AMD A10 "Kaveri" APU Pictured, Battlefield 4 Bundles Planned

[Mysteoa]

I call it fake.

 

[msx68k]

So you mean to tell me kaveri gets beat by a 1.7ghz ivy bridge I3 in single thread? That's not right, even Richland did better than that

Nop. just look the 2nd image. The APU was running @848 MHZ.

 

[newtekie1]

Look at the 2nd image. The APU was running @848 MHZ.

Your point? According to CPU-Z my 6800k will drop down to an 8x multiplier too when idle. Though it only seems to ever go down to a 20x multiplier, maybe AMD finally fixed that bug with Kaveri. The CPU-Z shot was taken when the CPU is idle, so it will show the clock speed of the power saving state. This has been a common thing for ages now.

 

[JDG1980]

Look closely, he has the benchmark set to show single core performance. The fact that a single Kaveri core at 3.7GHz is scoring right about half of a single Haswell core at 4.4GHz is a pretty decent improvement.

Look closer. The Cinebench "screenshot" shows a multi-threaded score of 311 CB, as well as the single-threaded score of 88 CB. Not only would both of those be pathetic showings, but the fact that the multi-threaded score is only 3.53x the single-threaded score (that's the "MP Ratio") would mean that the CMT penalty we see on Bulldozer and Piledriver architectures is still in full force. Since Steamroller's architectural changes are largely about getting rid of that penalty, these results make no sense. In the unlikely event that these screenshots were accurate, it would mean that Steamroller is essentially no advance over Piledriver at all. That doesn't make sense, given that we know that Steamroller has independent decoding units for each core (BD and PD had to share within modules) and that there are various other improvements to the cache, branch prediction, and so forth.

It's fake.

 

[Ravenas]

They're using the newest Steamroller CPU and Radeon GCN cores from the R7 series and are capable of running BF4 at a usable rate of 25-35FPS.

Why they'd supply BF4 with this I don't know.

You can pair these processors with a dedicated GPU in hybrid crossfire and get great performance.

 

[NeoXF]

What are you talking about bullshit? APUs are NOT for enthusiasts, period. You don't see All-In-One (monitor+pc in one) setups being touted as enthusiast platforms; stock cars are not enthusiast until you modify just about everything under the hood to make it top-notch ready for racing or for whatever reason; I can go on all day w/ examples, but I won't.

The point is that APUs are designed with simplicity in mind. Simplicity mixed with a bit of power, enough to do most tasks and then some (light gaming). Yes it addresses the paradigm shift in CPU computing, but it doesn't mean that there's going to be a high-end APU capable of playing BF4 w/ all settings maxed and high resolution, at least not anytime soon. Every product has it's place. If you convince yourself otherwise, then your just lying to yourself.

Computing =! Graphics
Modern iGPUs =/= Graphics Only

Okay, let's take this another way, just to prove how clueless you people are, I'm gonna ask you guys what does APU stand for and why does it stand for that.

To be fair, APUs, at least in AMD's true "vision", don't really exist yet, Kaveri should change that. But AMD is doing a pretty good job at explaining what they are (that CGI video was awesome), but not as good a job in showing it, yet.


The images are fake, as stated before, the multiplier is weird (measures of 0,5? Really? What is is this, Athlon 64?), the trademark/copyright sign is not where it's supposed to be, scores makes zero sense (A10 6800K/DDR3-2133 gets 328 points) and a lot of other shady stuff that I won't bother going into.

Not really, Richland in single threaded performance was way behind IvyBridge.

He never said Richland is faster in IPC than Ivy Bridge... he said Richland lost to a smaller margin in IPC (to whatever) than those images make out Kaveri to. Read again.

 

[OneCool]

2 cores 4 threads?

Is that right?

 

[newtekie1]

So you mean to tell me kaveri gets beat by a 1.7ghz ivy bridge I3 in single thread? That's not right, even Richland did better than that

He never said Richland is faster in IPC than Ivy Bridge... he said Richland lost to a smaller margin in IPC (to whatever) than those images make out Kaveri to. Read again.

I never even came close to saying Richland was faster in IPC than IvyBridge. I know exactly what he was saying and I'm saying Kaveri looses by a smaller margin than Richland. Richland wasn't even half as fast as Ivybridge in single threaded performance. Kaveri is now right about half of Haswell, that is a significant gain.

2 cores 4 threads? Is that right?

Cinibench reads the "module" as a core for some reason. It does this with all of AMD's recent CPUs.

Look closer. The Cinebench "screenshot" shows a multi-threaded score of 311 CB, as well as the single-threaded score of 88 CB. Not only would both of those be pathetic showings, but the fact that the multi-threaded score is only 3.53x the single-threaded score (that's the "MP Ratio") would mean that the CMT penalty we see on Bulldozer and Piledriver architectures is still in full force. Since Steamroller's architectural changes are largely about getting rid of that penalty, these results make no sense. In the unlikely event that these screenshots were accurate, it would mean that Steamroller is essentially no advance over Piledriver at all. That doesn't make sense, given that we know that Steamroller has independent decoding units for each core (BD and PD had to share within modules) and that there are various other improvements to the cache, branch prediction, and so forth.

I'm aware of that, however there definitely has been an significant improvement over Piledriver and Bulldozer. The MP Ratio on my Piledriver 6800K is 3.20x. So going up to 3.53x is a pretty significant jump. And remember, with a 4T processor the best you can really expect is 4.0x, so 3.53x is decent.

 

[EpicShweetness]

2 cores 4 threads?

Is that right?

Yes, Cinebench is reading it how it is, mostly. There are 2 modules 4 cores, but since it's just 4 integer cores, not 4 complete cores Cinebench for that probability posted it as such.

 

[eidairaman1]

I really want to see how this fares against the 5800K, 6800K, FX 4350, 6300/6350, 8300/8350, 9000 series.

im pretty split between this and a AM3+ CPU

 

[Arjai]

I have a 3317U i5, It truly is impressive. I ran a program, a few weeks ago, that showed me the current wattage, as I was doing tasks.

I was Crunching WCG at 100% cores and 100% CPU. 9W...9!

I have 8 Watts to play with that Crunching at 100% hasn't even touched!!

Now, I see that r15 where my little i5 beats the bestest A10? Wild.

I have a couple AMD desktops, in storage. One is an old Socket A, another is a dual core 64, don't remember. I also have a P4 intel, 2.8 GHz that crunches well. The Dual core puts up decent numbers but, is a wattage whore. The Socket A? That's my toy. It crunches a bit but mostly just a fun thing to OC and check Facebook on.

Sorry, kinda wandered off topic. Sorry, I seem to do that sometimes...

 

[HisDivineOrder]

I'm sure they expect Mantle to make the APU capable of running BF4 well enough and since Mantle is going to be pushing APU's in a big way, I'd expect this is a wise play.

 

[NeoXF]

I never even came close to saying Richland was faster in IPC than IvyBridge. I know exactly what he was saying and I'm saying Kaveri looses by a smaller margin than Richland. Richland wasn't even half as fast as Ivybridge in single threaded performance. Kaveri is now right about half of Haswell, that is a significant gain.



Cinibench reads the "module" as a core for some reason. It does this with all of AMD's recent CPUs.



I'm aware of that, however there definitely has been an significant improvement over Piledriver and Bulldozer. The MP Ratio on my Piledriver 6800K is 3.20x. So going up to 3.53x is a pretty significant jump. And remember, with a 4T processor the best you can really expect is 4.0x, so 3.53x is decent.

Then what was the point of contradicting him if you know what he meant and believe so as well.
Anyway, how do you know how Kaveri performs? LOL

If by any chance any of you are doing the math in relation to this "article", I gotta ask you to do a reality check and notice it's fake and to urge you to stop wasting your time.
Also, Cinebench is a pretty piss-poor measurement of performance between Intel and AMD (to begin with), as are all other ICC compiled software.

 

[newtekie1]

Then what was the point of contradicting him if you know what he meant and believe so as well. Anyway, how do you know how Kaveri performs? LOL

Umm...Do you even read?

His claim was that Kaveri is worse than Richland.

My points:

Richland isn't even half as fast as IvyBridge.
Kaveri is now half as fast as Haswell.

How do you get that I believe what the same thing he does from that?

 

[TheinsanegamerN]

I was impressed...until I realized the i5-3317U is an Ivy Bridge processor for notebooks.

a sub 2 GHz, 15 watt part at that. cpu performance is what killed trinity in notebooks. it always held back the graphics part of the apu is pretty much any game.

 

[agent00skid]

A sub 2GHz part with turbo up to 2,6 in a single threaded test...

Also, this is only 1 benchmark, which is too small a test set to give reliable information about performance

 

[TheoneandonlyMrK]

Not a great move IMO. If AMD was trying to tout it's new APU as being able to run BF4, people who don't know any better will expect it to run "good". AMD should've put a little more umph into it's graphics portion of the APU in order to make it more appealing at a smaller price bump which I'm sure some people wouldn't mind paying.

Its likely that demo video is not using Mantle, If Mantle does what they believe it will ,then they wont need more oomph just more mantle compatible games.

 

[TheinsanegamerN]

A sub 2GHz part with turbo up to 2,6 in a single threaded test...

Also, this is only 1 benchmark, which is too small a test set to give reliable information about performance

and the richland a10 has a turbo rating of 3.5 GHz in single thread testing. this chip is probably clocked even higher. clock rate isnt everything.

 

[sergionography]

Look at the 2nd image. The APU was running @848 MHZ.

Then you mean to tell me kaveri is twice as fast as ivy bridge? Lol

On a more serious note don't forget turbo, as those ivy cores when one core is running end up running well over 2.6-2.7ghz which is what they probably were running the test at, and the fact that a haswell clocked at 4.4ghz is not more than twice as fast as the ivy 1.7ghz is proof of that

 

[NeoXF]

Then you mean to tell me kaveri is twice as fast as ivy bridge? Lol

On a more serious note don't forget turbo, as those ivy cores when one core is running end up running well over 2.6-2.7ghz which is what they probably were running the test at, and the fact that a haswell clocked at 4.4ghz is not more than twice as fast as the ivy 1.7ghz is proof of that

You mean Richland?

Anyway, good point, low power Intel CPUs do turbo very high on one core especially.

But it's irrelevant, since the screeshots are 99% fake.

[...]

LMAO, all three of us are saying pretty much the same thing, just that he started by saying the scores in the images can't be right, and they aren't, since they are made up.

Can we STFU about this now?

 

[sergionography]

You mean Richland?

Anyway, good point, low power Intel CPUs do turbo very high on one core especially.

But it's irrelevant, since the screeshots are 99% fake.



LMAO, all three of us are saying pretty much the same thing, just that he started by saying the scores in the images can't be right, and they aren't, since they are made up.

Can we STFU about this now?

Well it's a rumor which like u say is most likely fake, but I'm.hypothetically speaking assuming this is real. so according to this we can say a 3.7ghz single steamroller core matches a 2.8ghz ivy bridge core which is reasonable if you think about it. And those bulldozer derived badboys do clock much higher at the same tdp than the Intel counterparts so amd might be on to something If they can clock the mobile parts like Richland or even higher then they can be much closer to Intel with the added benefit of better multithreaded performance since most Intel mobile is dual core.
but then again that multicore score is almost exactly what Richland scores if not less if I remember correctly, I can check reviews but I'm too lazy right now lol. And I'm using my cellphone to write this