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 ...

 

[esrever]

run on a hamster wheel duh

 

[amd955be5670]

and that's true but this time with Piledriver can it at least look the Phenom II X4 Deneb in the eye.?
because all I have to do is raise the clock on my 965BE and the FX-4100 especially is toast; don't even mention the FX-6 series..
the 980BE is just better stock.
disappointing Bulldozer shouldn't even had been released the way it is/was...

what should have happened is another tweak/stepping revision and call it Phenom III
while still developing Bulldozer back at the lab trying to get it right.
at the same time getting the administrative house in order as well..

Amd pulled a rabbit out of its hat with Phenom II series, if there was a Phenom III which had the Phenom II series clocked to 6ghz, sounds insane though, today would have been a better day, maybe by miracle if Amd decides to super enhance the Phenom II architecture by throwing Bulldozer away, then Piledriver would be named Phenom III and I really hope Amd does that, its not too late to change, atleast thats what I believe.

 

[amd955be5670]

I just think bulldozer was just designed to run at 5 ghz before they actually put it to silicon.

just a query, what if they used titanium?

 

[esrever]

just a query, what if they used titanium?

they'd need to use uranium before they get anything going

 

[amdfangirl]

low intelligence appliances as in toasters, but mixing blenders or microwaves are far too advanced still...

BD TDP = insta-cook.

 

[wieschie]

I honestly think the temperature sensors are broken in mine. Claims to run at 43 C at full load and overclocked from 3.3 to 4.0. I have a pretty decent air cooler, but.. that doesn't seem right.

But really, if Piledriver is just a clocked-up version of BD with no significant architecture changes.. AMD deserves to go out of business.

 

[mlscrow]

We all know that BD needs to perform at least 30% better in order to provide competition for the market. Most people have seen the AMD "Roadmap" that they released many months ago, stating that Piledriver will be approximately 10-15% faster than BD. Obviously, this isn't enough, but since then, AMD has been completely hush-hush in terms of leaking any information regarding PD. This could be a good thing (they did this with the HD4000 series of GPU's and then shocked the industry), so hopefully they have something up their sleeves. Obviously, out of anyone in the world, AMD knows exactly what they did wrong and what they need to do to stay alive, which is to create a product that actually performs.

We also know that one of the problems with BD was and remains that it is ahead of it's time and that current operating systems aren't designed to utilize it's architecture, which is why we're already going to see a 5-10% improvement using Windows 8 alone. There are also rumors that AMD is dumping Global Foundries to go with TSMC and running with a 28nm fab, which could further improve performance. If we look back, AMD was able to increase performance/efficiency over the course of the Phenom II lineup through improved process yielding by around 30%, so, hopefully, with all things considered, if AMD can execute this the right way, I do believe that Piledriver will become competitive and that later implementations (Excavator), will continue to keep the company going.

5-10% from Windows 8 Scheduler design
+
10-15% improvement by design (AMD roadmap)
+
0-??% improvement through revisions/stepping?
+
Hopefully clocks up to 5Ghz and up = possible success story.

 

[truegenius]

hope

 

[esrever]

pilldriver is just gonna be a bulldozer with higher clock probably.

isn't it just for trinity?

 

[AbdullahG]

5-10% from Windows 8 Scheduler design
+
[strike]10-15% improvement by design (AMD roadmap)[/strike]
+
0-??% improvement through revisions/stepping?
+
Hopefully clocks up to 5Ghz and up = possible success story.

Don't take AMD's word for it.
Remember the BD leaked slides which claimed BD beat/matched an i7 980X in performance for less money?
It wasn't even an improvement over the Phenom IIs.

 

[mlscrow]

dude what are you smoking...?
( "ahead of it's time, HD 4xxx shocked the industry" - if they knew what they did wrong then they need to step up there QA team..)

keeping hope alive huh.?
and the GF to TSMC info is now old news..

Uh...do you somehow (not sure how) believe that BD isn't ahead of it's time, or that in 2008 when the Radeon HD 4870 came out and then finally took the performance crown after not having it for many many years, making a statement that AMD/ATi/Radeon was back for good, wasn't a shock to the industry? Did you just start getting into computer hardware?

As far as the GF to TSMC move...uh, first of all, it's not old, only having been surfaced a couple months ago with just about nothing surfacing since and secondly, it's not even "news", it was an unofficial rumor and AMD hasn't and most likely won't officially announce anything until Financial Analyst Day in February, but gg.

Don't take AMD's word for it.
Remember the BD leaked slides which claimed BD beat/matched an i7 980X in performance for less money?
It wasn't even an improvement over the Phenom IIs.

I wouldn't strike out the 10-15% improvement by design figure. Not that we have any official benchmarks, but, on top of what AMD is saying, there were some Trinity benchies leaked out of Germany recently and although they were a bit fuzzy and could have easily been fabricated, the results looked realistic, with CPU performance up around 7-13%, and GPGPU performance up around 50%, which makes sense, considering the graphics core they're using for it (cayman 6900).

 

[mlscrow]

- your entire statement and reasoning...

this is for your HD 4870
http://www.anandtech.com/bench/Product/304?vs=318

By your rebuttal, you have made it painfully clear that you are not very familiar with the times when these products came out. To provide you with a better idea of what it was like...how about reading something a bit more relative to my words:

http://www.tomshardware.com/reviews/radeon-hd-4870,1964.html

I do not see why people try to argue with facts. It's rather pointless. Can we also please bring this thread back to the point, which is the Piledriver core? Thank you.

 

[mlscrow]

you said it was dominant for years....

anyway, moving on..
http://techreport.com/articles.x/21865

Never said this. Please read more carefully.

 

[AbdullahG]

I wouldn't strike out the 10-15% improvement by design figure. Not that we have any official benchmarks, but, on top of what AMD is saying, there were some Trinity benchies leaked out of Germany recently and although they were a bit fuzzy and could have easily been fabricated, the results looked realistic, with CPU performance up around 7-13%, and GPGPU performance up around 50%, which makes sense, considering the graphics core they're using for it (cayman 6900).

Those numbers seem realistic for Trinity, but I wouldn't count on such a jump right away considering they CAN be decisive as you had said-though I am not implying it isn't possible for such a boost, it's just plausible.
What I am saying is to not expect a 30% performance boost right away [for Piledriver].
The Windows 8 Scheduler Design performance boost is probably the only legit evidence for a performance ATM (official word and benchmarks which can done by others; the only problem to some being mostly synthetic rather than real world).
Everything else really doesn't have much credibility TBH, and raw power simply isn't everything.
Let's not forget power consumption, heat, price, etc., which we have no info on ATM, so we can't conclude anything on that.

Basically, we can't expect Piledriver to deliver competitive performance just yet until we see more credible info.
The only thing we can conclude on is that Windows 8 will mostly likely introduce features that will increase BD/PD's performance to some extent.

 

[mlscrow]

Those numbers seem realistic for Trinity, but I wouldn't count on such a jump right away considering they CAN be decisive as you had said-though I am not implying it isn't possible for such a boost, it's just plausible.
What I am saying is to not expect a 30% performance boost right away [for Piledriver].
The Windows 8 Scheduler Design performance boost is probably the only legit evidence for a performance ATM (official word and benchmarks which can done by others; the only problem to some being mostly synthetic rather than real world).
Everything else really doesn't have much credibility TBH, and raw power simply isn't everything.
Let's not forget power consumption, heat, price, etc., which we have no info on ATM, so we can't conclude anything on that.

Basically, we can't expect Piledriver to deliver competitive performance just yet until we see more credible info.
The only thing we can conclude on is that Windows 8 will mostly likely introduce features that will increase BD/PD's performance to some extent.

Agreed. I am not "expecting" anything from Piledriver. I've learned what happens when you expect something [queue Hitler's response to Bulldozer video]. I am, however, hopeful that AMD can pull it off. They need to if we all want to see relatively low CPU prices into the future, otherwise, we'll all be paying out whatever Intel wants us to. As a consumer, I always pull for competition. Unfortunately, and in my opinion, AMD only has 1, maybe 2 tries left, before we can kiss them goodbye.

 

[AbdullahG]

Agreed. I am not "expecting" anything from Piledriver. I've learned what happens when you expect something [queue Hitler's response to Bulldozer video]. I am, however, hopeful that AMD can pull it off. They need to if we all want to see relatively low CPU prices into the future, otherwise, we'll all be paying out whatever Intel wants us to. As a consumer, I always pull for competition. Unfortunately, and in my opinion, AMD only has 1, maybe 2 tries left, before we can kiss them goodbye.

I hope Piledriver is at least a significant improvement over the Phenom IIs.
Would be nice if it matched Bloomfield; I'd settle for that.
I shouldn't expect too much though; one thing at a time.
If Piledriver doesn't turn out well, they still have their APUs which are fairly cheap and offer desktop users an affordable, decent route into gaming.
They can also do well in the mobile department with their APUs (for gaming laptops I assume for under a grand).
AMD doesn't necessarily NEED to control the enthusiast market; they just need to hold up against something Intel has to offer (probably in the under $200 zone with the i3s and non-k i5s; still competition).

 

[jimmysmitty]

I hope Piledriver is at least a significant improvement over the Phenom IIs.
Would be nice if it matched Bloomfield; I'd settle for that.
I shouldn't expect too much though; one thing at a time.
If Piledriver doesn't turn out well, they still have their APUs which are fairly cheap and offer desktop users an affordable, decent route into gaming.
They can also do well in the mobile department with their APUs (for gaming laptops I assume for under a grand).
AMD doesn't necessarily NEED to control the enthusiast market; they just need to hold up against something Intel has to offer (probably in the under $200 zone with the i3s and non-k i5s; still competition).

Problem, Piledriver is the basis for Trinty. Another problem, we sold our first AMD Pro, a system that uses the Llano APUs last month. They sell the least out of all AMD systems, even the Athlon IIs sell more but it sucks as soon we wont have that as a choice for consumers as AMD is phasing them out.

I understand but with BD being meh and Llano being meh CPU with ok GPU, its not going to give consumers much to rave about or go for.

If PD does bad, Trinity will also do bad. AMD needs to focus more on a good product than maketing. I say fire marketing and funnel the extra funds into R&D. Hype only hurts the product more when it doesn't deliver.

 

[AbdullahG]

Problem, Piledriver is the basis for Trinty. Another problem, we sold our first AMD Pro, a system that uses the Llano APUs last month. They sell the least out of all AMD systems, even the Athlon IIs sell more but it sucks as soon we wont have that as a choice for consumers as AMD is phasing them out.

I understand but with BD being meh and Llano being meh CPU with ok GPU, its not going to give consumers much to rave about or go for.

If PD does bad, Trinity will also do bad. AMD needs to focus more on a good product than maketing. I say fire marketing and funnel the extra funds into R&D. Hype only hurts the product more when it doesn't deliver.

UGH!
I forgot that piece :\
Okay, so AMD is in a position where it needs to get Piledriver right or it will not only lose it's image in the enthusiast market-or the whole market for CPUs-but for their APUs as well.
Now things are even more difficult and complex...

 

[amd955be5670]

They should try manufacturing a double layered processor, fitting/perfecting the required components.

Thrice the thickness might save them, if that is possible, LOL.

 

[gamerk316]

Hopefully clocks up to 5Ghz and up = possible success story.

Stop it. Transistor leak is and will be the primary limiting factor on clocks for some time yet. You will NOT see chips clocked much above 3.6GHz at stock, and I doubt any chip in the near future will reliably clock above 4.6GHz or so [the current limit for most chips].

I said way back in the BD thread that there was no way in hell that BD would ever clock that high, despite all the news comming out that it would. And the reason is simple: Because it defies physics as we know it.

Problem, Piledriver is the basis for Trinty.

That scares me, because the world of Trinity is the lower power, high efficency world that BD/PD doesn't seem fit for. Looks like a very ugly match to me. Trinity may be worse then llano because of using the PD core, due to heat/power performance deficencies.

 

[amdfangirl]

On the other hand, CPU performance really isn't all that important I guess. Sacrifice the CPU, make a more simple design, better battery life and etc.

Making the GPU faster doesn't make sense, since it is bottlenecked by the RAM. We could end up with a beefed up Brazos.

 

[bawchicawawa]

On the other hand, CPU performance really isn't all that important I guess. Sacrifice the CPU, make a more simple design, better battery life and etc.

Making the GPU faster doesn't make sense, since it is bottlenecked by the RAM. We could end up with a beefed up Brazos.

Trinity will have a memory bandwidth of 2133. Get 2133 ram for trinity, just like you should get 1866 ram for llano now...


Personally, i have high hopes for trinity.

Win7 patch
piledriver will be an improvement from bulldozer
that 2133 ram will be helping out the memory frequency hungry APU's of course
and being able to crossfire with a better dedicated graphics card


though, piledriver isn't expected to be a super amazing step up from bulldozer, it IS a big step up from the athlon's in the llanos, helping tremendously in cpu bound games like bf3. Also, look at how well a a8 3870k+6670 can perform already, playable frame rates at 1080p+med/high settings, with modest overclocks, of course.

 

[AbdullahG]

RAM @ 2133MHz is expensive.
Just 4GB @ 2133MHz is a bit more expensive than 8GB @ 1600MHz.
Better off just getting cheaper ram and OCing it.

 

[-Fran-]

Trinity will have a memory bandwidth of 2133. Get 2133 ram for trinity, just like you should get 1866 ram for llano now...


Personally, i have high hopes for trinity.

Win7 patch
piledriver will be an improvement from bulldozer
that 2133 ram will be helping out the memory frequency hungry APU's of course
and being able to crossfire with a better dedicated graphics card


though, piledriver isn't expected to be a super amazing step up from bulldozer, it IS a big step up from the athlon's in the llanos, helping tremendously in cpu bound games like bf3. Also, look at how well a a8 3870k+6670 can perform already, playable frame rates at 1080p+med/high settings, with modest overclocks, of course.

As an A8 3850 owner (desktop, BTW), I can tell you that they will have to be very careful with Trinity. The CPU inside the Llanos are 32nm as well, remember that, so they won't be able to clock them that much higher unless they have a better 32nm process (which they should), but even so, if a Trinity CPU core doesn't beat steadily a Llano CPU core clock per clock, it's a lost battle. Plain and simple. They won't clock them high enough to make up for it. And with the just-released "k" series, it will be even harder to position Trinity as a better buy if it doesn't deliver.

Not even the stronger GPU will make up for it, since the CPU within Llano is very good. It's on par or even better against Athlon IIs and even Phenom IIs clock per clock. Clock an A8 to 3.2Ghz and you'll have a very powerful standalone CPU.

Cheers!

 

[esrever]

ram oc's don't work well and your performance won't increase much as you'd need to turn down the timings for the ram. Ram OC also make the system unstable at times and will cause BSOD when things get intense. I'd not recommend ram OC of more than 50mhz.