Discussion in 'TechPowerUp Club Forum' started by ShadowFold, Dec 22, 2008.
Are you going to go the whole OC way in choosing speed, or will you be happy with 1333/1600?
I am going to AIM for 1600 and see if I can work it up from there. But if the price is right I'll get some 1333 and overclock it.
I can't say that I agree with you on that. I do agree that having the reference clock higher does help with overall read write copy times and transfer rates. But upping the NB regardless of what reference clock is set to will always benefit.
Now I'm going by my testing with my current setup, as I always ran that everest cache/mem benchmark. And there is always an improvement. Whether it be tighter timings, higher NB, higher RAM frequency, whatever, there is always an improvement.
Now how it "feels" at keyboard level is a whole 'nother story. Something that might be fast in the benchmark might actually "feel" slower than a lower setting.
You should also check out the Patriot Viper series, they use the same IC's as the Crucial Ballistix that the guys over @ XS have been getting some crazy frequencies/timings out of. BTW, the IC's are the Micron D9GTS's.
I'll check those out as well, thanks for the heads up.
Be sure to look over at ZipZoomFly, I've gotten some awesome deals on some awesome RAM over there. The current kit I have I bought for ...... $50 where as everywhere else it was $90+. And today I just bought an OCZ trichannel kit (3x1GB) for $59. That was the cheapest 3 x 1GB kit anywhere online, and I spent about an hour looking. As a matter of fact it was $33 cheaper than newegg's cheapest kit.
I'll head on over there shortly and see what deals I find. Although I'm not ready yet, I'll still check them out to see how the prices are and stuff. Thanks again!
Not a problem .
I'd have to agree with erocker's statement, as I was wondering what caused those 'micro-stutters' in GTA IV. My observations concluded much the same as what erocker stated.
So, even if Everest shows an improvement, it isn't a true reflection on apps/games that actually use it. If your RAM has to wait for the L3, or vice versa, means little to nothing to the benchmark results that you see in Everest. It does, however, mean a world of difference for the very things you'd be increasing the speed for in the first place
exactly which is why i noticed a 6fps increase in GTA IV before i got the dreaded PnP error at that time my l3 and ram were reletively close in there perfromance numbers
There is absolutely a point when you're running too high or too low a northbridge speed in comparison to the system ram. Everest of course will indeed show gains on the various things, however actual usage will suffer if you're running too much or too little Northbridge.
There is a delicate balance that has to be maintained for optimal performance between the ram and memory controller/northbridge. erocker is right about the approximate speeds you need to run though you can get away with a variance of say +/- 150-200Mhz either way (which of course he already knows). I usually try to stay on the - side of things, because it appears to have less of a performance hit than if you go too high. Another benefit to staying on the - side of that limit is that it takes less voltage/heat so you're more stable that way as well. You definitely want to try to keep your Memory Read and L3 Read speed as close as possible for optimal performance. To "make up" for that fictional loss in performance that you feel you'd gain clocking the Northbridge as high as you can get it, you should instead get the ram timings as tight as possible which will give you that speed right back. This time you won't have any issues though with the two components running out of sync.
Depending on memory timings the motherboard will allow me to get away with, I usually run 1333Mhz ram cas6 with a 2200Mhz northbridge which seems to work out BRILLIANTLY quick with the stock 3.2Ghz processor speed. I'm still working on the 'perfect' speed for 1600Mhz ram as it's slightly harder to get the timings right depending on the board. The speed will end up being between 2200-2500Mhz depending on the timings I settle on. For a daily clock I don't really like to overclock anthing if it requires above minimal voltages (3.2Ghz @ 1.168v cpu/1.6v ram/1.075-1.100v cpu/nb) so that normall leaves the htt bus speed ouf of the question.
I'm hoping that I can find a balance similar to my last board (I REALLY miss the 1333 cas6 for daily clocks), but this new board great as it is doesn't like to run timings quite so tight without lowering the divider and clocking up the htt bus speed.
Oh well, I'll find something I'm happy with soon enough.
I'm not totally sure I follow why your mobo is limiting your CAS latency?
Nitpick: HT=AMD's HyperTransport. HTT=Intel's Hyper Threading Technology.
Well how close should the reads be?
Because mine are pretty dang close and I'm above my ideal by 200 MHz (2600). I can actually take the NB higher but at this speed is where I can run it overall the best.
Ideally, identical, or as close to it as you can. That said, however, it's not something that's completely set in concrete.
If your current settings are the best for your system, then that's exactly what you should run.
Your system's sweet-point, if you will
I'm going to fiddle around with it some more when I get a little bit more time for stability testing. The thing that's bugging me about the board is that there aren't a lot of RAM dividers.
This makes getting everything just right a pain in the butt.
There is no problems at all running a 2600mhz north bridge with 1600mhz ram. Perhaps one might run into some sort of problem at 3000mhz? The real concern is not having enough north bridge bandwith.
Not a pain. It's a challenge!
Isn't that one of the reasons you're doing this?
LOL, I totally didn't even realize that I kept adding that extra 'T' not sure what I was thinking about at the time haha.
Now then, as I was saying my motherboard is not a fan of running super tight ram timings without having to clock the HT bus speed (lol). Not every board will have the same compatability with whatever ram you choose so you can be limited in what you can do (bios reasons basically).
I know my ram is capable of running timings much tighter than what this board is allowing me to do right now. Regardless of voltage I can't run cas6 1333Mhz without using a different ram divider to do so...yet it will allow me to run 1600Mhz cas6 if I lower the ram divider and clock the bus. On the previous motherboard I could run 1333Mhz 6-6-5 @ 1.50v through every stress test known to man, and it wouldn't even blink! On that board however I couldn't clock the ram quite as high as this one can so it's a trade off. I'd much rather have the super tight timings over the higher speed (don't need the extra bandwidth of 1800+ really). I've tried a bunch of different bios versions on this board, and no matter it won't let me run cas6 without lowering the divider even on 1333Mhz speed.
My northbridge clock and especially voltage are also limited on this board compared to the last. I was able to achieve supreme northbridge performance levels on that board (2.4Ghz @ stock 1.10v no problem...2.6Ghz would even run for a bit at that voltage!) as well, I can still get 2.8Ghz northbridge if I need it on this board, but it takes 1.30v to do it whereas the previous board only needed 1.26v (and would boot it at 1.20v whereas this won't even think about it under 1.27v). 3Ghz on the other board I could run @ 1.26v though it wasn't 100% totally stable, it was still doable...on this board anything under 1.40v and it won't even think about it....and it's still not stable then.
The motherboard and especially it's bios has a large influence on what you're going to be able to do with your ram and northbridge, just like it does with your processor.
I'm aware of the limitations mobos can have in regards to RAM, I just wasn't sure how much this would translate into the CAS settings of the proceedings.
Makes sense though.
This is the setting that I TOTALLY miss...it was blazingly quick and as smooth as butter!
Well yes, but this is my first build since coming from 939 and using DFI boards which have LOTS of dividers. Bar-none the best 939 overclocking boards made, IMO.
Technically it's called the reference clock. It's not called HT, HT reference clock, FSB, HT bus or any of those things, just simply reference clock. It's like the base clock on the i series intel CPUs'. I researched this a while back.
At any rate, you're absolutely right about the motherboard, that's why I never skimp on my board. I'll get a slower/cheaper CPU before I get a mediocre board.
It's the same reason I lap most of my CPUs', because I don't want temps to be an issue. 45C under load after 16 hours of P95.
I'm ALLLLLLLLL about getting the right motherboad (basis of any system in my opinion), I was just in a tight situation which is how I ended up picking up this board. My ASUS M4A78T-E had some questionable things about it's voltage and temp readings (among other things) so I didn't want to risk sticking it out if the board was just going to die right after the 30 day return period. When I went to exchange the board they didn't have any more in stock so I would've been without my machine for another 2 weeks or so (after already being down prior to that board roughly 2 weeks)...I was not in the mood for that so I decided to give the Gigabyte MA770T a shot until I pick up the next board.
Thankfully besides a little quirk here and there, this board has been nothing but MIND BOGGLING in performance ability. It somehow is able to do at $80 what other boards of a much much higher price tag can claim, this board took me totally by surprise.
It's still only going to be a stopgap to the next board which will likely still be the Crosshair III (I'm a big ASUS man...though now Gigabyte is my #2 all because of this magical board). I'd really really love it if ASUS were to put 8-pin power connectors on all their boards (like Gigabyte seems to do this round even on a 'lowly' 770T), because I'd just pick up another GX board or maybe even 790X. I don't care about the four way slot configurations because I never ever use those anyway, two slots is plenty for me...if I need more than that then I should've just bought a more powerful gpu in the first place lol.
Because of this awesome little board I've even thought about picking up another Gigabyte board like the 790X or something, but the biggest thing I miss from my beloved ASUS boards is the bios layouts and updates...I'm not sure if I can really get over that lol. It's been a loooong while since I've owned an MSI board (though the board I owned was good), but I've still considered the GD-70 because well....
1) It's GORGEOUS!!!!!
2) IT'S GORGEOUS!!!!!!!!²
3) It's a beast of a motherboard (it's gorgeous³)
I've recently installed the ASUS M4A79T Deluxe mobo. I've got the PII X4 955 (C3 rev) installed.
Are any of u guys running this board? If so, could u pls share ur bios settings?
I'm currently OC'ed my ram (kingston) from 1333 to 1600. Just need to tighten the timings.
CPU @ 3.95mhz.
Core Speed = Reference Clock * CPU Multiplier
Northbridge Speed = Reference Clock * Northbridge Multiplier
HyperTransport Link Speed = Reference Clock * HyperTransport Multiplier
For all intents and purposes, the reference clock is still the good old FSB
So, using my X4 965 as a very basic example:
Core Speed: 200MHz (FSB) * 17 (CPU's native multiplier) = 3400MHz
HT: 200MHz (FSB) * 10 (HT's native Link Speed multiplier) = 2000MHz (which translates to 4000MT/s (MegaTransfers/second)).
Physically the CPU's memory controller still needs to communicate with the RAM, which is done over the FSB, which is why changing the HT/NB does not affect the FSB, but the HT/NB are affected by a change in FSB.
This also explains why increasing the CPU's multiplier gives a greater through-put (bandwidth), than does raising the FSB to get to the same Clock Speed.
To see a simple demonstration of that, run IntelBurnTest, with the CPU @ 4GHz.
At 20x 200FSB, a 1920mb calculation will take less than 55 seconds on the X4 965 (roughly 42 GFlops).
At 16x 250FSB, that same calculation will take more than 55 seconds (roughly 40/41 GFlops).
Your mileage may vary, of course, but the disparity will be there.
I just used 4GHz as an example, as it's a relatively easy (..) to get speed on the X4 965. Use whichever speed you like; the effect is still the same regardless
The whole point of AMD utilising HT, is to remove the [potential/encountered] inadequacies that the FSB has with the modern architectures, which can cause bottlenecks and/or other slow-downs.
The way things are going, I can't imagine FSB disappearing completely in a fairly short time-frame.
It is bound to be scrapped though.
BTW, HT is (a few technical exceptions aside) the same as Intel's QPI - both negate the need for many tasks that used to go through the FSB, hence the on-board memory controllers (thus including the NB) of the respective camps' CPUs.
This is what I remember from AMD's & the HyperTransport Consortium's respective white papers on the subject.
My memory is not as reliable as it used to be, which is why we have people like erocker & Kei, to name but a few, to help us steer through the murky waters
Welcome to this thread!
I have that board, though I'm using a 965 C3.
And different RAM
What are the native timings for your RAM?
Thanx for the welcome.
I think the native timings are crap ( 9 9 9 24) not sure but will check tonight. Will see if i can tighten up my timings.
Separate names with a comma.