1. Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

CPU and VGA question

Discussion in 'General Hardware' started by macbeth, Nov 11, 2006.

  1. macbeth New Member

    May 25, 2006
    128 (0.03/day)
    Thanks Received:
    1) What do Cache L1/L2/L3 size really effect? Will there be a lot of performance difference if Cache L1/L2/L3 is larger? If so how much performance difference will there be? What application will benefit from a larger Cache L1/L2/L3? And what application will not be effected by it? Comparing between AMD X2 3800+ and AMD X2 3600+.

    2) How much performance difference will there be between 1950 pro 256MB and 1950pro 512MB?
    10 Year Member at TPU
  2. cjoyce1980 New Member

    Apr 8, 2006
    714 (0.18/day)
    Thanks Received:
    L1 & L3 cache doesn't really effect performance, its the L2 cache you should be looking at.
    Intel's newer processors use 2MB and 4MB and AMD use 512KB and 1MB. looking at that you would guess the intel processors would be better than the AMD's. This is wrong, Intel and AMD have different memory management systems.

    if you are comparing the AMD X2 3800+ (L2 512KB) and AMD X2 3600+ (L2 256KB) performance different isn't that much. The AMD X2 3600+ is a AMD X2 3800+ just with half the cache, and from what i have been reading the AMD X2 3600+ is the better overclocker reaching performance levels near the AMD X2 5000+.

    See here http://www.xbitlabs.com/articles/cpu/print/athlon64-x2-3600.html

    The performance differences between 1950 pro 256MB and 1950pro 512MB using DX9 gaming will be minimum but as DX10 (or DX9L) 512 memory will become more paramount, but to future proof your self the 1950pro 512MB would be the better option of the 2.
    10 Year Member at TPU
  3. Aevum

    Oct 26, 2005
    642 (0.15/day)
    Thanks Received:
    Madrid, Spain
    L1 cache is on die cache, basicly integrated in to the inner workings of the chip,
    L2 is also on die but not integrated in to the iner workings, its next to the die but still on chip,
    L3 is cache external to the chip,

    L1 is important, but usualy a small amount is put in, 128KB at max, and devided between units on the chip,
    L2 is also important, general use cache for all on die components, for general use, 256k-512k is enough, you might want more if you work in photo/video/audio editing or large databases, and some games might run smoother with 1-2MB of l2 cache, but the differance is usualy misrable,
    L3 cache is basicly a joke and isnt really used alot, its almost the same use as L2 cache but since its off die/chip, the latency is greatly incresed, unless you use software that benefit from large ammounts of cache memory, L3 is a waste,

    another issue is the balance between cache size and cache latency, as always, the size and density of a memory unit will affect the latency, a block of memory is basicly a storage grid of 1 and 0, the bigger the grid the more time is needed to find and access an individual position. so under some conditions a processor with less cache will function faster then a processor with more cache (another reason to get a 2MB conroe insted of a 4MB one),

    so make sure you dont just let the cache blind you and take the other aspects of the processor in to account, best way ? look at what you do most, and then start looking at benchmarks from different websites, and depending on what you do, check how the processor does in the benchmarks relevant to your plans for the processor,

    as for your 2nd question, graphics card memory is usualy used to stora texture and rendering data, which means, that it depends on your monitor and the resolution you play at, also the game you play,

    at low resolutons, 640x480, 800x600, 1024x768, alot of cards with less memory are actually faster then cards with more memory, why ?, same reason as the cache latency issue, more memory, takes longer to fish stuff out of it, especialy when your not using most if it,

    but at higher resolutions, you get bigger textures, more detail, bigger maps and renders, which take up more video memory, so if the textures and the rendering is bigger then your graphics memory, it has to be dumped to main memory or even swap memory and then the rest has to be rendered again in the video memory, meaning you have more processor load becuase your short on video memory, so once again, its a balancing act between the memory you need depending on the games you play and at what resoluton,

    if you play at 1280x1024 or higher, go for 512mb,
    10 Year Member at TPU

Currently Active Users Viewing This Thread: 1 (0 members and 1 guest)