Thoughts on 2x512 Hyper-X PC3700 or PC3500?

edited December 2003 in Hardware
I was just thinking that if I were to sell my 2x256 of PC3500 I could get the 2x512 for roughly $100 more. I was checking out the Egg and they have 2x512 of Hyper-X PC3700 for $225. Then they have 2x512 of PC3500 for $239. The only difference I can tell is the CAS Latency. I'm not knowledgeable when it comes to timings and what not so any ideas about this would be nice.

For the PC3500 kit the CAS Latency is 2.
For the PC3700 kit the CAS Latency is 3-4-4-8-1T.

What's the difference? Would it make more sense for me to just get a couple more sticks of 256? If I wouldn't be able to sell the 256 then I'd obviously just get a couple of more. Any thoughts would be appreciated. Thanks-DaK
«1

Comments

  • ketoketo Occupied. Or is it preoccupied? Icrontian
    edited December 2003
    You *should* be able to continue with dual channel mode if you populate your 2 open DIMMs with 2 more sticks of PC3500. Best bet is to do that and run the 5:4 divider to get maximum overclock on cpu, tho your memory will probably be then running UNDER spec. This way, you can keep tightest timings which ARE important, I run my P4 rig at 253fsb 5:4 divider = 202 memory speed 2-2-2-5 timings and it's MUCH faster than 230 fsb 1:1 at looser 2.5-4-4-8 timings, even tho I can get Turbo at 1:1 230 and NOT at 202 5:4.

    That PC3700 will allow you higher 1:1 clockspeeds AND gross bandwidth, but likely not as fast overall.
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited December 2003
    Your current pair...can they do 6,2,2 cas 2 at 200 fsb?

    Between those sticks you mentioned, the 3500 would be Winbond CH5 chips and the rated timings on those would probably be 6,3,2 cas 2. What settings do you use now? Fsb and divider?
  • edited December 2003
    I followed a mini guide from another post, I've had this set for awhile. Not sure what each means so you guys can tell me.

    CPU:DRAM 3:2
    OC=250 (I'm assuming FSB)
    CAS Latency-2
    ACT to Precharge-5
    DRAM RAS # to CAS Delay-2
    DRAM RAS # Precharge-2

    I haven't tried any of those turbo settings or anything. When I changed the timings I did a Sandra benchmark and got the pic below. I honestly don't know if this is good or not. Any input would be great.
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited December 2003
    Just change to 5:4 and use turbo settings with the same timings. I think you have Winbond BH5 chips on those sticks. Best there is.

    Change to 5:4 without the turbo first though. Use the same timings. Then watch your memory bandwidth in Sandra. You'll be happy. 3:2 is 166 fsb on the memory at 250. 5:4 is 200 fsb at 250.

    Those sticks can very well handle that without problems.
  • ketoketo Occupied. Or is it preoccupied? Icrontian
    edited December 2003
    What mack said 100%. The only thing to add is that if you buy newer PC3500, it's going to *VERY* likely be CH5 not BH5 chips. They don't *quite* get as tight, 2-3-2-6 instead of 2-2-2-5 but that's a very small difference in real world speed.

    **EDIT** your mileage may vary but as noted above, I cannot use Turbo on my AsusP4C800-E Deluxe with any memory dividers. 1:1 works fine. Different board but same 875 chipset.
  • edited December 2003
    Here's with just 5:4. Do I want higher numbers or lower numbers? I'll try turbo now, it'll go up to F1 whatever the hell that is.
  • edited December 2003
    It hangs at the PCI Device list when I have it on turbo mode. There's also street racer and F1.
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited December 2003
    Try another one then. Street racer.
  • ketoketo Occupied. Or is it preoccupied? Icrontian
    edited December 2003
    Higher is better in the Sandra memory benchmark, ie, your more recent score is much better than the first.
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited December 2003
    I just tested a pair of HyperX 3500 Cas2 myself. As usual, i can't find memory worth a damn. One stick is bogus and fails memtest at any fsb. This is the fifth time i get bad sticks. What is it with memory manufacturers theese days? This kind of activity must be illegal at this price.
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited December 2003
    I´ll bet those sticks does 250 at 1:1 with 7,4,4, cas 2.5 at least. Just give them some juice and that system will fly.

    Winbond LOVES voltage, no doubt about that.
  • edited December 2003
    Mackanz had this to say
    I´ll bet those sticks does 250 at 1:1 with 7,4,4, cas 2.5 at least. Just give them some juice and that system will fly.

    Winbond LOVES voltage, no doubt about that.

    The voltage currently is 2.8v, that's as high as I can put it. I had to change it in order to install Windows with these sticks. So, in the same order as those other timings I posted, what do I change it to?

    Cas Latency-2.5
    ACT to Precharge-7
    DRAM RAS # to CAS Delay-4
    DRAM RAS # Precharge-4

    Like that? Also if I were to change the timings like that, still change it to street racer?
  • ketoketo Occupied. Or is it preoccupied? Icrontian
    edited December 2003
    The timings you had in your first post were fastest. Keep them there as long as possible. With memory timings, lower = better. Not carved in stone but a very good *general* rule.

    CAS Latency-2
    ACT to Precharge-5
    DRAM RAS # to CAS Delay-2
    DRAM RAS # Precharge-2
  • edited December 2003
    keto had this to say
    The timings you had in your first post were fastest. Keep them there as long as possible. With memory timings, lower = better. Not carved in stone but a very good *general* rule.

    CAS Latency-2
    ACT to Precharge-5
    DRAM RAS # to CAS Delay-2
    DRAM RAS # Precharge-2

    What's Mack talking about then?
  • ketoketo Occupied. Or is it preoccupied? Icrontian
    edited December 2003
    Ahhh he's saying that your memory probably won't run timings that good if you change from 5:4 to 1:1, which is correct. If you want to see how high the system will go at 1:1 ratio, do the changes as he suggests.

    I would start at 220 and go up by 5 fsb at a time at 1:1. My BH5 modules (Corsair PC3500) crap out at 230 fsb, 2.85v (board max), regardless what memory timings.
  • edited December 2003
    keto had this to say
    Ahhh he's saying that your memory probably won't run timings that good if you change from 5:4 to 1:1, which is correct. If you want to see how high the system will go at 1:1 ratio, do the changes as he suggests.

    Will there be any noticeable differences. I can't tell a difference between the 3:2 and the 5:4.
  • ketoketo Occupied. Or is it preoccupied? Icrontian
    edited December 2003
    Well, *IF* you could run the memory at 1:1 at your maximum cpu overclock speed, yes, you would have a substantially faster system - at least as far as framerates and benchmark scores that aren't video card limited.
  • edited December 2003
    keto had this to say
    Well, *IF* you could run the memory at 1:1 at your maximum cpu overclock speed, yes, you would have a substantially faster system - at least as far as framerates and benchmark scores that aren't video card limited.

    So I should try this and switch it to 1:1?

    Cas Latency-2.5
    ACT to Precharge-7
    DRAM RAS # to CAS Delay-4
    DRAM RAS # Precharge-4
  • ketoketo Occupied. Or is it preoccupied? Icrontian
    edited December 2003
    Well, I would point to the differences in your Sandra scores between 3:2 and 5:4 - 5:4 was much more bandwidth and if you ran any other benchmarks you should see improvements.

    The same trend will happen if you go 1:1 but probably not as dramatic an increase as you will *likely* lose something by going to 2.5-4-4-7 (or whatever looser timings the memory needs) instead of your current (and very fast) 2-2-2-5
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited December 2003
    dak125 had this to say
    keto had this to say
    Well, *IF* you could run the memory at 1:1 at your maximum cpu overclock speed, yes, you would have a substantially faster system - at least as far as framerates and benchmark scores that aren't video card limited.

    So I should try this and switch it to 1:1?

    Cas Latency-2.5
    ACT to Precharge-7
    DRAM RAS # to CAS Delay-4
    DRAM RAS # Precharge-4

    Correct. Start without "turbo, F1,dragster or volkswagen" first.

    If this setting works. Then we will get on further. But first things first. You will se what i mean in the end.

    What Keto is saying is very correct. Lower settings on the timings gives better bandwidth but not as much on Intel as it does on Amd. Off course, both eating and having the cake is the best but we have to start somewhere to get there. For example, the DRAM CAS to CAS Delay have a huge impact on Intel setups at high fsb and clocks. Cas latency don't.
  • edited December 2003
    1:1 and
    2.5
    7
    4
    4

    Didn't work at all. I had to reset it. 1:1 has never worked for me in the past.
  • ketoketo Occupied. Or is it preoccupied? Icrontian
    edited December 2003
    If you start out at about 215-220 fsb and max vdimm voltage it will work, then you work your way up from there. You could even try setting memory timings 'By SPD' instead of changing them manually.

    But, my guess is that your experience will be similar to mine - biggest bang is at 5:4 ratio because of a) memory speed limitation (higher cpu overclock) and b) tighter timings
  • edited December 2003
    I can't tell a difference between how it is now and how it was at 3:2. I doubt I'd be able to tell a difference if it were to run at 1:1. I don't really care about benchmarks too much, it's all about if I can notice.
  • EnverexEnverex Worcester, UK Icrontian
    edited December 2003
    What the hell motherboard is this? I can't brake 2600 with maxed timings and 192mhz FSB on this machine....

    NS
  • edited December 2003
    NightShade737 had this to say
    What the hell motherboard is this? I can't brake 2600 with maxed timings and 192mhz FSB on this machine....

    NS

    Huh? :confused2
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited December 2003
    Almost a gig/second better bandwidth between the tests. That's A LOT in one step. About 20% if my maths serves me right. Can't argue with that.
  • edited December 2003
    Mackanz had this to say
    Almost a gig/second better bandwidth between the tests. That's A LOT in one step. About 20% if my maths serves me right. Can't argue with that.

    Indeed. Thanks very much for the help :thumbsup:
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited December 2003
    It's in his sig NS.
  • EnverexEnverex Worcester, UK Icrontian
    edited December 2003
    Mackanz had this to say
    It's in his sig NS.

    Sigs are off as they take up too much real estate.....

    NS
  • edited December 2003
    NightShade737 had this to say
    Mackanz had this to say
    It's in his sig NS.

    Sigs are off as they take up too much real estate.....

    NS

    I'm confused. But, I have an Abit IC7, if that's what you were asking.
Sign In or Register to comment.