AMD A10 5800K Launch Review



Product: AMD A10 5800K / ASUS F2 A85-M Pro
Company: AMD
Author: James Prior
Editor: Charles Oliver
Date: October 1st, 2012

Summary and Conclusions

Contrasting the AMD A10-5800K to Intel's Core i3 is a hard task, because they don't exist alone in a vacuum like add-in boards do. A mainboard is required, as is memory, and these choices can significantly affect the value proposition and use cases of the final platform. That said, the fundamental basis for comparison is price, looking to find which platform gives you the best value, the best bang for your buck. To fully do that, we have to define what we're looking at.

Intel's Hyperthreading (HT) and AMD's Cluster-based multithreading (CMT) are both technologies designed to increase utilization and thread execution count out of a processor. Intel doesn't market HT as a full extra core due the highly shared nature of the design, and tells people that is offers applications upto 20% more performance per core. AMD claim CMT offers applications up to 80% of the performance of running on a single module for each thread, and markets modules as dual core modules because while there are shared components there are big concurrent processing abilities. Is it a dual core with CMT, or is it a quad core? Does it matter given the accurate price positioning?

Both AMD and Intel give their products meaningless names that don't inform the consumer of their capabilities, nothing new for Intel but for AMD another departure from the recent X2/X4/X6 processors. The A10 5800K is a quad thread CPU with a 3.8GHz clock, 4.2Ghz Turbo, and has Radeon HD 7660D graphics inside. There are four things wrong with that, if you don't have AMD's CPU naming decoder ring on you in the store, not that there is one - we need an app for this, like Intel does.

The most egregious of the marketing slip ups is the graphics branding, the 7660D doesn't perform on par with 6600 series graphics. Mostly this is due to the memory, discrete 6600s have 128-bit GDDR5 and the A10's 7660D has 64-bit DDR3, representing a big difference in bandwidth. The branding should reflect this more accurately, being a 7550D. The performance compared to Intel's HD 2000 or 2500 graphics is monstrous; the experience is literally game changing going from a Core i3 to AMD A10 APU. To exceed the abilities of the A10 5800K in all areas requires a more powerful product than an i3 plus a discrete graphics card, which increases initial spend considerably. If you respect the price for price comparison model, you've got to add that cash to the A10 system and then you start boosting the whole platform with SSD caches or more RAM or just much better GPU performance again, enriching the whole user experience beyond what you get for the Intel based system.

AMD's A85 platform is lacking in a couple of respects, there is no desktop switchable graphics like technology for running a single large discrete GPU alongside the APU graphics, a shame given the Zero Core Power options in the AMD Radeon 7700/7800/7900 series and the existence of said option on AMD's notebook platforms ostensibly using the same core dies. ASUS pick up the slack here with included Virtu software but a hardware based solution would be preferable. CrossfireX is also short of the mark, with dual discrete GPUs supported but tri-fire of the APU with two discrete not supported - for desktop gaming enthusiasts this would be an excellent way to scale graphics performance, but perhaps the overhead of triple GPU's just isn't worth it on that level of discrete card - there is no tri-fire on anything less than the top product lines anymore. The update to eight native SATA-3 6gbps ports is class leading, and an excellent idea. Lack of support for SAS drives is disappointing, even if it's not a consumer feature that workstation boards using the A85 chipset and the FirePro APU's ought to be able to use SAS drives without buying more hardware. A85's USB 3.0 performance is much better than Intel's Cougar Point although Panther Point brings back parity, but then surges ahead with something AMD's 2012-2013 platform lacks - PCI-Express Gen 3.0.

The ASUS F2 A85-M Pro mainboard is an excellent platform for the AMD second generation A-series APUs, perfect for combination home theater movie and gaming machines, office workhorses and enthusiast value computing. The introductory price appears to be around $140USD which is a little high but should drive down nearer to the ~$122 price point of the APU quickly. The features we'd like to see - capability to run dual discrete GPUs plus mSATA or SSD on a stick will likely be added in on other mainboard models soon, and aren't requirements for this price point. The added value software is comprehensive and mature, excellent values add for the board although it doesn't include a decent antivirus short subscription, just Norton. The board is fully Windows 8 ready with UEFI support for secure boot, TPM and drivers to match. It's a five star product, and we award it accordingly.

ASUS F2 A85-M Pro Mainboard

The A10-5800K results show how much graphics performance matter now to general computing, x86 performance alone doesn't solve the user experience question. AMD wins for video quality thanks to their enhancements like Steady Video, Picture Perfect HD make watching videos no matter the source more pleasant. AMD loses in Video processing, where they should win - the Video Codec Engine and GPU Compute power ought to give more of an advantage. Applications that meaningfully use the VCE and OpenCL compute power aren't working properly yet, disappointingly still leaning only on the CPU for processing power which is a problem given Intel's significant lead in x86 IPC and extended instruction performance. AMD's strategy here is to develop common and open APIs for vendors to use how they see fit, trying to break lock in of proprietary API's currently in use and move to a level playing field, OpenCL. Alongside this they have made API's for their UVD and VCE hardware, plus Eyefinity and AMD HD3D, so anyone can program to the features offered. Tools like Code XL and CodeAnalyst, and libraries for OpenCL and different types of workload are freely available, but adoption is not at the pace Enthusiasts want - yet. As Heterogeneous System Architecture (HSA) continues to evolve, AMD need to push harder for developers to get code to the public quickly that shows what it can do, preferably before launch - no feature left behind.

AMD is preeminent in the gaming benchmarks, there's no competition for the A10 at this price point and there won't be for some considerable time. On top of which, AMD's anisotropic filtering quality is near perfect, whereas Intel HD graphics provide shimmering and blurring textures. Intel got to DirectX 11 with the release of their Ivy Bridge codnamed 3rd generation Core I-series in April 2012, three and a half years after AMD launched their first DirectX 11 part and 10 months after AMD launched DirectX 11.1 parts for the upcoming Windows 8 launch. Add in Enhanced, Supersampling, Adaptive, and Morphological Anti-Aliasing and you really can't compare the two, Intel are in the dark ages of graphics compared to AMD.

The power consumption results were a surprise, AMD's in use results show that despite the 35W TDP chasm between our i3 2120 (65W) and the A10-5800K (100W) the only real difference in use comes when AMD are leveraging that monster GPU side of their processor to increase performance and quality over the Intel offering. Add the fact that the A10-5700 (65W) exists (at the same price) and you see the obvious potential of AMD's HSA implementation, it is going to be successful. This is largely in part due to AMD's power gating and dynamic boost technologies, which in our opinion aren't aggressive enough. There's plenty of clock headroom in both sides of the APU but AMD's overly conservative TDP caps don't exploit that enough. The technology and lessons learned this year with AMD's Southern Islands Boost revision using their GPU PowerTune technology can't come to the APU soon enough.

Likewise, PCI-Express 3.0 should be there, to support the updated data and command capabilities as well as offer more bandwidth – our A85 platform could have had full PCI-E 2.0 x16 bandwidth from each PCI-Express Graphics slot to the APU if the APU had PCI-E 3.0 capabilities, something that dual graphics and CrossfireX would both benefit from. Launching as they do so close to the Windows 8 release, we're disappointed in the lack of DX11.1, too, only owners AMD performance and enthusiast discrete graphics get DX11.1 goodies. However, all of that must come second to lighting a fire under the memory controller development, the biggest bottleneck for performance of both the CPU and GPU. Intel's memory performance is astonishing compared to AMD's, and contribute to their dominance in some areas. AMD are hopefully preparing a supremely optimized CPU and GPU shared memory interface for the recently announced DDR4 specification, but in the meantime there's plenty of work to be done on the DDR3 one.

We award the A10-5800K 5 stars, and hope that AMD will double down on their investment in relationships with developers to get their unique abilities supported faster and continue to improve the Bulldozer family. With the new technical leadership appointments over the last 12 months, such as Mark Papermaster, John Gustavson, and Jim Keller , the question is not 'can AMD do this', but 'how quickly'?

AMD A10 5800K