GPU vs. Memory Clock

Overclocking newbie here. Please be patient :o)

(I posted this on OCing forum too, but I see people are being referred from there to this forum ...)

Can someone knowledgeable help me understand the relationship between GPU clock and video card memory clock? I have an nVidia GeForce FX 5700 w/256Mb. In the driver properties I have the ability to overclock both the GPU and memory.

The default GPU clock is 425MHz
The default Memory clock is 501MHZ

How do I raise these clocks optimally?
Is it necessary to raise memory clock for optimal performance or raising GPU clock is enough? (Increasing GPU to anything over 454MHz with memory set to default begins to create occasional visual artifacts.)

12 answers Last reply
More about memory clock
  1. Raise any of them, you'll get some increased performance.
    so the fastest oc will be: OC both of them until their limit
  2. I am not sure that this is the best way to go.

    For example, if I leave GPU at default (425 MHz) and push memory from the default 501 MHz to, say, 550 MHz - I begin to get artifacts, BUT if GPU is at 448 I can push memory to only 515 before I begin to get artifacts.

    Also, if I first push GPU to 550 MHz I get no artifacts unless I push memory to at around 510 MHz.

    So, there is an obvious relationship between GPU and Mem clock for the purposes of performance and stability.

    At this point I have it set up to GPU 448 MHz/Mem 515 MHz

    but I wonder if GPU 440/Mem 540 or something like that would be faster overall while remaining stable as it is now.

    So, the question stands: what is the relationship between GPU and Memory clocks?
  3. THE GPU clock is the processor clock. It will determine how many polygons the video card can handle, and how fast the video card will process advanced effects like shaders.

    The memory clock determines the memory bandwidth available between the GPU and the system. This will have an effect on how fast the GPU can be fed new frames & textures, and how fast a video card can use implement antiailiassing, etc.

    Both are needed for good performance, although individual games will sometimes depend on one more than the other.

    <b>Radeon <font color=red>9500 PRO</b></font color=red> <i>(hardmodded 9500, o/c 322/322)</i>
    <b>AthlonXP <font color=red>2600+</b></font color=red> <i>(o/c 2400+ w/143Mhz fsb)</i>
    <b>3dMark03: <font color=red>4,055</b></font color=red>
  4. Cleeve, thanks. Your reply is clear and to the point.

    Now, I have another question if I may:

    Based on what you say (and it makes a lot of sense), the relationship works like this (just making sure I got it right):

    1. If GPU is too fast but memory bandwidth is insufficient, the result is that memory is the bottleneck since it does not matter how much the GPU can process if memory cannot feed the GPU fast enough.

    2. If the situation is reversed and memory is too fast while the GPU is slow it does not matter how much the memory is pushing through since the GPU is overwhelmed and and the memory has to wait,

    Question: all else equal, how should I overclock memory and GPU? The default clocks are 425 GPU and 501 memory. Common sense would suggest that for optimal OC boost both should be increased the same percent. But with hardware common sense is not always applicable.

    What is your advice/opinion?
  5. Honestly, I wouldn't worry about the relationship of GPU/Memory so much. The overclocker's goal is the max performance, so you're going to want to overclock both of them. Overclocking either will give you gains in almost every situation. SO:

    If you're overclocking this will extend the life of your whoile computer, and give you higher overclocks to boot. At minnimum you should have an intake fan (in the front bottom of the case) and an exhaust fan (in the rear of the case under the power supply) IN ADDITION to the fan everyone has in their power supply.

    2. Overclock your memory to the max (without producing any artifacts. A good test for artifacts is 3dMark 2003. If you see ANY little black dots randomly popping up on screen, or texture tearing and funny stuff, you've pushed it too far. Back up 5 Mhz and try again.)

    3. Overclock your GPU to the max (without producing any artifacts. See above)

    4. Pull them both back 10 mhz or so (whatever you feel comfortable with), to provide a little longer life for the card. (It's not good to run it at max all the time, it's asking for trouble really.)

    5. Play lots of games and enjoy your free performance

    <b>Radeon <font color=red>9500 PRO</b></font color=red> <i>(hardmodded 9500, o/c 322/322)</i>
    <b>AthlonXP <font color=red>2600+</b></font color=red> <i>(o/c 2400+ w/143Mhz fsb)</i>
    <b>3dMark03: <font color=red>4,055</b></font color=red>
  6. You see? I was right !! :tongue:
  7. Yes you were! I just did not quite understand why ...

    Okay, one more thing, please:


    Can overclocking AGP bus slow down the card?

    I have two options for overclocking the CPU: via BIOS by increasing the bus speed and via jumpers.

    Currently, the CPU is overclocked via BIOS with a 100MHz bus at 120MHz and stable. This, with 1.75V speeds up my P-4/ 2.4GHz to almost 2.9GHz but PCI and AGP bus are also overclocked.

    Using jumpers I can push system bus to any level I want while keeping PCI and AGP bus at their defaults. Should I do this or not? At this time the system is stable except I am a bit disappointed in the VGA performance.
  8. lmao, youre an [-peep-], you didnt have to rub it in his face....


    <A HREF="" target="_new">-={Apostalic Alcoholic.}=-</A>
  9. Overclocking the AGP bus will speed up the card a smidge if anything.

    Although you might find the card less stable, and your overclocks might not get as high without causing instability, so it's a tradeoff.

    <b>Radeon <font color=red>9500 PRO</b></font color=red> <i>(hardmodded 9500, o/c 322/322)</i>
    <b>AthlonXP <font color=red>2600+</b></font color=red> <i>(o/c 2400+ w/143Mhz fsb)</i>
    <b>3dMark03: <font color=red>4,055</b></font color=red>
  10. I am a new guy here so all of this info was helpful,
    I have a question. At my specs according to the overdrive software that ATI handsout my GPU clock is overclocked to about 801mHz and the memory clock is at 1131 mhz. First it sounds to me like my memory is not processing fast enough, so if that is the case how do I improve this apart from buying a new video card? Second is that my gpu speed sounds a little fast so maybe there is a little insight for someone who really knows what they are talking about here.
  11. well 801mhz for the gpu is pretty darn quick imo, but that memory is actually 2262 effective Double Data Rate (DDR), so its not actually slow at all. what video card do you actually have?
  12. plz USE something like riva tuner for the temps, i burned my 6600 GTs with a 10% OC cus i didnt notice my temps were 2 high b4 OC (thx god i was planning to get a new comp anyway)
Ask a new question

Read More

Graphics Cards Overclocking GPUs Memory Graphics