Piledriver not coming until 2013 now?

[janas19]

http://i749.photobucket.com/albums/xx132/rpisano01/Will_Ferrell_Old_School_Streaking.jpg


:hmm:

Not trying to derail here, but... rofl.

 

[Torquemada]

This thread is officially unsafe for children

 

[janas19]

This thread is officially unsafe for children

Lulz. :thumbup:

 

[Ventanni]

So there are two schools of thought when it comes to this Bulldozer architecture and our thoughts of AMD actually improving it. They certainly took a credibility nosedive when it comes to its performance, but those two schools of thought that I gather from these threads are:

1. No matter how much you polish a turd, it's still a turd.
2. AMD has a track record of releasing poor first generation products, but has historically rebounded with relatively high levels of success.

The Radeon 2600 and the Phenom 1 come to mind. The Radeon 2600 was an average performing power hog, but later resulted in some amazing products in both performance profitability. The Phenom I suffered from the same problem, but you have to admit that the Phenom II was overall a design success. It may not have lead AMD to outstanding profitability levels, but it did lead to the range of Stars cores, which were used in Fusion products (where future AMD overall profitability will come from).

Unlike Bulldozer though, the Radeon 2600 and Phenom II successors enjoyed the availability of a fresh die node. And, from my understanding, Piledriver won't have that option. Even still, AMD seems pretty good at getting the most out of a die node, so may not all is lost.

With that said though, I believe the main advantages to Piledriver will be an overall reduction in power consumption and not necessarily performance. As the previous posts show, Bulldozer cores do scale pretty well in multi-threaded applications, so even small increases in individual core IPC should result in reasonable overall performance gains. I think we might actually see rather large performance increases in a select few benchmarks as a result of fixing a few bottlenecks, but I think it won't be until Streamroller that we see any reasonable IPC performance gains.

I don't think AMD will be the CPU of choice for future gamers though.

Overall, I'd rate my expectations of Piledriver to be moderate, but not over-reaching. If AMD has learned anything, it's that you can be profitable and not be the fastest CPU on the block. All AMD needs to do is release a CPU architecture that's low power, has strong graphics capabilities, and provides strong enough CPU performance to provide users with a quality user experience.

 

[janas19]

So there are two schools of thought when it comes to this Bulldozer architecture and our thoughts of AMD actually improving it. They certainly took a credibility nosedive when it comes to its performance, but those two schools of thought that I gather from these threads are:

1. No matter how much you polish a turd, it's still a turd.
2. AMD has a track record of releasing poor first generation products, but has historically rebounded with relatively high levels of success.

The Radeon 2600 and the Phenom 1 come to mind. The Radeon 2600 was an average performing power hog, but later resulted in some amazing products in both performance profitability. The Phenom I suffered from the same problem, but you have to admit that the Phenom II was overall a design success. It may not have lead AMD to outstanding profitability levels, but it did lead to the range of Stars cores, which were used in Fusion products (where future AMD overall profitability will come from).

Unlike Bulldozer though, the Radeon 2600 and Phenom II successors enjoyed the availability of a fresh die node. And, from my understanding, Piledriver won't have that option. Even still, AMD seems pretty good at getting the most out of a die node, so may not all is lost.

With that said though, I believe the main advantages to Piledriver will be an overall reduction in power consumption and not necessarily performance. As the previous posts show, Bulldozer cores do scale pretty well in multi-threaded applications, so even small increases in individual core IPC should result in reasonable overall performance gains. I think we might actually see rather large performance increases in a select few benchmarks as a result of fixing a few bottlenecks, but I think it won't be until Streamroller that we see any reasonable IPC performance gains.

I don't think AMD will be the CPU of choice for future gamers though.

Overall, I'd rate my expectations of Piledriver to be moderate, but not over-reaching. If AMD has learned anything, it's that you can be profitable and not be the fastest CPU on the block. All AMD needs to do is release a CPU architecture that's low power, has strong graphics capabilities, and provides strong enough CPU performance to provide users with a quality user experience.

*nods head in agreement*

After the BD hotfix, an AT benchmark showed it pulling close to the SB processors in HEAVILY THREADED APPS.

If more game developers made their games in such a way that processing tasks were dynamically scaled into lots and lots of really small threads, then you would see BD doing a lot better.

But, I don't think that's how it's done

 

[ViRGE]

Then VIA bought Cyrix in 1999, introduced the VIA C3 processor, it was pretty cool that you didn't have to use a heatsink with those CPUs but they were pretty slow, especially when it came to gaming, AMD K6-2 would smoke it. Then after the first generation of the VIA CPUs, they disappeared. VIA seems to be a pretty silent company as of late.

Ironically VIA's biggest business right now seems to be selling ARM SoCs for craplets and set-top boxes.

 

[Maximilian]

*nods head in agreement*

After the BD hotfix, an AT benchmark showed it pulling close to the SB processors in HEAVILY THREADED APPS.

If more game developers made their games in such a way that processing tasks were dynamically scaled into lots and lots of really small threads, then you would see BD doing a lot better.

But, I don't think that's how it's done

C&C generals 2 should do this (should if EA dont turn around and botch it completely somehow). Its using the BF3 engine and since its an RTS a single player bench vs AI should actually prove fruitful. BF3 engine uses 8 threads but all the benchs are done in largely GPU limited single player for "fairness".

By the time that comes out we will be looking at haswell though, hopefully a 6 core consumer version not just enthusiast.

 

[Olikan]

About PD, yes if they will keep the higher scaling with 10-15% higher single thread performance then PD could be really fast in multithreaded apps. But they will really need to lower power usage a lot too.

if you look amd trinity pdf, piledriver have 16% better IPC than bulldozer at Productivity suite in pcmark vantage.

well, if you trust amd words

 

[guskline]

if you look amd trinity pdf, piledriver have 16% better IPC than bulldozer at Productivity suite in pcmark vantage.

well, if you trust amd words

Funny, I bet the same marketing group that did the PR for the Bulldozer pre-release did this.

 

[piesquared]

Funny, I bet the same marketing group that did the PR for the Bulldozer pre-release did this.

They must come from the same family as the marketing group that claims IntEl's graphics are 'visualy smart'. Always get a laugh out of how they con consumers and then troll forums to pump up that message.

 

[lifeblood]

Is Piledriver a server or desktop architecture?

PD cores are being put into Trinity. That means they need to be reasonably efficient for desktop use because Trinity is a desktop CPU. Looking at the power envelope Trinity is being targeted at, performance per watt is critical. BD sucked at desktop use, it took to much power and didn't give enough performance. It was worse than the stars architecture used in the Phenom II and Llano CPUs. BD cores in Trinity would sink it as BD cores are not very efficient at desktop work. For PD cores to improve over the Stars in Llano, they MUST have improved PD for desktop use.

Given that, PD should offer much better desktop performance than BD did. If not, Trinity is going to bomb. That gives me some hope that Piledriver is going to actually be competitive.

Is my logic good or am I missing something?

 

[Yuriman]

Is Piledriver a server or desktop architecture?

PD cores are being put into Trinity. That means they need to be reasonably efficient for desktop use because Trinity is a desktop CPU. Looking at the power envelope Trinity is being targeted at, performance per watt is critical. BD sucked at desktop use, it took to much power and didn't give enough performance. It was worse than the stars architecture used in the Phenom II and Llano CPUs. BD cores in Trinity would sink it as BD cores are not very efficient at desktop work. For PD cores to improve over the Stars in Llano, they MUST have improved PD for desktop use.

Given that, PD should offer much better desktop performance than BD did. If not, Trinity is going to bomb. That gives me some hope that Piledriver is going to actually be competitive.

Is my logic good or am I missing something?

Performance per watt is enormously important in the server market.

 

[formulav8]

While PD IPC will increase some, the real increase will be performance per watt. GF just might be getting the hang of 32nm now... That along with some small core changes could make for quite a decent chip.

 

[Chiropteran]

Performance per watt is enormously important in the server market.

And servers very rarely run overclocked CPUs. The outrageous power consumption numbers of overclocked FX-8150 isn't representative of the power usage of bulldozer based server CPU.

http://www.anandtech.com/show/5058/amds-opteron-interlagos-6200/8

Nothing terrible.

 

[lifeblood]

Performance per watt is enormously important in the server market.

Correct, but server workloads are different from desktop workloads. Just because it has good performance per watt doing server workloads doesn't mean it will be the same for desktops. That's why AMD and Intel have desktop CPUs and server CPUs.

How much difference their is I have no idea.

 

[Olikan]

While PD IPC will increase some, the real increase will be performance per watt. GF just might be getting the hang of 32nm now... That along with some small core changes could make for quite a decent chip.

well, PD seems to have the same IPC as phenom 2, yet amd says that trinity laptops are going to be more eficient...from 240 min to 320 min in 3d mark 06..i think, can't remeber very well

 

[exar333]

And servers very rarely run overclocked CPUs. The outrageous power consumption numbers of overclocked FX-8150 isn't representative of the power usage of bulldozer based server CPU.

http://www.anandtech.com/show/5058/amds-opteron-interlagos-6200/8

Nothing terrible.

I agree with you 100% there. The problem with BD is that to be competitive, it needs to really dial-up the Mhz. Unfortunately, that brings a huge increase in power consumption after a certain point.

For servers that are fine with many cores running at 2.4ghz (for example) BD is great! If you need or rely on raw clockspeed as well, its not as good nor as efficient as other offerings (stars, Intel).

 

[Torquemada]

BD sucked at desktop use, it took to much power and didn't give enough performance. It was worse than the stars architecture used in the Phenom II and Llano CPUs. BD cores in Trinity would sink it as BD cores are not very efficient at desktop work. For PD cores to improve over the Stars in Llano, they MUST have improved PD for desktop use.

Part of BD/Orochi's clockspeed/power problem may have been the immature GloFo 32nm process. Hopefully PD/Trinity gains a "natural" benefit there, in addition to any design improvements.

 

[Rvenger]

Part of BD/Orochi's clockspeed/power problem may have been the immature GloFo 32nm process. Hopefully PD/Trinity gains a "natural" benefit there, in addition to any design improvements.

I just hope Piledriver doesn't end up with some insane clocks instead of a nice IPC gain.

 

[Olikan]

I just hope Piledriver doesn't end up with some insane clocks instead of a nice IPC gain.

from the ES, trinity will have 4.1Ghz max turbo

Vishera probably will go more

 

[lifeblood]

I just hope Piledriver doesn't end up with some insane clocks instead of a nice IPC gain.

Why not? If increasing the clocks gives the same benefits compared to better IPC, who cares how they do it? Unless, of course, higher clocks means more heat or higher power requirements compared to higher IPC.

 

[guskline]

From what I have read, the problem is that to achieve higher OCs the Bulldozer has to consume too much power, thus too much heat is generated. Also, the original idea for Bulldozer was to run at a much higher frequency but the thermal envelope was just too high. Let's hope power management significantly improves and as a result much higher OCing occurs.

 

[Rvenger]

Yes, power consumption was my concern.

 

[tweakboy]

Makes sense Bulldozer is their processor of 2012 .... Well get 8970 soon.... gl

 

[ButtMagician]

I hope Piledriver improves upon Bulldozer as much as Phenom II improved upon the original Phenom. It was a good affordable CPU, in fact thanks to core unlocking it's the best processor in price vs performance I had ever bought :biggrin:

This thread did not need to be bumped
-ViRGE