What do these overclocking numbers mean?

TimTim Southwest PA Icrontian
edited November 2003 in Hardware
I read the beginners article in this forum, and it didn't answer my question. I see people talking about how they set up their system, and it always has a set of numbers like 6/2/2/2 or 6/2/2/3/ or 5/4/3/2 or whatever. What are those numbers for?

Comments

  • edited November 2003
    They are cas numbers (the number of clocks it takes for the memory to do a specific task such as read or write to put it in simple terms) so lower numbers mean that the memory is a bit faster but sometimes you'll get errors from those lower numbers due to the fact that the memory needed longer to get all the information into que.
  • panzerkwpanzerkw New York City
    edited November 2003
    Those are the timings of the RAM, latency timings that is. Lower is better. If two RAM's are rated the same (PC2700 for example) the one with the lower latency numbers is (generally) better.
  • CreepCreep Hell Icrontian
    edited November 2003
    You read that beginners guide? I didn't think ANYONE read that! Did ya like it?
  • profdlpprofdlp The Holy City Of Westlake, Ohio
    edited November 2003
    Creep had this to say
    You read that beginners guide? I didn't think ANYONE read that! Did ya like it?
    I had read it before, and just read it again. It got "stickied" for a reason: It's a great resource!
  • CreepCreep Hell Icrontian
    edited November 2003
    Wow, maybe I should go back over it and fix the mistakes.... Someone want to write up an updated text for the newer CPU's out there? Maybe toss in the Mem timings too since I just go with the good rule of "Lower is better" and just set them low...
  • ketoketo Occupied. Or is it preoccupied? Icrontian
    edited November 2003
    I'll go into a little more detail, though everyone above is absolutely correct. Normally RAM timings are expressed in reverse order to what you posted, ie. 2-2-2-5, 3-3-4-8 etc.

    Using 2-3-4-5 as an example, they are found in your bios (sometimes under 'chipset configuration') and represent:

    2 = CAS latency, lower is better, 2 is usually the lowest option

    3 = RAS to CAS delay, lower is better, 2 is usually the lowest option

    4 = RAS precharge, lower is *usually* better, 2 is usually the lowest option

    5 = DRAM precharge delay, 4 is usually the lowest option but most memory modules won't run at 4. *generally* lower is better. Some AMD systems with nForce2 chipset run fastest at 2-2-2-11.

    Sometimes the numbers are in different order in the bios but this is the most common configuration.

    Again generalizing, 2-2-2-5 is usally the best timings for a fast system but it normally takes high quality RAM to run these timings. Unless you take charge of these timings, they are set automatically by your computer, which reads the programming module on your RAM and uses default timings, aka SPD. SPD timings for generic RAM usually look like 2.5-4-4-7, tho it varies wildly and that's yet another generalization.

    The biggest gain (you guessed it, I'm generalizing again) is usually found by reducing that first number - CAS latency. going from a '3' to a '2' can greatly increase the memory bandwith available to your system and speed things up measurably. Sometimes a voltage increase on generic or cheap RAM can enable tighter (better, lower) timings. I've never heard of anyone frying a memory DIMM by applying more voltage.

    This is far from definitive, I suggest doing some searches on 'RAM timings' and doing more reading if you'd like to learn more.
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited November 2003
    RAS to CAS delay has a huge impact on Intel rigs while it hasn't that very big effect on Amd rigs. Keto, since you are running both, when you have the time, could you do a few sandras with different RAS to CAS settings at let say 200 fsb on both rigs?
  • CreepCreep Hell Icrontian
    edited November 2003
    Hey Mac, could you write up an updated CPU/RAM OC guide for me? I've been out of the loop for so long I dont know what's going on.
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited November 2003
    There are many more peeps here that knows MUCH more than i do but sure, give me a few days and i´ll write something up.
  • ketoketo Occupied. Or is it preoccupied? Icrontian
    edited November 2003
    Mack (and other interested parties) (did he say parties?):celebrate

    P4 results first, Asus P4C800-E Deluxe motherboard, 875 (dual channel DDR) chipset:

    **EDIT all settings are UNBUFFERED (see the bottom post on this page for Sandra memory bandwidth benchmark unbuffered settings.)
  • ketoketo Occupied. Or is it preoccupied? Icrontian
    edited November 2003
    RAS to CAS has a significant effect in both my AMD and P4 systems
  • ketoketo Occupied. Or is it preoccupied? Icrontian
    edited November 2003
    Sorry, reserved the space above not realizing I couldn't attach an image to an 'edit'.
  • ketoketo Occupied. Or is it preoccupied? Icrontian
    edited November 2003
    One more quickie experiment on the AMD, changing timings from 2-2-2-5 to 2-2-2-11, same 200fsb. Your mileage may vary.

    UNBUFFERED (see the bottom post on this page for Sandra memory bandwidth benchmark unbuffered settings.)
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited November 2003
    Keto,

    Somethings VERY wrong with those amd numbers. Check the bios settings and make sure it runs in sync. Also have cpu int enabled.

    I get that bandwidth at 150 fsb on the same multi/fsb/cpu. It might not matter in this case, but just a heads up.

    MANY thanks for that comparison. That should be in the sticky part. Great stuff!

    Edit//** As usual i should learn to read before post. UN-buffered you used, buffered i mixed. My bad.
  • ketoketo Occupied. Or is it preoccupied? Icrontian
    edited November 2003
    No problem. So, being unbuffered, these numbers are uhhh reasonable? Yes, I am in sync (4/4) and with CPU interface enabled. Should I be using 6/6? does it make a difference? It was my understanding that sync is the same regardless which ratio, as long as both same.

    UNBUFFERED (see the bottom post on this page for Sandra memory bandwidth benchmark unbuffered settings.)

    Feel free to cop these for your upcoming article, full unqualified permission, text and or graphics as you see fit.
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited November 2003
    It SHOULD be the same numbers but you know how things are in this business :\
    Infact, i do get SOME difference between 4/4 and 6/6 and that is proof enough for me. Those scores look just fine i would say.
    APIC 1.4 enabled or disabled in bios?
  • ketoketo Occupied. Or is it preoccupied? Icrontian
    edited November 2003
    APIC 1.4 = yes. Haven't messed with it.

    Changed sync mode from 4/4 to 6/6, got statistically unimportant difference in results results, though just on 1 run each way which really makes the whole thing unscientific as hell :buck: :

    4/4 Sandra UNbuffered 2-2-2-11 = 1542/1642

    6/6 " " " = 1544/1646
  • ketoketo Occupied. Or is it preoccupied? Icrontian
    edited November 2003
    Here is an excellent article on PCStats that discusses memory timings vs. bandwidth. It is very specific to Intel systems on 865/875 dual channel chipsets and whether it's better to run high/sync fsb at loose timings or async at tighter timings, including benchmarks. It does also have a lot of useful data pertaining to this discussion.

    The author, Colin Sun, is one of my favourite Canucks - real friendly guy and good enthusiast/writer.:thumbsup:
  • croc_croc_ New
    edited November 2003
    Nice article keto, I was looking for something like that!
  • edited November 2003
    What exactly is the 5:4 and 1:1 and such. When it was set to 1:1 on my pc, my CPU could not be overclocked. When it's set to 5:4 it can. I'm deffinetly going to mess with the timings when I get back to school, thanks for explaining.
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited November 2003
    It means how fast the memory/cpu runs towards each other. If you have your fsb at 200 and run 1:1, it means that you run your cpu at the multi it has times 200. If you have it at 5:4, the cpu speed is the same, but the memory is running at 200/5X4=160.
    On your rig, my bet is that your memory outruns the cpu.

    Is 3 gigs the max it can do? What's the mutli on a 2.4? 15?
    If 3 gigs is the max, you have memory that is far better than the cpu. In your case, i would use 200 1:1 at the timings 5,2,2 cas 2 or 5,3,2 cas 2. Try to add some vcore and see if that cpu can do more.
  • edited November 2003
    The highest I had the CPU was 3.2. I still use the stock heatsink though. My biggest problem is the fact that at school my dorm is incredibly hot (91 degrees in my RA's room next door at one point). When I get back to school I'll check to see what the multi is. I couldn't overclock the CPU at all when it was set to 1:1. I really would rather not mess with the CPU voltage. The RAM voltage is set to 2.8.
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited November 2003
    If you are afraid of voltages, stay away from overclocking.

    Try to run 1:1 as a goal. With 3 gigs on the cpu, 200 at 1:1 is no problems.
  • edited November 2003
    2.4c=12x200 so 3.00=12x250 so, no his 3500 or ddr 433 won't swing it unless he raises the cas numbers and raises the ram voltages.
Sign In or Register to comment.