AMD Piledriver rumours ... and expert conjecture

[Reynod]

We have had several requests for a sticky on AMD's yet to be released Piledriver architecture ... so here it is.

I want to make a few things clear though.

Post a question relevant to the topic, or information about the topic, or it will be deleted.

Post any negative personal comments about another user ... and they will be deleted.

Post flame baiting comments about the blue, red and green team and they will be deleted.

Enjoy ...

 

[truegenius]

if it is compatible, then wow a bonus.
Only thing needed is a bios update from gigabyte (my board)

 

[palladin9479]

if it is compatible, then wow a bonus.
Only thing needed is a bios update from gigabyte (my board)

Well if your board is only PCIe 2.0 then dropping a PCIe 3.0 CPU won't give you 3.0, it should just default back to 2.0 speeds. That's what I mean by compatible. Again I'm not absolutely positive about this, so don't quote me on it.

 

[truegenius]

oh my bonus, it is 50-50 probability 🙁

But strangely, i can overclock my pcie to 150mhz from 100 😱 and fully stable, even it provides me another 50 mhz overclock headroom in videoram (means 1430mhz (1470mhz for c11.5) is stable now while with 100mhz pcie 1380 was stable, default is 1200mhz)

means my pcie is like a v2.5 slot (50% faster than v2.0)

 

[sarinaide]

Is it forseable that they will make revisions to AM3+ such as to cater for the USB 3.0 and PCI-e 3.0. Lucid MVP etc.

 

[esrever]

AM3+ isn't dead. AMD was going for AM4 with steamroller with ddr4 memory but since ddr4 is coming slower than expected, AM4 won't be there for a while.

PCI-e 3.0 can be done on the same pins, just need the CPU and the motherboard to support it. There is a lot of unused pins on AM3 that can be used if needed.

 

[sarinaide]

Are you refering to Bios flashing, don't want to sell my motherboard, I like it a lot.

 

[noob2222]

i have am3+, now where is my steamroller

Here

But as for the hint towards AM3+ lasting a bit longer:

I was told that delaying the socket migration beyond the AM3+, C32 and G34 to new socket is a good move,since AMD can design more aggressive, rather than stop gap,

Read more: http://vr-zone.com/articles/amd-to-survive-and-thrive-still-/15564.html#ixzz1sO5CcLcA

Who wants a stop-gap before DDR4? only thing that may work as a stop-gap would be a 1090FX chpset with pci-e 3.0 support. wouldn't need to change the full socket for it since its just a controller issue.

btw, Intel's pcie controller is on chip, AMD's is still on the motherboard, so the cpu doesn't have to have it.

 

[esrever]

I always thought they were planning AM4 with the new server sockets.

I guess they were going to have fusion be their entire desktop line.

 

[diablo34life]

ROFL NOOB

 

[mayankleoboy1]

has anyone tried the newest Winzip 16.5 that supports openCL on AMD CPU, GPU ,APU ?

want some benchmarks....

 

[-Fran-]

7zip here as well.

And if I remember correctly, the next FM socket (FM2+ or FM3 was it?) will have quad channel DDR3.

Also, what were the new things our new 1090FX chipset have? Has that info been leaked/showed?

Cheers! 😛

EDIT: More good news for BD in Linux: http://www.phoronix.com/scan.php?page=article&item=amd_fx8150_compilers&num=1

 

[Cazalan]

EDIT: More good news for BD in Linux: http://www.phoronix.com/scan.php?page=article&item=amd_fx8150_compilers&num=1

Impressive gains there with different compiler options.

No comparison with Sandy to see how much it helped close the gap though. 🙁

 

[noob2222]

I never payed attention to that before, very interesting, you just gave me something to Google.....
much appreciated.

Its also one of the reasons SB has an advantage in graphics, lower latency to the gpu itself. http://www.anandtech.com/show/2773/2

Instead of crippling graphics on its mainstream platform, Intel did the next best thing and integrated 16 PCIe 2.0 lanes onto the die of its mainstream Nehalem part.

The result: lynnfield was faster in games than its daddy cpu, nahalem with quad-channel memory.

AMD didn't move to on-chip pci-e controller till Llano.

 

[palladin9479]

I've said this before.....
there is no and never was an AM4 in the plans...
you guys thinking of what you want and not of actual facts.
I even corrected 'jimmysmitty' a long time ago.
THERE WAS NEVER TO BE AM4....
period..

the one socket moving forward is to be FM2
and then after that, who knows now...

Not exactly. FM2 does not replace AM3+, never has.

FM1/FS1 / FM2/FS2 is for the APU class of chips. They needed another socket type due to the on-board GPU needing external connections, Intel had already integrated those connections into the platform.

The "high performance" *cough cough* class desktop chips won't have an integrated GPU, at least not one that actually puts a display out. Now AMD could choose to either put everything into FM(3??) or build a new socket (AM4???) when DDR4 comes around. DDR4 ~WILL~ require a new socket, its not like DDR2 ->DDR3 which was a natural evolutionary upgrade, it's a serious redesign to how CPU's talk to memory.

Right Now
High Performance Desktop *cough cough* = AM3+
Mobile Devices = FS1/FS2
Low Performance Integrated Desktop = FM1/FM2

Will they converge into a single socket? Who knows. Regardless of what AMD's press reports about "not competing in high end" might say, their still going to be producing PileDriver / SteamRoller chips which are not Trinity / Kalamari (whatever name) chips.

 

[palladin9479]

http://www.phoronix.com/scan.php?page=article&item=amd_fx8150_compilers&num=1

Reposted from Yuka in the Intel Thread.

Shows the sheer difference compile options can make in performance, specifically the difference between K8 (SSE2 instructions, what Intel's newest compiler does for non-Intel CPUs) and BDVER1 (SSE4.1 / ect..).

This is why I question all benchmarks that do not list their compile options.

 

[jimmysmitty]

Well AM3+ already has all the interconnects for PCIe and dual channel memory, only reason Llano needed a new socket was for the external display for the GPU. AMD could in theory mount steamroller onto AM3+ and it should be compatible with a BIOS update. The only issue I see is PCIe 3.0 vs PCIe 2.0. Does PCIe 3.0 need a new pin out or is it electrically compatible with PCIe 2.0. I think it's compatible but don't quote me on that.

I think it is compatible, thats why the current PCIe 3.0 cards will work in PCIe 2.0, or 2.1 slots.

That said, PCie 3.0 is using 128b/130b encoding instead of 8b/10b encoding and instead has a loss of 1.5% vs 20% for the 8b/10b that PCIe 2.0 uses and does require a wire setup for it.

If it was just plug and go then you could plug a Ivy Bridge CPU into a P8Z68-V mobo without the Gen3 and get PCIe 3.0 but instead you have to have the Gen3 which is equipped with PCIe 3.0 lanes and a switch that switches from PCIe 3.0 to PCIe 2.0.

Not sure of the complete specs behind that but thats my understanding of it.

But I do agree its probably electrically compatible since it still gives the same amount of power as PCIe 2.0 to the bus (150W altough some reviews of GPUs using more than the alloted 450W from the bus and two 8Pin connectors still worked fine so the bus may be able to do more than 150W).

With that said, for Steamroller it may depends. Currently the PCIe controller is still on the chipset, unlike Intels on the CPU die so that means it might depend on the chipset unless AMD moves the PCIe controller to the CPU die itself.

 

[esrever]

http://www.brightsideofnews.com/news/2012/4/19/opinion-are-benchmarks-worthless.aspx

some interesting info in there saying the trinity is more smooth than the i5.

 

[ctbaars]

http://www.brightsideofnews.com/news/2012/4/19/opinion-are-benchmarks-worthless.aspx

some interesting info in there saying the trinity is more smooth than the i5.

It better say that, being that "... Trinity Reviewers Day, AMD did a series of surveys ..." This came up a while back already. Ironic that Theo Valich says that, after saying "... BAPCo: The one-sided marching band that wins billions for one side ..." He does make a very good point about filtering and analyzing data presented by reviewers.

 

[esrever]

I never trusted sysmark and now I know why anandtech uses it so much.

 

[g4114rd0]

AMD Reports First Quarter Results..."

http://www.techpowerup.com/164503/AMD-Reports-First-Quarter-Results.html

 

[noob2222]

I love how sysmark was tested on a via chip (doesn't have cpu branding id), spoofed it an intel and gained 10% higher numbers. can't find the link right now.

 

[fazers_on_stun]

http://www.brightsideofnews.com/news/2012/4/19/amd-releases-first-quarter-2012-results-24159b-revenue2c-24592m-loss.aspx

AMD releases First Quarter 2012 Results: $1.59B Revenue, $592M Loss

AMD released its financial results for the first quarter of 2012. The company reported 1.59 billion dollar revenue, -6% QoQ and -2% YoY, while recording a massive loss of $590 million (GAAP). The company actually achieved $138 million operating income and a profit of $92 million, but the GAAP results took a beating with a $703 million charge to GlobalFoundries as the company secured exclusivity for a certain 28 nanometer APU products for 2012, as well as established a baseline for wafer pricing agreement in 2013.

The charge clearly shows that the companies are battling to secure precious 28nm wafers, as they are expected to be the key growth driver during second half of 2012 and 2013.

Rory Read, CEO of AMD said that the company "delivered solid results in the first quarter as we remain focused on improving our execution, delivering innovative products, and building a company around a strategy to deliver strong cash flow and earnings growth. A complete top-to-bottom introduction of new APU offerings, combined with ample product supply resulting from continued progress with our manufacturing partners, positions us to win and grow."

When we take a look into released highlights, we see that the GPU revenue was actually down from 2011 by 7%, with operating income increasing to $34 million. It is obvious that AMD is now selling less graphics cards for higher selling price (ASP), which is a significant increase from $19 million from a year ago.

Still, the key aspect of this financial results is amending the Wafer Supply Agreement with GlobalFoundries, based on "take or pay" agreement. The company transferred the remaining share to GlobalFoundries in exchange "waiving the exclusivity agreement for AMD to manufacture certain 28nm APU products for a specified period." To us, it is a stark evidence that AMD cannot get enough 28nm wafers at TSMC for its upcoming APUs and the company had to react.

Maybe they really should get Intel to fab their CPUs for them, like Jhen-sun Huang at NV joked about, since it appears Intel will be adding Altera as a foundry customer on 22nm.. I wouldda stuck that into the settlement agreement a few years back 😛..

 

[noob2222]

That way intel could charge you for a wafer and only give you 2 chips? Looks like amd is working on giving gf the boot, as is nvidia to tsmc.

 

[jimmysmitty]

http://www.brightsideofnews.com/news/2012/4/19/opinion-are-benchmarks-worthless.aspx

some interesting info in there saying the trinity is more smooth than the i5.

This sounds like another "Phenom is smoother than Core 2 Quad argument made up when a CPU that came out almost two years after C2Q couldn't beat it per core or per clock.

Smoother is a point of view more than anything. And I can say that after 5 years with a C2Q on both HDDs and SSDs, it felt plenty smooth when doing anything, even multitasking (multiple browsers, gaming movies etc).

I've said this before.....
there is no and never was an AM4 in the plans...
you guys thinking of what you want and not of actual facts.
I even corrected 'jimmysmitty' a long time ago.
THERE WAS NEVER TO BE AM4....
period..

the one socket moving forward is to be FM2
and then after that, who knows now...

There will still be a "AM4". It may not be called that but when DDR4 comes out, there will be a major change and not even AMD will be able to escape it since the way it works will be much different from current DDR.

From what I remember, each DIMM will be connected to the IMC individually, so basically if you have 4 DIMMs each is connected to the IMC and not seperated via channels.

As well, it will be faster and probably have a different pinout.

AMD might call it FM4 or they may call it AM4 or even super socket 7 for all we know.

 

[jimmysmitty]

^I like Super Socket 7. Bring back some old school into this modern age....