most functionality of a desktop can easily be done on a laptop. This includes running multiple displays.

VR is probably not going to be easily accessible for the time being but I would imagine, you could just as easily plug the VR headset into your smartphone. Computation could massively be done on cloud and consumer devices will become very portable.

 

[palladin9479]

The biggest threat this poses is to Via's market and it has nothing to do with x86 dying overall.

Yeah Via is currently the king for embedded industrial / medical applications. Small niche market but very large profit margin, just long cycle time. The world of DOS, Windows CE, even specialized Linux distro's. Lots of GPIO and legacy communication methods used for automation. Completely outside the consumer space, unless your doing home automation or automobile computing. Intel's been trying to make some headway but it's not an area they focus on and their chips aren't as accommodating of harsh environments.

 

[palladin9479]

I was more taking about consumer space, which most people seem to be content with laptops and prefer them to desktops. And with the introduction of smartphones and tablets, even more people have moved away from desktop computing. Desktops will definitely be here for a long time for business and work environments because most people don't need to carry their work around with them. It is also cheaper for the business to supply desktops than laptops. But in the end, most functionality of a desktop can easily be done on a laptop. This includes running multiple displays.

You need to divide up the market a bit. People aren't "computing" on tablets / phones, that term is entirely too generic and can be interpreted to validate any argument position. When people utilize technology they do it so as to do things. We have to divide up what those things are to get an idea of markets and target demographics. The thing people are using the tablet / phone for is communicating and media consumption. What has moved to phone was what we used MSN Messenger, AIM, IRC, Skype and other communication services for. Forming social circles and keeping in constant contact with them. Humans naturally form tribes, technology allows us to maintain our tribal nature in a modern world where we are constantly moving around. Research, Email, Office Automation, Financial banking, and other such services are staying on the DT due to ease of use. Those may have lightweight applications to allow some implementation of that service on a mobile platform but they will never be as efficient or as useful as what exists on the DT.

 

[esrever]

I was more taking about consumer space, which most people seem to be content with laptops and prefer them to desktops. And with the introduction of smartphones and tablets, even more people have moved away from desktop computing. Desktops will definitely be here for a long time for business and work environments because most people don't need to carry their work around with them. It is also cheaper for the business to supply desktops than laptops. But in the end, most functionality of a desktop can easily be done on a laptop. This includes running multiple displays.

You need to divide up the market a bit. People aren't "computing" on tablets / phones, that term is entirely too generic and can be interpreted to validate any argument position. When people utilize technology they do it so as to do things. We have to divide up what those things are to get an idea of markets and target demographics. The thing people are using the tablet / phone for is communicating and media consumption. What has moved to phone was what we used MSN Messenger, AIM, IRC, Skype and other communication services for. Forming social circles and keeping in constant contact with them. Humans naturally form tribes, technology allows us to maintain our tribal nature in a modern world where we are constantly moving around. Research, Email, Office Automation, Financial banking, and other such services are staying on the DT due to ease of use. Those may have lightweight applications to allow some implementation of that service on a mobile platform but they will never be as efficient or as useful as what exists on the DT.

Those are all true but the expectation of what a person generally wants from a "PC" and how much a smartphone can do has really been getting closer and closer. It just means the market is shifting and a less percent of things would need a desktop. This means the size of the market would likely shrink. It will never completely go away but its not going to be the dominant computing device in consumer's lives.

Im thinking of a device that can wirelessly connect your entire home, can access all your data where ever you go. Can wirelessly connect to displays and input devices and more. All this will be possible from devices the size of a smartphone in the next 10 years or so. All of this is actually possible now but the infrastructure isn't there yet. I would expect it to get there over time.

 

[juanrga]

Excellent summary of MANTLE http://techreport.com/review/25683/delving-deeper-into-amd-mantle-api

 

[juanrga]

I didn't even mention Temash. Why what I never said is being discussed, whereas my points are ignored?

I have said and repeated and repeated and repeated that this ARM assault is not happening tomorrow. This is not happening during 2014, because it is not planned for 2014. Why are 2014 roadmaps for mobile mentioned then?

I have said and repeated and repeated which will be the possible roadmap: first servers and embedded, then HPC, next laptop, and finally desktop. Why is this ignored again and again? Who said that desktops are planned for 2014? Nobody Who said that all the markets will be taken in 2014? Nobody Who said 2015? Nobody

Is this sequence so difficult to understand? First servers and embedded, then HPC, next laptop, and finally desktops.

Nvidia has a similar roadmap. First supercomputers (they already ported CUDA to ARM), then servers, and finally PCs.

I repeat, is this so difficult to understand?

I also like the double standards. When AMD claims in a slide that Seattle is 4x faster than Opteron-X. This is denounced as marketing hype and ARM is (wrongly) labeled as low-performance. LOL

When AMD claims in a slide that Berlin APU is 7x faster than 16-core Opteron. This is denounced as marketing hype and APUs are (wrongly) labeled as cheap gaming. LOL

Now AMD claims in a slide that Mullins is 2x faster than Temash and nobody mentions "marketing hype". LOL I am not refuting AMD claims about Mullins, no, I am merely noticing the double standard used here regarding slides.

 

[juanrga]

This was years ago when ARM 64-bit was not ready. Next year, AMD presents Seattle and Hierofalcon with the first ARM 64-bit cores and AMD already admits that are better than jaguar cores used in both consoles. By 2015 Nvidia comes with its custom ARM64 core, which will be faster. If all that "bake-off" were made today, probably the new PS4 and XboxOne would be ARM-based.

I am convinced that next gen consoles (PS5...) will be build over the ARM64 architecture.

ARM lost the round. Seatle/Hiero being faster than a part on the market for 1 year already doesn't say a whole lot. NVidia having a part faster than a 2 year old product doesn't say a whole lot. If AMD sits on their thumbs and lets Jaguar/Beema tank then that's just self prophecy.

5 years is a long time in CPU world. The performance/watt disparity gap has already collapsed. If ARM doesn't create a significant advantage they won't switch.

This has gotten way off topic and needs to be moved to a different thread.

What round? Consoles? Sure. But it was lost because it was played years ago when ARM64 was not ready. It was a ARM 32-bit vs x86 64-bit. If you read the quote that you snipped. Both Sony and Microsoft are confident that ARM will outperform x86.

Seatle/Hiero are not only faster, but more efficient.

2 year old product? 2015-2014 = 1. Also we don't know how much faster it will be. I am sure Nvidia doesn't spend millions on research to get a custom core that will be only a 10% faster than the standard. I expect huge gains.

There are many rumors. Several claim that Nvidia is designing a chip that will be faster than any Opteron/Xeon. Time will say.

 

[etayorius]

This was years ago when ARM 64-bit was not ready. Next year, AMD presents Seattle and Hierofalcon with the first ARM 64-bit cores and AMD already admits that are better than jaguar cores used in both consoles. By 2015 Nvidia comes with its custom ARM64 core, which will be faster. If all that "bake-off" were made today, probably the new PS4 and XboxOne would be ARM-based.

I am convinced that next gen consoles (PS5...) will be build over the ARM64 architecture.

ARM lost the round. Seatle/Hiero being faster than a part on the market for 1 year already doesn't say a whole lot. NVidia having a part faster than a 2 year old product doesn't say a whole lot. If AMD sits on their thumbs and lets Jaguar/Beema tank then that's just self prophecy.

5 years is a long time in CPU world. The performance/watt disparity gap has already collapsed. If ARM doesn't create a significant advantage they won't switch.

This has gotten way off topic and needs to be moved to a different thread.

What round? Consoles? Sure. But it was lost because it was played years ago when ARM64 was not ready. It was a ARM 32-bit vs x86 64-bit. If you read the quote that you snipped. Both Sony and Microsoft are confident that ARM will outperform x86.

Seatle/Hiero are not only faster, but more efficient.

2 year old product? 2015-2014 = 1. Also we don't know how much faster it will be. I am sure Nvidia doesn't spend millions on research to get a custom core that will be only a 10% faster than the standard. I expect huge gains.

There are many rumors. Several claim that Nvidia is designing a chip that will be faster than any Opteron/Xeon. Time will say.

What happens is that a lot of people here are way too much of a fan boy, i think they don`t like to be wrong, i seen you prove them wrong over and over and over and yet they keep asking for the same "proof" you already posted tons of pages back, also they just cannot accept that AMD will not release Steamroller FX, anyone remember when i posted an article about that? most the usual users of this forum went APE MODE on me, claiming i was spreading rumors and that i was an Intel fanboy... then i posted CHW.net article about AMD plans for 2014 and they again went APE MODE because how can a chilenian site be right? i don`t really feel surprised about AMD plans, we already got tons of sources about AMD plans.

I think you already proved your Kaveri estimations were spot on and i also believe ARM will beat x86, but like you already said a thousand times it will be in the LONG RUN, which means not anytime soon. We already saw TONS of slides and comments from AMD claiming the same, and also from others like nVidia, i don`t know why people don`t get it and refuse to accept it, but we all need to stop talking about ARM in this thread none the less.

On a different Note, I have to say all this you said he said is getting in my nerves... this goes to Juanrga and also the others who gang against him, i am getting tired of all these huge walls of texts demanding proof, and replies to the reply of the reply which was posted 10 pages back... JUST STOP IT you all.

 

[Rum]

Excellent summary of MANTLE http://techreport.com/review/25683/delving-deeper-into-amd-mantle-api

Great read, thanks for the link Juan!

 

[Cazalan]

What round? Consoles? Sure. But it was lost because it was played years ago when ARM64 was not ready. It was a ARM 32-bit vs x86 64-bit. If you read the quote that you snipped. Both Sony and Microsoft are confident that ARM will outperform x86.

Seatle/Hiero are not only faster, but more efficient.

2 year old product? 2015-2014 = 1. Also we don't know how much faster it will be. I am sure Nvidia doesn't spend millions on research to get a custom core that will be only a 10% faster than the standard. I expect huge gains.

There are many rumors. Several claim that Nvidia is designing a chip that will be faster than any Opteron/Xeon. Time will say.

I was merely repeating your words about ARM losing the last console round. Why ultimately doesn't matter because it won't be revisited for another 5 years. The playing field will be quite different. Even IF ARM has a slight advantage in 5 years will Sony/MS throw away 5 years of the retooling they made to transition to x86? That's a hard pill to swallow if they do it every console generation. What performance advantage percentage is acceptable to lose backwards compatibility?

Jaguar started sampling in early 2013, so yes by 2015 that's closer to 2 years. NVidia says a lot of things but they haven't made much traction. They're getting crushed by Qualcomm and MediaTek in the ARM space even after 4 iterations of Tegra. You think Qualcomm doesn't want a bigger piece of that infotainment pie? What about Samsung? Lots of things can happen in 5 years.

 

[blackkstar]

This was years ago when ARM 64-bit was not ready. Next year, AMD presents Seattle and Hierofalcon with the first ARM 64-bit cores and AMD already admits that are better than jaguar cores used in both consoles. By 2015 Nvidia comes with its custom ARM64 core, which will be faster. If all that "bake-off" were made today, probably the new PS4 and XboxOne would be ARM-based.

I am convinced that next gen consoles (PS5...) will be build over the ARM64 architecture.

ARM lost the round. Seatle/Hiero being faster than a part on the market for 1 year already doesn't say a whole lot. NVidia having a part faster than a 2 year old product doesn't say a whole lot. If AMD sits on their thumbs and lets Jaguar/Beema tank then that's just self prophecy.

5 years is a long time in CPU world. The performance/watt disparity gap has already collapsed. If ARM doesn't create a significant advantage they won't switch.

This has gotten way off topic and needs to be moved to a different thread.

What round? Consoles? Sure. But it was lost because it was played years ago when ARM64 was not ready. It was a ARM 32-bit vs x86 64-bit. If you read the quote that you snipped. Both Sony and Microsoft are confident that ARM will outperform x86.

Seatle/Hiero are not only faster, but more efficient.

2 year old product? 2015-2014 = 1. Also we don't know how much faster it will be. I am sure Nvidia doesn't spend millions on research to get a custom core that will be only a 10% faster than the standard. I expect huge gains.

There are many rumors. Several claim that Nvidia is designing a chip that will be faster than any Opteron/Xeon. Time will say.

What happens is that a lot of people here are way too much of a fan boy, i think they don`t like to be wrong, i seen you prove them wrong over and over and over and yet they keep asking for the same "proof" you already posted tons of pages back, also they just cannot accept that AMD will not release Steamroller FX, anyone remember when i posted an article about that? most the usual users of this forum went APE MODE on me, claiming i was spreading rumors and that i was an Intel fanboy... then i posted CHW.net article about AMD plans for 2014 and they again went APE MODE because how can a chilenian site be right? i don`t really feel surprised about AMD plans, we already got tons of sources about AMD plans.

I think you already proved your Kaveri estimations were spot on and i also believe ARM will beat x86, but like you already said a thousand times it will be in the LONG RUN, which means not anytime soon. We already saw TONS of slides and comments from AMD claiming the same, and also from others like nVidia, i don`t know why people don`t get it and refuse to accept it, but we all need to stop talking about ARM in this thread none the less.

On a different Note, I have to say all this you said he said is getting in my nerves... this goes to Juanrga and also the others who gang against him, i am getting tired of all these huge walls of texts demanding proof, and replies to the reply of the reply which was posted 10 pages back... JUST STOP IT you all.

I still don't believe ARM will take over DT. ARM's selling point is efficiency and it matters not at all on DT.

My issue with ARM taking over like some of you are suggesting is that it doesn't solve any problems with x86 in some markets.

Yeah, x86 is big, uses a lot of power, etc. but that doesn't matter in some markets like HEDT.

I already suggested that ARM will do fine in mobile toys (phones, tablets, things you can't actually do real work on) and in situations where the client can compile things themselves like (HPC), but that's it.

ARM might end up with more marketshare/chips sold with that path but it's not going to replace x86 in anything other than HPC. And like I said the real power in HPC comes from GPGPU now, like Tesla cards. The x86 CPU at that point is just some legacy cyst sitting on the back of something nearly an order of magnitude more powerful. HPC will turn into something where the GPUs do all the work and the CPU is a co-processor, which is absolutely perfect for what ARM is good at.

x86 hasn't stood a chance in mobile devices and it never will. ARM will win in tablets and phones but it won't win in laptops. The one thing people love to forget about is advances in battery technology and by the time 2016 comes around the LiON will be obsolete like Nickle Hydride batteries are today. And x86's problems will make up for the fact that it has a really good battery attached.

As for Steamroller FX, yes it's not coming, but there's more than one possible reason as to why it's not coming yet the viva ARM RIP x86 crowd just loves to think that HEDT CPUs from AMD are finished, completely ignoring the fact that there is no good SOI process for AMD to release a new big CPU on. Or the fact that Kaveri is SteamrollerB, implying that SteamrollerA got cancelled because it was designed for SOI and AMD had to port SR to bulk.

FX CPUs still make up a significant portion of AMD's overall sales and they should continue to with AMD getting props for every Xbone and PS4 that gets sold. FX can be made into a more profitable chip on a better node, but 32nm isn't going to be it. So there's no reason to make an SR chip on 32nm that's bigger than 4m when it's going to be massive yet again.

I've seen you ARM guys make the following claims that I simply can't believe

1. Warsaw being listed as for "transitioning" clients means that it's to transition to ARM or APU, not that it's to transition from an old dCPU platform to a new one with HSA

2. Marketing slides that state AMD believes ARM will win in the long run, but without declaring which markets it's going to win in. So you just assume it means every single market that exists

3. Since there's no SR FX chip coming, AMD is done with big FX style chips forever, completely ignoring the fact that there's no good place to make an SOI FX chip. Even if AMD wanted to ignore ARM, APUs, and just make traditional x86 CPUs, they couldn't even do that right now because there's no place to make an high end SOI chip.

4. Since ARM is good at mobile and very efficient, it's going to take over markets despite it's software incompatibility with the largest software library on earth where being good at mobile and efficiency don't matter at all.

The only way ARM is going to take off is if the traditional desktop completely dies and everything is pushed into the cloud, and Valve, EA, and every gaming company has a ton to lose if every device sold is some sort of locked down ecosystem.

Which is why AMD is pushing Mantle so hard and pushing Linux so hard, the game companies don't want to go for ARM and they feel like Windows being pushed into a direction where we're looking at weak portals to a cloud to buy content that isn't what companies like Valve and EA like to deliver.

And it's why the game companies are eating it up like crazy. They know that MS wants to sell weak toys used to consume content instead of a workstation class OS that also is great at playing games, and they want out of it as fast as they can. So they can stick with traditional x86 HEDTs and x86 APUs.

 

[etayorius]

I do think there will be an Excavator CPU in 2015-16, and i even think there MAY still be a chance for a Steamroller CPU, just not in 2014... but i never claimed NEVER will there be another High performance CPU, but we got TONS of proof that AMD may not be releasing any Steamroller CPU in 2014, unless GlobalScrewndries get their act together.

There is a difference between "Not in 2014" and "never more"... and i don`t remember stating the former, i do think AMD will EVENTUALLY move to APUs only, but i think it will be several years later (2017+?), and seriously... who cares as long as the CPU is frigging fast and have a extremely powerful GPU?

 

[iceclock]

We'll at this rate we won't have a intel killer anytime soon sadly

 

[iceclock]

We'll at this rate we won't have a intel killer anytime soon sadly

 

[-Fran-]

I was merely repeating your words about ARM losing the last console round. Why ultimately doesn't matter because it won't be revisited for another 5 years. The playing field will be quite different. Even IF ARM has a slight advantage in 5 years will Sony/MS throw away 5 years of the retooling they made to transition to x86? That's a hard pill to swallow if they do it every console generation. What performance advantage percentage is acceptable to lose backwards compatibility?

Jaguar started sampling in early 2013, so yes by 2015 that's closer to 2 years. NVidia says a lot of things but they haven't made much traction. They're getting crushed by Qualcomm and MediaTek in the ARM space even after 4 iterations of Tegra. You think Qualcomm doesn't want a bigger piece of that infotainment pie? What about Samsung? Lots of things can happen in 5 years.

I'm pretty sure 7 years ago (2006-ish), when the consoles where fresh, no one would have agreed to someone saying that consoles would be turning to X86. They had to turn their software to X86 and specially Sony, had the hardest time I'm sure, having to dump pretty much all the tools they had for the Playstation. And not even with ARM-ish Cell (humor me 😛) and the XBox with the Pentium-ish CPU for the 1st gen offered backwards compatibility last gen. They crammed the previous consoles into the new ones to offer it, haha.

I wouldn't use Consoles for either argument since they're money-making machines with very special/specific uses, so they're not like PCs (or the PC market as a whole) IMO. Its like talking about Desktop, Server and... Consoles, haha. I would say smartphones and handhelds are their own special cases as well.

And I want DDR4, dammit! Well, not really for an FX/Phenom replacement, but for APUs.

Cheers!

 

[Cazalan]

@Yuka - Can't say no one would guess x86 in future consoles because they already had x86 in the first XBox. IBM made some hefty claims and people got on the Cell bandwagon. PowerPC (not ARM-ish) has been hit or miss for a long time.

Forget DDR4. I want HBM/HMC stacked memory and MUCH faster "near" memory on package.

 

[noob2222]

Broken website. Mobile version is getting worse and worse.

I agree with cazalan that we need an juan loves ARM sticky so he can go there and rant all he wants about how great ARM is. His ever-changing stories about he said this but meant that are getting old.

This is about Steamroller not ARM.

 

[palladin9479]

Those are all true but the expectation of what a person generally wants from a "PC" and how much a smartphone can do has really been getting closer and closer. It just means the market is shifting and a less percent of things would need a desktop. This means the size of the market would likely shrink. It will never completely go away but its not going to be the dominant computing device in consumer's lives.

Im thinking of a device that can wirelessly connect your entire home, can access all your data where ever you go. Can wirelessly connect to displays and input devices and more. All this will be possible from devices the size of a smartphone in the next 10 years or so. All of this is actually possible now but the infrastructure isn't there yet. I would expect it to get there over time.

Not really. People don't buy a different PC for every task they do, that's not how technology is used. As long as there is one thing that needs to be done on a PC, then the user will still have a PC. Tablets / Phones didn't take market share away from PC's, this is a common misconception. Correlation doesn't equal causation. PC's hit a saturation point at about the same time that Apple made smartphones stylish which launched the market into the big time. You had a new market open up while an old one had become saturated and thus growth stopped. Financial numbers are usually representative of growth of a market not the actual size. A market can be at zero growth or even slight negative growth and still be alive and active. Remember growth is not size, it's just the measure of purchase's. If I bought a PC last year, and didn't buy one this year but did buy a disposable ARM based smartphone, have I stopped using my PC and converted to ARM computing? No, I'm using the phone for what it does and still using my PC for what it does. This is concurrent usage which is what we are seeing now.

 

[esrever]

Those are all true but the expectation of what a person generally wants from a "PC" and how much a smartphone can do has really been getting closer and closer. It just means the market is shifting and a less percent of things would need a desktop. This means the size of the market would likely shrink. It will never completely go away but its not going to be the dominant computing device in consumer's lives.

Im thinking of a device that can wirelessly connect your entire home, can access all your data where ever you go. Can wirelessly connect to displays and input devices and more. All this will be possible from devices the size of a smartphone in the next 10 years or so. All of this is actually possible now but the infrastructure isn't there yet. I would expect it to get there over time.

Not really. People don't buy a different PC for every task they do, that's not how technology is used. As long as there is one thing that needs to be done on a PC, then the user will still have a PC. Tablets / Phones didn't take market share away from PC's, this is a common misconception. Correlation doesn't equal causation. PC's hit a saturation point at about the same time that Apple made smartphones stylish which launched the market into the big time. You had a new market open up while an old one had become saturated and thus growth stopped. Financial numbers are usually representative of growth of a market not the actual size. A market can be at zero growth or even slight negative growth and still be alive and active. Remember growth is not size, it's just the measure of purchase's. If I bought a PC last year, and didn't buy one this year but did buy a disposable ARM based smartphone, have I stopped using my PC and converted to ARM computing? No, I'm using the phone for what it does and still using my PC for what it does. This is concurrent usage which is what we are seeing now.

Most people don't need to do the things PCs are needed for.

 

[de5_Roy]

some detailed looks into console chips. by detailed, i mean with a big D and a giant tail.

A Look at Sony’s Playstation 4 Core Processor
http://www.chipworks.com/en/technical-competitive-analysis/resources/blog/ps4-core-processor/
Inside the Sony PS4
http://www.chipworks.com/en/technical-competitive-analysis/resources/blog/inside-the-sony-ps4/
Inside the Xbox One
http://www.chipworks.com/en/technical-competitive-analysis/resources/blog/inside-the-xbox-one/

i wish i knew more about chip design so that i could make remarks like "ooh that channel looks so integrated..." and stuff. 🙁