Thursday, August 18, 2011

AMD Desktop Llano A8-3850 Review

 
AMD has today launched a new processor in the desktop arena that radically alters the competitive environment. Fusion (or Desktop Llano) merges one or more of AMDs X86 cores with hundreds of Radeon cores on the same silicon. They now refer to the CPU section as an APU and the graphics as a GPU creating a fusion of the two. Llano is the successor to Zacate which gave Intel a huge headache when competing against their Atom processors in the Notebook and Netbook market over the past several months.

Traditionally Intel and AMD have launched new processors at the top of their range and adjusted the others accordingly, allowing these premium CPUs to trickle down over time. Llano is different in that AMD are targeting the low end even if it means the risk of cannibalising their own market - expect to see a phasing out of low end AMD processors and possibly some good baragins for socket AM3 owners. The idea is to have Llano grab the bottom of the market while the forthcoming Bulldozer (Zambezi) does the same for the top end. This is a bold move and we'll see over the coming months how it affects the marketplace.

Processor Architecture

To support the new processors, we are getting a new architecture and two new chipsets, the A75 and A55. Readers may need to click on the diagrams below to see the full size ones to be able to make out details.



The A75 is the high end option with 6GBps SATA ports and 4 USB3.0 ports (the first time we don't need a 3rd party chip for USB3.0 support).

The A55 uses less power, loses all the USB3.0 ports and has the SATA-3 ports cut back to SATA-2. While this may not be much of an issue for mechanical hard drive users, even low end SSDs are now able to make use of the SATA-3 headroom so anyone planning on having an SSD (or even one of the newer USB3.0 thumb drives) may want to go for an A75 motherboard.

Here are the specs and comparisons for AMDs A8 and A6 range:

AMD Llano Desktop Lineup
GPU Total TDP (GPU + CPU) CPU Cores CPU Clock (Base/Turbo) GPU Cores GPU Clock Price
AMD A8-3850 Radeon HD 6550D 100W 4 2.9GHz 400 600MHz $135
AMD A8-3800 Radeon HD 6550D 65W 4 2.4/2.7GHz 400 600MHz $??
AMD A6-3650 Radeon HD 6530D 100W 4 2.6GHz 320 443MHz $115
AMD A6-3600 Radeon HD 6530D 65W 4 2.1/2.4GHz 320 443MHz $??

Turbo core is done away with on the XX50 models but remains on the low power versions. This seems to be a conscious effort and probably is to avoid exceeding a thermal envelope, even briefly (AMD advised us to use lower maximum capability power supplies to really see the power efficiency of Llanos. The A8 has GPU cores running at 600MHz and seems to have its GPU part based on the Radeon 5570. Since system memory has to be allocated for the GPU, the choice of memory becomes even more important.

CPU Specification Comparison
CPU Manufacturing Process Cores Transistor Count Die Size
AMD Llano 4C 32nm 4 1.45B 228mm2
AMD Thuban 6C 45nm 6 904M 346mm2
AMD Deneb 4C 45nm 4 758M 258mm2
Intel Gulftown 6C 32nm 6 1.17B 240mm2
Intel Nehalem/Bloomfield 4C 45nm 4 731M 263mm2
Intel Sandy Bridge 4C 32nm 4 995M 216mm2
Intel Lynnfield 4C 45nm 4 774M 296mm2
Intel Clarkdale 2C 32nm 2 384M 81mm2
Intel Sandy Bridge 2C (GT1) 32nm 2 504M 131mm2
Intel Sandy Bridge 2C (GT2) 32nm 2 624M 149mm2

As far as die sizes go, the 4 core Llano is almost as big as the 6 core Intel Gulftown, mainly due to the space taken up by the GPU

part of the chip. Llano is about the same size as Sandy Bridge.

The CPU 
 

The processor looks like a normal socket-3 CPU but this is not the case. To accommodate the GPU functionality AMD have had to add a few pins.



 
Socket FM-1 is 905 pins and reminds us of the days of Clawhammer. Fortunately, AMD have stuck to using the same heatsink mounting so anyone upgrading can keep their existing processor coolers.

To test the Llano CPU we used a Gigabyte A75M-UD2H motherboard launched recently.


The board is a square Mini-ATX design and we installed it into our Antec Minuet 350 HTPC case using a low profile Scythe Shuriken heatsink/fan. This is a fully featured board with dual channel memory (up to 1866MHz) and SATA3 as standard allong with 4 USB3.0 connectors (two on the back and headers for 2 case connectors). The single lug on each side of the retention mechanism is a step backward but may be due to space restrictions around the CPU. Two PCI-E slots are available for Crossfire and there is the possibility of using a Hybrid Crossfire with a single external card and the onchip graphics.

Turbo Core

Turbo core is not available on the high performing parts and we don't have a clear answer as to why..

Overclocking

Since this si the first time that 32nm technology is being used for AMD desktop processors we were keen to see how well the A8-3850 would overclock:



The upshot is that AMDs new processor is a delight for overclockers - particularly poignant given the difficulties Intel users have overclocking the new Sandy Bridge processors.

AMD's new manufacturing process should have overclockers rubbing their hands with glee especially given the price. AMD estimate that an entire system based around the A8-3850 including monitor would be between $600 and $700. We've done some online price checking and can put one together for under $500 so depending on the graphics performance (which we'll see later in this review) AMD may be offering a lot of power for not much money.


The Problem with Multi-Tasking


Since this review is primarily about multi-core efficiency it is worth explaining the inherent problems with multi-tasking. This may surprise some readers as we already have supercomputers made up of thousands of Intel or AMD processors and if they did not scale well then research institutions would not buy them to predict climate change, where minerals are buried and so on. The reason they work so well is that it is easy to split millions of operations among thousands of cores. Splitting one thread across multiple cores is actually quite difficult.

The problem involves concurrency, monitors and semaphores and is too involved to go into here although interested readers are encouraged to read the Wikipedia article on “Dining Philosophers” which explains the whole problem in easy to visualize terms. It can be found here.

Until Quantum Computing is viable we will have to rely on programmers making allowances for multiple cores and programming accordingly. Some games and applications are already optimized to a limited degree for multiple cores and theoretically every application will get a boost with a second core, even if just by offloading the usual Windows background processes to the other unused core.

It has been clear for some years that frequencies cannot continue to increase due to manufacturing limits and have remained roughly constant around the 3GHz mark for about 6 years. Instead it seems that the future gains will be attained by increasing the number of cores in a CPU, whether physical or also virtual (as with HyperThreading).

More importantly for AMD is how to make use of the 400 Radeon cores on the A8-3850. Thet have teamed up with software vendors such as Corel, Arcsoft and Cyberlink to speed up many applications in much the same wat that Intel did with their Sandy Bridge launch. In terms of Compute Power AMD claim the A8-3850 can deliver over twice the performance of an Intel i7-2600K - the most powerful Sandy Bridge processor on the market today.

Test Setup

Test Configuration

System Hardware

CPU

Intel Core i7-870, i7-2600K and i5-2500K
AMD A8-3850 with onchip Radeon HD 6550D
Motherboard
ASUS Maximus III Gene, P67, BG67
GIGABYTE A75M-UD2H
CPU Cooler
Corsair H50
Scythe Shuriken 2
RAM
Kingston KHX2133C8D3T1K2/4GX 4GB 2133MHz DDR3 Non-ECC
CL8 (Kit of 2) Intel XMP Tall HS CAS 8-8-8-24
Kingston KHX1600C8D3T1K2/4GX 4GB 1600MHz DDR3 T1 Series Non-ECC
CL8 DIMM (Kit of 2) XMP CAS 8-8-8-24
Graphics
ATI Radeon 5850 HD
ATI Radeon 6550D HD
Hard Drive
Maxtor 300GB SATA-2
Maxtor 300GB SATA-2
Sound
SupremeFX X-Fi built-in and Intel HD Audio
Realtek® 1200 8 -Channel High Definition Audio CODEC
Network
Gigabit LAN controller
Realtek® 8112 Gigabit LAN controller
Chassis
Antec 902 Midi Tower Case
Antec Minuet 350 HTPC Case
Power
Antec TruPower 750W
Antec 350W PSU

Software

Operating System
Windows 7 Professional
Windows 7 Professional
Graphics
ATI Catalyst 10.3
ATI 8.862 RC1
Chipset
Intel P55, P/BG67
AMD A75
Applications
  • SiSoft Sandra 2009
  • 3DMark Vantage Pro
  • PCMark Vantage Pro
  • Everest Ultimate
  • CPU-Z
  • Far Cry 2
  • HAWX
  • Resident Evil 5
  • SiSoft Sandra 2009
  • 3DMark Vantage Pro
  • PCMark Vantage Pro
  • Everest Ultimate
  • CPU-Z
  • Far Cry 2
  • HAWX
  • Resident Evil 5

All games are tested at the maximum available settings and initially at 1024x768 so we can be sure of hitting CPU limitations before bandwidth or fill rate ones related to the GPU. We selected Far Cry 2 (first person shooter), HAWX (air combat) and Resident Evil 5 (horror) for our tests as they are newish titles that are suited to benchmarking and make most systems struggle. We've also started benchmarking a number of DX11 titles and will include these when we have enough results.

Test Results - SiSoft Sandra
SiSoft Sandra 2011 is a comprehensive tool to diagnose and benchmark all areas of a PC system. Here we focus on some processor specific benchmarks.

 
The Llano beats the low end AMD processors as it was designed to do. This is a CPU only test so does not benefit from onchip graphics.
 
Another CPU test that shows the Llano as being almost as good as AMDs top quad core processors.
 
Memory performance tends to vary little little in benchmarks and the real-world impact can be unpredictable depending on the application. This may not be the case with onchip graphics that rely on system memory. Ultimately, the 2000MHz DDR3 of the Intel platform makes all the difference over the 1333MHz DDR3 the AMD systems have.

Test Results - Everest Ultimate Edition 
Everest (now superceded by AIDA64) is a very comprehensive benchmark suite that is set to take the synthetic crown from SiSoft Sandra. We limited our testing to the CPU and FPU benchmarks provided.
 
CPU Queen is a simple integer benchmark which focuses on the branch prediction capabilities and the misprediction penalties of the CPU. It finds the solutions for the classic "Queens problem" on a 10 by 10 sized chessboard. CPU Photoworx is an integer benchmark that performs different common tasks used during digital photo processing. CPU Zlib is an integer benchmark that measures combined CPU and memory subsystem performance through the public ZLib compression library. CPU ZLib test uses only the basic x86 instructions, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware. CPU AES is an integer benchmark that measures CPU performance using AES (a.k.a. Rijndael) data encryption. It utilizes Vincent Rijmen, Antoon Bosselaers and Paulo Barreto's public domain C code in ECB mode.
The A8-3850 does well to hold its own here. There is no way to match the performance of the Sandy Bridge processors especially with hardware encryption to boost the AES test.

The FPU Julia benchmark measures the single precision (also known as 32-bit) floating-point performance through the computation of several frames of the popular "Julia" fractal. The code behind this benchmark method is written in Assembly, and it is extremely optimized for every popular AMD and Intel processor core variants by utilizing the appropriate x87, 3DNow!, 3DNow!+ or SSE instruction set extension.

The FPU Mandel benchmark measures the double precision (also known as 64-bit) floating-point performance through the computation of several frames of the popular "Mandelbrot" fractal. The code behind this benchmark method is written in Assembly, and it is extremely optimized for every popular AMD and Intel processor core variants by utilizing the appropriate x87 or SSE2 instruction set extension.

The FPU SinJulia benchmark measures the extended precision (also known as 80-bit) floating-point performance through the computation of a single frame of a modified "Julia" fractal. The code behind this benchmark method is written in Assembly, and it is extremely optimized for every popular AMD and Intel processor core variants by utilizing trigonometric and exponential x87 instructions.

The Llano processor can beat the low end AMD CPUs but not any of the Intel ones.

Test Results - PC Mark Vantage Pro
PC Mark Vantage tests a whole range of activities from web browsing to photo manipulation and music conversion.

The gaming results are low because we are using the onchip Radeon HD 6550D on the Llano and a Radeon HD5850 on the others but apart from that the A8-3850 does reasonably well.

Test Results - 3D Mark Vantage Pro
Of much more interest to gamers is 3D Mark Vantage and this is the de facto standard for synthetic 3D graphics benchmarks for a wide variety of gaming types.
 
The CPU figure is not important. The Radeon HD 6550D on the A8-3850 outperforms the Radeon HD 6450 and, most importantly, the Intel HD3000 graphics by a huge margin of over 2x.

Test Results - Overall Gaming Performance
Now for a variety of different games at popular resolutions. It should be noted that our standardised testing is to use the highest possible settings and this may cause framerates to dip below playable levels on some entry level hardware. The solution is to lower the resolution and/or reduce quality settings. For example, in Resident Evil 5, just switching off the 8x anti-aliasing gives a 30% speed boost.
 
While soundly beating the HD3000 the Llano does need to make some sacrifices to achieve playable frame rates.
 
 
For some strange reason HAWX tends to favour the HD3000 graphics at lower resolutions. Just turning off the 8AA allows this game to go above 30fps at all resolutions. 
 
 
On a par with the Radeon HD 6450, the Llano once again beats the Intel HD3000 graphics of the i5-2500K.

Power Consumption and Temperatures
One of the key things on buyer's minds these days is power efficiency. As global fuel prices have risen sharply in recent years consumers are looking for greener PCs.
 
With idle power useage of only 47 watts, the Llano is exceptionally power efficient. Even at full speed it uses far less power than a system with a Radeon HD 6450.

Temperatures are even more surprising with idle readings of only 31 degree Celsius. 46 degrees under FurMark is very cool and may be down to the fact that onchip graphics can make use of good quality CPU coolers to benefit the GPU parts of the architecture.

Conclusion
It may not seem like it, but we are witnessing a fundamental transition on the desktop space by AMD. For those looking to build a system using only onboard graphics then Llano offers the only DX11 solution and the best DX9/10 performance. Asymmetric Crossfire is an interesting option and allows, for example, a Radeon HD 6450 to be added to give a combined performance equal to a Radeon HD 6670. This not only saves $40 on a more expensive GPU but can be done at a later stage for the cash strapped.

Priced cheaper than an Athlon X4 and discrete GPU combined, Llano is exceptionally good value for money and easily beats Intels HD3000 integrated graphics in almost every test.

For existing socket AM3 users it may be better to wait until Bulldozer comes along (unless you have a very old processor). For anyone building a new budget system today the choice can only be Llano and Desktop Fusion.

No comments:

Post a Comment