IkarugaAgain, please don't get me wrong, Trinity APUs are really useful nice products and they can be the better choice in many builds. All I said is that they are hopeless against performance monsters like the SB/IB-K series.

Perhaps in a CPU intensive benchmark, but in Graphics, Trinity rules hands down. And we are talking about overall performance with no outside interference; Trinity is a better solution in terms of price/performance.

AquinusRCM makes the CPU power usage most efficient at the RCM's resonant frequency. It uses charge pumps to use otherwise wasted energy to supply the CPU again. It recycles CPU power, but the further the clocks get from the resonant frequency the more power it will consume. It will run best to the clock speed that the RCM is tuned to. Higher temperatures could also dampen the RCMs resonant frequency. I'm really curious to see how PD works out. I could care less about it outpacing Intel, the architecture is pretty neat and is a step in the right direction.

This design should eventually pay off in the long run. Hopefully by the middle of 2013, people will say WOW, good job AMD, now we understand :D

Dent1Why is everyone always on AMD's case?
Intel has 10x the resources and they promote their processors with unstable LN2 overclocks too but nobody says nothing.

well said...
Nobody say's nothing in articles regarding VIA while them too,boasting their capabilities.

bencrutzadded feature.
amd just provide options.
if there's no crossfire support, i think people would then start complaining why the heck there's no crossfire available for FM2? :slap:

and yet people still whining about lack of PCIe 3.0 and 16x16x bandwidth :p
PS:thanks for invitation :toast:

1d10twell said...
Nobody say's nothing in articles regarding VIA while them too,boasting their capabilities.
and yet people still whining about lack of PCIe 3.0 and 16x16x bandwidth :p
PS:thanks for invitation :toast:

True, and Welcome to TechPowerUP :)

ITT: people bitching that Intel pulls the same LN2 overclocking stunts as AMD.

If that's true, why do I never hear about this from Intel PR; while every time AMD releases a new CPU, they claim it overclocks like a demon on LN2?

Maybe it's because - once again - the CPU part of their APU, aka the most important part, isn't competitive and they're frantically trying to distract people from this fact?

Just throwing that out there... use it, don't use it...

So @ 6.5 GHz with LN2 Trinity is Cinebench R11.5 benchable?

lets see some of these LN2 jockey`s actually bench at these clocks then.....

seronxwww.sweclockers.com/image/red/2012/09/27/AMD_Trinity_A-series_0011_Layer+7.jpg?t=paneBanner&k=9c75e774

So @ 6.5 GHz with LN2 Trinity is Cinebench R11.5 benchable?

If the 68% more in cinebench is true, doesn't it mean it scales VERY well?

More CPU benches are needed to make a proper judgment on this but it sure does look good.

Super XPPerhaps in a CPU intensive benchmark, but in Graphics, Trinity rules hands down. And we are talking about overall performance with no outside interference; Trinity is a better solution in terms of price/performance.

No, that's absolutely not what the context was. Look up what I wrote. The choice between an easy solid 4.7Ghz SB performance on air vs an AMD APU @6.5Ghz on LN2(!) and the reactor you need, (not to mention the fairness of the comparison since the 2500K was launched almost two years ago). It's clearly not the better solution in that context.

IkarugaNo, that's absolutely not what the context was. Look up what I wrote. The choice between an easy solid 4.7Ghz SB performance on air vs an AMD APU @6.5Ghz on LN2(!) and the reactor you need, (not to mention the fairness of the comparison since the 2500K was launched almost two years ago). It's clearly not the better solution in that context.

Oh, so you were stating a joke.

Super XPOh, so you were stating a joke.

I found it funny indeed, but it wasn't me who stated it;).

Wow....:laugh::laugh::laugh::slap::slap: :laugh::laugh::laugh:

Super XPTrue, and Welcome to TechPowerUP :)

thank's mate :)

AssimilatorIf that's true, why do I never hear about this from Intel PR; while every time AMD releases a new CPU, they claim it overclocks like a demon on LN2?

Maybe it's because - once again - the CPU part of their APU, aka the most important part, isn't competitive and they're frantically trying to distract people from this fact?

no,but the board maker does.
you can see here if they haven't changed the slides...

www.msi.com/product/mb/#/?sk=Socket%202011%20(Intel%20i7)
www.asus.com/Motherboards/Intel_Socket_1155/Intel_Z77
www.asrock.com/microsite/OCFormula/

they bashing each other claiming can reach 7 Ghz.it is good for them or good for intel?

CPU might important,but not the most.on to question,which software or games that any mediocre quad-core can't handle?if you have an opt,what would you like to choose,upgrade to i7 3770k or GTX 660Ti?
let just leave PR be,it's their job :)

Next AMD Headline involving clocks:
AMD plans to introduce 10 GHz chips by 2016.

Edit: Added 1 year to be super accurate.

I thought the goal of an apu was to not have a cpu and gpu in one die, but the performance of a gpu working with a cpu. So that gpu performance becomes cpu performance. I know it isn't that way yet, but one day maybe.

james888I thought the goal of an apu was to not have a cpu and gpu in one die, but the performance of a gpu working with a cpu. So that gpu performance becomes cpu performance. I know it isn't that way yet, but one day maybe.

The goal of the APU is to:
Have the CPU do serial tasks efficiently.
Have the GPU do parallel tasks efficiently.

The idea is to place parallel tasks on the GPU because they work great on parallel tasks. While placing serial tasks on the CPUs because they work great on serial tasks.

A10-5800K FP SP GFlops: 4 Cores * 8 Flops * 3.8 GHz = 121.6 FMA SP GFlops
7660D FP SP GFlops: 6 Cores * 128 Flops * 0.8 GHz = 614.4 FMA SP GFlops

seronxThe goal of the APU is to:
Have the CPU do serial tasks efficiently.
Have the GPU do parallel tasks efficiently.

The idea is to place parallel tasks on the GPU because they work great on parallel tasks. While placing serial tasks on the CPUs because they work great on serial tasks.

Much better said.

seronxA10-5800K FP SP GFlops: 4 Cores * 8 Flops * 3.8 GHz = 121.6 FMA SP GFlops
7660D FP SP GFlops: 6 Cores * 128 Flops * 0.8 GHz = 614.4 FMA SP GFlops

So you can combine those numbers for total of 736 FMA SP GFlops? If so how does that compare to a 2500k.

sooo tl;Dr amd's latest chip clocks 6Ghz but loses to a 3750k that clocks 3.3 yeaaaaa thats some quality product you got there AMD
good lord they think they would have learned from bulldozer
moar clock speed != MOAR performance
the only thing moar clock speed grants you is higher power consumption and heat
the reason we see AMD driving to push the clock speeds up is they they fired all the engineers that told them otherwise .... and they now lack the talent to design efficient chips they are so far in the hole that its gonna take one or two more generations for them to climb back-out for all of AMD's pr bullshit we have yet to see ONE CHIP from them that lived up to expectations
\
bottem line is when there top end part is getting soundly beaten by offerings from intel that are only ~25-50 bucks more
the people that are looking for performance tend to look REALLY hard at there wallets.

if I went out and built another pc today and was gonna spend >600 bucks. it would be intel because I know in ~4 years that THAT 3750k I bought today will still be-able to offer good (gaming)performance. with amd its a crap shoot if you don't believe me pull of some benchmarks of a Q8400 Vrs a Phenom II 940 yes most titles are very playable and there is less then a 20Fps difference but 20Fps can make the difference between playable and a slideshow ...
---------------------------------------------------------------------------------------------------------------
I used to be a AMD guy like you but then .... I got a 2500k ..... and a wakeup call

Am I the only one that finds it weird that A8-5600K and A10-5800K are so close in GPU performance? It's 256 760MHz shaders vs 384 800MHz shaders after all... I smell bottleneck...

OneMoarsooo tl;Dr amd's latest chip clocks 6Ghz but loses to a 3750k that clocks 3.3 yeaaaaa thats some quality product you got there AMD
good lord they think they would have learned from bulldozer
moar clock speed != MOAR performance
the only thing moar clock speed grants you is higher power consumption and heat
the reason we see AMD driving to push the clock speeds up is they they fired all the engineers that told them otherwise .... and they now lack the talent to design efficient chips they are so far in the hole that its gonna take one or two more generations for them to climb back-out for all of AMD's pr bullshit we have yet to see ONE CHIP from them that lived up to expectations
\
bottem line is when there top end part is getting soundly beaten by offerings from intel that are only ~25-50 bucks more
the people that are looking for performance tend to look REALLY hard at there wallets.

if I went out and built another pc today and was gonna spend >600 bucks. it would be intel because I know in ~4 years that THAT 3750k I bought today will still be-able to offer good (gaming)performance. with amd its a crap shoot if you don't believe me pull of some benchmarks of a Q8400 Vrs a Phenom II 940 yes most titles are very playable and there is less then a 20Fps difference but 20Fps can make the difference between playable and a slideshow ...
---------------------------------------------------------------------------------------------------------------
I used to be a AMD guy like you but then .... I got a 2500k ..... and a wakeup call

QQ some "moar" maybe?

NeoXFAm I the only one that finds it weird that A8-5600K and A10-5800K are so close in GPU performance? It's 256 760MHz shaders vs 384 800MHz shaders after all... I smell bottleneck...

Bottleneck by what? The CPU performance?

The GPU is only a 7660D, not powerful at all. People are running slower Athlon IIs and Phenom IIs and Bulldozers on faster dedicated GPUs and they are not being bottlenecked.

Dent1Bottleneck by what? The CPU performance?

The GPU is only a 7660D, not powerful at all. People are running slower Athlon IIs and Phenom IIs and Bulldozers on faster dedicated GPUs and they are not being bottlenecked.

Uhm... IDK... drivers (yes, I do want to remind people that APUs, like GPUs, are very susceptible to underperforming drivers), system/iGPU memory bandwidth?






VERSUS



And these are just some random pics I picked up to exemplify with.


Yeah, sure looks like all those extra 128 graphics cores (and MHz) are being fully tapped out in most cases. :rollseyes:

NeoXFUhm... IDK... drivers (yes, I do want to remind people that APUs, like GPUs, are very susceptible to underperforming drivers), system/iGPU memory bandwidth?

images.anandtech.com/graphs/graph6332/50163.png

And these are just some random pics I picked up to exemplify with.

Yeah, sure looks like all those extra 128 graphics cores (and MHz) are being fully tapped out in most cases. :rollseyes:

And you are comparing games? How about none game applications. Anyhow this is but a small number of games you've selected. Though I do agree the GPU specs are almost identical. There should have been more of an overall and distinctive performance difference between them. :D

james888Much better said.
So you can combine those numbers for total of 736 FMA SP GFlops? If so how does that compare to a 2500k.

Within the APU, both CPU & GPU can communicate a lot more efficient.

NeoXFYeah, sure looks like all those extra 128 graphics cores (and MHz) are being fully tapped out in most cases. :rollseyes:

But as you pointed out earlier the difference is 256 760MHz shaders vs 384 800MHz shaders between the A8-5600K and A10-5800K. More than likely the GPU just isnt scaling well in general.

We've seen situations with dedicated video cards where increasing the shader count doesn't scale e.g. the 9800GT and 9800GTX or the 4830 and 4850. Similar performance despite the increased shader count. But back in 2009 nobody said CPU bottleneck we just excepted scaling isnt linear.

I was hoping for a more powerful Trinity APU such a based on the HD 7800 series but a tuned down version so they reach there power requirements. I guess for HDTV and mild gaming, AMD's tapped heavily in the OEM with its current offerings.

Dent1But as you pointed out earlier the difference is 256 760MHz shaders vs 384 800MHz shaders between the A8-5600K and A10-5800K. More than likely the GPU just isnt scaling well in general.

We've seen situations with dedicated video cards where increasing the shader count doesn't scale e.g. the 9800GT and 9800GTX or the 4830 and 4850. Similar performance despite the increased shader count. But back in 2009 nobody said CPU bottleneck we just excepted scaling isnt linear.

As far I've played with APU, the bottleneck is the crappy memory controller, therefore the margin between those gpus isn't very big. It wasn't with G92 and RV770 like that. No magic involved. If AMD had something similar as Intel's memory bandwidth, then it would become more fun.

Ferrum MasterAs far I've played with APU, the bottleneck is the crappy memory controller, therefore the margin between those gpus isn't very big. It wasn't with G92 and RV770 like that. No magic involved. If AMD had something similar as Intel's memory bandwidth, then it would become more fun.

You can also add AMD's lack luster cache structure that they've been messing with per each generation but choose not to fix. If AMD can re-design both the IMC and cache, they'll be way better off in both efficiency and performance.