Intel may dismiss tick-tock after Haswell, no performance CPUs (LGA based) anymore?

[boxleitnerb]

http://translate.google.de/translat...umn/ubiq/20121122_574440.html?ref=rss&act=url

The googlish is a bit hard to read, but here are what I believe to be the main points/speculations:

• Only SoCs with Broadwell and later CPUs. Mainboards and CPUs will no longer be separate.
• More focus on tablets/notebooks/...
• Lower TDP (10W and possibly less), cannibalizing Atom CPUs/APUs product line

They are going on saying that Intel might abandon their tick-tock philosophy and develop a different architecture to compete with ARM at some point.


For us enthusiasts and DIY PC builders, that might mean having to order relabeled Xeons instead as there might not be dedicated desktop LGA SKUs in the future.

 

[ShintaiDK]

Never trust google translate. Specially not from asian languages to english.

Broadwell will be LGA1150. Skylake and Skymont will be LGA based too: last 2 will have the PCH on the CPU package. Either ondie or shared like the Haswell/Broadwell ULT. Basicly meaning Mbs become downgraded connector boards only. We might see 10-20$ MBs then.

And tick-tock continues. Intels server roadmap also confirms it. (Note server versions are behind desktop/mobile.)

Haswell and Rockwell needs to be reserved tho. Since Rockwell is the platform name.

 

[_Rick_]

The "only SoC" is nothing new really, after all the only thing that remains to be integrated is the PCH and its IO-functionality, and that has been on the road map for a while. The VRMs are already disappearing with Haswell.
I doubt that SoC means obligatory low power, and I also doubt that they will stop making desktop CPUs, simply as the market won't disappear, and the cost of transforming a server CPU into a desktop CPU is relatively low. "Relabeled Xeon" is what Intel has been selling as desktop CPUs for quite a while, as their two desktop plattforms are very compatible with their server platforms.

Even with Broadwell, I doubt that current 70W workloads will be able to fit into 10W, so there will remain a market for bigger dies/higher voltages/clocks.

Atom is on its way into phones, and there Broadwell's 10W are pretty gigantic. But I doubt we will ever see Atom in anything else again. Currently Intel excels at iterating over x86, and bombed when it came to develop a new architecture. I'm not sure they're willing to go that way again, especially as there is zero gain to be had.

 

[dma0991]

Even without Tick Tock, it wouldn't mean that there wouldn't be performance CPUs. The method of separating a new architecture from a process shrink doesn't determine that performance will rise/fall. With it being SoCs, it doesn't mean that desktop CPUs would be a BGA.

Full sized laptops CPUs are still PGA socket, with the exception for Ultrabooks. I think BGA is only used where weight and size(thickness) is a concern. For full sized laptops it will remain as PGA and LGA for desktops.

 

[Nemesis 1]

http://translate.google.de/translat...umn/ubiq/20121122_574440.html?ref=rss&act=url

The googlish is a bit hard to read, but here are what I believe to be the main points/speculations:

• Only SoCs with Broadwell and later CPUs. Mainboards and CPUs will no longer be separate.
• More focus on tablets/notebooks/...
• Lower TDP (10W and possibly less), cannibalizing Atom CPUs/APUs product line

They are going on saying that Intel might abandon their tick-tock philosophy and develop a different architecture to compete with ARM at some point.


For us enthusiasts and DIY PC builders, that might mean having to order relabeled Xeons instead as there might not be dedicated desktop LGA SKUs in the future.

Haswell can already run in 8 watts so this is a hatchet article nothing more .

 

[boxleitnerb]

The most interesting point was that there might not be LGA parts for the desktop anymore after Haswell. But of course that is only speculation right now.

 

[Nemesis 1]

The 14nm shrink to broadwell is after haswell . It will use the same socket as haswell.
I know its crappy when little things like Shrinks get in the way of made up stories .

 

[GreenChile]

Why would Intel give up on one of the best serving strategies they've ever come up with. Tick-Tock has been a huge success.

 

[boxleitnerb]

Because at some point it might become too expensive to shrink x86 further and further. Maybe a good analogy is the combustion engine. It has been around for a long long time, but it will get replaced by fuel cells and/or electric engines eventually. If you still stick with the old tech, you might get overrun in the long run.

 

[GreenChile]

Because at some point it might become too expensive to shrink x86 further and further. Maybe a good analogy is the combustion engine. It has been around for a long long time, but it will get replaced by fuel cells and/or electric engines eventually. If you still stick with the old tech, you might get overrun in the long run.

I'm sure that is true for the traditional silicon transistor. At some point a wall will certainly be hit which prevents economical scaling but do you think that will be the end of development? When that happens everyone will need to shift to more exotic transistors. Hopefully the technology will be ready when it's needed.

Believe me Intel, along with every other silicon manufacturer, is actively researching this. Mark Bohr recently stated silicon has a path down to 5nm.

Tick-tock is what has given Intel such a large process lead on the competition and their future depends on it. It's not going anywhere.

 

[boxleitnerb]

Of course Intel will work with other technologies. But one has to question how efficient x86 can be for future applications, even with extensions and improvements. If they really want to compete with ARM, they might need something completely new.

 

[GreenChile]

Of course Intel will work with other technologies. But one has to question how efficient x86 can be for future applications, even with extensions and improvements. If they really want to compete with ARM, they might need something completely new.

Perhaps. But that is not tied to tick-tock.

 

[mikk]

Of course Intel will work with other technologies. But one has to question how efficient x86 can be for future applications, even with extensions and improvements. If they really want to compete with ARM, they might need something completely new.

Medfield doing pretty good considering it's just an old 32nm in-order Atom based singlecore chip. x86 is surely not the problem.

 

[boxleitnerb]

Perhaps. But that is not tied to tick-tock.

Why not? As I see it they need tick-tock to maintain their tech advantage over the competition. If they were to develop better CPU/GPU tech, they could slow down a bit and save quite a bit of money.

 

[blastingcap]

For us enthusiasts and DIY PC build...s more cosmic radiation than you might think.

 

[GreenChile]

Why not? As I see it they need tick-tock to maintain their tech advantage over the competition. If they were to develop better CPU/GPU tech, they could slow down a bit and save quite a bit of money.

Tick-tock is the cadence for shrinking process technology and creating new microarchitecture. It is not specifically tied to x86. If they decided to develop a new microarchitecture or even utilize Arm microarchitecture they could still continue the tick-tock cadence.

I won't be surprised to see that cadence spread out to an longer time period though. It appears Intel has already fallen behind on the 2 year cadence. It seems more like 2 1/4 years now.

 

[Sheep221]

The tick-tock cadence is a business model but they are probably no longer able to follow it, it's nowhere said there will be any change in CPU assembly or package.

 

[meloz]

• Only SoCs with Broadwell and later CPUs. Mainboards and CPUs will no longer be separate.

Makes sense, considering Intel are finally getting serious about reducing the platform idle power consumption as well, and not just aiming for pure performance / watt under load.

• More focus on tablets/notebooks/...

Makes sense, considering this has been their weak area. They will not ignore Atom anymore. Good.

• Lower TDP (10W and possibly less), cannibalizing Atom CPUs/APUs product line

Market forces will decide that. There's room for both powerful and 'good enough' SoCs in mobile / portable gadget space.

For us enthusiasts and DIY PC builders, that might mean having to order relabeled Xeons instead as there might not be dedicated desktop LGA SKUs in the future.

Does not make any sense, and I do not know how you came to that conslusion. Contrary to Apple and mainstream media propoganda, plenty of people still buy desktop PCs. And will continue to buy them, because a desktop PC gives a level of performance and flexibility igadgets do not.

Intel already split the performance and enthusiast segments with their large socket and small socket split of x86-64. The small socket is not going anywhere. It might get more focused towards lower TDP range -say, 5-50 watt- as opposed to 15-75 watt today, but it is here to stay.

And to make sure people continue to upgrade and buy new PCs, Intel will have to give them a compelling reason. So they cannot stall on performance, either. I expect decent progress on performance, and great leaps in performance per watt from Intel.

The fact that they are focusing on their weak areas does not mean they are also going to surrender their strengths.

 

[boxleitnerb]

But what if Intel only sells products like AMDs E-series? A mainboard with an integrated CPU. The one diagram seems to imply that LGA might die with Broadwell. Not only would this kill the mainboard manufacturers, it would most certainly reduce available choices for us consumers.

 

[ShintaiDK]

But what if Intel only sells products like AMDs E-series? A mainboard with an integrated CPU. The one diagram seems to imply that LGA might die with Broadwell. Not only would this kill the mainboard manufacturers, it would most certainly reduce available choices for us consumers.

Are you searching for something?

 

[Phynaz]

they could slow down a bit and save quite a bit of money.

Which is exactly what AMD did. How'd that work out for them?

 

[Sheep221]

But what if Intel only sells products like AMDs E-series? A mainboard with an integrated CPU. The one diagram seems to imply that LGA might die with Broadwell. Not only would this kill the mainboard manufacturers, it would most certainly reduce available choices for us consumers.

It does not make sense, the only boards that come with soldered CPUs are in mini ITX or mATX format and most of the time intended for embedded, industrial or super silent computers with specific use. There are so alot of configurations of motherboards and CPUs that is virtually impossible to sell the stuff this way. Every industry, including semiconductor industry over a time developed practical standards, this include CPU to be removable part and not bound to the board.

 

[Arachnotronic]

Guys...the "no more tick tock" means that they're going to be doing more drastic changes even with the die shrink. This is an ACCELERATED tick tock, not a slowing down.

 

[tweakboy]

Never trust google translate. Specially not from asian languages to english.

Broadwell will be LGA1150. Skylake and Skymont will be LGA based too: last 2 will have the PCH on the CPU package. Either ondie or shared like the Haswell/Broadwell ULT. Basicly meaning Mbs become downgraded connector boards only. We might see 10-20$ MBs then.

And tick-tock continues. Intels server roadmap also confirms it. (Note server versions are behind desktop/mobile.)

Haswell and Rockwell needs to be reserved tho. Since Rockwell is the platform name.

Nice 1150. But Haswell and Skylake are soo close to each other same year. BTW does anyone know if that chart said 2H 2013 for haswell. ????

 

[ShintaiDK]

Nice 1150. But Haswell and Skylake are soo close to each other same year. BTW does anyone know if that chart said 2H 2013 for haswell. ????

Its a SERVER chart. Not desktop/mobile. And it says H1 2014 for Haswell-E/EP/EX.

Haswell-DT is still for march 2013.