Testing Platform

For our Kaveri testing AMD sent us two APUs – the top 95W A10-7850K SKU and the configurable TDP version of the A8-7600 APU, the latter of which can be set at 45W or 65W through the BIOS. The A8-7600 was tested in both power configurations, ultimately the difference between them both being only a few hundred MHz. The 65W configuration is only 200 MHz off the A10-7700K base frequency, and incidentally they both turbo to the same frequency of 3.8GHz.

Kaveri will be the first APU put through the mangle in terms of my new 2014 benchmarking suite, focusing on more compute tasks, video conversion in different software, and more real world scenarios geared for the prosumer.

We must thank the following companies for their contribution to the test beds:

Our test setup for AMD is as follows:

AMD APU TestBed
  SKU Cores CPU /
Turbo
DRAM
MHz
Power IGP SPs GPU
MHz
Kaveri APUs A10-7850K 2M/4T 3.7 GHz
4.0 GHz
2133 95W R7 512 720 MHz
A8-7600 2M/4T 3.3 GHz
3.8 GHz
2133 65W R7 384 720 MHz
A8-7600 2M/4T 3.1 GHz
3.3 GHz
2133 45W R7 384 720 MHz
Richland APUs A10-6800K 2M/4T 4.1 GHz
4.4 GHz
2133 100W 8670D 384 844 MHz
A10-6700T 2M/4T 2.5 GHz
3.5 GHz
1866 45W 8650D 384 720 MHz
A8-6500T 2M/4T 2.1 GHz
3.1 GHz
1866 45W 8550D 256 720 MHz
Trinity APUs A10-5800K 2M/4T 3.8 GHz
4.2 GHz
2133 100W 7660D 384 800 MHz
A8-5500 2M/4T 3.2 GHz
3.7 GHz
1866 65W 7560D 256 760 MHz
Memory AMD Radeon 2 x 8 GB DDR3-2133 10-11-11 1.65V
G.Skill RipjawsX 4 x 4 GB DDR3-2133 9-11-11 1.65V
G.Skill RipjawsZ 4 x 4 GB DDR3-1866 8-9-9 1.65V
Motherboards ASRock FM2A88X Extreme6+
ASRock FM2A88X-ITX+
Power Supply OCZ 1250W ZX Series
Storage OCZ 256GB Vertex 3 SSDs
Operating System Windows 7 64-bit SP1 with Core Parking updates
Video Drivers Graphics Driver Build 13.300 RC2 for Radeon R7
Catalyst 13.12 for all others

Unfortunately we were not able to source a 65W Richland part in time, however a midrange 65W Trinity part was on hand. The important thing to note is that within each power bracket, both the CPU frequencies and the supported memory changes depending on the architecture and the binning process AMD uses. The benchmarks in this review are run at the processors' maximum supported frequency, rather than any AMD Memory Profiles which the processor can also support via overclocking. This has implications in conjunction with the IPC or MHz difference.

For this review we also took a few Intel processors of varying TDPs:

Intel TestBed
  SKU Cores CPU /
Turbo
DRAM
MHz
Power IGP SPs GPU
MHz
Sandy Bridge i5-2500K 4C/4T 3.3 GHz
3.7 GHz
1600 95W HD 3000 12 850
Ivy Bridge i3-3225 2C/4T 3.3 GHz 1600 55W HD 4000 16 550
i7-3770K 4C/8T 3.5 GHz
3.9 GHz
1600 77W HD 4000 16 550
Haswell i3-4330 2C/4T 3.5 GHz 1600 54W HD 4600 20  
i7-4770K 4C/8T 3.5 GHz
3.9 GHz
1600 84W HD 4600 20  
i7-4770R
+ Iris Pro
4C/8T 3.2 GHz
3.9 GHz
1600 65W HD 5200 40  
Memory ADATA XPG 2 x 8 GB DDR3L-1600 9-11-9 1.35V
Motherboards ASUS Z87 Gryphon
Power Supply OCZ 1250W ZX Series
Storage OCZ 256GB Vertex 3 SSDs
Operating System Windows 7 64-bit SP1 with Core Parking updates
Video Drivers 15.28.20.64.3347 for HD 3000
15.33.8.64.3345 for HD 4000+4600

Unfortunately our stock of i5 and i3 processors is actually rather limited – Intel prefers to source the i7s when we review those platforms, but I was able to use a personal i3-3225 from my NAS and we sourced the Haswell i3 as well. Given that Ganesh has the BRIX Pro in for review, I asked him to run as many benchmarks from our gaming suite as I could, to see how well Intel's Haswell eDRAM (Crystalwell) equipped processors stand up to Kaveri’s GCN mêlée.

For reference we also benchmarked the only mid-range GPU to hand - a HD 6750 while connected to the i7-4770K.

Overclocking and Underclocking the A10-7850K

As part of the final testing for this review we did some basic overclocking on the A10-7850K processor. Despite our processor being an engineering sample, we would assume that it is as close/identical to the retail silicon as you can get, given that this is meant to be a review on which people make purchasing decisions.

Our A10-7850K CPU starts out with a peak voltage under load of 1.24 volts when running OCCT. From this point we clocked back to 3.5 GHz and 1.100 volts, with a full-on CPU load line calibration and adjusted turbo mode to equal the base clock. Our standard overclocking test applies – OCCT for five minutes, PovRay, and new for 2014, a run of LuxMark. At our settings, we test the system for stability by running these tests. If the system fails, the CPU voltage is raised 0.025 volts until the system is stable during testing. When stable, the system multiplier is then raised and our testing moves on to the new MHz range.

Our results are as follows:

There was an unexpected jump in the voltage required to move from 3.5 GHz to 3.6 GHz (likely hitting the limits of what we can easily attain on this process). The system would not remain stable until 1.225 volts as set in the BIOS.

We also did the power tests, measuring the power draw at the wall as the delta between idle and OCCT load:

As expected, raising the voltage has a significant effect on the power consumption of the processor. One thing I should point out is that even at stock, the power delivery VRMs were getting very hot to touch – so much in fact that the system generated significant errors without an active fan on them. This got worse as the system was overclocked. I am not sure if this is an effect of the platform or the motherboard, but it will be something to inspect in our motherboard reviews going forward.

The Kaveri Socket and Chipset Line Up: Today and Q1, No Plans for FX or Server(?) CPU Performance
Comments Locked

380 Comments

View All Comments

  • geniekid - Tuesday, January 14, 2014 - link

    Would've been nice to see a discrete GPU thrown in the mix - especially with all that talk about Dual Graphics.
  • Ryan Smith - Tuesday, January 14, 2014 - link

    Dual graphics is not yet up and running (and it would require a different card than the 6750 Ian had on hand).
  • Nenad - Wednesday, January 15, 2014 - link

    I wonder if Dual Graphics can work with HSA, although I doubt due to cache coherence if nothing else.

    While on HSA, I must say that it looks very promising. I do not have experience with AMD specific GPU programming, or with OpenCL, but I do with CUDA (and some AMP) - and ability to avoid CPU/GPU copy would be great advantage in certain cases.

    Interesting thing is that AMD now have HW that support HSA, but does not yet have software tools (drivers, compilers...), while NVidia does not have HW, but does have software: in new CUDA, you can use unified memory, even if driver simulate copy for you (but that supposedly means when NVidia deliver HW, your unaltered app from last year will work and use advantage of HSA)

    Also, while HSA is great step ahead, I wonder if we will ever see one much more important thing if GPGPU is ever to became mainstream: PREEMPTIVE MULTITASKING. As it is now, still programer/app needs to spend time to figure out how to split work in small chunks for GPU, in order to not take too much time of GPU at once. It increase complexity of GPU code, and rely on good behavior of other GPU apps. Hopefully, next AMD 'unification' after HSA would be 'preemptive multitasking' ;p
  • tcube - Thursday, January 16, 2014 - link

    Preemtion, dynamic context switching is said to come with excavator core/ carizo apu. And they do have the toolset for hsa/hsail, just look it up on amd's site, bolt i think it's called it is a c library.

    Further more project sumatra will make java execute on the gpu. At first via a opencl wrapper then via hsa and in the end the jvm itself will do it for you via hsa. Oracle is prety commited to this.
  • kazriko - Thursday, January 30, 2014 - link

    I think where multiple GPU and Dual Graphics stuff will really shine is when we start getting more Mantle applications. With that, each GPU in the system can be controlled independently, and the developers could put GPGPU processes that work better with low latency to the CPU on the APU's built in GPU, and processes for graphics rendering that don't need as low of latency to the discrete graphics card.

    Preemptive would be interesting, but I'm not sure how game-changing it would be once you get into HSA's juggling of tasks back and forth between different processors. Right now, they do have multitasking they could do by having several queues going into the GPU, and you could have several tasks running from each queue across the different CUs on the chip. Not preemptive, but definitely multi-threaded.
  • MaRao - Thursday, January 16, 2014 - link

    Instead AMD should create new chipsets with dual AMU sockets. Two A8-7600 APUs can give tremendous CPU and GPU performance, yet maintaining 90-100W power usage.
  • PatHeist - Thursday, February 13, 2014 - link

    Making dual socket boards scale well is tremendously complex. You also need to increase things like the CPU cache by a lot. Not to mention that performance would tend to scale very badly with the additional CPU cores for things like gaming.
  • kzac - Monday, February 16, 2015 - link

    Having 2 or more APUs on a logic board would defeat the purpose of having an APU in the first place, which was to eliminate processing being handled by the logic board controller. With dual APU sockets, there would need to be some controller interjected to direct work to the APUs which could create a bottle neck in processing time (clock cycles). This is the very reason for the existence of multi core APUs and CPUs of today.

    Its my expectation that we will start to observe much more memory being added to the APU at some point, to increase throughput speeds. Essentially think of future APUs becoming a mini computer within, the only limitations currently to this issue are heat extraction and power consumption.
  • 5thaccount - Tuesday, January 21, 2014 - link

    I'm not so interested in dual graphics... I am really curious to see how it performs as a standard old-fashioned CPU. You could even bench it with an nVidia card. No one seems to be reviewing it as a processor. All reviews review it as an APU. Funny thing is, several people I work with use these, but they all have discrete graphics.
  • geniekid - Tuesday, January 14, 2014 - link

    Nvm. Too early!

Log in

Don't have an account? Sign up now