2013 AMD A-Series APU Die

AMD Richland A10-6800K APU Review

Add Your Comments

Back in October of 2012, we introduced all of you to the AMD Trinity A10-5800K APU. Fast forward to today and we have a 2013 AMD A-Series APU to show you, this time called the Richland APU. While based upon many of the same attributes Trinity brought us, there are also important improvements that we’ll cover today. AMD was kind enough to send along their top dog in the Richland product offering for us to review, the A10-6800K. Without question, the Trinity APUs brought forth an impressive integrated GPU and the ability to build a wide range of good performing, low cost systems. How much has the A10-6800K raised the bar from our last review of the A10-5800K? Let’s dive in and find out!

Die_Right_FINAL_trans - Copy

Specifications and Features

Here is the specifications table for the 2013 A-Series APUs, as provided by AMD. Also included are brief A85 FCH specifications, which should give you a good idea of other technologies you’ll have when building a system around this platform.

amd_richland (27)

Our more astute readers will pick up on the only two differences between this specifications table and the one for Trinity APUs. The first difference is official support for DDR3 2133 MHz memory on A10-6800K APUs (up from Trinity’s 1866 MHz). Secondly, the new Enhanced Turbo Core features a Temperature Smart function, Bottleneck Detect Algorithms, and more frequency/voltage levels for CPU and x86 cores.

As we take a look at the features associated with the 2013 A-Series APUs, more improvements over Trinity make themselves known. The integrated GPU has had an upgrade from Trinity’s HD 7000 series to the HD 8000 series for the new Richland APUs. Specifically, today’s review sample has the HD 8670D iGPU. Along with this iGPU update comes a faster GPU clock speed of 844 MHz, up from the A10-5800K speed of 800 MHz. We also get a boost in maximum turbo frequencies across all the 2013 A-Series APUs when compared to the Trinity line. In our case, the A10-6800K has a maximum turbo frequency of 4.4 GHz, up from the 4.2 GHz the A10-5800K offered.

As we mentioned before, the basic architecture hasn’t changed from when Trinity was released. We still have two Piledriver modules that each house two Piledriver x86 cores and one shared FPU. The same dual integer and shared single floating point core per module that Trinity used carries on with Richland. L1 and L2 Cache are handled with each integer core getting its own scheduler and L1 Cache, but the 4 MB of L2 Cache is split between the two modules, same as Trinity. The die is still built on a 32 nm SOI, which contains 1.3 B transistor chips and remains 246 mm² in physical size.

On the GPU side, we still have 384 VLIW4 Northern Island Radeon Cores, but the GPU series is now HD 8000 with a faster core clock than the Trinity APUs offered.

Turbo Charged x86 architecture featuring “Piledriver” cores

  • Supports up to 4 cores and support for the latest ISA instructions including FMA4/3, AVX, AES, XOP
  • Up to 2MB L2 cache per dual-core module (up to 4MB total)
  • Max Turbo Frequencies up to 4.4 GHz
  • Configurable via AMD OverDrive1

AMD Radeon™ HD 8000 Series GPU Cores

  • Featuring VLIW 4 architecture
  • Up to 384 shaders
  • Up to 844MHz
  • Up to 8xAA and 16AF support
  • Controllable via AMD OverDrive1
  • DirectX®11 Support

New DDR3-2133 support on A10 APUs

  • A10-6800K Only

Enhanced AMD Turbo Core

  • More Frequency/Voltage levels for CPU and x86 cores
  • Temperature Smart Turbo Core
  • New bottleneck detect algorithms
  • Controllable via AMD OverDrive1

UVD and VCE

  • Video Encode and Decode Hardware to offload CPU
  • AMD Picture Perfect support with HD Post Processing technologies

Support for latest display technologies

  • AMD Eyefinity2 technology for 3+1 monitor support
  • Display Port 1.2 support

FM2 Platform

  • AMD CrossFire support with AMD A85X motherboards
  • AMD Memory Profile support (auto select memory timings in select DIMMs)
  • AMD Dual Graphics3 support with AMD Radeon™ HD 6450, 6570, and 6670 graphics cards

All slides are from the Product literature provided by AMD.

Module Design

Piledriver 32 nm Compute Module

AMD's 2013 APU Platform

AMD’s 2013 APU Platform

Richland APU Architecture

Richland APU Architecture

Benefits of DDR3 2133 MHz Memory

Benefits of DDR3 2133 MHz Memory

amd_richland (14)

2013 A-Series APU Benefits

Available SKUs

Available SKUs

A_Series_White_Box A_Series_BE_Box

Photo Op

Here are a few pictures taken from various angles. I know it’s not the most exciting computer component to look at, but you do need a proper introduction to the AMD 2013 A-Series Richland APU – A10-6800K style!

Meet the A10-6800K APU!

Meet the A10-6800K APU!

Top Side Angle

Top Side Angle View

Another Top Side Angle

Another Top Side Angle View

Side View

Side View

Pin Structure Angle View

Pin Structure Angle View

Pin Structure Bottom View

Pin Structure Bottom View

The FM2 Platform

AMD has been known to begrudgingly change socket types over the years, thus making the upgrade path an enticing part of the AMD experience. With the release of Trinity APUs in 2012, a rare socket change was implemented by AMD; but you’ll be glad to know socket FM2 lives on with the release of the Richland APUs. Motherboard manufacturers are rolling out BIOS updates as we speak to fully support Richland APUs. Chances are by the time you read this, there is already a BIOS available for any FM2 motherboard you may own.

The FM2 platform comes in three flavors – A55 (entry level), A75 (media and gamers), and A85X (high performance users).

FM2 FCH Options

FM2 FCH Options

In order to ease the financial burden often associated with upgrading a system, you can take advantage of the Dual Graphics feature to pair the iGPU with an inexpensive discrete graphics card for substantially better graphics performance. If you need a faster all around APU and would like to upgrade from Trinity to Richland, then you’ll find the entire line of Richland APUs very affordable as well.

Dual Graphics Recommended Pairings

Dual Graphics Recommended Pairings

Richland - Inexpensive to Upgrade From Trinity

Richland – Inexpensive to Upgrade From Trinity

Not only are these APUs affordable, AMD used the recently released PCMark 8 to illustrate the performance per dollar ratio when compared to more expensive Intel processors. These tests reflect typical home and office use.

amd_richland (55)
amd_richland (56)

The HD 8000 iGPU

AMD has been at the forefront of integrated graphics for a few years now, and their APUs have consistently outperformed anything from the competition on the iGPU performance front. The slides below show AMD’s in-house testing numbers to illustrate the progression since the Llano APUs and how image quality has been enhanced along the way.

iGPU - It Just Keeps Getting Better!

iGPU – It Just Keeps Getting Better!

Image Quality Enhancements

Image Quality Enhancements

Looks Better to Me!

Looks Better to Me!

Just because the HD 8000 graphics are integrated into the APU doesn’t mean you can’t benefit from some of the technologies found on discrete graphics cards. Case in point being AMD’s Eyefinity technology, which according to the slide below can work rather well on less demanding game titles. You’re probably not going to want to try this with such game titles like Metro 2033, but just having Eyefinity available is a good thing!

AMD Eyefinity

AMD Eyefinity

AMD performed some in-house testing to compare their iGPU gaming performance against that of Intel’s iGPU. It’s no great surprise AMD rules the roost here, we noticed similar results back when we tested the Trinity APU. AMD also has some in-house test results based on compute applications that take advantage of the OpenCL framework. Again, no surprise the AMD APUs far outperform the competition because of their superior iGPU.

AMD iGPU Gaming Compared to Intel iGPU

AMD iGPU Gaming Compared to Intel iGPU

Compute Application OpenCL Performance

Compute Application OpenCL Performance Vs. Intel

Adobe Premiere OpenCL Performance

Adobe Premiere OpenCL Performance

There are a couple more Richland APU features AMD would like you to know about, the first being wireless display support. As wireless technology becomes more mainstream across many different display devices, AMD APUs allow you to connect to such devices via the built in Miracast support. You’ll need a motherboard with an Atheros or Broadcom wireless adapter and a compatible monitor/TV to use this feature, but the option is there if you have the required hardware.

Miracast Support

Miracast Support

Partnering with Splashtop, AMD APUs can take advantage of personal cloud gaming technologies. This will require installing the AMD accelerated version of the Splashtop software on both the APU machine and the client device.

Splashtop

Splashtop

Richland Vs. Trinity Benchmarks

This section of the review will compare the Richland A10-6800K against the Trinity A10-5800K, with both set at their stock configuration. This should give us a good idea of how much performance levels have changed between the two APUs. Because the Richland is clocked higher and officially supports faster memory speeds, it should have an advantage in our tests. The comparisons between Trinity and Intel processors were well documented last year as it pertains to the CPU side of things. Because the Richland APUs are not a platform or technology change from Trinity, there isn’t much sense in beating that dead horse any further. However, when we get to the iGPU testing, we’ll include results from the Intel Ivy Bridge Graphics HD 4000 and the newly released Haswell Graphics HD 4600 iGPUs. We’ll see if AMD can continue to hold it’s dominance in this area. So, let’s dive into the benchmarks and find out if the performance gains are worth an upgrade, or choosing Richland over Trinity for your next AMD system build.

Test System

AMD Richland A10-6800K APU
ASRock FM2A85X Extreme6 Motherboard (Overclockers Approved)
G.Skill 2X8 DDR3-2400 MHz Gb F3-2400C10D-16GTX
Swiftech H220 LCS AIO CPU Water Cooler (Overclockers Approved)
Corsair HX1050 Power Supply
Toshiba HDS721050DLE630 500 GB SATA 6 Gb/s Hard Drive
Windows 7 x64 (Fresh Install)

Here is what CPU-z and GPU-z report with everything left at stock and the memory at the officially supported 2133 MHz.

CPU-z and GPU-z Readings

CPU-z and GPU-z Readings

I always like to run a stability test before diving into benchmark runs, so a quick 20 minute pass of AIDA64’s system stability test was in order. No problems encountered here!

System Stability Test

System Stability Test

CPU Side Benchmarks

AIDA64 was used to test CPU and FPU performance using a total of nine different tests. I pulled the Trinity A10-5800K scores directly from the list of comparisons built into AIDA64. The CPU tests show anywhere between a 2% to a 7% performance gain. These gains are consistent with the increased processor speed advantage the Richland has over the Trinity. The FPU tests showed similar percentage gains, between 2% and as high as 10%.

AIDA64 CPU Tests

AIDA64 CPU Tests

AIDA64 FPU Tests

AIDA64 FPU Tests

Next up, we have SuperPi and wPrime. I ran the shorter tests first, which included SuperPI 1M and wPrime 32M. In these tests, we recorded minimal gains of 1.5% and 3.5%. The longer SuperPi 32M and wPrime 1024M tests showed gains of roughly 2.5% to 4.5%. Again, these gains are in line with the increased processor and memory speed.

SuperPi 1M & wPrime 32M Results

SuperPi 1M & wPrime 32M Results

SuperPi 32M & wPrime 1024M Results

SuperPi 32M & wPrime 1024M Results

Moving on to some real world testing, we’ll start with Cinebench R10 and R11.5. Both of these test runs showed right at a 4.5% increase over the Trinity scores. Nothing earth shattering here, but an increase is an increase!

Cinebench R10 Results

Cinebench R10 Results

Cinebench R11.5 Results

Cinebench R11.5 Results

PoV Ray 3.7 and x264 benchmarks held true to the previous test results. Here we had the A10-6800K Richland showing anywhere from a 3.6% to 6.5% increase over the A10-5800K Trinity.

PoV Ray & x264 Benchmark Results

PoV Ray & x264 Benchmark Results

Some of you may be wondering how the A10-6800K and A10-5800K compare if clock by clock testing is performed. Well, I’ll tell you…. after running several of the above benchmarks to test this, they are about as identical as they can be! As I mentioned earlier, the base platform and technology hasn’t changed, so it should come as no surprise a clock for clock comparison yields no distinguishable differences.

iGPU Side Gaming Benchmarks

As promised, we’ll not only compare the Richland APU against the Trinity, but we’re going to throw the Intel Graphics HD 4000 and HD 4600 iGPU results in the mix. Let’s see if the A10-6800K can continue AMD’s stranglehold on iGPU performance in our gaming benchmarks.

If you’re not familiar with our gaming benchmark procedure, I invite you to visit our video card testing procedure page. Here is the down and dirty version of the procedure.

  • All Synthetic benchmarks were at their default settings
  • Unigine Heaven (HWbot) was run using the “extreme” setting
  • Aliens vs. Predator – 1920×1080 with highest settings offered (4x AA, textures set to highest)
  • Battlefield 3 – 1920×1080 at Ultra settings (4xAA/HBAO by default)
  • Dirt 3  – 1920×1080 with 8x MSAA and all settings enabled and at Ultra where possible
  • Metro 2033 – 1920×1080, DX11, Very High, 4x MSAA/ 16x AF, PhysX OFF, DOF enabled, Scene: Frontline
  • Civilization V – 1920×1080, 8x MSAA, VSync OFF, High Detail Strategic View: Enabled, Other Settings: High, using full render frames value ( / 60)
  • Batman: Arkham City – 1920×1080, VSync off,  8xMSAA, MVSS and HBAO, Tessellation set to high, Extreme Detail Level, PhysX Off

Because we are cross platform testing between AMD and Intel, we’ll adhere to the testing procedure for the Intel processors, which basically means the CPU set at 4.0 GHz and the memory set to 1866 MHz. For the AMD CPUs, we’ll leave those at stock speed with the memory set to the officially supported values of 1866 MHz and 2133 MHz respectively. Most of our benchmarks are GPU intensive, so the minor differences between the Intel and AMD settings probably won’t amount to a hill of beans anyway.

Beginning with our synthetic benchmarks, there are some nice gains to be had in 3DMark03. The Richland APU outperformed everything in the chart quite substantially and continues to beat up on anything Intel has to offer. The 3DMark11 results again show the the Richland APU the clear winner. I see a trend developing here.

3DMark03 Results

3DMark03 Results

3DMMark11 Results

3DMMark11 Results

3DMark Vantage shows a little resurgence on behalf of the Intel Graphics HD 4600 as it manages to top the Trinity APU, but still falls a few hundred points behind the Richland APU. Our HWBot Heaven results again show the Intel Graphics HD 4600 putting up a good fight by beating out the Trinity and falling just a tad behind the Richland APU. The end result of our synthetic benchmarks have the Richland A10-6800K with a clean sweep, and the new Haswell Graphics HD 4600 beginning to gain some ground on the AMD iGPUs.

3DMark Vantage Results

3DMark Vantage Results

HWBot Heaven Results

HWBot Heaven Results

Moving over to our game benchmarks, the Aliens vs. Predator benchmark shows the Richland APU darn near doubling the Intel Graphics HD 4600, Ouch! In Batman: Arkham City, we see pretty much the same domination by the AMD APUs – by a substantial margin.

AvP Results

AvP Results

Batman: Arkham City Results

Batman: Arkham City Results

Moving over to Battlefield 3 and Civilization V, we see the pecking order remain the same. Even though it appears Intel has come up a notch from their Graphics HD 4000 iGPU, so far they are still quite a ways from matching the gaming results we’re seeing so far.

Battlefield 3 Results

Battlefield 3 Results

Civilization V Results

Civilization V Results

Our final pair of game benchmarks includes Dirt3 and the GPU stomping Metro 2033. The Dirt3 benchmark had less than 1 FPS difference between all the competing samples, which is no surprise as it’s probably the less demanding of all the games in our test suite. Metro 2033 has no business running at max settings on any iGPU or even a lot of discrete graphics cards! Even so, the AMD APUs won this battle too.

Dirt3 Results

Dirt3 Results

Metro 2033 Results

Metro 2033 Results

None of our testing reveals what we would call playable frame rates. However, if you’re willing to lower a game’s graphic settings, you should be able to play most of these games easily.

Overclocking/Pushing the Limits

Typically the “Pushing the Limits” section involves ramping a CPU or GPU speed as far as it will go, stability be damned. The Richland APUs on the other hand are brought to market as a CPU and GPU all-in-one solution, which requires finding three areas that need to work in unison. Those three areas being CPU, iGPU, and memory. So, that will be what we will do here. Once we find the fastest CPU speed that will run the memory at its supported 2133 MHz, we’ll add the iGPU overclocking on top of that and see what we come up with by running a few benchmarks.

The maximum stable overclock I could achieve with the memory set to 2133 MHz was 4.9 GHz. Any attempt at 5.0 GHz was fruitless, regardless of the voltage I threw at it. Even at this, it’s quite an improvement over the Trinity A10-5800K. The best stable overclock I ever achieved on Trinity was 4.6 GHz, but that was with the memory at 1866 Mhz.

20 Minute Stress Test @ 4.9 GHz

20 Minute Stress Test @ 4.9 GHz

Just because I was curious (and I’m sure you are too), I wanted to find out  happens if we drop the memory speed down to 1866 MHz? Well, with a bump in CPU voltage to 1.450 V, we managed to get a little more out of it and hit the 5.0 GHz mark! At this point, our thermal capacity was reaching its limit, so we stopped here. Still a nice result and certainly nothing to complain about. If you’re willing to run memory a little below the officially support speed, you’ve got a 5.0 GHz chip on your hands! We won’t be using these settings for our benchmark runs below because slower memory speed can result in iGPU bottlenecks, which is probably why AMD upped the supported memory speed on this APU. Still, it was worth finding out how far raw CPU MHz would go on our setup.

As a side note: If you remember back to the Trinity review we brought you last year, we complained about the lack of any meaningful temperature monitoring. It seems an attempt was made to fix that problem, but it’ll be hit and miss as to what utility can read the temperatures correctly. Luckily, the latest version of AIDA64 seems to be up to the task as you can tell by the picture above and the picture below.

20 Minute Stress Test @ 5.0 GHz

20 Minute Stress Test @ 5.0 GHz

Ok, back to our overclocking trifecta, since we now know the 4.9 GHz is the maximum CPU speed that will still run the memory at the supported 2133 MHz, let’s drop back to that overclock and start upping the iGPU. The reason for this method is simple really. Just like I mentioned above, anytime you use an iGPU, you’ll be sharing the system memory with it; and faster memory speed will reduce bottlenecks.

AMD tells us they have seen the HD 8670D iGPU be able to run at over 1000 Mhz, and they’re right! We were able to get a very impressive 1083 MHz iGPU overclock as witnessed by the HWBot Heaven screen shot below. That’s a pretty impressive number considering the stock iGPU clock is 844 MHz.

HWBot Stable 1083 MHz iGPU

HWBot Stable 1083 MHz iGPU

Let’s run a few more benchmarks now that we’ve pushed the combined limit of the CPU, iGPU, and memory. First up was a run of wPrime 32M and SuperPI 1M. Some very nice gains were recorded when compared to the previous stock results above.

wPrime 32M Overclocked

wPrime 32M Overclocked

SuperPI 1M Overclocked

SuperPI 1M Overclocked

3DMark Vantage and 3dMark11 put up some pretty good numbers too, not bad at all for an iGPU, eh?

3DMark Vantage Overclocked

3DMark Vantage Overclocked

3DMark11 Overclocked

3DMark11 Overclocked

Lastly, a pretty impressive 3DMark Firestrike score for an iGPU.

3DMark Firestrike Overclocked

3DMark Firestrike Overclocked

All and all, a pretty darn impressive overclocking experience on both the CPU and iGPU side of things. I like the fact that higher overclocks are now obtainable with memory set to 2133 MHz, which wasn’t always the case with Trinity APUs. So, we can safely say there are definitely some nice performance gains when comparing the Richland A10-6800K to the Trinity A10-5800K.

Conclusion

I gave the pricing structure away in one of the pictures above, but it appears the Richland APUs will start around $69.00 at the lower end and top out at around $149.99 for the A10-6800K. When we compare that to the cost of the Trinity APUs, I think you will agree the price increase is modest at best. With this modest increase in price, you get a higher CPU clock and faster memory speed support. Couple that with a faster iGPU and you can see the value is definitely there. Overclocking? Yea, we got that too. Drop the memory speed a little and a 5.0 GHz CPU speed is pretty easy to obtain – who’s not liking that? As you go along your overclocking journey, you’ll be able to reliably monitor the temperatures, which is one of my favorite improvements over the Trinity APU. It’ll be hit and miss as to what utilities will correctly monitor the temperatures, but at least there are some options this time around.

It really is pretty astounding at how inexpensive a mainstream PC can be built around the Richland APUs with the end result being a pretty nice performing system. Whether you’re a gamer on a tight budget or simply looking for a more balanced approach where your processor is concerned (think combined CPU and GPU performance), you definitely want to consider the Richland APUs for your next system build.

AMD’s catch phrase of “the sum is worth more than the individual parts” was used to describe the Trinity APUs, and that slogan is again used to describe the Richland APUs. And the good news is…. that Sum just got better!

Overclockers_clear_approvedClick the stamp for an explanation of what this means.

-Dino DeCesari (Lvcoyote)

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Discussion
  1. majorleague
    Here is a youtube link showing 3dmark11 and windows index rating for the 4770k 3.5ghz Haswell. Not overclocked.

    This is around 10-30fps slower than the 6800k in most games. And almost twice the price!!

    Link to yoube video:

    http://youtu.be/k7Yo2A__1Xw


    Haswell isn't in the same product class as AMD's apu's are. Haswells top end chips are not only out of the price range, but they are also target at much higher integer performance. They also have l3 cache, something that is dropped from APU's in the interest of reduced manufacturing cost.

    As for these new apu's

    The real question is what is the new norm for performance gonna be. With the 5800k's with a good aftermarket cooler you routinely could push to around 1200mhz iGPU. While 2400mhz on memory was out of the questions for chips that where not of a high quality bin level. 2300mhz in most situations was not. 2800-3000mhz was not unheard of running single stick under a cold chip either.

    Some might of complained about the core temp montioring on earlier apu's I had pretty much just learned to deal with it, since it wasn't to different compared to dealing with some of the more temperamental phenoms out there. 4600-4800 was considering to be core norm for a high quality 5800k with after market cooling before. So if that holds true with the new revision hopefully we see 4900-5100 on air. Then cold performance of 5500-5800 on phase change. Speculation of course.

    The biggest OC problem still with these chips will still bethe a85x chipset. Thats still gonna cause major limitations. Due to its unstable PLL clockspeed. Till that gets fixed its gonna continue to be somewhat of a pain in the butt when really pushing these chips.

    Gonna try to pick one up soon, since my inbox is already blowing up with "have you played with these yet".
    Nice review Dino,

    The only thing I think would be good to see is what settings for the games would provide a playable framerate (if it was possible).
    @Robert17 - Depending on what demands you put on a HTPC I think it would be a good choice. For media playback it would work great, but if you do heavy gaming at high resolutions a discrete video card might be in order.

    @ATMINSIDE - The iGPU shares the system memory, so naturally you would benefit from higher memory speeds as far as that side of things go. The iGPU overclocked pretty well in it's own right too.

    @xsuperbgx - IMC limitation most likely.
    Lvcoyote
    Hi xsuperbgx,

    The more you overclock the CPU side, the lower you will have to set the memory. The motherboard I used for the review had a BIOS selectable memory speed of up to 2400 MHz, and it runs fine at that speed with the CPU at stock speed, although the CPU overclocking suffers when set that high.


    Do you feel that this limitation is due to temperatures or some kind of power limiting function?
    So, on Haswell, core clocks are primary over memory clocks.

    If you're using the iGPU on Richland though, would you benefit more from core clock or memory speed?

    Edit: No higher hybrid crossfire card? *Insert sadface here*
    This is a pretty nice chip when all is said and done. Thanks for your review, well done. :cool:

    I've been pondering an HTPC re-build around something like this. Your opinion?
    Hi xsuperbgx,

    The more you overclock the CPU side, the lower you will have to set the memory. The motherboard I used for the review had a BIOS selectable memory speed of up to 2400 MHz, and it runs fine at that speed with the CPU at stock speed, although the CPU overclocking suffers when set that high.
    Thanks for the great review! I have been thinking about picking one of these apu rigs just to play around with. Did you try to push your memory to see how fast you could get it?