Fully Enabling the A10-7850K while Utilizing a Standalone GPU

Subject: Processors | July 9, 2014 - 05:42 PM |
Tagged: nvidia, msi, Luxmark, Lightning, hsa, GTX 580, GCN, APU, amd, A88X, A10-7850K

When I first read many of the initial AMD A10 7850K reviews, my primary question was how would the APU act if there was a different GPU installed on the system and did not utilize the CrossFire X functionality that AMD talked about.  Typically when a user installs a standalone graphics card on the AMD FM2/FM2+ platform, they disable the graphics portion of the APU.  They also have to uninstall the AMD Catalyst driver suite.  So this then leaves the APU as a CPU only, and all of that graphics silicon is left silent and dark.

View Full Size

Who in their right mind would pair a high end graphics card with the A10-7850K? This guy!

Does this need to be the case?  Absolutely not!  The GCN based graphics unit on the latest Kaveri APUs is pretty powerful when used in GPGPU/OpenCL applications.  The 4 cores/2 modules and 8 GCN cores can push out around 856 GFlops when fully utilized.  We also must consider that the APU is the first fully compliant HSA (Heterogeneous System Architecture) chip, and it handles memory accesses much more efficiently than standalone GPUs.  The shared memory space with the CPU gets rid of a lot of the workarounds typically needed for GPGPU type applications.  It makes sense that users would want to leverage the performance potential of a fully functioning APU while upgrading their overall graphics performance with a higher end standalone GPU.

To get this to work is very simple.  Assuming that the user has been using the APU as their primary graphics controller, they should update to the latest Catalyst drivers.  If the user is going to use an AMD card, then it would behoove them to totally uninstall the Catalyst driver and re-install only after the new card is installed.  After this is completed restart the machine, go into the UEFI, and change the primary video boot device to PEG (PCI-Express Graphics) from the integrated unit.  Save the setting and shut down the machine.  Insert the new video card and attach the monitor cable(s) to it.  Boot the machine and either re-install the Catalyst suite if an AMD card is used, or install the latest NVIDIA drivers if that is the graphics choice.

Windows 7 and Windows 8 allow users to install multiple graphics drivers from different vendors.  In my case I utilized a last generation GTX 580 (the MSI N580GTX Lightning) along with the AMD A10 7850K.  These products coexist happily together on the MSI A88X-G45 Gaming motherboard.  The monitor is attached to the NVIDIA card and all games are routed through that since it is the primary graphics adapter.  Performance seems unaffected with both drivers active.

View Full Size

I find it interesting that the GPU portion of the APU is named "Spectre".  Who owns those 3dfx trademarks anymore?

When I load up Luxmark I see three entries: the APU (CPU and GPU portions), the GPU portion of the APU, and then the GTX 580.  Luxmark defaults to the GPUs.  We see these GPUs listed as “Spectre”, which is the GCN portion of the APU, and the NVIDIA GTX 580.  Spectre supports OpenCL 1.2 while the GTX 580 is an OpenCL 1.1 compliant part.

With both GPUs active I can successfully run the Luxmark “Sala” test.  The two units perform better together than when they are run separately.  Adding in the CPU does increase the score, but not by very much (my guess here is that the APU is going to be very memory bandwidth bound in such a situation).  Below we can see the results of the different units separate and together.

View Full Size

These results make me hopeful about the potential of AMD’s latest APU.  It can run side by side with a standalone card, and applications can leverage the performance of this unit.  Now all we need is more HSA aware software.  More time and more testing is needed for setups such as this, and we need to see if HSA enabled software really does see a boost from using the GPU portion of the APU as compared to a pure CPU piece of software or code that will run on the standalone GPU.

Personally I find the idea of a heterogeneous solution such as this appealing.  The standalone graphics card handles the actual graphics portions, the CPU handles that code, and the HSA software can then fully utilize the graphics portion of the APU in a very efficient manner.  Unfortunately, we do not have hard numbers on the handful of HSA aware applications out there, especially when used in conjunction with standalone graphics.  We know in theory that this can work (and should work), but until developers get out there and really optimize their code for such a solution, we simply do not know if having an APU will really net the user big gains as compared to something like the i7 4770 or 4790 running pure x86 code.

View Full Size

In the meantime, at least we know that these products work together without issue.  The mixed mode OpenCL results make a nice case for improving overall performance in such a system.  I would imagine with more time and more effort from developers, we could see some really interesting implementations that will fully utilize a system such as this one.  Until then, happy experimenting!

Source: AMD
July 9, 2014 | 06:14 PM - Posted by AMDBumLover (not verified)

now if games could use this, scott michaud how feasible is it to send a render target to different hardware and drivers?

July 9, 2014 | 06:54 PM - Posted by Scott Michaud

Games can use this like any other software. That said, probably not for render targets. I'm thinking more for quick bursts in memory (AI visibility, pathfinding, audio, collision, etc.)

July 13, 2014 | 11:25 PM - Posted by DerekR (not verified)

Scott is pretty much right, unless the developer does some really fancy dancing.

To flesh out the situation a bit more, in DirectX you need to select a singular device, primary, secondary, or other to use as the device that is used to create the games DX device/device context. That device and device context pair is where the game's engine would be based on to run/render the game.

A developer could possibly create a second DX device and device context pair and utilize it for another purpose such as physics, AI, or any other task. If rendering an image, the rendered output would still have to be sent to the primary DX device instance to render to the screen. SLI/Crossfire does a good job this, and this is where Frame Rating comes in.

July 9, 2014 | 07:11 PM - Posted by JohnGR

Very interesting article.

July 9, 2014 | 07:36 PM - Posted by Anonymous (not verified)

This should be the job of the OS, the OS should be able to utilize all of the computational resources on the computer, and the GPU manufactures should be forced to supply the drivers that play nice together, otherwise the Makers of OSs should not certify the GPU hardware for use with the OS. All this integrated GPU off to be able work with a discrete GPU nonsense would stop fast, if GPU makers and CPU/SOC makers were forced to provide working drivers for their products, no working drivers to allow the OS to be able to use the GPU(Integrated and Discrete) hardware(All of the Time), then no certification for use with the OS, and the GPU hardware is not whitelisted to work with the OS, and this includes APUs/Intel/AMD/whoever's integrated GPUs also! Selling GPU products that intentionally, or unintentionally do not work (All of the Time) on computing platforms should be grounds for a lawsuit, or a class action lawsuit, and any product that has any GPU limitations with respect to Internal and discrete GPUs should be declared unfit for the intended purpose. OpenCL has be available for a while, and should be required by any up to date OS, there is no excuse going forward for GPGPU to not be available, as well a graphics workload sharing between GPUs of any make and model, and this "ether or" crap has gone on much too long.

July 9, 2014 | 08:12 PM - Posted by Scott Michaud

Eh, I disagree. This should the software developer's job. The OS vendor should just make it easy for developers (in collaboration with hardware vendors).

July 11, 2014 | 09:48 AM - Posted by Anonymous (not verified)

The OS should be responsible for controlling the hardware, and providing the abstraction layer/s, the software developer should not have to worry about the hardware, that should be the job of the driver/API/SDK programmers, and any OS that can not fulfill the job of using all available computing resources[ALL OF THE TIME] is a poorly constructed and dysfunctional OS. The GPU industry has been shortchanging the public with their poor drivers and product rushed to market, and a lot of this driver incompatibility is just ploy, to keep the users needing to buy the newest hardware in order to obtain more computing or gaming power, HSA when it finally arrives, and not just AMD's version of HSA, will put an end to the wasted hardware resources and allow all computing resources to be utilized, all of the time, for gaming or computing, and HSA aware OSs will be needed. The OS vendor should not certify any hardware that cannot be utilized, all of the time, for the hardware's intended purpose, and the GPU industry should be put on notice by the OS industry, to get the GPU driver resources to a point that all GPU hardware will be available all of the time, no more excuses. This includes the SOC makers, the integrated GPUs need to be available for use all of the time, if needed for demanding tasks that require all available resources, and only be switched off or switched to, only when the extra computing/graphics resources are not needed, this unavailable dew to lack of driver/lack of driver cooperation among the GPU industry's players has to end.

July 11, 2014 | 01:47 AM - Posted by Anonymous (not verified)

zzzzzzzzzzzzzzzzz

July 10, 2014 | 01:23 PM - Posted by Anonymous (not verified)

be sweet to do this in a laptop

July 11, 2014 | 05:52 PM - Posted by fade2blac

I would be careful about trying to do this in a thermally or power constrained environment such as a laptop. One could inadvertently cripple CPU core performance by heavily loading the IGP at the same time. I would expect it to be a bit of a balancing act where one ideally assigns work to the most efficient resources until they are saturated or some important performance metric starts to suffer.

A typical APU (or any processor with an IGP) has TDP limits for the whole chip. By design, the CPU and IGP portions are often capable of using more than half the total available power envelope when stressed independently. If you try and max them both out at the same time, then some sort of throttling is likely to occur to keep total power and temperatures in check. Today that usually means dynamically dropping out of turbo or even to lower than base/stock clocks.

July 10, 2014 | 01:40 PM - Posted by Geekavenger (not verified)

Does anyone know if this would work with an older APU? I have an A8-5600k, and also do you think this would work in Linux, I am running steamOS right now.

July 10, 2014 | 03:56 PM - Posted by Josh Walrath

Good question, I don't have an answer for you.  It should work, but sounds like you are in a good position to experiment!

July 12, 2014 | 06:19 AM - Posted by Mordorson (not verified)

no, it won't. The HSA architecture is only present on the new APUs with R7 graphics units, which is why it's such a big deal.

July 12, 2014 | 11:10 AM - Posted by ZoA (not verified)

I think this software used in this experiment is not HSA compatible, only OpenCL compatible, so it should work even on APU older then Kaveri

July 10, 2014 | 07:46 PM - Posted by eyeoncomputers

Lets make a list of software that can utilize AMD GPUs/iGPUs.

NOTE: List refers to latest versions for each application, so previous revisions of each software might have support as well.

List last update: 2014-07-10

List of OpenCL / AMD® APP (Accelerated Parallel Processing) / ATI-Stream applications:

Adobe Photoshop CS6 and up
Adobe Premier Pro
Adobe After Effects
Adobe SpeedGrade
Corel VideoStudio Pro
Corel PaintShop Pro
Corel AfterShot Pro
Corel WinDVD Pro
CyberLink MediaExpresso
CyberLink PowerDirector
CyberLink Director Suite
Wondershare Video Converter Ultimate
Wondershare Video Editor
Sony Vegas Pro
Final Cut Pro X
Motion 5
Compressor 4
Indigo Renderer
Octane Renderer
GIMP
LuxMark ..... (benchmark)
Bandicam ..... (game recording)
mirillis Action! .... (live game/screen capture/streaming)
Mathematica 8
VLC media player
Handbrake
x264 encoders
Nero Recode 2014
ArcSoft MediaConverter (AMD® VCE)
Xilisoft Video Converter's
ImTOO Video Converter's
Aiseesoft Video Converter's
iSkysoft Video Converter's
Brorsoft Video Converter's
Acrok Video Converter's
AnyMP4 Video Converter's
UFUSoft HD Video Converter's
4Media Video Converter's
ConverterXpress Video Converter's
Pavtube Video Converter's
Leawo Video Converter's (including the Free one)
A's Video Converter (Free)
MediaCoder (Free)

Please add to the list, it would be useful for everyone.

July 10, 2014 | 05:27 PM - Posted by hoxlund

yep using a r9 280x in my kaveri system. my specs in my signiture

Corsair Graphite 760T White Case
Corsair RM1000 Power Supply - Corsair custom braided white cable replacement set
MSI A88X-G45 Gaming - FM2+ A88X Mobo
AMD A10-7850K Kaveri APU @ 4.8GHz
Silverstone TD03 120mm Radiator CPU Cooler
G.Skill Trident X DDR2400 4x8GB - 32GB
Gigabyte R9 280X Windforce Rev. 2
AuzenTech X-Fi Forte 7.1 PCI-E Sound Card
Samsung 840 Pro 256GB SSD
2x3TB 2x2TB Mechanical Hard Drives - 10TB internal storage

July 10, 2014 | 08:33 PM - Posted by fade2blac

I've wondered about this sort of concept for a couple of years now. It's like adding an nVidia card for PhysX acceleration, but without being a proprietary one-trick pony that usually only adds superficial complexity while also artificially creating exclusivity. /rant I had the HD3000 enabled on my i7-2600K along with my Radeon 7950 but I never really found a good use case for the HD3000 resources other than maybe QuickSync and perhaps extra monitor outputs (think full-screen 3D on main display via discreet GPU while using integrated GPU to drive secondary display for web browser or video playback). Intel has since released OpenCL drivers.

It's great to see an article give a concise demonstration of how practical heterogeneous compute can be with a mixed build. I have always thought that having any sort of integrated/secondary GPU allows potential for GPGPU acceleration of all sorts of tasks using OpenCL, etc. There has to be real potential to benefit from accelerating things like AI, physics, audio effects, weather, or whatever else one can fit to the strengths of GPU compute and/or shared memory access via HSA.

Any idea if this hybrid build supports TrueAudio via the A10-7850K while using the GTX 580 to drive the display?

My understanding is that AMD has separate dedicated hardware (ASIC?) embedded in its APU/GPU chips for TrueAudio. I doubt this would work to allow FreeSync, but maybe that could be a possibility too if you somehow pipe the discreet video output through the display controller of the APU. I had actually proposed a similar question when PCPer hosted the Richard Huddy interview but there were many more questions than time would allow.

July 10, 2014 | 08:50 PM - Posted by Scott Michaud

Yeah, OpenCL on Intel HD graphics wasn't useful (as in not supported on the GPU at all) until HD 4000 (with Ivy Bridge).

July 10, 2014 | 08:56 PM - Posted by Josh Walrath

Perhaps this weekend I can test out the TrueAudio functionality.  I have the TA demos as well as Thief... just gotta see if it works!  That is a very good point to make... especially if it works!

July 15, 2014 | 10:47 AM - Posted by Josh Walrath

TrueAudio seems to work in this particular configuration.  That is some nice information to know when building a gaming machine.

July 15, 2014 | 12:29 PM - Posted by fade2blac

Thanks Josh for taking the time to confirm that. Sorry if I gave you extra homework over the weekend. You have my permission to consume an additional adult beverage during the next podcast. I will await more discussion of discharge buttons....good times!

AMD talked about the TrueAudio support being implemented by integrating specialized and dedicated hardware DSP's so it seemed reasonable to expect that TrueAudio could be used as long as one has the proper drivers are installed and the device is otherwise enabled. It may not be a big selling point, but at least it brings a little something extra to the table when paired with a discreet card.

July 10, 2014 | 08:50 PM - Posted by fkr

I run 2 7950's on an i5 2500k.

I use lucid logic to bypass the gpu's for most SW like vlc and web browsers. I was curious if enabling this lux SW to run on lucid would make any difference. It did not. I ran the test 9 times with and without lucid enabled and no effect.

I love the idea of leaving nothing on the table as far as performance goes. I have everything OC'd (gpu CPU Ram).

thanks for the article as it is a good read

July 11, 2014 | 11:16 AM - Posted by obababoy

This is interesting and cool. Id love to be able to use my HD4000 on my 4770k as a "physx" GPU and my R9290 for the big stuff.

As it stands now I keep the Intel GPU going and use it for my 50ft HDMI (to projector for movies), and second monitor (for monitoring). This works perfect and lets the R9 290 lower its memory speed when not gaming because with dual monitors on AMD your memory wont lower its frequency.

July 11, 2014 | 01:01 PM - Posted by Bernardblack (not verified)

Interesting, I was first running it this way. I was so perplexed on as to why when I disabled the IGP, the display on my card would go dark at Windows load. I couldn't get into Windows. After much fuss, I unplugged my card and went in with the IGP-only, uninstalled the drivers for the IGP, installed my card and then disabled IGP and I was okay and could use my card without the IGP having to be enabled. I initially did this to save on power. I figured, if the IGP was enabled with my card, there was potential wattage going to waste and unnecessary heat. I did not know you could use the APU with another card. The card I am running is the R9 280X.

July 11, 2014 | 05:01 PM - Posted by David E. (not verified)

Thank you, my buddy does not believe me that my A10 7850 GPU will still work with a stand alone GPU. He believes that the stand alone will take over everything and the GPU on the APU wont do anything. But I just sent this to him to prove him wrong. Cuz my CPU has brand new tech in it and his doesn't so we are going on a cpu war now but once I buy a graphics card for my A10 7850 I will blow him away even more :) I can play battlefield 4 on recommend high settings and I can play warthunder on movie settings but I can only play World of Tanks on low cuz they only use one core. That's why I want a graphics card for that sole purpose and maybe if I get into other games but I don't really need one. Its just to play every game in maxed out settings

July 14, 2014 | 05:29 PM - Posted by Anonymous (not verified)

Can I not use HSA and Opencl if I have a Dual graphics setup? In my case A10-7850 r7 250

July 15, 2014 | 10:46 AM - Posted by Josh Walrath

In a dual graphics setup, you still can use HSA and OpenCL.

July 15, 2014 | 12:04 AM - Posted by Anonymous (not verified)

The current APUs are kind of an intermediate step in the technology. A large number of laptops are being sold with only integrated graphics these days. This doesn't have to be slow. They could make an APU with a more GPU like memory interface, and hook some GDDR5 directly to the APU. Connect something like m.2 SSD devices directly to the APU and this would make a powerful gaming laptop since this is essentially what a PlayStation 4 is. They would need power management on the memory though to down-clock and lower voltage when the speed is not needed. The cpu performance is not really as important any more. Many games, even at 1080, are gpu bound showing very little difference between different cpus.

The form factor just has not caught up with the current situation. We still have about half of an ATX motherboard dedicated to the cpu with a relatively small space for each gpu even though the GPU may take up to 5x as much power.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.