Bookmark and Share


Advanced Micro Devices said on Monday that it would start selling its next-generation A-series “Kaveri” accelerated processing units with next-generation Steamroller x86 cores, significantly improved graphics processing  units and heterogeneous system architecture in mid-January. The company also revealed the first official performance numbers for the new chip.

General specifications of AMD A-series “Kaveri” are pretty much well known by now: the new APUs feature two dual-core x86 Steamroller modules (two integer units per module, one FPU with two 128-bit FMACs, one MMX unit), Radeon R7 graphics processing unit with 512 stream processors (eight GCN compute units) as well as HUMA [heterogeneous unified memory architecture] shared memory controller. The total compute performance of AMD A10-7850K (3.7GHz x86 core clock-rate, 720MHz graphics engine frequency) “Kaveri” APU will be around 856GFLOPS, which is quite a lot.

The new series of APUs support all the technologies that AMD introduced along with its latest family of graphics processors – which feature improved second-gen GCN architecture – including TrueAudio as well as DirectX 11.2 and Mantle application programming interfaces.

One of the key innovations of Kaveri is its HUMA shared memory controller that fetches data to both x86 and GCN processing units as well as acts like a dispatcher/hub. The new APUs also support heterogeneous queuing that defines how processors interact equally and enables uniform visibility into the entire memory space by both GPU and CPU.

Thanks to HUMA and heterogeneous queuing, applications can easily run processes on different types of cores, such as x86 CPU or GCN GPU compute units. As a result, AMD claims that its new Kaveri features twelve compute units in total. While the claim is technically correct, it should be kept in mind that not all applications can offload computing to stream processors of graphics adapters, hence, a number of programs will take advantage of only four x86 processors.

On Monday, AMD also released the first official performance benchmark results for AMD A10-7850K “Kaveri” accelerated processing units. Based on the company’s own tests in synthetic applications, the flagship APU from AMD will deliver 24% - 87% higher performance depending on the application when compared to Intel Core i5-4670K chip. Real-world performance will differ from synthetic programs, but it is good to see that the new APU is faster than its predecessor by quite a significant margin as well.

Starting January 14, 2014, desktop system builders will offer PCs with AMD A10-7850K and AMD A10-7700K accelerated processing units inside. The chips are compatible with FM2+ mainboards and should be released to the channel shortly. Unfortunately, their exact prices are unknown. Later this year AMD intends to reveal other versions of Kaveri aimed at lower-end desktop PC segment as well as special versions of the chips for notebooks and even servers.

Tags: AMD, Kaveri, Fusion, Steamroller, Radeon, GCN


Comments currently: 39
Discussion started: 01/07/14 08:20:23 PM
Latest comment: 01/11/14 06:52:12 PM
Expand all threads | Collapse all threads


Conspicuous lack of a CPU based benchmark while still claiming "vastly superior performance". Hell the conspicuous lack of any independent pre-release benchmarks is reminiscent of Bulldozer.
4 4 [Posted by: AnonymousGuy  | Date: 01/07/14 09:24:32 PM]
- collapse thread

You can be sure that graphics benchmarks will be vastly superior on AMD. No doubt single thread CPU benchmarks will continue to be dominated by Intel. Both take the approach, or at least are forced by circumstance to take the approach, that good enough is, well, good enough; Intel with regards to graphics, AMD with pure CPU, and most of the time they are both correct - with the exception of course, of comparison benchmarks. This is true, not simply due to expertise or technology, but at least as importantly due to the choice of how many transistors and die area to allocate to CPU and GPU.

An increasingly important factor this time around, and going forward, is how much GPU compute application software optimisations affect the balance of power in the benchmarks. And perhaps even more interesting will be how well HUMA is able leverage performance gains, with or without those specific software optimisations.
3 0 [Posted by: mjv1121  | Date: 01/08/14 06:09:55 AM]

Till benchmarks are used without the help of a separate video card the Intel will usually win. Please compare like with like. If a chip has an in built video then it should be tested with its native in-built video not a 3rd party video card. This will then give a truer picture of the chips performance. If the results don't suit some people and or advertisers its too bad.
4 1 [Posted by: tedstoy  | Date: 01/08/14 12:07:48 AM]

No one should expect miracles here, probably some 10+% computing (CPU power) and more than that in GPU. Will bury Intel in graphic department, but then again, it was always like that.

More important is that new Windows can use HSA, and that major players supports HSA/OpenCL.

With Mantle, GPU is expected to be up to 50% faster than best Intel offering (which, by the way, comes on 300$ CPU, price range in what people buy separate GPU), making a number of games playable, for a decent price. Appealing to a part of market, but then again, 300+ CPU buyers are just a part of market, too
2 0 [Posted by: snakefist  | Date: 01/08/14 05:57:28 AM]

I would care if it was the CPU part that was faster.
0 0 [Posted by: Harry Lloyd  | Date: 01/08/14 08:32:02 AM]
- collapse thread

It is, or can you not read.
0 0 [Posted by: KeyBoardG  | Date: 01/08/14 11:18:31 AM]

Seems taht these Kavery chips really are delivering the promissed performance. Congrats AMD! Can't wait to see the chips complete tests.
3 4 [Posted by: MHudon  | Date: 01/08/14 08:43:09 AM]

Actually there is up to 20% IPC improvement clock-for-clock over Richland and up to 50% GPU improvement before using Mantle which is worth another up to 45% GPU improvement.

AMD got it right with Kaveria and the mobo and system makers are all onboard because of it. Kaveri will really shine in laptop where it is at least a year or more ahead of Intel's APUs in system performance.

The haters will continue to hate because they are clueless fanbois. Kaveri is very good for consumers who want great performance at a good price.
4 3 [Posted by: beenthere  | Date: 01/08/14 08:56:49 AM]
- collapse thread

Thinking Mantle is actually going to be adopted by devs: cute.
1 0 [Posted by: AnonymousGuy  | Date: 01/08/14 06:20:43 PM]
The 20% IPC improvement is only with multi-threaded software, when both "cores" of a module is used. With single-thread active per module, the IPC improvement is only 5-10%.

And single-threaded performance is where AMD has the biggest bottleneck and furthest away from intel.
0 0 [Posted by: hkultala  | Date: 01/11/14 02:43:33 AM]

I could love to see that promises to be truth so we can finally enjoy some competition and lower prices in the market, but since the first benchmarks hit the web I will not celebrate yet.

Also if this is truth I want to see what they will do with the athlon CPUs (who so far where APU with GPU disabled and in lower price than the APU counterpart) because that unification of the CPU and the GPU worry me a bit.

I wonder why however they don't release a FX chip as well, if the steamroller architecture is that good some high performing FX chips could possible catch up with intel in more fields than the Vishera did (because really where they did, it didn't matter for the majority of the people out there).
0 0 [Posted by: nitro912gr  | Date: 01/08/14 01:42:42 PM]

Unfortunately, the benchmarks published by Puget show that the new iGPU is only 10% faster than the iGPU on Richland.

Sure .. the frequency is now 10% lower and the power consumption is also 5% lower, but for the 173 USD that we're going to pay for the best Kaveri APU at launch, we'll only get 10% better graphics.

That's most unfortunate for AMD, but better Mantle adoption could possibly transform that 10% graphics improvement into 50% and that might sweeten the deal.

It's great to see that the x86 performance is up 10% also, but the clockspeed downgrade almost negates that.

Kaveri will be great once it get down to current Richland prices.

Overall this is the best AMD CPU launch since they've announced the Jaguar cores that brought them so many big deals.

It's unfortunate that they are not using Soitec's wafers anymore as SOI could have brought us 4.3 GHz base frequencies with probably 5 GHz Turbo clocks.

SOI also brings important improvements in the low power consumption department, but on the other hand, SOI wafers are more expensive.

It is interesting that, from what I hear, since AMD moved away from SOI, Soitec is now a bit in trouble, as they've lost a very important customer. I'm not sure that IBM still buys their wafers from them, but AMD is a big loss for Soitec.
0 0 [Posted by: East17  | Date: 01/08/14 02:41:59 PM]

it doesnt matter we will all be going back to the abacus soon any ways, then i bet you will fight over who makes the fastest abacus
0 1 [Posted by: Ted Feasel  | Date: 01/09/14 02:27:23 PM]


Add your Comment

Related news

Latest News

Wednesday, November 5, 2014

10:48 pm | LG’s Unique Ultra-Wide Curved 34” Display Finally Hits the Market. LG 34UC97 Available in the U.S. and the U.K.

Wednesday, October 8, 2014

12:52 pm | Lisa Su Appointed as New CEO of Advanced Micro Devices. Rory Read Steps Down, Lisa Su Becomes New CEO of AMD

Thursday, August 28, 2014

4:22 am | AMD Has No Plans to Reconsider Recommended Prices of Radeon R9 Graphics Cards. AMD Will Not Lower Recommended Prices of Radeon R9 Graphics Solutions

Wednesday, August 27, 2014

1:09 pm | Samsung Begins to Produce 2.13GHz 64GB DDR4 Memory Modules. Samsung Uses TSV DRAMs for 64GB DDR4 RDIMMs

Tuesday, August 26, 2014

10:41 am | AMD Quietly Reveals Third Iteration of GCN Architecture with Tonga GPU. AMD Unleashes Radeon R9 285 Graphics Cards, Tonga GPU, GCN 1.2 Architecture