AMD - Piledriver chitchat

[CAT-THE-FIFTH]

Test of the new AMD stock cooler:

https://www.youtube.com/watch?v=ocvZt3ATIl8

I suspect these will come with Zen and look a good improvement over the older ones.

[abaxas]

Test of the new AMD stock cooler:

https://www.youtube.com/watch?v=ocvZt3ATIl8

I suspect these will come with Zen and look a good improvement over the older ones.

God that bloke drones on and on and on and on and on.

[DanceswithUnix]

Zeppelin??? Really AMD, you call a CPU Zeppelin???

https://lkml.org/lkml/2016/1/29/78

That has to be an engineer's little joke, yes?

[shaithis]

Zeppelin??? Really AMD, you call a CPU Zeppelin???

https://lkml.org/lkml/2016/1/29/78

That has to be an engineer's little joke, yes?

As long as they don't go all "Hindenburg"!

[CAT-THE-FIFTH]

It looks like AMD has gotten another 15% extra performance out of Bristol Ridge when compared to Carrizo due to better power and thermal management:

http://www.eetimes.com/author.asp?se...&page_number=2

Its been pretty impressive what AMD has managed to get out of an old process node.

It makes me wonder once they get to 14NM,they will have a whole toolbag of tricks they can apply to Zen!! I wonder if GPUs like Polaris are being helped out too??

They appear to be on the same process node.

Zeppelin??? Really AMD, you call a CPU Zeppelin???

https://lkml.org/lkml/2016/1/29/78

That has to be an engineer's little joke, yes?

Could be a Led Zeppelin reference??

[CAT-THE-FIFTH]

Another example of AMD PR screwing up AGAIN:

http://www.anandtech.com/show/10000/...roughly-tested

We had four Carrizo devices on hand to test for a week, along with a single Kaveri system. These devices were sourced by AMD, and I put in requests for a variety of price points, hardware configurations and styles, along with some specific testing equipment to which we don’t have access.

All the Carrizo laptops which were sent out were running in single channel and the Kaveri system in dual channel.

What a joke!!

Edit!!

It also appears some of the Carrizo laptops still run their RAM in single channel even if two DIMMs are used,if the OEM has skimped on the motherboard.

Second Edit!!

It appears that laptop is an early prototype so probably why it had issues:

http://forums.anandtech.com/showpost...&postcount=114

LMAO,so not only do they not configure the laptops they send probably,one of the models is not even a production equivalent example which probably lacks the issues AT encountered.

[CAT-THE-FIFTH]

CERN leaks some details about Zen:

http://wccftech.com/amd-zen-opteron-processor-32-core/

One key point explained by Liviu is that AMD might go with the MCM design as they currently do with their Opteron 6300 series processors.

[watercooled]

Interesting that it should come from CERN - I wonder if there's been some interest there?

Also with regard to the MCM comment, I wonder if 32 cores means a 4 die package, each with two memory channels? I'd expect some sort of new inter-chip fabric if that's the case - maybe something based on the SeaMicro tech? Or it could just be something like Hypertransport of course, but unless it's been improved it might make the processor more like a multi-socket system with the usual NUMA concerns.

[scaryjim]

... I wonder if 32 cores means a 4 die package, each with two memory channels? ...

There were rumours of a 16 core / quad channel / greenland graphics Zen APU not so long ago, targeting the HPC market, so this leak would be consistent with two of those chips. Could be a 2 socket workstation, or some kind of MCM (perhaps 2.5D on an interposer, even).

Alternatively, given AMD's modular approach, it could easily be a semi-custom HPC x86 processor that switches the graphics portion of such an APU for more cores and memory channels in a single chip. AFAIK Piledriver is still fabbed at 32nm, so a shift to 14nm finfet is a massive process node leap: combining the node leap with space-optimised layouts could easily give you space to jam in that many cores , and massively parallel processing won't mind sacrificing some clock-speed to the high density libraries if it gives you another 16 threads to play with....

[CAT-THE-FIFTH]

Looks like CCL Computers have listings for the new AMD chips recently released:

http://www.cclonline.com/product/202...ooler/CPU0400/
http://www.cclonline.com/product/202...ooler/CPU0402/
http://www.cclonline.com/product/202...ooler/CPU0403/
http://www.cclonline.com/product/202...ooler/CPU0407/
http://www.cclonline.com/product/202...ooler/CPU0409/
http://www.cclonline.com/product/202...ooler/CPU0410/

The X4 845 is only £50.

[DanceswithUnix]

I was interested to read this article about 10nm and 7nm cpu releases http://www.fudzilla.com/news/process...knifes-samsung

There is no mention of x86 devices there, all Samsung, Apple and Qualcomm.

So TSMC are mass producing 10nm in June? AMD are going to be releasing Zen on a generation behind in silicon process?

[kalniel]

So TSMC are mass producing 10nm in June?

Presumably memory and/or small SoC things rather than GPU/CPUs?

[scaryjim]

So TSMC are mass producing 10nm in June? AMD are going to be releasing Zen on a generation behind in silicon process?

*shrug* AMD has generally been at least a process behind for quite a long time now. Having a quick google it certainly looks like TSMC expect to be shipping 10nm products this year, and it wouldn't be entirely surprising if they landed before AMD manage to kick a 14nm part out the door. OTOH Intel won't be putting x86 on 10nm this year (AFAIK, anyway) so there will be at least some time where AMD have node parity with Intel. That last happened in October 2011 with the release of Bulldozer on 32nm, and lasted until April 2012 when Ivy Bridge went 22nm.

Presumably memory and/or small SoC things rather than GPU/CPUs?

The article basically straight-up states that Apple's next application processor will start production in June for an autumn release of whatever number iPhone we're up to now. To me that looks optimistic, but I wouldn't rule it out. So mobile SoC, but not what I'd call a small thing...

[HalloweenJack]

GloFo have had 14nm in volume on FinFet LPE for a year now

[DanceswithUnix]

GloFo have had 14nm in volume on FinFet LPE for a year now

14nm has been around long enough that Qualcomm are starting to use it for the Snapdragon 625 budget processor:

http://www.anandtech.com/show/10030/...nd-lowend-socs

The article basically straight-up states that Apple's next application processor will start production in June for an autumn release of whatever number iPhone we're up to now. To me that looks optimistic, but I wouldn't rule it out. So mobile SoC, but not what I'd call a small thing...

Yes, those Apple chips are pretty large for a mobile device so interesting that Apple always push the envelope on silicon processes given their reputation for getting the highest profit margins they can.

[watercooled]

Interestingly, Apple's A9X is actually larger than the quad core Skylake!