Below is the Trident X version of the Ripjaws.Quote:
Originally Posted by ZX2Slow
http://www.edgeofstability.com/image...deluxe\mem.jpg
Printable View
Below is the Trident X version of the Ripjaws.Quote:
Originally Posted by ZX2Slow
http://www.edgeofstability.com/image...deluxe\mem.jpg
Your memory read and copy throughput is significantly higher, also the cache speed is much higher. Is this the result of having more cores when compared to my 4820K?Quote:
Originally Posted by Praz
EDIT: Answering my own question, yes. Disabled two cores and results appeared to scale:
http://i294.photobucket.com/albums/m...psac93ba1b.jpg
I'm going to install my 4930K soon, I believe that I will see a bump to 59K+ MB/s read and copy.
Thanks for pointing me in the right direction, I'm getting back into the hobby after being away for a while and have been a little overwhelmed.
I know this is way outside the scope of this thread but I just thought I'd mention that with Mac OS X Mavericks this board works really really well. No manual file editing or forum scouring for kexts. Sound, Networking, Storage (even the Asmedia chips) USB 3.0 = all works with installers like Tony's Multibeast post-OS X Installer. This is even on the very latest BIOS from Asus website. No DSDT necessary unless you want speedstep and turboboost but I expect most of us here will be overclocking anyway.
On the r4be, the 3rd set of timings cannot be changed unless you change the divider, regardless they cannot be adjusted manually.
I kinda need this bug fixed, it's setting my lower divisors to aggressively to boot windows.
2nd set of timings when auto set are to aggressive too, but at least I can adjust those back to were they are supposed to be.
I'm at a wall right now, without those 3rd set of timings I have to rely on slow boot, but that's only to post at higher speeds then what I'm actually messing with at the moment.
If I could manually adjust the 3rd set it might help get me 100% stable instead of like 98% (110mhz bclk).
And besides, it might help higher dividers, I get a 50% write speed boost with a lower one, say 23xxmhz vs 26xxmhz same timings, the 23xx is 50% faster...
That's at 10-12-12, and that's 50% faster as well compared to 9-11-11 2400mhz at 100mhz bclk...
Use Aida V4.00 http://www.xtremesystems.org/forums/...=1#post5221015Quote:
Originally Posted by ZX2Slow
Try all three latency boundary settings first, nearer setting will be too tight for some rams or need more voltsQuote:
Originally Posted by NEOAethyr
I appreciate the help.Quote:
Originally Posted by Dumo
I tried nearer and farther on my 110 bclk setting, 2646mhz mem I think it was.
Also tried it on the lower mem divider at that same bclk (normally unstable though).
Farther helped on the lower mem divider, made it more stable, did not check for that long though, but it was for sure more stable.
It brought the speed back down somewhat to compare to 2400 c9, 40k writes in aida64 instead of 60k+.
Tried the same on the higher mem divider and it made it alot less stable.
Nearer was the most stable.
However the other day playing with cas9 at that speed, farthest helped a bit, it helped me pass aida64 without a ton of vmem (the ram gets hungry at those timings with those speeds...).
Very slow though, alot like that raw mhz setting, cool for playing around but not something you wanna use for 24/7.
I know the mem wants more volts at cas9, I don't have any fans on it right now.
So I've using cas 10 for the time being, 10-12-12, to see if I can't get 110mhz bclk stable.
Prob is, if I lower the mem divider to try to get rid of the mem out of the equation, it gets less stable, way less stable...
It changes those 3rd set of timings and I cant adjust them, I set advance and delay for each setting that needs it and they don't change.
Anyways whatever, I wanted to make sure I said something about it :).
I've tried alot of things and it does seem like it's the memory.
As for higher bclks, like booting into windows with say 113mhz, or posting 114+, I don't know what the deal is yet.
I think its the mem but I'm not sure.
I can post at least 115, 120 didn't post..., with the slow mode switch on.
It's not a huge deal though right now, it's almost christmas and all that, busy time of the year.
And I'm stable anyways at 100mhz, 2400mhz mem cas9.
I'm just messing around from time to time, trying for higher bclks without the straps.
One odd thing though...
I can boot 190mhz bclk np, using the 166mhz strap (I used the profile for that today for the heck of it).
I would figure the dmi/pcie and so on would be at a higher speed with that setup then say 113mhz - 100mhz strap.
Odd thing is, obviously it's more stable. :\
I'm staying away from the straps for actual usage (messing around is something else) though because I can't idle my cpu, I don't really wanna use my cpu at 100% power 100% of the time lol :).
2 more things I wanna mention, loading my 110mhz bclk, either from a cleared cmos or from my 100mhz bclk profile works fine.
Loading it vise versa though, from 110mhz 2600mhz+ mem, to the 100mhz blck setting, 2400mhz mem, doesn't post.
Never does.
Though it will if I back the ram down to around 2400mhz 1st before loading the profile (it's actually like 23xxmhz something or another).
The other thing, when I crash, on my stability testing on 110mhz, I don't bsod or anything, but I just noticed this, my game crashed and my cpu cooler, the led logo on it, it's powered by usb.
The pump isn't, just the light thing.
It was flickering when it crashed...
That was different, never seen that before, I just plugged it in the other day though on this board so I didn't notice it until now.
1st time I saw it a few mins ago, I'll keep on eye out on that when I crash from now on (no software installed, just usb power...).
Now that I remember lol, I got one more thing to mention, not that this matters any.
But my phone flips out on the efi bios when I got it plugged in.
It keeps turning it on, from standby to turning the screen on by it's self.
It does it at least 3-4+ times.
It also can't see the storage as a usb stick, even though that's how my phone is setup.
Then I got this usb stick, one of those cdrom emulated crap sticks..., I hate those things...
But anyways, it comes up as 2 drives on post, I can understand that, but in the bios boot menu and in setup it comes up as 3 drives.
And only ones boots, it's not efi compatible that's for sure (efi method won't boot).
I got memtest86+ on it.
Every single boot I have to change the boot order manually, otherwise it tries efi mode for some reason.
I found it best just to say screw it and hammer on the f8 key when booting to select right one each time.
Haven't tried booting from my phone yet, but it won't save bios pics to it, maybe because it's formatted as exfat, that could be the prob...
I should try fat32 in the afternoon.
I gotta get some sleep though :).
I know I'm rambling, but I figure more info the better at this early stage of the boards release.
The only info I figure I just posted that is important though is the 3rd set of timings not being adjustable, and "maybe" the profile loading no post prob I get personally, 2600+ -> 2400 (not many would have this prob, it just kinda sucks for me).
Thought it was it was that the 2600mhz profile had enhanced training enabled and the other doesn't but that wasn't it.
I really suspect its the 4 vs 6 core results. I don't see much of a change:Quote:
Originally Posted by Dumo
http://i294.photobucket.com/albums/m...ps7531f4fa.jpg
Am I the only one having higher read/write/copy and lower latency, when running 1866 8-9-9 or 2133 9-11-10 or are it my early RipjawsX that scale better with tight timings?Quote:
Originally Posted by Praz
And yes, I`m old school:p:
sorry looks old thread. this is XMS3 double sided 5.11v. i think is hynix bind. not very impressed with read write performance. got no idea whats wrong with write result. ram is fully stable with memtest overnight @2400 though. any idea whats the cause maybe?
http://i61.tinypic.com/mc64hu.png
Seems to be a common problem with the 2400 divider on X79. My recommendation would be to use the 125BCLK strap and 1866 RAM divider. 125 x 18.66 = 2333(2332.5 actual). If you're dead-set on 2400, you can increase the BCLK to 128.625. 128.625 x 18.66 = 2400(2400.1425 actual).Quote:
Originally Posted by xpower
so this is the problem with x79. is ASUS knows this problem and willing to solve. 4804 bios does not cure to this problem. what a shame "ASUS". i asked this very same problem in the asus ROG forums but no one has the answer :(
This is not a problem fixable by anybody but Intel. Asus has nothing to do with it.Quote:
Originally Posted by xpower
I already gave you a workaround. Something else you can do if you're unwilling to play with BCLK is to simply run the RAM at a slower speed with tighter timings.
@Random how are u doing with Aida score wise at 2400?
sorry ASUS wrongly said is asus :D ram is just not cooperating when set tighter timings. setting relax timings in result higher latency. will get some other ram kit to play with.
Hell if I know. Haven't run it since I was on a 3820, and even then the scores wouldn't be comparable as I was running 129 BCLK, not 100.Quote:
Originally Posted by xpower
I would have to reinstall AIDA, but I could probably post a couple of stable runs early next week. I was planning on doing some RAM clocking this weekend anyway, as I have a 4x8GB kit on its way that should be here tomorrow or Saturday. For comparison, though, I have posted a run by a friend over on TPU using the same kit of RAM I'm expecting:
http://img.techpowerup.org/140328/trident32gbaida.jpg
As you can see, different board, different CPU, different RAM, and still bad write performance using the 2400 divider.
We should probably take this to PM, as we're getting a bit off-topic.