AMD Kaveri APU Successor Named, Carizo Coming In 2015

[Over_Lord]

Even though AMD's next-generation Kaveri APU, based on the Steamroller x86 CPU, GCN (Graphics Core Next) GPU, and HUMA memory architecture might not arrive until late in 1H of 2014, its successor is already named and being worked on by AMD. Carizo, successor to Kaveri, will arrive sometime in 2015, we are betting before the end of 1H 2015, and will feature further improvements in AMD's big APU push.

While nothing is confirmed, Carizo most probably packs AMD's next-in-line Excavator CPU core and perhaps an improved GCN based GPU core. Memory architecture will undoubtedly feature further improvements, in line with AMD's big push for heterogeneous computing and unified memory.




"In the first half of 2014, AMD plans to release two more HSA-based APU series: Berlin for server products and Beema, to replace the existing Kabini APU series. In 2015, AMD will release Carizo and Nolan APU series to replace Kaveri and Beema, respectively, both featuring HSA."

There's no word on the manufacturing process, but AMD may move on to a smaller 20nm node by then, only to compete with Intel's 14nm Skylake platform.

View at TechPowerUp Main Site

 

[Ghost]

We just don't want AMD's APUs manufactured on a 28nm process go up against Intel's 14nm Skylake platform, which is scheduled to launch in mid-2015.

Who's we, you and that Turd in your pocket?

 

[Dent1]

but we sincerely hope that AMD adopts something smaller than 28nm by then. We just don't want AMD's APUs manufactured on a 28nm process go up against Intel's 14nm Skylake platform, which is scheduled to launch in mid-2015.

Not impressed. This is supposed to be news, not your own bias opinion on what "we want". - Just report the news don't add your own views and thoughts and pass them off the consensus amongst the community.

 

[Over_Lord]

Edited article as per complaints.

 

[jagd]

Nice , also i am agree news and comments must be seperate . AMD has no control over production node , it is globalfoundries area now and might confuse people not watching industry . If global foundaries has better node AMD will use if not whatever available will be used.

Edited article as per complaints.

 

[p3ngwin1]

Carizo might most probably...

what the hell does that mean ?

You either want to convey "possible", or "probable".

pick one.

no need for redundancy by saying it's both probably AND possible, because how would something be probable if it wasn't possible ?

 

[Jorge]

Some folks still don't understand that smaller die skrinks below 32nm only show tangible gains in power reduction, NOT in compute performance. They will eventually get the tech memo...

Since Read took command AMD has been working on bringing their products to market faster than originally scheduled as has GloFo with 28nm/20nm concurrent development. This is a bit tricky as there are many aspects of CPU/APU design that are dependent on the development of other technologies. In the end I see AMD continuing to deliver top notch products in all segments and at great prices.

Computing power is so good these days that you can buy a very good system from AMD for hundreds of dollars less than a comparable performing Intel product, so it's not a difficult decision for those who want the most bang for their buck. Excavator should be a big performance jump as will be Steamroller. It's all good for consumers.

 

[TheinsanegamerN]

Some folks still don't understand that smaller die skrinks below 32nm only show tangible gains in power reduction, NOT in compute performance. They will eventually get the tech memo...

Since Read took command AMD has been working on bringing their products to market faster than originally scheduled as has GloFo with 28nm/20nm concurrent development. This is a bit tricky as there are many aspects of CPU/APU design that are dependent on the development of other technologies. In the end I see AMD continuing to deliver top notch products in all segments and at great prices.

Computing power is so good these days that you can buy a very good system from AMD for hundreds of dollars less than a comparable performing Intel product, so it's not a difficult decision for those who want the most bang for their buck. Excavator should be a big performance jump as will be Steamroller. It's all good for consumers.

die shrinks NEVER result in compute performance increases, but being able to include more transistors because of those die shrinks DOES. and I'm still waiting to see amd give the same performance as intel in the mobile market (their cpu performance is a joke, and lower than any core i5 or i7, and barely able to keep up with the i3) or the desktop market (same performance, but over twice the power consumption.)
I want AMD to succeed, but richland launched in march. only the msi gx series uses it. kaveri wont come out for another 9 months at the earliest.....and by then, intel will be done with their haswell refresh/broadwell chips.

 

[Shinshin]

First article - AMD Kaveri is delayed to 1H 2014.
Second article - AMD Carizo coming in 2015.

That's wrong... Here, I fixed it for you:

 

[Prima.Vera]

Some folks still don't understand that smaller die skrinks below 32nm only show tangible gains in power reduction, NOT in compute performance. They will eventually get the tech memo...

Is easy. Smaller transistors means less power and also you can add more in the same space, meaning more performance on the same area.

 

[Dent1]

Is easy. Smaller transistors means less power and also you can add more in the same space, meaning more performance on the same area.

It's too simplistic view.

What would AMD do with the extra space specifically. Add more cache? I can't see the performance boost being significant enough to warrant the financial burden. The instructions per second and instruction per clock will be relatively the same.

Yes, smaller transistors will benefit from the CPU using less voltage and output less heat, which could translate to higher stock clocks and higher overclocks, but AMD is doing decently in that department already.

 

[TheGuruStud]

If intel's 22nm is actually 26-27, then 14 is more like 18? (I must be a broken record)

If the foundries can get 20nm working well, then AMD will have caught up in process to intel in 2015. Hopefully, they won't be using bulk, either. What a steaming pile that would be.

 

[RejZoR]

I just wish AMD would step up and be able to make the same breakthroughs as they did with Athlon T-bird, AthlonXP and Athlon64. Those were the cool times when AMD was really poking Intel in the eyes. And i really miss those days, eagerly waiting for new hardware releases to see the fight of the giants. But these days, nothing. Not even graphic cards world is as interesting as it used to be. It's realyl boring, everyone releasing hardware with tiny bumps in performance but with the same price as they use to in the past. Ugh...

 

[Dent1]

I just wish AMD would step up and be able to make the same breakthroughs as they did with Athlon T-bird, AthlonXP and Athlon64. Those were the cool times when AMD was really poking Intel in the eyes. And i really miss those days, eagerly waiting for new hardware releases to see the fight of the giants. But these days, nothing. Not even graphic cards world is as interesting as it used to be. It's realyl boring, everyone releasing hardware with tiny bumps in performance but with the same price as they use to in the past. Ugh...

Don't confuse breakthrough with performance.

AMD did make a breakthrough with Bulldozer, it's radical design with two cores per module was considered revolutionary prelaunch. On Paper AMD has taken a bigger gamble and make a bigger stride for breakthrough than Intel. Their breakthrough just didn't translate into performance.

It's like electric cars are a breakthrough, their top speed is slower. At one time powering an entire vehicle without a diesel engine was thought of as impossible, but breakthrough is a breakthrough regardless of its end performance.

 

[WaroDaBeast]

Don't confuse breakthrough with performance.

AMD did make a breakthrough with Bulldozer, it's radical design with two cores per module was considered revolutionary prelaunch. On Paper AMD has taken a bigger gamble and make a bigger stride for breakthrough than Intel. Their breakthrough just didn't translate into performance.

It's like electric cars are a breakthrough, their top speed is slower. At one time powering an entire vehicle without a diesel engine was thought of as impossible, but breakthrough is a breakthrough regardless of its end performance.

C'mon! Give this man a cigar!

 

[Frick]

Don't confuse breakthrough with performance.

AMD did make a breakthrough with Bulldozer, it's radical design with two cores per module was considered revolutionary prelaunch. On Paper AMD has taken a bigger gamble and make a bigger stride for breakthrough than Intel. Their breakthrough just didn't translate into performance.

It's like electric cars are a breakthrough, their top speed is slower. At one time powering an entire vehicle without a diesel engine was thought of as impossible, but breakthrough is a breakthrough regardless of its end performance.

But but but SuperPI is faster on Intel.

Does this mean Steamroller comes then too? I have no idea when it's supposed to come anyway..

EDIT: Originally it was supposed to come 2H 2014, then the Opterons were supposed to come late 2013, but now??

 

[midnightoil]

But but but SuperPI is faster on Intel.

Does this mean Steamroller comes then too? I have no idea when it's supposed to come anyway..

EDIT: Originally it was supposed to come 2H 2014, then the Opterons were supposed to come late 2013, but now??

It's very probable that Steamroller FX / Opteron has been shelved entirely, in favour of an expedited Excavator launch, either at the end of '14 or early in '15.