my first kit had one stick dead,Quote:
Originally Posted by eva2000
second kit was a poor clocker 2k C7 at 1.67v
Printable View
my first kit had one stick dead,Quote:
Originally Posted by eva2000
second kit was a poor clocker 2k C7 at 1.67v
Quote:
Originally Posted by LardArse
What IC's week peter ?
Thanks mate. Not that is really matters now with the recall, but it would at least be interesting to see a list of the year/week codes of Corsair kits that died and ones that are still going strong after some use.Quote:
Originally Posted by chispy
im not usre i dont have the allen key to unscrew the top screws yet.Quote:
Originally Posted by Pt1t
http://www.pcworld.fr/files/61031-g-...F4LTk4MHgw.jpg
Those Gskill PS seems to dont die too , some month i m benching with it. IC's are mix of 848 and 849.
There still seems to be a bit of confusion as to exactly what chips are on the terminally flaky Eplidia Hyper kits...Quote:
Originally Posted by Bei Fei
It seems to me the affected Hypers have the code (on the actual chip) of BASE
My cheapo G.Skill tridents also have Elpida chips but the code is BBSE, hopefully this means they are either not Hypers or are differently binned / a different batch.
Either way I can boot into Windows at will :D (for now...)
lol Hipro been stressing them in memtest all night, still tuning. I did bin the sticks though, one seemed rather flaky and got tossed to side as a spare.Quote:
Originally Posted by hipro5
Maybe if you went green this wouldn't happen :p:
BBSE are different, and much less dying.. tough, less overclockable if Im correct..Quote:
Originally Posted by Biker
You might get lucky on green also. Given that P is a derivative of both F and I; AMD simply won't stress those modules like i7 will :ROTF:Quote:
Originally Posted by chew*
Quote:
Originally Posted by Leeghoofd
Quote:
Originally Posted by hipro5
I think I can stress them quite well ;)Quote:
Originally Posted by raju
Still a work in progress. I will say I found some oddities, exploring further before I comment.
http://chew.ln2cooling.com/elpida/elpida%201.JPG
More Chew, we want more :) MM is testing the blades on the P55 right now...
heh already running second round of 32m as we speak a tad faster ;)Quote:
Originally Posted by Leeghoofd
Also in no way stating that there isn't a problem just maybe the problem might be a combination of circumstances.
I've tested the Blades on the AM3 a long time ago - 1400CL5 and 1700CL6 1,65v.
But ssshhttt: Leeghoofd doesn't want to know :p:
Massman, you run into any peculiar issues on AM3? Try with a tad more volts per chance? Were they 2x2g?Quote:
Originally Posted by massman
Only issue was the poor quality of the IMC on this 720BE, so I was limited by cpu anyway. Didn't try much more voltage, never went over 1.7v iirc. 2x2GB is correct.
No issues, to be honest, never had any so far. I really love this kit :D
Interesting, run 32m by any chance @ 1.7v?
Reason I'm asking is at same speed above @ 1.7v = fail, Memtest passes @1760 +.
I think I'm gonna break out my mach 1 to get to the bottom of this, seems all to familiar.
Fail due to too high clocks or voltage?
Nah Dropped cpu speed and nb speeds and voltages down......still fails 32m with 1.7v at same speed, passes at 1.65v fails higher than 1720.Quote:
Originally Posted by massman
It gets stranger I can do whatever I want at 1720, Like crank down on all sub timings but can't pass 1720 and can not give more than 1.65v.
Now we had a 3x1 g kingston kit that we could clock higher with and give rather excessive voltage to.....
I think you know what i'm getting at here. I'm on air atm, will bring some cold into equasion in a bit.
Tightening subs. Like 10th run so far I think. Still alive and kicking,
http://chew.ln2cooling.com/elpida/elpida%202.JPG
EDIT, think I figured out a way around part of the problem....
http://chew.ln2cooling.com/elpida/elpida%203.JPG
I was actually going to suggest that: drop the NB frequency :).
The big problem with AM3 memory overclocking really is the quality of the northbridge. First of all, it's not necessary to have high speed memory if you're not using a high-speed northbridge and, secondly, the stress put on the northbridge by running high clocked memories is enormeous. Have you seen Titon's 940MHz+ memory speeds? That's only possible because he was running a 955 (better silicon quality) ánd, very important, had his cpu run under phase.
You'll notice, by the way, that there's not a lot of scaling in performance going up even higher in the memory frequency. You might want to try CL5.
Not sure what this has to do with the Elpida's dieing on everyone.
Well nah it is off topic, but I'm hammering them with 32m to A: make sure they are going to hold up, and B: break the myth that they will die if you run 32m...I'm sure they have issues, but I don't see any AMD guys with said issues.......Quote:
Originally Posted by massman
I tried 1500 5-5-5 in pi and a few other benches when doing the review for the 955, utterly useless at 6 gig +........my gskills smoked them 9-9-9 1700+
These are going to be for 3d only.
The NB can handle that speed, Just dropped for now to eliminate it from equasion, ram is ramping up now due to some adjustments was still failing with lowered NB prior to adjustments.
How high was your northbridge running at 6G+ runs? I bet higher than in the screens you're posting now. That's what I'm talking about here: low NB frequency and you don't need high frequencies on the memory to have good performance.
Can you run 864MHz CL6 with 2.8G NB on air cooling (as in same temps)?
As for the myth; it's obviously a myth. There's no such thing as 32M killing memory :p:. Has anyone made a list of the motherboards that killed the memory?
Quote:
Originally Posted by massman
Was 3800 + NB
Ahh, yah at 4 gig and 2600 NB I saw no benefit over ram speeds. 5-5-5 kicks ass.
I'm not tuning these for 4 gig though :p:
Yah thats what I'm getting at is it certain boards and or just plain intel boards, I'm on my 15th 32M run still no dead sticks.... I think its a combo, elipdas design flaw, intel board voltage implementation? Maybe a voltage spike at boot?
You're wasting your time with 32M: it's just not the reason why the memory gets killed.
As far as I can see (and I've heard), it seems that Evga, Dfi and Asus are mentioned the most. Then, the obvious question would be: why not MSI, Gigabyte or other manufacturers?
Because the first 3 brands you mention are used by the majority of the Elpida hyper users perhaps? I do not think the board matters, its the dimms it seems. I have had several sets on Evga, Asus, DFI and GB board, First dead dimm this week :( New kit, installed OS at rated clocks (2000mhz @ 8-8-8 1,65v) reboot, set 1600mzh for test and no boot. 1 dimm = dead
It's possible that the explanation is indeed that simple.
What board were you on?