PileDriver Performance only 10% -15% (maybe)

[eternalone]

http://www.xbitlabs.com/news/cpu/di...ure_Peculiarities_to_Software_Developers.html

AMD has a rather strong and clear roadmap for the period of the following three years with the aim to increase performance-per-watt of its high-performance cores by approximately 10% - 15% every year. In practice, this transforms into 33% - 52% speed boost of Excavator micro-architecture compared to Bulldozer. Technologically, AMD wants to improve IPC (instructions per clock) performance and reduce power consumption at the same time, which transforms into higher clock-speeds.

WTH????



So what they are saying is essentially that Piledriver will basically finally catch up to its older phenom 2 /Thuban architecture performance level LOL . So basically Piledriver will be a Phenom 2 level cpu am I wrong???. Also what Im understanding is that in 3 years time AMD will be at the same IPC and performance as what intel is today in 2012 (Before Ivy Bridge) which will bring even better performance. Tell me Im wrong about this , please tell me they are joking. Isnt the Gap to great too stay in the game at this point??? One thing I do see as a positive is that I see alot of free motherboard offers from Microcenter and such in the future.

 

[God Mode]

AMD should concentrate their efforts on better APU's. This is the only way their cpu business will survive.

 

[BD231]

Wow. After all these years of intel kicking their ass with a simple linear road map they decide to follow suit. I've been saying this for years, baby steps AMD!! Hopefully this means no more multi-core madness with ridiculous time frames between release's.

I don't have to much faith in that road map but if they do manage to make the switch to intel progression methodology and actually pull it off it'll be the best thing to happen to AMD in a long time. No more of this hit or miss nonsense.

 

[Ancalagon44]

So, Bulldozer, Piledriver, and Steamroller. Take away the second part of each word, we have
Bull, Pile, and Steam.

Now we put Steam first, because it shares a name with an awesome gaming service
Steam Bull Pile

Now we insert some fun words
Steaming Pile of Bullshit!

Anyway, back to your regular programming. If they can keep to 10-15% per year, then maybe they will eventually catch up to Intel. For PileDriver, they need to increase IPC by 10-15%, increase clock speed headroom by 10-15%, while simultaneously decreasing power drain by 10-15%, even at the elevated clock speeds.

 

[richaron]

I like to bring up old news & jump on the current bandwagon.

It makes me look cool.

 

[Olikan]

38.67% better than LLANO at same TPD for integer

5.47% better than LLANO at same TPD for floating

http://www.nordichardware.com/news/69-cpu-chipset/45627-amd-trinity-a8-4500m-benchmarked-in-folding.html

 

[blckgrffn]

This is old news.

I guess it is fun to bring it up again?

There are some of us who are hoping they will deal with specific issues (ie, FPU performance hamstring in certain cases by the Write-Back L1 cache) that might bring solid gains in some applications that may not be easily reflected in that 10-15% number.

Hoping, not expecting

I am entirely confident that I speak for at least one member of this site...

 

[pelov]

They pushed up the expectations for Trinity > Llano from 10-15% to 20-30% uplift. If you go by that nordichardware benchmark then it also depends on the workload. Either way things are looking pretty good for Trinity but I've learned not to expect too much and then become disappointed.

 

[RavenSEAL]

This is a 3 month old article and that slide has been around for longer. Pretty much EVERYONE already knows this.

 

[aaksheytalwar]

AFAIK AMD X6 1100 was somewhere btw a Q9550 and I5 750. So it basically means that after about 1 year they will be nearly equal to I5 750.

That means in 3 years time they will be about equal to a I7920 to I5 2500k but which time Intel would be at least 2x+ of 2600k.

There is no way they are catching up with Intel without a miracle, unlikely in the next 3-5 years at the very least.

 

[Olikan]

A10 APU with everything set on high and 2xMSAA
seems to have good fps

http://www.youtube.com/watch?feature=player_embedded&v=WD6GaFEpfC4

 

[Maximilian]

 

[blckgrffn]

pic

I lol'd

You come up with that? Looks like an epic paint masterpiece...

 

[Maximilian]

I lol'd

You come up with that? Looks like an epic paint masterpiece...

Not my work of art unfortunately. Stolen from google image search

 

[KingFatty]

So what they are saying is essentially that Piledriver will basically finally catch up to its older phenom 2 /Thuban architecture performance level LOL . So basically Piledriver will be a Phenom 2 level cpu am I wrong???

The chart you cited shows performance per watt.
That is not the same as raw performance.

In fact, it is possible that a new processor could have a performance per watt that potentially stays the same or even get WORSE compared to earlier chips, while having BETTER raw performance.

For an easy analogy, think of how the GTX480 video card was having pretty poor power efficiency (e.g., the performance per watt) and made a lot of heat, but it was still a beast in raw performance and outperformed the previous generation.

It seems you are concluding that raw performance is the same as performance per watt, and I think that is a wrong conclusion to make.

I like more efficient chips, but the chips efficiency doesn't tell me its raw performance, and I think I'm willing to accept a modest increase in efficiency if I get the raw performance increase I'm willing to pay for.

 

[blckgrffn]

The chart you cited shows performance per watt.
That is not the same as raw performance.

In fact, it is possible that a new processor could have a performance per watt that potentially stays the same or even get WORSE compared to earlier chips, while having BETTER raw performance.

For an easy analogy, think of how the GTX480 video card was having pretty poor power efficiency (e.g., the performance per watt) and made a lot of heat, but it was still a beast in raw performance and outperformed the previous generation.

It seems you are concluding that raw performance is the same as performance per watt, and I think that is a wrong conclusion to make.

I like more efficient chips, but the chips efficiency doesn't tell me its raw performance, and I think I'm willing to accept a modest increase in efficiency if I get the raw performance increase I'm willing to pay for.

Similarly, if IVB doesn't OC as well as SB, I think many will be disappointed even if it uses less power at a given clock speed - or the power usage scales poorly at high clocks due to high core temps (lower speeds for given power consumption). Hopefully that is not the case! There might be a reason Intel didn't increase the clock speeds much, though...

 

[Paul98]

LMAO

 

[thilanliyan]

That means in 3 years time they will be about equal to a I7920 to I5 2500k but which time Intel would be at least 2x+ of 2600k.

Intel will more than double 2600K performance in 3 years?

 

[Stayfr0sty]

From what I heard piledriver is also gonna be a new socket. Which means even if you have an AM3+ board your gonna have to upgrade mobo as well if you want a piledriver.

 

[kdubbs]

I for one am fascinated by the fact that their core-to-core ratio does not increase linearly. I think this accurately reflects their module strategy. :hmm:

 

[Puppies04]

I for one am fascinated by the fact that their core-to-core ratio does not increase linearly. I think this accurately reflects their module strategy. :hmm:

Actually I think it has more to do with them being batpoo crazy (you can tell by the eyes).

 

[pelov]

From what I heard piledriver is also gonna be a new socket. Which means even if you have an AM3+ board your gonna have to upgrade mobo as well if you want a piledriver.

No. And more accurately no for 2 reasons:

1 - Piledriver isn't the name of a processor but rather the core design. Bulldozer > Piledriver > Steamroller. The desktop processor is called Vishera and will feature Piledriver cores just as Trinity will feature Piledriver cores. There might be some small differences between them but the only difference that was hinted was the L3.

2 - Vishera is an AM3+ part.

 

[guskline]

No. And more accurately no for 2 reasons:

1 - Piledriver isn't the name of a processor but rather the core design. Bulldozer > Piledriver > Steamroller. The desktop processor is called Vishera and will feature Piledriver cores just as Trinity will feature Piledriver cores. There might be some small differences between them but the only difference that was hinted was the L3.

2 - Vishera is an AM3+ part.

Pelov: That is my understanding also. If Bulldozer to Piledriver is similar to Phenom I to Phenom II I will be happy. I think, hope and pray that the AMD engineers and most importantly the front office realize that Bulldozer was hardly a homerun with CPU enthusiasts despite a very few reviews to the contrary. Using the baseball analogy if Piledriver is a line shot that just stays inside the foul pole for a homer( or even off the wall for a triple) it will keep AMD in the CPU enthusiast game. Bulldozer was a blustering pop up.

 

[SlowSpyder]

Absolute performance isn't AMD's real issue. AMD simply needs a chip that is competitive somewhere... a pretty good argument can be made that in many situations an i3 is faster than what AMD sells.

But even so, not having the most absolute performance doesn't have to be a problem. But having decent performance per watt as well as doing something about single and lightly threaded performance is where AMD should be concentrating, in my opinion.

AMD may very well never beat Intel in all out performance again, it is difficult to out do a company that spends many times what your company does on R&D. First AMD needs to get competitive in areas that it can actually compete.

 

[formulav8]

Up to 15% has been known for a long time. This is definitely nothing new.