Kaveri A10 spotted in the wild!

[Kai Robinson]

http://www.hardware-360.com/amd-10-k...ed-and-tested/

Dubious source? How does it stack up compared to other AMD CPU's?

[xVeinx]

Sooo, the desktop chip almost catches the ultra book chip? *sigh*

[informal]

2 more weeks, just hold on. :P
Most interesting will be the price for 7850K and 7770K. How much will AMD ask for these "next" gen APUs? We already have Richland so all comparisons must be made against these parts (both perf. and perf./$).

PS That "ultrabook" chip costs a ton and turbos much higher than its "rated stock" clock. It is actually on par with x5650 (3.06Ghz ST Turbo Nehalem part)...

[xVeinx]

Oh, I know the specs on the processor and about turbo. Still, one would hope... Anyway, I'm actually still interested in power consumption. Price should remain at or slightly above where the pricing is now.

[Mech0z]

I want to see how well that configurable TPD affects power consumption

I want to use it in my HTPC/NAS server, so power consumption is key, atm there is a A4-4000 in, but my A88X is FM2+ ready, so just waiting for a power effecient and cheap Kaveri!

[[XC] Lead Head]

Can someone explain to me why Cinebench says 3.7GHz, while CPU-Z says the maximum clock can only be 2GHz? Can Kaveri dynamically change bus speed? Or is CPU-z or Cinebench wrong?

Are these GCN 2.0 cores, or GCN 1.1 like Hawaii? Does Steamroller "B" imply there was an "A" revision that never saw the light of day?

[xVeinx]

They are GCN 1.1 cores, like hawaii. B means no A, yes. And CPU-z is wrong in this case.

[PatRaceTin]

Oh new Internet Cafe CPU is coming

[informal]

xVeinx is correct.

[Heinz68]

"More details to come at CES 2014" from AMD’s annual developer conference (APU13) slides published at AnandTech on November 11, 2013

First and foremost, AMD has confirmed that Kaveri will be shipping in Q4’13, with a launch/availability date of January 14th, 2014. For those of you keeping track of your calendars, this is the week after CES 2014, with AMD promising further details on the Kaveri launch for CES.

EDIT
From the same AnandTech article

Furthermore, AMD has also confirmed that their new low-level API, Mantle, will also be supported on Kaveri – it is after all a GCN based GPU.

From Charlie article about Battlefield 4 Mantle patch delay

SemiAccurate does expect the Mantle engine version to be demo?d all over the place at CES next week though, stay tuned.S|A

[mAJORD]

That cinebench score is slower than the 6800K, so unlikely to be correct.

Should score in the high 300's if AMD's perf targets are to be believed.. which are quite modest anyway.

[informal]

That cinebench score is slower than the 6800K, so unlikely to be correct.

Should score in the high 300's if AMD's perf targets are to be believed.. which are quite modest anyway.

You forget the clock speed that 7850K runs C15 at... The score is perfectly in line with what we know about SR: MT scaling is better and this is what makes SR outperform PD in overall score at the same clock. FP resources are in fact not changed (as per AMD) so in "legacy" FP workloads (SSE and not AVX/FMA) SR has no advantage over PD, it even has one pipeline less. It's a feat it doesn't perform worse in the overall score per clock...
Excavator core is bringing the SIMD pipeline widening(native 256bit paths) but I think only the tuned workloads (AVX2/FMA) will show big gains VS SR/PD.

[FlanK3r]

its a bit slower, but the reason is, A10-6800K working in hard load between 4100-4300 MHz clocks. Kaveri around 3700-3800 MHz.

[mAJORD]

You forget the clock speed that 7850K runs C15 at... The score is perfectly in line with what we know about SR: MT scaling is better and this is what makes SR outperform PD in overall score at the same clock. FP resources are in fact not changed (as per AMD) so in "legacy" FP workloads (SSE and not AVX/FMA) SR has no advantage over PD, it even has one pipeline less. It's a feat it doesn't perform worse in the overall score per clock...
Excavator core is bringing the SIMD pipeline widening(native 256bit paths) but I think only the tuned workloads (AVX2/FMA) will show big gains VS SR/PD.

I misread it saying up to 4Ghz. I thought that was the base speed.

So correcting for that, not so bad, but its still only showing a ~5% incease in IPC in this bench, Multithreaded. That surely can't be right, and if they're the final clockspeeds, then, well it's slower than the 6800K in this bench.

If it is right, then hopefully its integer perf is a lot higher , or this is an outlier benchmark

[To(V)bo Co(V)bo]

I think AMD is just missing all the opportunities to capitalize on Intel's laurels. If they would release a cutting edge platform as a whole, people would be far more interested in buying it over Intel. Right now they have a awful chipset. Back when phenom II came out they had a edge in chipset technology that many people did choose over Intel. I cant believe we still dont have native usb3 and sata3 for every port. I refuse to invest in a new platform till this is standard. AMD server boards are even worse as far as chipset is concerned. If AMD cant compete in cpu performance, then find another angle to compete. Give me a better computing experience, and you would be surprised by the interest it will create.

[FlanK3r]

there is my answer:
http://www.xtremesystems.org/forums/...=1#post5221941

[Piledriver]

there is my answer:
http://www.xtremesystems.org/forums/...=1#post5221941

IPC is better around +300-350 MHz.

But clock is lower and clock is part of the equation. Never understood the people that kept whining about bulldozer\piledriver "bu-bu it has more more 800mhz" like being faster because of higher clockspeed was some sort of disability.

[Dimitriman]

I can't believe AMD is releasing a processor that's 500 Mhz slower while being in a smaller node...

What the is going on?

[Olivon]

125W TDP (Richland) Vs 95W TDP (Kaveri), that's all.

[FlanK3r]

of yourse, u have right in this. The CPU performance is +- the same as product to product. I take it from my point of view with OC. I believe Kaveri can hit around 4500-4600 MHz stable, maybe more.

Dimitriman: based at one benchmark we can say nothing...In Cinebench is performance +- the same. But we dont know nothing about next benchmarks and tests.

Olivion, you are wrong, Richland TDP is 100 W not 125 W

[informal]

I can't believe AMD is releasing a processor that's 500 Mhz slower while being in a smaller node...

What the is going on?

It's called 28nm bulk process node...

[Olivon]

Olivion, you are wrong, Richland TDP is 100 W not 125 W

You're right FlunK0r, it's 100W TDP, my bad.

[Ace123]

FlunK0r

No reason to be rude. Please change it

[Olivon]

The day Flanker will begin to write my name correctly, I will change, I promise. (I told him tenth times even more, but you seem new on XS and didn't noticed it).

[FlanK3r]

Hope I remember it now Olivon (but Olivion seems as some hero from Greek mythology )