K8N
Yeah it's your memory-cpu subsystem that's failing in Prime. If it's not 3D&Prime-stable.. it's not.
Printable View
K8N
Yeah it's your memory-cpu subsystem that's failing in Prime. If it's not 3D&Prime-stable.. it's not.
Ivan
Well, I knew it wasn't stable. And I wasn't fishing for a full-pass for getting through one of the Prime95 tests. I just re-read my post, though, and I have to admit I can see how it might have appeared that way. :)
What I was really looking for was maybe what some of you guys would do as your next move if you were faced with the same situation I described.
Well, here's what I did and it worked!
Since Prime95 was pointing to memory, I knew I'd have to start there. I also knew I didn't want to cut back on mem timings if I didn't have to. And since the CPU (and chipset) seemed to be passing Prime okay, I thought it might have to do with cooling. Both my CPU and chipset have nice heatsinks/fans, but my memory only has heat spreaders and they get pretty hot. So, I went out and bought an 80 mm / 2700 RPM / 35 CFM airlfow fan, made a bracket out of aluminum stock and mounted it directly beside (actually touching) the 80mm Thermatake Venus CPU fan. The new fan is directly over the memory slots; when looking in the side panel window of the case, it now looks like I might have two CPUs mounted side by side based on the appearance of the twin fans. I also replaced the stock Lian Li rear-exhaust fan while I was at it.
Memory is cool to the touch now, and I pass all Prime95 tests.
You water-cooling guys are probably laughing, but hey, it worked. :)
K8N
Hmm.. interesting. Never had anyone have an improvement by cooling the memory. Maybe since it's so hot right now? Can you maybe test what's the highest stable with and without memory cooling? That would be VERY interesting data.
Ivan:
Yea. I can test that. I'll do tests lowering the HTT (which will lower memory speed), and I'll also do a test or two just relaxing the memory timings.
I'm pretty sure that the absolute stable HTT is gonna be around 250-255 with the memory timings set as they are now (2.5-2-2-10-1T @ 2.7v); a little higher with relaxed timings perhaps. I didn't do any 8 hour tests of Prime95 at 250-255, but I do recall that I passed at least a couple of hours before I stopped the tests.
The Prime95 tests that kept snagging me were the Lucas-Lehmer iterations using 1024k FFT length. When I first installed the fan and got passed that test, I was pretty sure I was home free ... and I was.
Anyway, it may be a couple of days before I get back with this info; now that I'm stable (even though I was "stable" before...just not Prime95 stable...meaning I was running Photoshop with 300MB+ files, even playing Doom 3, etc. and didn't notice any problems), I need to get some work done. Running all these multi-hour stress tests pretty much had my computer tied up for several days.
Later.
I am currently testing Asus K8N-E Deluxe and my FSB is stable up to 228mhz, but only in non 3D applications, if I try to run any 3D application PC restarts immediately. 3D stable FSB is around 215MHz. Will setting HT multiplier to 3x solve my problems?
And there is another strange thing. I set mem frequency to 400MHz (that should mean 1:1 FSB:RAM), when I check mem frequency in CPU-Z it says 228MHz 1:1 multi. But if I test mem bandwith with Sandra it is around 3050MB/s. And that is lower than bandwith of non overclocked FSB at 200MHz. Are there any known problems like mine? Is there any solution?
Thanks
TribesMan
1) 228*4=912 HTT, I don't think the K8N-E is capable of that, switching to *3 should definetly solve your problem, with no noticeable loss in performance.Quote:
Originally posted by TribesMan
I am currently testing Asus K8N-E Deluxe and my FSB is stable up to 228mhz, but only in non 3D applications, if I try to run any 3D application PC restarts immediately. 3D stable FSB is around 215MHz. Will setting HT multiplier to 3x solve my problems?
And there is another strange thing. I set mem frequency to 400MHz (that should mean 1:1 FSB:RAM), when I check mem frequency in CPU-Z it says 228MHz 1:1 multi. But if I test mem bandwith with Sandra it is around 3050MB/s. And that is lower than bandwith of non overclocked FSB at 200MHz. Are there any known problems like mine? Is there any solution?
Thanks
TribesMan
2) Bandwidth depends on frequancy, of course, but also on timings. Try switching to 1T command rate or reduce timings if possible.
Quote:
Originally posted by K8N
Ivan
Well, I knew it wasn't stable. And I wasn't fishing for a full-pass for getting through one of the Prime95 tests. I just re-read my post, though, and I have to admit I can see how it might have appeared that way. :)
What I was really looking for was maybe what some of you guys would do as your next move if you were faced with the same situation I described.
Well, here's what I did and it worked!
Since Prime95 was pointing to memory, I knew I'd have to start there. I also knew I didn't want to cut back on mem timings if I didn't have to. And since the CPU (and chipset) seemed to be passing Prime okay, I thought it might have to do with cooling. Both my CPU and chipset have nice heatsinks/fans, but my memory only has heat spreaders and they get pretty hot. So, I went out and bought an 80 mm / 2700 RPM / 35 CFM airlfow fan, made a bracket out of aluminum stock and mounted it directly beside (actually touching) the 80mm Thermatake Venus CPU fan. The new fan is directly over the memory slots; when looking in the side panel window of the case, it now looks like I might have two CPUs mounted side by side based on the appearance of the twin fans. I also replaced the stock Lian Li rear-exhaust fan while I was at it.
Memory is cool to the touch now, and I pass all Prime95 tests.
You water-cooling guys are probably laughing, but hey, it worked. :)
Hi
Well I have to admit i also have a 92 mm lownoise fan, at 5volts blowing on mem, just so I know its not boiling :), mem gets real hot these days hardtimings and all. Btw just pulled the IHS on my 3400+ CH sh7-cg, ill post some pics l8r, temp drop off 5 degrees, it carried no waranty anyways, I got real cheap from some guy who pulled 2 legs off it, I just soldered 2 new legs on it and kazoom it worked again. All it needs to cool down is a little flow, mine also feels cool to touch, bu before the fan was there very hot.
Menty
Vcore Mod + Vdimm mod has succeeded :D :D
Will hopefully post some VERY High Res pix later tonight. :D
Thx goes out to YoupY for helping me !!
Got Vdimm up to 3.4 :)
Mobo usses the 3.3 V for Vdimm
Hi
OHHH come on share :), hehe ill be studying those pics up closely I can assure u :).
Menty
Yeah :)Quote:
Originally posted by Mentaly
Hi
OHHH come on share :), hehe ill be studying those pics up closely I can assure u :).
Menty
My 8KDA3+ wont boot at all (boot = FF emidiately) its as good as dead ..... :mad:
Tomorrow ill build it in my pc and see what kind of potential this thing realy has :D ;) :banana:
Right, I've never done a volt mod before. So I'm going to need some very detailed information and pics, and exactly what to do.Quote:
Originally posted by SmokeyTheBandit
Vcore Mod + Vdimm mod has succeeded :D :D
Will hopefully post some VERY High Res pix later tonight. :D
Thx goes out to YoupY for helping me !!
Got Vdimm up to 3.4 :)
Mobo usses the 3.3 V for Vdimm
Looking forward to trying this now. Great work Smokey!!
Nick
Dont worry m8 youll get some VERY high quality pix :D :DQuote:
Originally posted by Nem
Right, I've never done a volt mod before. So I'm going to need some very detailed information and pics, and exactly what to do.
Looking forward to trying this now. Great work Smokey!!
Nick
Sound great mate :)
I'm sure I'm up to doing it, but just havn't done anything like that before.
Also a list of exact parts would be great.
I'm currently running 240 HTT 1:1 at 2.7v with this new Crucial Ballistix ram at 2.5,3,3,10, but also have some OCZ BH5 which does 250 - 260 HTT at 2,2,2,5 - but needs 3.2v to do that. So this might let me try that again.
Thanks
Nick
My ballistix does 2.5-2-2-5 at 240 (2.7 v), you should try ;-)
I'll give it a go....
:)
Nick
SmokeyTheBandit
Let's go photos :D I guess VDIMM is something I need so this Crucial Value stuff can go higher.
Got the pix @ my not working epox pc ...... cant get to them till tomorrow .... :mad: :stick:Quote:
Originally posted by IvanAndreevich
SmokeyTheBandit
Let's go photos :D I guess VDIMM is something I need so this Crucial Value stuff can go higher.
SmokeyTheBandit
No worries.. what resistor did you use to raise the VDIMM? Do the other volts track along with it or not?
Got 50 more mhz out of this board compared to Chaintech VNF3-250. Seem to allow a little bit tighter memory timing. (using DDR booster ATM. It seem to work with both board granted you are willing to go with 1 DIMM or heavy modding).
Just need some points for volt modding and measurings :)
anyone try K8N-E Deluxe Beta BIOS 1004.003? any differences?
Not tried it, but I've got the release 1004 installed and it shows up as 1004.003 anyway. So seems to be the same really.
Nick
CHR,
All my posts have been based on BIOS 1004.003. I never flashed to 1004, so I can't make a comparison, other than to say I'm getting similar results to other people in this thread.
It definitely does not have the change most people are wanting, namely a wider range of vdimm values beyond the current max of 2.7v.
Let's all email ASUS and complain for at lease 2.8V of VMEM?
Hi guys i havent got any great news im afraid i cant boot my pc up for a few days because the hose of my prometeia broke :( :(
Its going to be fixed but it will take a few days ........
Ill put the pix up right away when i see a chance of getting them.
Contact me and I'll send the ones you need asap ;)