Aceshardware

Trinity is getting closer. It'll be interesting to see if the actual performance is anything like what's being claimed.
-----------------------------------------------------
From AnandTech reports -> Trinity will be available in the middle of the year in three configurations: a 65W - 100W TDP desktop part, a 35 - 45W notebook part and a 17W ULV part.
....AMD claims the 17W Trinity should offer similar aggregate CPU/GPU performance to existing Llano notebook APUs at ~35W. The standard voltage notebook Trinity APU will offer a 25% increase in CPU performance and a 50% increase in GPU performance over the A-series Llano APUs available today. Finally the desktop Trinity will be 15% faster on the CPU side and 25% faster on the GPU. Although AMD didn't disclose details, it's likely that these numbers are comparing a two-module Piledriver based Trinity to a quad-core Llano.

The CPU gains seem modest on the desktop Trinity, but the standard voltage notebook part is pretty interesting as the gains should be enough to mostly bring it up to mobile Sandy Bridge performance (if AMD's numbers are correct).

Trinity is likely going to maintain the integrated GPU performance advantage AMD currently holds, even when Ivy Bridge arrives.
The rest is at: http://www.anandtech.com/show/5411/amds ... in-mid2012

After looking back at Anandtech's own Llano review, I'm afraid that's a bit optimistic. The gap between Llano and Sandy Bridge is consistently over 25%, sometimes well over that. Still, it would close the gap and bring AMD to a CPU performance level most people should be OK with. Of course, there's always Ivy Bridge.

Some performance numbers: http://phx.corporate-ir.net/External.Fi ... lwZT0z&t=1

A4-4355M (17W):
PCMark Vantage: 3525
3DMark Vantage: 1700

A6-4455M (17W):
PCMark Vantage: 4200
3DMark Vantage: 2012

Fastest 17W Sandy bridge (Core i7-2677M):
PCMark Vantage: 9939

Cheap 17W Sandy bridge (Core i3-2367M):
PCMark Vantage: 5894

I couldn't find 3DMark scores with performance preset for Sandy bridge CPUs.
http://www.engadget.com/2011/11/13/leno ... 0s-review/

The CPU performance of the best Trinity is less than a half compared to Sandy Bridge, even the cheap Sandy bridge 17W CPU manages to beat Trinity by a wide margin! And Ivy Bridge will increase this gap further. Trinity will have about 50% lead in GPU performance though.

Despite the improvements, Bulldozer-based CPUs aren't suited for mobile chips. I wonder how low AMD had to clock Trinity to reach 17W target?

Just for clarification/addition to jack's post: A4 is dual core(1 module) and A6 should be quad core(2 modules). Clocks should be ~25% higher than Llano in the same TDP envelope and same core count.

edit: "cheap" version of 17W SB costs 250$ as per intel's recommended price. "Expensive" version is at 317$
http://ark.intel.com/products/59798/Int ... _40-GHz%29
http://ark.intel.com/products/54617

Both are 4T parts. BTW from here you can see i3 2100 @ 3.1Ghz and SMT scores lower than this 17W SB,so that score is pretty abysmal in my opinion(max turbo state of 2.9Ghz is a lot lower than i3 2100 def. clock of 3.1Ghz)

PC Mark score may includes all subsystens, including graphics card and disks, if the systems aren't equal these scores aren't comparable.

Yes PC Mark isn't a best benchmark, I hope other scores become available.

Core i3-2367M is cheap enough to be included in <600$ laptops: http://www.cdw.com/shop/products/Lenovo ... VP-233828U

Now that Trinity is just around the corner,how about Kaveri(SR/GCN based) APU specs.
From AMD official PDF ,here is what they say about Kaveri:


8CU means ~7750 level. Of course the problem is memory BW so the part will most likely land ~15-20% below 7750. Nevertheless this level of performance will be very high and Haswell probably won't be able to touch it (perf. or image quality wise). Then we have SR cores which hopefully will bring native AVX(2) support with much much better SIMD performance versus BDver1. This would imply 2x256bit FMAC/SSE units per core pair,instead of 2x128bit ones in current BD core.