AMD AM4 CPU, APU And Motherboard Spotted On Zauba –Spring 2016 Launch Imminent

[monstercameron]

It´s actually 15W chip, which can be configured up to 35W Currently there are no > 15W TDP models available, but it will change soon. All currently available Carrizos should allow cTDP configuration 12-35W, or at least I´ve not seen any which wouldn´t have.

I must admit that I don´t like the way AMD has been marketing Carrizo (business as usual...). For instance there is a huge confusion about the TDP:

Carrizo (8K series or 6th gen) defaults to:

- 15W TDP
- 18W BBB (Boost TDP limit on battery)
- 25W STAPM PPT (Boost TDP limit on AC, availability depends on timer and temperature)
- 12-35W cTDP (ODM configurable, if not set then 15W)

Yeah...

its...flexible?

 

[LTC8K6]

It´s actually 15W chip, which can be configured up to 35W

Oh brother...

 

[Arachnotronic]

Oh brother...

It's not too much different from the Intel U series processors which can scale basically from 15W all the way to 28W.

 

[LTC8K6]

It's not too much different from the Intel U series processors which can scale basically from 15W all the way to 28W.

U chips can have TDP configured down or up, from a middle normal setting, also.

6500U for example is 15W with a TDP UP of 25W and a TDP DOWN of 7.5W.

http://ark.intel.com/products/88192/Intel-Core-i7-6600U-Processor-4M-Cache-up-to-3_40-GHz

 

[Dresdenboy]

Right, but it's pretty well-known that Kaveri (for example) has a wall on air/water of around 4.7 GHz. LN2 barely gets it past 5.2 GHz. If Bristol Ridge can at least reach the same clockspeeds, it'll be interesting.

Of course a done design has its design specific wall. I meant that a new design (modified or from scratch, same or new process, libs, design methods, RCG, etc.) has a new wall depending on all these factors. It's not that easily comparable.

 

[Shehriazad]

Now correct me if I'm wrong...but wasn't the 4.6-4.8 Ghz wall on Kaveri stemming from the specific 28nm process they used?


And....isn't Bristol Ridge made on the same process? So in theory wouldn't the wall still be the same...with the only differences being that Bristol will consume quite a bit less power and have DDR4 ram.


So in the end you'd end up with a 4.7 Ghz Bristol Chip that has about 5-10% more IPC than Kaveri but doesn't really do much more than be the placeholder for Zen release. Of course I wouldn't mind to be wrong and would love it if that thing actually hit 5+Ghz on air/water...because starting with Carrizo there is now a very clear performance gap between the FX chips and their APU-line.
I mean sure...Kaveri was basically always equivalent to an FX chip that was 300-400 Mhz faster...but that was about it. Add another 5%-10% IP on top of that and the performance difference is clear enough...sadly Bristol Ridge is still only 4-core... I'd really like it if they made some 6core versions at least...since now they could afford it as it's gonna be on the new platform, anyway.


Seriously though...6 core with what, 15% more IPC than FX X300 series? At least some people would take a look at it if the price was alright...especially with a Zen upgrade in sight.

 

[Dresdenboy]

Now correct me if I'm wrong...but wasn't the 4.6-4.8 Ghz wall on Kaveri stemming from the specific 28nm process they used?


And....isn't Bristol Ridge made on the same process? So in theory wouldn't the wall still be the same...with the only differences being that Bristol will consume quite a bit less power and have DDR4 ram.

If BR comes with XV cores made with a 28SHP process, you have eliminated one variable, still leaving out design tunings, power management changes, process evolution (even the "same" 65nm process got several improvements over time), used libs (they changed between SR and XV), etc. This might still cause a 20% "OC wall" error.

 

[NostaSeronx]

They could port to 22FDX w/ ULP-UHP extensions. Then, optimize for efficiency rather than for speed. Which would give a hefty shrink. Minor performance loss, which can be replaced with more units within same area. So, there would be no performance loss, but a gain. Alas, it is not the case NEW NEW NEW architecture!

 

[DrMrLordX]

It´s actually 15W chip, which can be configured up to 35W

Okay, but what AMD showed on their reference design in previews (3dmark etc) was the chip operating in a 35W configuration. That's more-or-less what we all expected from the FX8800P in the real world.

Now correct me if I'm wrong...but wasn't the 4.6-4.8 Ghz wall on Kaveri stemming from the specific 28nm process they used?

Pretty much. Well not exactly. The L2 design had something to do with it as well . . .

And....isn't Bristol Ridge made on the same process?

No. It's the same process as Carrizo, which is GF28A which is . . .28spp or whatever the heck it is. GF28a was just the code for the process.

They could port to 22FDX w/ ULP-UHP extensions.!

Not gonna happen, at least not for desktop/mainstream mobile use anyway. 22FDX is for:

http://globalfoundries.com/newsroom...ustry-s-first-22nm-fd-soi-technology-platform

in other words, maybe smartphone SoCs, wearables, and some embedded stuff.