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 ...

 

[viridiancrystal]

but how do the HD 7 series Fold.?
because it seems Fermi is better than Kepler in Folding.
as for the CPU side, more cores the better for Folding but the GPU needs to be stout.

dunno why but FaH doesn't see my 7850. All it seems to see is Cuda... Unless I'm missing something?

 

[viridiancrystal]

I have a fix for that, or should I say I know where there is one..
guys were talking about it the other day.

start towards the half-way point of this page:
http://www.tomshardware.com/forum/page-26158_41_14600.html

ah, no luck. I really want to figure this out now.

 

[Haserath]

Perhaps you can try a 'client-type:beta' flag on your GPU slot?

 

[guskline]

Since the banter at this point seems mostly speculation, what is the best info on when PD will begin to be manufactured? I ask for this reason. How much time does AMD need to make tweeks etc to the PD core before it must go to production ( I assume Global Foundries?).
I've seen a lot of negative reviews, but is there actually a review that talks about realistic improvements that can be achieved over the BD core in the timeline before production?

 

[gamerk316]

The bit JS said about Software not scaling with CPU frequency is nonsense, we're no longer in the days of DOS where software handled I/O interrupts and requires a raw boot disk to work. Software doesn't always scale with core count, this is where software limits come into play.

I/O waits still cause programs to halt in their tracks though, and in Windows, this will cause another non-I/O blocked thread to be run instead. Hence why its impossible to scale perfectly with frequency, as I/O waits remain a major bottleneck.

As for core count, the main factor is how parallel your processing is. If its very parallel, you can get decent scaling. If not, you are going to use one core.

 

[viridiancrystal]

Since the banter at this point seems mostly speculation, what is the best info on when PD will begin to be manufactured? I ask for this reason. How much time does AMD need to make tweeks etc to the PD core before it must go to production ( I assume Global Foundries?).
I've seen a lot of negative reviews, but is there actually a review that talks about realistic improvements that can be achieved over the BD core in the timeline before production?

Well, remembering an article from a while back, AMD was going to their version of "tick-tock" by linking the same cpu arch in their mainstream apu's and their high end processors. They have stated many times that they expect the full PD to be a little better than the Trinity cores, which implies they allotted themselves a little bit of time to make tweaks. I wouldn't expect anything major, but small little adjustments to add a little more IPC.

So, giving their engineers a little bit of time to do so, starting with Trinity's mobile release, I expect PD to be finalized around mid to late July, and production to begin soon after.

 

[jdwii]

http://www.fiercecio.com/techwatch/story/amd-integrate-arm-technology-x86-processor/2012-06-15

+1 to AMD!

Great news this means on Windows 8 we could just use Arm when were doing email, or just online on tomshardware complaining Amd is so behind on a piledriver forum. No but if they can do this it will enable longer battery life on apps that don't need x86.

The next 10 years well be interesting in this career with HSA and Arm.

 

[-Fran-]

http://www.fiercecio.com/techwatch/story/amd-integrate-arm-technology-x86-processor/2012-06-15

+1 to AMD!

Great news this means on Windows 8 we could just use Arm when were doing email, or just online on tomshardware complaining Amd is so behind on a piledriver forum. No but if they can do this it will enable longer battery life on apps that don't need x86.

The next 10 years well be interesting in this career with HSA and Arm.

Well, if they allow the ARM ISA into the APU, sure, why not. I wonder if they can do that in one simple package though... And by package, I mean software layer (driver and API).

I'll wait on Palladin's opinion on that, lol.

Cheers!

 

[jdwii]

http://www.legitreviews.com/news/13413/

Thank's to TSMC, whose 28nm process manufacturing quality is becoming better and better, AMD is able to release the second revision Tahiti XT2. For the record you will be able to find the Tahiti XT2 on the PCB of the new upcoming Radeon HD 7970 GHz Edition but not only. According to our source the Tahiti XT2 chip will be also powering the Radeon HD 7970 X2 for sure and maybe a new version of the Radeon HD 7950. About the new version of the Radeon HD 7950 nothing is sure but it might be possible.

7970 GHz Edition will supposedly come clocked at 1100 MHz a full 175 MHz faster then the original reference HD 7970 at launch, not to mention it will do these clocks at 1.020 volts compared to the 1.175 volts previously.

OMG i still can only fine 1 680 this is so sad its been like this since release. If this is true about the 7970ghz edition Nvidia basically failed this gen real bad.

 

[noob2222]

I don't know. Looking at those benchies I'm really not happy. 3.8GHz vs 2.9? You're talking about a ~33% advantage in clock speed alone. And considering it was clocked so much faster it wasn't faster by much. Temps were also pointed out but isn't Llano 40nm? I am really not happy with where AMD is taking "my" company.

The only people trying to compare clock speed to performance are the ones trying to justify why it sucks. Clock speed comparisons need to just go away other than just for information.

Performance is all that matters to sales, and to businesses cost matters just as much. Trinity is faster than LLano at the same power draw, thats all they care about. They don't go shopping for IPC, they could care less about hmz, they want to save time and money

Run the averages, and you actually see the trio end up within 4 W of each other. A8-5600K averages 101 W total system power use, while the A10 lands at 105 W. The Llano-based APU is in between.

25% faster for 2 watts.

Compared to BD, 3.3 ghz PII was = power draw to 3.6 ghz BD.


On other notes, meant to post this one yesterday.

http://www.anandtech.com/show/6007/amd-2013-apus-to-include-arm-cortexa5-processor-for-trustzone-capabilities

x86 +ARM ... There was some information on this "blank spot" on the APU earlier this year, this is the first implementation.

 

[gamerk316]

http://www.fiercecio.com/techwatch/story/amd-integrate-arm-technology-x86-processor/2012-06-15

+1 to AMD!

Great news this means on Windows 8 we could just use Arm when were doing email, or just online on tomshardware complaining Amd is so behind on a piledriver forum. No but if they can do this it will enable longer battery life on apps that don't need x86.

The next 10 years well be interesting in this career with HSA and Arm.

Uhh...no, thats not how it works. AMD chips suddenly won't have the ability to run ARM versions of windows. Its an extra security layer, nothing more.

http://en.wikipedia.org/wiki/ARM_architecture#Security_Extensions_.28TrustZone.29

 

[Haserath]

It's unfortunate that ARM might be mere cheese on AMD's chili cheese fries. With Windows as the beans to make you gassy and bloated.

Don't worry, you won't miss the arm cores, because even on less beans they perform slowly. Probably worse than the first atom, the hydrogen kind.

It would be cool if in a few years the OS could switch between the two, after arm gets a little more performance, and we could have a low power 'companion core' for desktop browsing.

 

[gamerk316]

It's unfortunate that ARM might be mere cheese on AMD's chili cheese fries. With Windows as the beans to make you gassy and bloated.

Don't worry, you won't miss the arm cores, because even on less beans they perform slowly. Probably worse than the first atom, the hydrogen kind.

It would be cool if in a few years the OS could switch between the two, after arm gets a little more performance, and we could have a low power 'companion core' for desktop browsing.

That would require that the OS, all its underlying components, would be able to switch seemlessly between two separate kernels. So...no, not happening.

 

[Haserath]

That would require that the OS, all its underlying components, would be able to switch seemlessly between two separate kernels. So...no, not happening.

How about boot both at startup. Freeze one while the other is working and save its state on the hard drive, then just let the user switch between them with a quick boot?

Scratch that, only I would be crazy enough to switch between them for the low power tasks.

 

[lilcinw]

I think that would be too disruptive for the average end user. Besides that the user base that would want that feature would be too small for MS to spend the money developing it.

It reminds me of an idea I heard a while back that MBs could include a small flash drive that allowed the PC to boot to Chrome OS for basic web browsing and then could boot to Windows for other tasks. It sounds like an interesting concept but I don't know how many people would actually use it.

Maybe we will see some unique Linux distros that can exploit the ARM core in a useful fashion.

 

[r0ck3tm@n]

That would require that the OS, all its underlying components, would be able to switch seemlessly between two separate kernels. So...no, not happening.

Well, I think that's what VM Ware does.

 

[jdwii]

That would require that the OS, all its underlying components, would be able to switch seemlessly between two separate kernels. So...no, not happening.

I'm sorry to say but if they can find a benefit they will do it, if not they wont. Arm is more efficient then x86 under certain tasks and under others we need x86. Their really is no other reason to put Arm inside of a Amd chip unless this was some what true in the future.

 

[-Fran-]

I'm sorry to say but if they can find a benefit they will do it, if not they wont. Arm is more efficient then x86 under certain tasks and under others we need x86. Their really is no other reason to put Arm inside of a Amd chip unless this was some what true in the future.

Yes, they have another very good reason: the proprietary ARM security stuff.

If they don't have an actually ARM core, they can't run it AFAIK, so that's why they're slapping one in the next APUs.

And regarding the kernel swapping. In Linux you can do that with no issues at all, so I'm sure that in Linux we'll be able to hack the ARM module with no much hassle. Windows on the other hand... Don't have details how they load the kernel, but AFAIK it would be a PITA to do.

Cheers!

 

[bawchicawawa]

Yes, they have another very good reason: the proprietary ARM security stuff.

If they don't have an actually ARM core, they can't run it AFAIK, so that's why they're slapping one in the next APUs.

And regarding the kernel swapping. In Linux you can do that with no issues at all, so I'm sure that in Linux we'll be able to hack the ARM module with no much hassle. Windows on the other hand... Don't have details how they load the kernel, but AFAIK it would be a PITA to do.

Cheers!

Do you think it might be like tegra? 5 cores, the ARM core for low power browsing with the 4 other cores being idle and changing to the 4 cores when an application needs it?


This is what i'm imagining, and i hope it comes true.

 

[Cazalan]

Do you think it might be like tegra? 5 cores, the ARM core for low power browsing with the 4 other cores being idle and changing to the 4 cores when an application needs it?


This is what i'm imagining, and i hope it comes true.

It's just a Cortex-A5 core which is loads weaker than anything cell phones are using today. It's for security only. The powerful cell phones and tablets like iPhone/iPad, Samsung SIII, are using multiple Cortex-A9 cores.

If AMD want's more low power cores they can add more power performance states and get the clock speeds down. P0, P1, P2, P3, .....

Intel was able to take a running windows system with a Pentium core all the way down to 1Mhz. That's a bit extreme but people were happily checking emails with 133Mhz computers over a decade ago. AMD's slowest shipping chip (C-50) runs at 997Mhz and uses just 9 Watts.

 

[gamerk316]

I doubt it though, since only AMD CPus will hace an ARM core and even then, not all AMD CPus will have it for a long time, I doubt developers will waste their time to include an ARM code into their software when pure x86 works just fine.

Well, that too. 😀

 

[sarinaide]

And so I return from "tommy johns" to rather pleasing Trinity initial benchmarks, so the anxiety of whether or not trinity was a step up seems to be eased, now it will be a issue of how much.

 

[vishalaestro]

hey guys when will the consumers can buy fx piledriver cpu's is there any date confirmation

 

[de5_Roy]

as far as i know, no confirmed date. my latest rumored release date (LaRRD) is october, when windows 8 comes out (another larrd) and also the time when bulldozer launched last year. that way amd gets to keep up with their promise of yearly release cycle.
you'll see piledriver when amd launches it.

 

[JAYDEEJOHN]

The vast majority of times its PEBKAC, but just like anyone, they have some glitches.
Havnt used nVidia in awhile (old 8800), but it was OK, just like anything it takes some getting used to.
Looks like PD is where I thought it would be, and the little birdie who told me deserves a cracker