DDR3-2400 on A10-6700

Discussion in 'Computer Hardware, Devices and Accessories' started by PityOnU, Jul 17, 2013.

  1. PityOnU
    OP

    PityOnU GBAtemp Advanced Fan

    Member
    907
    327
    Jul 5, 2012
    United States
    The A10-6700 is not unlocked, so can it make use of DDR3-2400 even though it only supports DDR3-1866?
     


  2. trumpet-205

    trumpet-205 Embrace the darkness within

    Member
    4,363
    542
    Jan 14, 2009
    United States
    Locked or unlocked refers to the multiplier, one variable in calculating CPU clock speed (base clock * multiplier = clock speed).

    It has nothing to do with memory speed. As long as BIOS supports it, you will be fine.
     
  3. PityOnU
    OP

    PityOnU GBAtemp Advanced Fan

    Member
    907
    327
    Jul 5, 2012
    United States
    Is not the memory speed calculated as a multiple of the base clock as well? With the memory controller being integrated into the APU which has all its multipliers locked I'm not really seeing how you could set it faster...

    Please explain in more detail.
     
  4. Fishaman P

    Fishaman P Speedrunner

    Member
    3,241
    507
    Jan 2, 2010
    United States
    Wisconsin
    The multiplier and the base clock are two different things. The base clock is never locked, but it can be difficult to ramp it up more than 5MHz on modern systems.
    The multiplier only affects the CPU: it's specifically there so you can over/under-clock the CPU independently from the rest of the system.
    The base clock, however, affects the entire system. The CPU and RAM use it, and I believe things like USB and PCIe do as well.

    In order to support high memory speeds, the CPU's IMC must be able to handle it.
    Since I'm not an expert in AMD CPUs (because they suck), I would conservatively estimate they'd handle up to 2000MHz stably.

    Also, once you hit 1600MHz, the diminishing returns... diminish. We're talking a ~1.5% performance boost at most going from 1600MHz to 2400MHz.
     
  5. PityOnU
    OP

    PityOnU GBAtemp Advanced Fan

    Member
    907
    327
    Jul 5, 2012
    United States
    AMD can handle (by default) much higher memory clocks than Intel. Mostly because of their emphasis on iGPU, which benefits greatly from greater bandwidth.

    In purely CPU related situations. The A10-6700 is an APU.
     
  6. trumpet-205

    trumpet-205 Embrace the darkness within

    Member
    4,363
    542
    Jan 14, 2009
    United States
    No. Multiplier has nothing to do with memory speed. Memory speed is regulated by integrated memory controller, which does not depend on multiplier. You set the memory speed in BIOS, and you also adjust voltage and timing until you find one that is stable by doing stress test.

    Multiplier only affects CPU clock speed. You do NOT want to mess with base clock, since everything is tied to it (CPU, IMC, SATA, PCIe, USB, etc). Not only is it hard to increase base clock, increase too much and you can face all sorts of instability (data corruption, BSoD, etc).
     
  7. Fishaman P

    Fishaman P Speedrunner

    Member
    3,241
    507
    Jan 2, 2010
    United States
    Wisconsin
    Oh... a standalone AMD proc in a system...
    If you have the money for ridiculous 2400MHz RAM, I would much rather put it to use buying an Intel system.
    I could only ever justify getting an AMD APU if it was a dedicated video encoding rig.
     
  8. PityOnU
    OP

    PityOnU GBAtemp Advanced Fan

    Member
    907
    327
    Jul 5, 2012
    United States
    This will be an HTPC in this case.

    Why are processors rated as having a maximum memory speed then if you can just set the speed as fast as the board allows?
     
  9. Fishaman P

    Fishaman P Speedrunner

    Member
    3,241
    507
    Jan 2, 2010
    United States
    Wisconsin
    An AMD proc? In an HTPC? While that's fine on paper, you haven't taken into account just how much power AMD procs consume, and by extension, how much heat they put out.
    Seriously, IIRC, AMD is challenging Intel's 73W procs with 125W procs. Wut.

    On the subject, I would highly recommend a ULV Intel i3, or maybe even an Atom. I would go with the i3 because it should be able to cope with the upcoming HEVC (H.265) codec, but if you won't be using this after a few years, an Atom should serve you well.

    Processors list a rated maximum memory speed because the manufacturer guarantees that speed of memory to work with the IMC. IMCs do have limits, you know; just because your mobo and RAM can zoom to 4000MHz, that doesn't mean the IMC can handle it.
     
  10. PityOnU
    OP

    PityOnU GBAtemp Advanced Fan

    Member
    907
    327
    Jul 5, 2012
    United States
    It's a gaming HTPC.

    Because why? Because I said so. Deal with it.
     
    Thanatos Telos likes this.
  11. Fishaman P

    Fishaman P Speedrunner

    Member
    3,241
    507
    Jan 2, 2010
    United States
    Wisconsin
    Another win for the i3.

    EDIT: I should also mention the AMD A10-6700 has a 65W TDP. Your case is rated for 65W at maximum, and the fan would be running pretty loudly. That's pushing it.
     
  12. trumpet-205

    trumpet-205 Embrace the darkness within

    Member
    4,363
    542
    Jan 14, 2009
    United States
    That is the speed AMD guarantees it'll work assuming the memory itself is stable. You can use higher speed than rated one.

    Now why in the world would you want to do video encoding on an APU? Core i3 will be a lot faster than APU or FX-43xx regards to video encoding. APU is very slow on video encoding, even if you use OpenCL to facilitate it. Ironically OpenCL only accelerates video decoding, encoding portion is still done by CPU.
     
    Thanatos Telos likes this.
  13. PityOnU
    OP

    PityOnU GBAtemp Advanced Fan

    Member
    907
    327
    Jul 5, 2012
    United States
    HD4000 GFX are poop.

    Story of my life, man. Story of my life.
     
    Thanatos Telos likes this.
  14. Fishaman P

    Fishaman P Speedrunner

    Member
    3,241
    507
    Jan 2, 2010
    United States
    Wisconsin
    Actually, HandBrake can utilize OpenCL for cropping and other video filters, and x264 itself can do the intra prediction with OpenCL, which is actually a substantial speedup on APUs.

    Anyway, I wasn't referring to the built-in GPU, I was saying that while AMD's "many weak cores" approach doesn't normally work, it's decently cost-effective for encoding.
     
  15. trumpet-205

    trumpet-205 Embrace the darkness within

    Member
    4,363
    542
    Jan 14, 2009
    United States
    Right, still not for the encoding portion. Just decoding and pre-processing tasks. Other encoders (such as paid software) do use OpenCL for encoding, but its coding efficiency is inferior. You need higher bitrate to compensate video quality.

    If you want speed, you really should go for Core i3. Especially now since Handbrake now has Quick Sync support. APU is not a cost effective solution, since the extra time it takes to do video encoding could've been used else where.
     
  16. Thanatos Telos

    Thanatos Telos random stuff

    Member
    842
    187
    Sep 13, 2009
    United States
    Yeah, this forum is very Intel biased (Especially FishamanP), even though the 8350 has matched an i5 in various games. (Really, check Tomshardware's newer benchmarks.) (Often having a difference of 2 FPS in higher games) Personally, I see where you're coming from. That APU would be great for watching movies and the A10 6800K (Which is the same APU, just unlocked IIRC), trumped every i7 using its iGPU until Haswell, where it still challenges it pretty well and still beats it out ~75% of the time. (Note: I'm talking about all of their iGPUs, not the i7 paired with a 7970 or something.) Also, I like your style, PityOnU.
    For people requesting the benchmarks, here: http://www.tomshardware.com/reviews/core-i7-4770k-haswell-review,3521-4.html

    And that's with the weaker 5800K.
    For the FX-8350, give me a sec.

    EDIT: Boom and boom. NOTE: 8350 trades blows in the first because of the frame latency, and read on how they say it's smoother in the second one.
     
    Fishaman P likes this.
  17. Fishaman P

    Fishaman P Speedrunner

    Member
    3,241
    507
    Jan 2, 2010
    United States
    Wisconsin
    Damn right! :lol:

    EDIT:
    Since when is intra prediction not part of encoding?
    It's a very good portion of it, especially when you encode using Placebo like I do.
     
  18. PityOnU
    OP

    PityOnU GBAtemp Advanced Fan

    Member
    907
    327
    Jul 5, 2012
    United States
    To interject, OpenCL and GPU oomph really don't play a very large role in encoding at all. At least last time I checked.

    Which sucks because you think it's be something that would be easily parallelizable. Apparently not.

    Back on topic, I won't be doing any encoding because I just rip Blu-Rays and don't down sample. Purist master race all the way.
     
    Thanatos Telos likes this.
  19. Fishaman P

    Fishaman P Speedrunner

    Member
    3,241
    507
    Jan 2, 2010
    United States
    Wisconsin
    Let's say the OpenCL speedup is horribly slow, something like a 2% improvement.
    That's still shaving an hour off my encode.
    I thought I remembered the speedup being 10% at the very least when OpenCL is done on the iGPU.
     
  20. trumpet-205

    trumpet-205 Embrace the darkness within

    Member
    4,363
    542
    Jan 14, 2009
    United States
    Sorry, a bit not clear. Intra prediction has yet to be included in x264, and x264 developers have said that speedup will be less than 5% (consistent with online benchmark). OpenCL and GPGPU are limited in usefulness when it comes to video encoding because GPU somewhat operates like RISC CPU, simple operations at fast speed. That is why a lot of advanced features used by desktop CPU (CISC CPU) in video encoding couldn't be ported to GPU.

    And seriously, if you are taking several hours to encode a video (so that 2% improvement shaves of a couple hours), you seriously need to upgrade your machine. You cannot just look at it from money perspective when it comes to being cost effective. Time is a form of cost as well. Extra time it takes to encode a video could've been used else where. It could've been used on next video encoding, could've been used on gaming, etc. Also, there is electricity cost too. Taking too long to encode a video using a slow processor consumes more electricity.

    I moved from i3 to i5 then i7. Each step yielded improvements to video encoding. To me it is still slow (I do 10-bit x264 encoding 24/7), and I'm eager to see how Haswell-E will perform.

    PityOnU
    Do you still have any question on RAM speed?