1. OCAU Merchandise is available! Check out our 20th Anniversary Mugs, Classic Logo Shirts and much more! Discussion in this thread.
    Dismiss Notice

REVIEW AMD A10-5800K /Gigbayte GA-A85X-UP4 Platform

Discussion in 'AMD x86 CPUs and chipsets' started by mAJORD, Jan 20, 2013.

  1. mAJORD

    mAJORD Member

    Joined:
    Jun 4, 2002
    Messages:
    14,351
    Location:
    Griffin , Brisbane
    Introduction:

    Given the large number of Trinity reviews already on the web, rather than bombard everyone with CPU and game benchmarks, I decided to put some effort into exploring the capabilities of AMD's latest FM2 Platform. In particular extracting the most performance / Watt out of it, to take advantage of what I consider an advantage of the 'fusion' concept that's now well and truly a part of both the CPU Makers products. By Eliminating discrete components, one can save money, and thermal/power requirements

    So Thanks to AMD I have the opportunity to test the following Parts

    A10-5800K

    [​IMG]

    Sitting in....

    Gigabyte GA-F2A-A85X-UP4 ATX, FM2 Motherboard

    [​IMG]


    -----


    Platform first impressions.


    This review is mainly about the APU, but given the lets take a look at this motherboard in some detail.. Really I think t's worthy of its own review!

    It's been a little while since I've gotten my hands on a higher-end gigabyte board, and let me just say from the outset, if looks could kill, I wouldn't be here to write this by now. The above pic doesn't really do it justice, but its simplistic in design, yet fully populated, finished off in a nice Matte black solder mask.. The sight of it was enough to make want to start building a rig around it.

    Anyway, apart from looks it also packs a fairly decent 6+2 phase Driver-mosfet VRM, using International Rectifier IC's, as do all their P series boards AFAIK.

    Another standout feature of all the P series AFAIK are the on-board power, reset , RTC-reset buttons which are par the course for enthusiast boards these days, but nice to find at this price point. It's a lot a board for sub- $150 , and would have to be one of the cheapest of the "P" series.

    This board uses the latest A85X Chipset, which launched alongisde Trinity, and as the Table below show's gives one extra SATA 6GB/s Ports and supports a new RAID mode.

    With PCI-e functionality locked inside the APU, you're left with the standard 1 x 16 , or 2 x 8 Crossfire.

    [​IMG]

    Test Setup:

    CPU: AMD A10-5800K
    Motherboard: Gigabyte F2A85X-UP4
    RAM: Patriot IEM 2133
    Power Supply:
    Vantec 440W (OC)
    12V Regulated Supply + PICO PSU 120W (underclocking/ consumption tests)


    Stock (Out of box) Performance:



    I'm keeping the number of benchmarks to a minimum, as we don't need a huge set of data to look at the effects of Overclocking (or underclocking) on performance..

    To get performance baseline we have a couple of the usual suspects. Cinebench 11.5, and the old faithful CPUmark99.. useful for comparing single threaded, and also reflects memory subsystem performance.


    CPU:

    Whilst AMD doesn't have the best rep at the moment regarding high end CPU performance, in this class there's enough performance on tap to compete no problem with the main price competitor - Intel's Core i3


    [​IMG]

    GPU (IGP)


    Settings:

    APU : Stock Frequency
    Memory: DDR3 @ 1866 10-10-10-20

    [SEE RESULTS FURTHER DOWN REVIEW]

    I've selected only a few Games (Along with 3DM11) . 2 of which have built in benchmarks to give consistent / reliable results. Battlefield 3 is the Game I'm focusing on for performance impressions, as it still a benchmark title IMO. and quite demanding. utilizing the majority of a 4 core processor. and is also quite mean to GPU's.

    as the results further down the page show, all but Battlefield 3 are playable in performance. Certainly not butter smooth with those FPS, but BF3, is what I'd consider borderline playable at 1080P.

    There are times when FPS remain in the playable 30FPS region,.but any complexity on screen and it will dip down to the dreaded 25FPS region, which most of us would agree becomes pretty 'unplayable' . The casual gamer, and / or someone with no idea what FPS even means would play all day long though. As enthusiasts though, it's not really enough.

    One thing is a standout though, and was also for Llano. There is absolutely no noticeable system lagging , or inconsistent FPS 'drop's associated with the shared RAM systems of the past. Game menu's, alt tabbing, all without lag, or sluggishness whatsoever.


    ----



    Over Clocking




    1. The "No brainer" OC

    What is a No brainer Overclock? Well I consider it something you can achieve out of the box, without touching voltages, and without too much fiddling.

    CPU:

    Lets start with the CPU. To be honest. Without increasing cooling capacity, and/or fiddling with voltage there isn't much to be gained from a CPU OC.

    Unfortunately I had to disable Turbo functionality to get it to play nice. and the end result was 4.3Ghz. Given the stock turbo mode often pushes to 3.9-4Ghz anyway even under full load.. this only yields less than 10% performance improvement.


    [​IMG]

    GPU:

    The GPU though is a different story.

    I ended up just setting the GPU at 1000Mhz, not really expecting stability,but found it was completely stable.

    With a bit of extra pushing and testing I found 1050Mhz to be completely stable and settled on that. I didn't run into any obvious stability issues until 1070Mhz.


    RAM:

    Trinity Officially supports DDR3 1866, like it's predecessor. I had hopped this would make a bump to the DDR3-2133 ratio painless on the IEM RAM, and I wasn't disappointed. In fact, upon installing this RAM, I had XMP profile already on, and it booted straight up @ 2133 without any intervention.

    A full memtest run later it was considered good to go..


    Final Settings:

    CPU: 4.3Ghz
    GPU: 1060Mhz
    RAM: 2133Mhz

    [​IMG]



    High end OC - AIR

    Time to Strap on a Noctua U12 and get into it.

    CPU:

    Raising the Voltage to 1.5V brings us to the pointy end of Frequency scaling. At this voltage 4.5Ghz was stable. 4.6 - a bit iffy especially when OCing the GPU as well.. this all adds heat, and there's not too much point of a stable CPU clock only when it's loaded on its own! So in effect, if I was testing the CPU portion on it's own, perhaps with a discrete GFX, 4.6 Would be quite doable.

    [​IMG]


    GPU & NB


    Increasing the CPU-NB voltage allowed further scaling of the GPU Clock. It also seemed to aid further RAM overclocking, after it too had its voltage raised from 1.5 to 1.62.

    Final Settings:
    CPU: 4600Mhz
    NB clock of 2.3Ghz
    RAM clock 2400
    GPU: 1120Mhz


    [​IMG]

    As you can see, 2K 3DM was broken without too much fuss. This was a fun goal, and I'll probably load this OC profile up again soon and aim for 2100 :)

    The effect on Gaming FPS is also significant as the results below show.

    Performance Results

    CPU Benchmarks:

    [​IMG]


    Gaming Benchmarks

    [​IMG]


    All in all, Overclocking yielded a ~10% increase in CPU performance, and ~29% in Gaming. Obviously there's more thermal and frequency headroom on the GPU side. This will become clear in the next section.

    [​IMG]

    -----



    Analyzing BF3 performance




    Since this is the game with actual in game- testing, lets look at the performance in a bit more detail.

    Using FRAPS, I've recorded general Multiplayer conquest game play across several maps (at each clockspeed) and graphed FPS over time . Thankfully performance between maps is fairly consistent, and i think you'd be hard pressed to pick there's multiple maps joined in that data.

    [​IMG]

    The thing to take from this is as I mentioned, FPS never really dips to anything below around 25FPS, there's no hesitations or lags (which sould show up as massive dips in FPS).
    ----



    Taming the CPU power consumption - Under Volting / clocking!



    With the K series Quad core APU's all coming with a 100W TDP, power consumption is a potential

    Issue. The large majority of this TDP is aimed at the CPU, in order push the frequency high enough

    So, when it comes to increasing the overall Performance/watt it make sense to me at least to focus on bringing CPU consumption down.

    Looking at the 65W Models tells a lot about what should be achievable.


    Looking at the A8-5700 one can see with a 3.4Ghz base clock vs 3.8 for the 5800K and only 40Mhz difference in the max GPU clock, the GPU is not the limiting factor.


    So the goal of this exercise ? Sacrifice a small amount of CPU performance, and hopefully gain more GPU performance, all while lowering power consumption at the same time.

    Target Settings:

    CPU Frequency: 3.4Ghz
    RAM Frequecny: 2133Mhz
    GPU: 1000Mhz


    Voltages:

    CPU: 1.15 set
    CPU-NB: 1.17 set
    DDR3: 1.5v (default)


    Power consumption:

    The Following are all Platform power consumption. I.E Motherboard, CPU RAM (not HDD)

    To measure this I'm using a 120W PicoPSU , and measuring DC Current and voltage on the input. This gives us Accurate readings for the all 12V consumption, as it is passed through directly. The 3.3 and 5V rails run through regulators, so there are some losses here, but these ar e minor rails, and the efficiency of these regulators is quite high..

    In a nutshell, it's simply a far better method than measuring AC wall power.

    The only downside? I cannot measure Overvolted consumption, as it exceeds the Pico PSU's rating.


    [​IMG]


    Performance Results:


    [​IMG]




    Performance / Watt

    [​IMG]

    The results speak for themselves here.. A 53% increase in CPU performance / Watt.. and with the extra gaming performance over stock, a massive 73% higher during a game of BF3

    A total system power draw of just 71W during BF3 - which uses most of the CPU resources is

    quite impressive at this level of performance.

    ------



    Conclusions:




    Platform impressions:

    Firstly it has to be said, that this is a very competent platform. These days there's really no argument to be had between Intel vs AMD at a platform quality level.

    While Intel has traditionally set's the bar. AMD have refined their platform to an distinguishable level in stability. Whether SATA3 and USB performance is also now a match, I haven't been able to compare YET, but In the second part of this article I intend to do this. The indication is that the performance here has improved enough to compete, which is promising.

    It's also a lot of fun to play with. The unlocked K series gives you free reign with everything, and there's twice as much to play with now too.

    There are some issues with Overclocking and Turbo functionality. It would be nice to see these rectified, as I think having to sacrifice this functionality when moving from default settings isn't that great. After all It works on AM3 platforms (most of the time)

    The APU Itelf is a mixed bag. On the CPU front, whilst performance is adequate, and fully inline with its completion (if not faster - at least in Multi-threaded loads), it does really demand too much power at stock settings. Whilst i understand the reasoning behind the 100W TDP class, It's clear the main contributor to this is the CPU, and when it isn't offering significantly more performance than Intel it doesn't seem worth pushing to these frequencies. Sacrificing just a bit of performance, gives you a 65W TDP, in the form of the A8-5700.

    I haven't tested that APU, and its availabilty here is seemingly non existent. but obviously, it's CPU consumption would be a lot closer to competition, and it has the benefit of fitting into a 65W TDP out of the box.

    Nevertheless, the undervolting exercise showed the potential, and what you can achieve with just a little time in the BIOS if a lower TDP is what you're after.

    The GPU side of things more than makes up for any shortfalls in the CPU. It's common knowledge its performance is in a different class to both the equivalently priced, and highest offering from the competition. But regardless of this, for the most part, perhaps for the first time in the history of IGP's, is actually a plausible option for basic, low power gaming machine.. even at the popular 1080P resolution. It also achieves this with astonishingly low power consumption. Even out of the box, taking up a considerably smaller share of the APU's TDP.


    Finally. The motherboard has to get a special mention.

    If there's anything to complain about here. it has to be the fact I can't buy an 8 core (thread) FM2 CPU to put in it.. Overclocking was painless, everything functioned perfectly overclocked or not. I honestly can't think of anything worth a mention except perhaps the slightly confusing voltage

    There are 95W 8 core AM3 models launching at the moment that would go great guns in here even without an IGP.


    'Part II' of this will be a follow up looking at platform performance (USB, SATA) and also should have a Core i3 onboard, to look at some comparative performance. including GPGPU
     
  2. OP
    OP
    mAJORD

    mAJORD Member

    Joined:
    Jun 4, 2002
    Messages:
    14,351
    Location:
    Griffin , Brisbane
    RESERVED for Follow up
     
  3. dirtyd

    dirtyd Member

    Joined:
    Jan 4, 2006
    Messages:
    4,265
    Location:
    Melbs
    Nice one mate, I've got some different gaming results coming up and a dry ice run :)
     
  4. stmok

    stmok Member

    Joined:
    Jul 24, 2001
    Messages:
    8,882
    Location:
    Sydney
    Good article. I liked the stock voltage overclock and undervolting/clocking sections.
     
  5. bobbth

    bobbth Member

    Joined:
    Nov 1, 2010
    Messages:
    681
    Location:
    Hastings usually.
    Awesome review! Was holding off untill richland to upgrade my htpc but now I think I might go for a 5800k+mATX.
    Does anyone know much about FM2 mATX boards? There are heaps of motherboards but not many reviews.
     
  6. OP
    OP
    mAJORD

    mAJORD Member

    Joined:
    Jun 4, 2002
    Messages:
    14,351
    Location:
    Griffin , Brisbane
    Dirty.. Wouldn't mind a dry ice run either. I just hate doing anything sub zero in weather like this (too humid and hot!)

    I might push some more with some chilled water instead :)


    Bob: If the quality of this board filteres down to the cheaper / mATX GB boards, they might be worth a shot.

    At the least, find some model numbers and search for reviews on them specifically. :cool:
     
  7. TaroT

    TaroT Member

    Joined:
    Jan 18, 2002
    Messages:
    8,727
    Location:
    Hazelbrook nsw 2779
  8. bobbth

    bobbth Member

    Joined:
    Nov 1, 2010
    Messages:
    681
    Location:
    Hastings usually.
  9. kandrews

    kandrews Member

    Joined:
    Feb 9, 2006
    Messages:
    792
    Location:
    South Perth
    Yep the FM2 platform is indeed a very nice bit of kit for the money. Although I'm currently using an A85XA-G65 it still is the easiest to prep and run compared to any Intel setup.

    I think I might be falling in love with AMD all over again......:lol:

    (PS. mAJORD, what heatsink did you use, the standard AMD one or a custom job?)
     
    Last edited: Jan 23, 2013

Share This Page

Advertisement: