High Vdimm derate performance?

MJOMJO Denmark New
edited December 2004 in Hardware
I stumbled upon this old review.

http://www.lostcircuits.com/motherboard/abit_nf7-s12/2.shtml

I qoute:
Whichever voltage will give the best results in terms of stability and performance will depend on the memory modules used, as a rule of thumb, one can say with higher voltage, the actual performance of the DIMM in terms of output will derate. In other words, at lower voltage, most DIMMs will be faster and that will cause less retries and increase the benchmark scores. Keep in mind also that all but some outdated DIMMs are using components that are running internally at 1.8V, anything beyond this will only be converted into heat. Along these lines, we found the best stability with the Mushkin PC3500 at 200 MHz 2:2:2 (running 2 x 512 MB in dual channel mode) at 2.6V. 2.7V or higher caused system instabilities.

The Corsair XMS 3200 CL-2, on the other hand would not run stable below a 2.7V setting. Among other things, the design of the PCB and the resistor network account for some of these differences as well. Bottom line is, though, that it is hard to justify anything beyond a real voltage of 2.75 V for both VDD and VDDQ. Contrary to popular belief, there is neither any gain in operating frequency nor in performance at higher voltages beyond 2.8V unless some really old components are used.

Is that true?
Memory performance drops at high Vdimm?

Comments

  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited January 2004
    Brand new ram, PERHAPS but run those at 230 instead. That's just bs. Ram needs burn-in most of the time. I have never ran any ram below 3 volts, no matter the speed i choose.
  • MJOMJO Denmark New
    edited January 2004
    I was quite puzzled as well.
    My Corsair modules are running at 225 Mhz 2.0-2-2-11 at 2.8V.
    I tried lowering Vdimm to 2.7V, it wouldn't make it into windows.
    I had it at 2.9V before. I am trying to save a little power here. ;)
    I need to test stability at 2.8V though.
  • Straight_ManStraight_Man Geeky, in my own way Naples, FL Icrontian
    edited January 2004
    I do not run RAM at higher than 2.8 unless the system case is real cool inside. Where that is so, you can run higher voltages and the heat will be "sluffed off" (more fomally, sloughed off, air pulls heat off of the heat-adsorbing carbon\ceramic outer module coverings). BUT, in warmer climates, if case ambient is higher, the heat will not sluff off at a high enough rate compared to heat buildup as the air has the heat already in it (molecules excited).

    High end is temp related, in this case the "sluff" is air that is cooler with unexcited mocelules and the air structure is able to "absorb" more heat which leads one to think that RAM can run high voltage. In fact, if the case is at 24-28 C already you might not be able to run RAM at high voltages without heat retention and corrallary cascades of heat damage (progressive damage) in RAM. To those of you who have cases at lower temps inside case, try 2.9 to 3.0 volts, being aware you ARE pulling more low power leg power in watts and your PSU has only so much of a "pool" of wattage available for that leg adn your mobo only has so many CAPS with so much capactiy each. Good PSU, cool case, do the higher voltage thing. Not so good PSU, vary warm case, do NOT do the higher voltage thing.

    Part is diff between case ambients typical in Norway versus those in Florida, Mackanz, I have to stop at 2.7 volts as the RAM is up to working temp due to retained heat at lower voltages. If I go to 2.8 boxes here get more unstable. In your case, and I suspect MJOs, temp per se and solely will not prevent this, but in MJO's case this might be a PSU limiting the total low leg pool of stable wattage flow available or a limit in cap "storage pool" capabiliites on mobo, or both.

    In MY case, this is temp pure and simple-- I live literally in Northern part of Tropic zone of this earth. Room temp is 19.9 C now and this is unusually cool for here, and room is typically about 85 F for low median when in summer with AC on. Temps of 90 F to 110 F outside are common in summer. In summer, I have to roll back my OC, so I tend to go for more normal stock rates for everything, compared to those of you in MI and Norway and Denmark-- and need to without spending huge amounts of money to simply keep things cool. Climate demands this. Summer and Fall electric bill is typically $110.00 USD per MONTH simply to keep my room to about 85 F. THERMOSTAT is set at 79-80 F.

    John.
  • csimoncsimon Acadiana Icrontian
    edited January 2004
    John I would suffocate here in south LA at those room temps ...I try keep my house 70F. Probably due to the poor insulation technique in which I'm building a new house and will remedy that real quick.
    I run my mushkin at 2.8v for 11-2-2-2 timings and I know that of I set it to 2.7v it will fail ...simple as that ...but what have I got to lose but giving it a shot?
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited January 2004
    Room temp here is 21c, system/case temp is generally between 24c-26c. Lasse did some tempreadings via diode on the chips, inside of the spreaders of a HyperX @ 3.5 volts i think it was. He then measured another stick without spreaders at same voltage without a difference. Temp of the chips wasn't even 45c. DDR can take a huge amount of voltage without adding heat enough so that you notice it.
    In here, we need to spend the money to keep the room warm. Same amount of money.

    John, you are just plain wrong on all parts in that post as far as memory is concerned.

    Simon, you can add as much voltage as you want. You have nothing to worry about.
  • CryptoCrypto W.Sussex UK Member
    edited January 2004
    Well I've got a strange experience to throw in here.

    When trying to maximise my fsb overclock, the post speech reporter on my Asus A7N8X always gives me " System failed memory test" if I bump up the memory voltage above stock, when getting near the fsb limit.

    I am now running 190x11.5 at stock voltages but could only manage 185x11.5 with VDimm bumped up a couple of notches :scratch:

    Cheers

    Crypto :)
  • nomannoman pakistan
    edited December 2004
    hi..

    i am also facing this problem....my system is more unstable at 2.8v then at 2.6v(stock)....prime95 gives an error earlier at 2.8 than at 2.6v..... i kept all other things normal....no oc of processor..only fsb with lower multi on the cpu...
  • JimboraeJimborae Newbury, Berks, UK New
    edited December 2004
    Mackanz wrote:
    .....

    Simon, you can add as much voltage as you want. You have nothing to worry about.


    Mack, I'll have to disagree with you here but I'm talking at the extreme side of thing where as you may not have been. :)

    Over the last 4-5 days, using memtest, i've been trying to find the best combination of volts, timings & speed for my TwinMos BH-5 using a ram booster. Performance & speed started to drop off rapidly after 3.7volts to the point where at 4.2v so many errors occured that it wouldn't boot at any given speed. So it proved to me that you can pump to much juice through your ram at. However I'd point out that no damage was done to the ram even at this extreme voltage. I also found out that the optimum voltage for my BH-5 is 3.7volts @ 250fsb, 1T, 2,2,2,5
Sign In or Register to comment.