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 ...

 

[warmon6]

no.
AM4 socket was never in the plans.
FM2 will be the 1 socket moving forward AFAIK unless the future roadmap changes.
AM3+ is the last of it's kind.

and currently I have decided to run my 965BE and not go FX-4100.
I'm hoping that the revision and the FX-4170 will be a step in the right direction.

+1

There will be no AM4 as far as we know. FM2 is the socket that is meant to take over the desktop arena.

 

[jimmysmitty]

I took an AM3+ 990X Chipset board recently, which cost me quite a buck, I wonder if AM4 offers something good to make AMD's system of future processor compatibility the Leader in marketting.
So to speak, I really hope Piledriver does something to perform awesomely, yet cost around a Phenom II's original cost radius to get people to purchase it.

If PD performs "awesomley" then it wont be cheap. Just as I said about BD. The price is relative to performance. Always is, except in the ultra high end.

+1

There will be no AM4 as far as we know. FM2 is the socket that is meant to take over the desktop arena.

AM4 would be the AM socket on DDR4. So unless they plan to only have Llano on the DT, I doubt they would kill off a mjaor market that wants just the CPU, then they should have a AM4 when DDR4 hits.

But if FM takes over then any previous AM3+ socket CPUs will be SOL as would any who have a AM3+ mobo, which I would find hard toe believe since AMD is normally better at backwards compatibility by a bit.

Still they will have a new socket for DDR4. Possibly FM3 or FM4.

 

[noob2222]

Loks like AMD is still targeting Q3 for Piledriver.

http://www.fudzilla.com/processors/item/25158-amd-piledriver-to-launch-in-q3-2012

hopefully they can address some issues with B3, then improve PD by their 10-15% claim for C0 or C1 release.

Im curious wether or not Next Gen BD will be developed with AMD's new direction as its supposed to address the lackluster FPU performance (8 core - 4 FPU) that BD is being plagued with.

As for socket AM3+, should be good through most of 2013 cpus as the earliest to expect new cpus after PD would be Q3-Q4 2013, most likely 2014 release.

 

[jimmysmitty]

also the word is that RAM will be skipping DDR4 and going straight to DDR5.
I'll try to find links as well.

http://www.anandtech.com/show/4669/jedec-reveals-the-key-aspects-of-ddr4

JEDEC has specs already set for DDR4. I would imagine that DDR4 will hit in 2012 but we probably wont see it until Intels Haswell (Intel always jumps to the new DDR standard first).

Mainly because DDR4 is supposed to take away the multiple DIMMs per channel and have just a DIMM per channel but also support higher module sizes.

Of course we will have to see but I can easily see Haswell being DDR4 based IMCs.

 

[jimmysmitty]

http://translate.google.com/translate?js=y&prev=_t&hl=en&ie=UTF-8&layout=1&eotf=1&u=http%3A%2F%2Fpc.watch.impress.co.jp%2Fdocs%2Fcolumn%2Fkaigai%2F20100816_387444.html&sl=ja&tl=en

http://www.bit-tech.net/hardware/memory/2010/08/26/ddr4-what-we-can-expect/1

 

[jdwii]

http://translate.google.com/translate?js=y&prev=_t&hl=en&ie=UTF-8&layout=1&eotf=1&u=http%3A%2F%2Fpc.watch.impress.co.jp%2Fdocs%2Fcolumn%2Fkaigai%2F20100816_387444.html&sl=ja&tl=en

http://www.bit-tech.net/hardware/memory/2010/08/26/ddr4-what-we-can-expect/1

how about a article that's made in 2011 or late 2011.

 

[fazers_on_stun]

1.2 Nillion transistors makes a lot more sense now out of the performance that they have been getting with the product ... does it not?

This would tend to suggest (unfortunately) that no amount of tuning or spins is going to improve the performance of the product other than improvements in clockspeed.

Unfortunately as it has a significant IPC disadvantage then a redesign is needed.

Infact they should probably just ditch it and spend time on Llano ... it seems a more promising product.

Hmm, been in your part of the world this past week (Vietnam actually), and just now logged on to read about how AMD mistakenly said 2BN transistors instead of 1.2BN. Kinda big error or perhaps they no longer care that much. Anyway, it seems AMD and GF are not getting anywhere near the "50%" transistor size reduction they should have, going from 45nm to 32nm. IIRC Thuban has around the same number of transistors on a 350mm^2 die size, vs. 315mm^2 for BD..

This, along with the slower-than-expected performance, gives more credence to the rumour about AMD/GF using SoC automated design tools on BD and little or no hand-tuning. And the article MM linked about AMD head-hunting for experienced SoC engineers is yet more evidence. So I expect AMD to announce next month that desktop/server is going to be downgraded or phased out in the next several years..

 

[truegenius]

faster ram can increase performance of gpu in apu with huge margine
also amd had their hands on ram, and thus can introduce faster ram than ddr4
so skipping ddr4 can provide some more lead to amd (in portable computers)

 

[dogman_1234]

So, what about TDR? (Triple Data Rate?)*

 

[gamerk316]

Im curious wether or not Next Gen BD will be developed with AMD's new direction as its supposed to address the lackluster FPU performance (8 core - 4 FPU) that BD is being plagued with.

Too bad no one here predicted a drop in FP performance...

*whistles loudly*

Hmm, been in your part of the world this past week (Vietnam actually), and just now logged on to read about how AMD mistakenly said 2BN transistors instead of 1.2BN. Kinda big error or perhaps they no longer care that much. Anyway, it seems AMD and GF are not getting anywhere near the "50%" transistor size reduction they should have, going from 45nm to 32nm. IIRC Thuban has around the same number of transistors on a 350mm^2 die size, vs. 315mm^2 for BD..

This, along with the slower-than-expected performance, gives more credence to the rumour about AMD/GF using SoC automated design tools on BD and little or no hand-tuning. And the article MM linked about AMD head-hunting for experienced SoC engineers is yet more evidence. So I expect AMD to announce next month that desktop/server is going to be downgraded or phased out in the next several years..

I'll say it again: AMD is most likely going to abandon the desktop; it doesn't fit with their entire Fusion strategy.

 

[palladin9479]

Too bad no one here predicted a drop in FP performance...

*whistles loudly*

Hmm, been in your part of the world this past week (Vietnam actually), and just now logged on to read about how AMD mistakenly said 2BN transistors instead of 1.2BN. Kinda big error or perhaps they no longer care that much. Anyway, it seems AMD and GF are not getting anywhere near the "50%" transistor size reduction they should have, going from 45nm to 32nm. IIRC Thuban has around the same number of transistors on a 350mm^2 die size, vs. 315mm^2 for BD..

This, along with the slower-than-expected performance, gives more credence to the rumour about AMD/GF using SoC automated design tools on BD and little or no hand-tuning. And the article MM linked about AMD head-hunting for experienced SoC engineers is yet more evidence. So I expect AMD to announce next month that desktop/server is going to be downgraded or phased out in the next several years..

I'll say it again: AMD is most likely going to abandon the desktop; it doesn't fit with their entire Fusion strategy.

Hmm I'd say they abandon the high end desktop and have moderate dealings in the low power and mid tier OEM builds. The A8-3510/3530MX is actually quite nice for an incredibly small build.

 

[Reynod]

Hmm, been in your part of the world this past week (Vietnam actually), and just now logged on to read about how AMD mistakenly said 2BN transistors instead of 1.2BN. Kinda big error or perhaps they no longer care that much. Anyway, it seems AMD and GF are not getting anywhere near the "50%" transistor size reduction they should have, going from 45nm to 32nm. IIRC Thuban has around the same number of transistors on a 350mm^2 die size, vs. 315mm^2 for BD..

This, along with the slower-than-expected performance, gives more credence to the rumour about AMD/GF using SoC automated design tools on BD and little or no hand-tuning. And the article MM linked about AMD head-hunting for experienced SoC engineers is yet more evidence. So I expect AMD to announce next month that desktop/server is going to be downgraded or phased out in the next several years..

Well if you had popped in to Perth for a stopover I could have met you for a Redback beer down in Freo.

http://matildabay.com.au/

They don't sell CPU's ... though if they did I fear they would overclock well ... and cause dizziness.



Your point about ADT vs fine tuning the lithographic process seems to fit well with this story from charlie below:

http://semiaccurate.com/2011/11/22/intel%e2%80%99s-analog-shrink-is-for-real/

 

[truegenius]

i too want Ph3
in addition to 32nm, amd can bridge x4 cpu to make it a 8 core and since it is 32nm then power will be around 125w

 

[gamerk316]

If only AMD launched a 32nm “Phenom III”:
Sometimes, I wonder why AMD just didn’t do a 32nm die-shrink of their Phenom II architecture, tweak it to boost IPC performance, raise up the frequency and call it Phenom III.
I doubt that it could have been worse than Bulldozer…

Because its a dead-end arch. Theres only so much more they can do with PII. I mean, it isn't THAT much better then a 9000 series C2Q to begin with...Nevermind the time to develop for the new node, and the fact they would have to eat all those BD losses...

Still sad to see AMD sticking to that "8 core" nonsense though. 🙁

 

[-Fran-]

Hmm I'd say they abandon the high end desktop and have moderate dealings in the low power and mid tier OEM builds. The A8-3510/3530MX is actually quite nice for an incredibly small build.

An underclocked A8 would be amazing, but currently, the A8 is a lil' overkill for a small PC setup. It produces a LOT of heat, it's blowing like 120W or more constantly when on load. Plus, the RAM usage is damn intensive that the sticks are always hot. Not even with my main PC the RAM modules heat that much. The A6 is a better choice for a small build IMO.

Because its a dead-end arch. Theres only so much more they can do with PII. I mean, it isn't THAT much better then a 9000 series C2Q to begin with...Nevermind the time to develop for the new node, and the fact they would have to eat all those BD losses...

Still sad to see AMD sticking to that "8 core" nonsense though. 🙁

They won't tweak for Desktop usage anyway, so it's a lost case with a possible PhIII, haha.

And I agree with the 8 core marketing stunt, it's been dumb and horrible in it's execution for the FX line. They could have had a successful new "CMT Tech" that rivals Intel's HT, but no...

Looking forward to Trinity though. Wanna see what they come up with to compete against the Stars+HD6k in Llano.

On a side note; anyone else with an A8? Does your RAM modules get really hot? XD

Cheers!

 

[simon12]

The closest to a Phenom III is the Athlon II 631 on the FM1 chipset but it lacks the L3 cache (and a black edition) and at stock runs at only 2.6GHz. I did read someone overclocked it to 4.4GHz on stock voltage but others say its a lie. If true the Phenom version could be a Phenom III 955BE @ stock clocks of 4GHz+ that would at least let it universally beat an I3 2100 which at the moment nothing AMD has can @ stock.

 

[dogman_1234]

Actually, I feel a PIII would have to be 28 nm...

 

[dogman_1234]

Marketing.

 

[de5_Roy]

piledriver or piledriver's successor could be marketed as phenom iii or phenom fx like amd used fx again for zambezi.

 

[palladin9479]

An underclocked A8 would be amazing, but currently, the A8 is a lil' overkill for a small PC setup. It produces a LOT of heat, it's blowing like 120W or more constantly when on load. Plus, the RAM usage is damn intensive that the sticks are always hot. Not even with my main PC the RAM modules heat that much. The A6 is a better choice for a small build IMO.



They won't tweak for Desktop usage anyway, so it's a lost case with a possible PhIII, haha.

And I agree with the 8 core marketing stunt, it's been dumb and horrible in it's execution for the FX line. They could have had a successful new "CMT Tech" that rivals Intel's HT, but no...

Looking forward to Trinity though. Wanna see what they come up with to compete against the Stars+HD6k in Llano.

On a side note; anyone else with an A8? Does your RAM modules get really hot? XD

Cheers!

That doesn't make any sense, the A8-3130MX is 1.9 Ghz with a TDP of 45W. There is no way your getting "120W!!!!" of heat off it.

http://en.wikipedia.org/wiki/AMD_Fusion#.22Llano.22_.2832.C2.A0nm.29_2

It's memory bus is DDR3-1600, the exact same I'm using in my current gaming rig and those chips aren't melting holes through the case.

Those are the updates versions of the A8-3500M which is 1.5Ghz with the same TDP of 35W and use's DDR3-1333.

Having used multiple 15inch laptops based on these CPU's I can say they do not get very hot, their actually quite cool compared to my C2D laptop with it's discrete GPU. Their so nice that I've recently decided to order an HP Pavilion dv6z for my GF. She likes smaller laptops but wants to do some light gaming with me (it's a couple bonding thing).

Its under what a discrete GPU can provide but leaps and bounds better then Intel's integrated graphics. Especially when you go from DDR3-1333 to DDR3-1600.

An A6-3400M has the same 35TDP as the A8-3500M, same four cores at 1.4Ghz. A6-3410/3430MX has the same 45W TDP as the A8-3510/3530 and runs at 1.6/1.9.

For light mobile devices their beyond awesome.

Now desktop Llano is speced for 100W TDP at the highest. So again unless your overclocking somehow then it shouldn't be "128W!!!!". 100W for CPU + GPU is pretty amazing actually.

 

[Cazalan]

I think he was referring to the whole PC using 120W at load.

 

[we1shcake]

bulldozer hotfix? http://www.techpowerup.com/156844/AMD-Bulldozer-gets-an-Update-from-Microsoft..html

 

[-Fran-]

That doesn't make any sense, the A8-3130MX is 1.9 Ghz with a TDP of 45W. There is no way your getting "120W!!!!" of heat off it.

http://en.wikipedia.org/wiki/AMD_Fusion#.22Llano.22_.2832.C2.A0nm.29_2

It's memory bus is DDR3-1600, the exact same I'm using in my current gaming rig and those chips aren't melting holes through the case.

Those are the updates versions of the A8-3500M which is 1.5Ghz with the same TDP of 35W and use's DDR3-1333.

Having used multiple 15inch laptops based on these CPU's I can say they do not get very hot, their actually quite cool compared to my C2D laptop with it's discrete GPU. Their so nice that I've recently decided to order an HP Pavilion dv6z for my GF. She likes smaller laptops but wants to do some light gaming with me (it's a couple bonding thing).

Its under what a discrete GPU can provide but leaps and bounds better then Intel's integrated graphics. Especially when you go from DDR3-1333 to DDR3-1600.

An A6-3400M has the same 35TDP as the A8-3500M, same four cores at 1.4Ghz. A6-3410/3430MX has the same 45W TDP as the A8-3510/3530 and runs at 1.6/1.9.

For light mobile devices their beyond awesome.

Now desktop Llano is speced for 100W TDP at the highest. So again unless your overclocking somehow then it shouldn't be "128W!!!!". 100W for CPU + GPU is pretty amazing actually.

It is the Desktop model, actually. A8 3850 + 4xDDR3 1600 CL9 modules.

I couldn't really show you with images, but do trust me the RAM modules and the APU itself produce a LOT of heat. I'm using AS5 and the 965's HSF to cool it off and the heatpipes are hot to the touch (near burning on touch). And the stock HSF for the 965 is WAY better than the one that came boxed with the A8.

That's why I want to know if some other people have noticed the high temps.

For reference, in the UEFI (BIOS xD) CPU screen, it shows 70ºc and NB at 45ºc. That's quite quite a lot in my book for a 100W TDP part, that's why I'm guessing it's puttin' around 120W of heat.

My fear is that, since Llano is also a 32nm SOI, PD within Trinity will be even hotter. Hope the 32nm have matured enough to be around the same TDP with better perf, but still will be a high "temperature" part.

I think he was referring to the whole PC using 120W at load.

I really really hope that would be the case 8(

Cheers!

EDIT: Typo

 

[jdwii]

bulldozer hotfix? http://www.techpowerup.com/156844/AMD-Bulldozer-gets-an-Update-from-Microsoft..html

Yes i saw that i hope Toms does a article showing its improved performance and i hope they compare it to the Phenom II x6 1100t and the 2500K,2600K.

 

[metallifux]

Yes i saw that i hope Toms does a article showing its improved performance and i hope they compare it to the Phenom II x6 1100t and the 2500K,2600K.

+1