OCZ 4000VX on NF2 Lan Party B

[Guest Darien]

Wow.. I never checked or thought about my IDE drives...

 

I'm using a SATA HD... but the primary IDE is my DVD player.... which was LBA, I changed it to CHS..

put AGP aperture to 32M, latency 16 (was 32)

I left IDE prefetch mode on....

disabled my secondary IDE channel as it's not being used

 

my 6800LE has 128MB and a 4 pin molex for power.. I always run it at 1.8v, I overclock the card but when I get unstable I run it at stock to rule out video card issues

 

i'll test your BIOS sure... I tested your other one... sad to say I didn't get 1 more FSB our of my rig from using it... and it made my SATA more suseptible to corruption so I went back to Hellfire r3...

 

but I'd love to try the new bios... link plz =)

[NEOAethyr]

Ok then .

 

In the morning I'll upload it for ya.

I got 275x8.5 both superpi and s&m stable.

As well as 276, the same multi.

 

Right after I got 275 that way, I tried 9x, with 1.725v(instead of what I was using as a guess for the pervious, 1.65v).

Well it did'nt liek that, failed suerpi on the 6th loop, then the 4th, then the 3rd...

I then moved to 278x8.5, failed, and the memory bandwith was quite low compared to what 275 was, around maybe 100megs less :.

 

So I did the 276 thing, worked out.

However it was sorta slow, after suerpi and s&m, at 1.65v of course, I re-checked, I gained around 20megs or so.

 

I lso notcied the other day when I tried 1.85v for 266x9.5, my system freaked out.

Refused to boot at 1st, and afterwards I was plaqued with a ton of warm boot issues.

Even at 266x8.

Anyways to make a long story short, after adjusting rails and more stressing, the boot issues are going away.

 

I'm really starting to wonder about the higher vcore's.

I have my panaflo on my cpu right now, I know it's not the best, but it should be doing allmosy as well as my old 80mm tornado.

I guess I leave it at a max of 1.65v for now until I figuer what the heck is going on with that.

My 2500mhz thing was done at 1.725v, so 1.65v is not far away, I may beable to manage if I work on that, which is nothing short of amazing if it does work out :.

 

 

I'm priming for the heck of it on 276x8.5.

Using a nromal liek config of 1024-4096, like the old days, for me anyways.

I know not to bother with prime for breakin anymore, but I gotta do somehting while I'm idling away ^^.

Plus I need to get an 8hr run here somewhere so I can at least 100% confirm something.

 

I wonder if I should just jack it up to 275x9 anyways at 1.65...

[Guest Darien]

Well I'm defineately starting to think it's the new BH5 RAM...

 

I put my VX RAM back in yesterday to see if it would give the same problems I'm having at 255... well with the VX in the Rig is does 255 3dmark01 no problem

 

so 245 and above with the BH5 doesn't work, with the VX it does all the way up to 255.... pretty straight forward...

 

also for no apparent reason the VX scaled all the way up to 260 x 10 1t in memtest!!! it used to get hundreds of errors per pass at 258, and now it was doing 260 somwhat stable... it would error every few loops though on test 5.... I actually booted to memtest at 261 x 10, just to see my memory bandwidth at over 2000... hehehe was neat... but not stable at all... and as usual after a few hours of Incredible record breaking performance, the VX gets the looping 256MB error and is again useless.... so I took it out.... again...

 

anyways, I did the deeplogic + 5% FPU... and it helped cpu stability cuz then I could do all the S&M cpu tests at 100%.... but it didn't do a damn thing for my RAM stability... when at 250 RAM still failed S&M, Prime95, SuperPi, 3dmark01, and AquaMark3 (I found a free download for AQ3)...

now again passes S&M cpu at 255, fails everything else prone to BSOD's but does it fine with the VX RAM all at the same settings

 

Now check this out, I wanted to have something hammering on the RAM overnight... the only thing I could do was memtest... so I had memtest Hammer test 5 on the damn thing for like 6 hours overnight... I slept on the couch and check periodically if I woke up... it rarely errored.. it would fail a pass every 50-100 passes... failed once as short as 20 loops... went as long as a 121 loops wihtout failing as well...

 

but here is the thing... when the RAM does fail a pass, it throws like 8-10 errors at a time, normally when RAM is basically stable and fails a pass every so often, RAM throws just 2 errors (Dual Channel)... but this stuff is throwing 8-10 !!!! then it's fine for 50-100 passes.... and it always the same address failing... I forget the exact address but its like 1001.5MB or something just over 1000....

 

It just sucks! my mushkin blue, Twinmos Speed Premium and now the BH5 are all phenominal in memetest... but just won't do a damn thing correctly in windows unless I run it 10-15 FSB lower than what it can do in memtest....

 

I was thinking that may be the case with any 2-2-2-11 high voltage RAM on my board.... but the VX works just fine! (when it works)... so I don't know WTF 2 do :mad:

[kitfit1]

As you know Darien, we all have to back off a little bit from what Memtest says, but 10-15fsb sounds like a hell of a back off to me. I don't think i've ever had to drop more than 5fsb from Memtest to be stable in Windows. Something in the back of my mind says that this could well be a power fluctuation issue. To be looping Memtest clear for hours, then suddenly get a few errors, then run clear again. Surely the number 1 suspect has got to be a power spike or drop. Did you have a look around UPS, it could well solve your problem.

[NEOAethyr]

Yeah it very well might be power probs.

 

 

And yeah we all have to back off .

At times I've had to back off around 1-2mhz from what is stable in memtest.

However recently, like I was saying before, I found even if it's failing the 1st pass, you can make it memtest stable, sometimes.

 

 

 

So apperently looping memtest helps...

That I knew somewhat, but with the results you got it seems that my orignal thaughts were wrong.

I do know that errors are really bad, but apperently, from what you've seen and what I've seen, that they can be ok, that you can ignore them(some of them^^) and just hammer away and still get a good breakin.

 

Last night I played quite a bit of gtasa at 275x9, which is allmost 2500, @ 1.65v vcore of all things.

Pretty stable at 1st.

I'd drive around, after a long period of time it which exit out and I'd try again.

However when I started flying around, it got worse and worse.

To the point where S&m, deeplogic, superpi and capturing were resulting in bsods .

So I know not to fly around in gtasa when going for lower vcore...

 

I noticed thoday my superpi 32m inital value was up another sec :, so now it's up 2 secs from before.

It must be cpu heat related, because it's the day time.

Also I know that past 266 on my setup does nothign for superpi speed, no joke ^^.

It's the cpu speed that's holding me back on superpi, apperently I got more then enough memory bandwith to compete agais'nt superpi computers...(Quite interesting I think)

I started noticing it recently, and that's why I know, the cpu gets slower when it's hotter, I could actually prove it as my rig is(given I had 24hr's to do so), or at least I should still beable to.

 

 

 

Darien, was that 2k in memtest?

If so what sort of bandwith do you get in windows?

 

I'll attach a screen of my current suicide bandwith, nothing stable past 280 though.

I can game and such even during the day now at 280.

Also, lmao, I can boot and mess around with the aggressive interface at 275 .

I was s&m'ing it by accident just a few mins ago.

Even at optimal, I get the same amount of errors(1 in test 4) at 275 dc in the day .

Still one heck of an inpovement.

 

I should'nt beable to boot 275x8.5 or 9x at all in dc, but I can now, and even game it and etc.

8.5x is actually stable during night time of all things.

I even got 276x8.5 stable last night.

275x8.5 passes superpi during the day.

277x8.5 passes superpi in the night, but I did not have time to s&m it, I feel asleep.

When I woke up it had passed, but when I tried s&m'ing it, I got some errors, I redid superpi and instant hardlock right after the inital value or one of the 1st loops .

 

After messing around with that 279x9 and screwing my cpu up a tad, I messed up a few directories, so my old superpi install is dead and gone, as well as some windows files I guess.

So there goes all my times...

 

 

Anyways my last suicide shot so far, check it out.

Also I'm very glad to hear that the deeplogic and s&m trick worked out for ya , for s&m anyways.

Now the only thing you might need to worry about it actual priming at that cpu speed and vcore, but if that's good then it's all good to go .

I need to install 2k3 today sometime so I can get down and do that.

 

My temps during the night, last time I looked last night, system: 19c, cpu 28c.

Day time, cpu 36c system 30-31c.

Cpu = socket.

Psu temp is 37c right now .

[Guest Darien]

yeah, bandwidth was 2005 in memtest86, wasn't stable but it looked cool...

 

i didn't test 261 x 10 in windows / everest....

 

I'm not a bandwidth fanatic with everest or anything.. I don't have exact speeds for exact settings... but my best in everest for the memory read was around 3987... or just a hair under 4K... and latency was 62ns...

 

I don't push to get the best benches for those, but those were the highest and lowest I got... yours are much better with that 275+ FSB...

 

what I forgot to try last night, that I'm trying today is a different memory burn in..

 

I learned this from, a review i read for the 'new bh5' (twinmos speed premium) that I saw from VRforums...

 

it's just the burn in wizard from SiSandra... I select only the memory bandwidth test and it loops continuously in the background... they reported measurable improvements from doing this... I did it for 1 hour this morning and will do it for 6-7 today...

 

don't know if it will help any but, as always, I'll post results

 

@ Kitfit1: yeah thx for the input... I'm planning on getting a nice UPS here in a month or two... but I got an OCZ 520w powerstream so I'm getting pretty clean power here..

 

EDIT:

I don't know exactly what I did... may have been the SiSandra burn in for 1 hour this morning...

 

but now I can pass 3dmark01 at 250.... couldn't do that before... did check other stablility test as I'm off to work now.... but that's progress...

 

right now I'm continuing the SiSandra mem burn... gonna go 6+hours... bye 4 now

[kitfit1]

@ Kitfit1: yeah thx for the input... I'm planning on getting a nice UPS here in a month or two... but I got an OCZ 520w powerstream so I'm getting pretty clean power here..

 

I was'nt actually thinking about your psu Darien. What i was thinking was either a spike or a drop in the mains power before it gets to your psu, which then then has an effect on the psu, ram, cpu and so on. The psu is only as good as the mains that it's being fed.

[Guest Darien]

I was'nt actually thinking about your psu Darien. What i was thinking was either a spike or a drop in the mains power before it gets to your psu, which then then has an effect on the psu, ram, cpu and so on. The psu is only as good as the mains that it's being fed.

 

I think this is probably an issue.. this normally doesn't happen, but my stock computer that is always stable was at the windows login screen this morning and yesterday... and I didn't log out... so it must have restarted for no reason... could be the power is going out messing with my other computer too...

 

as for the SiSandra Memory Burn in... both times I left it ot run, either at work or overnight I come back to it and it's BSOD....

 

3dmark01 is still touch and go at 250, every thing else fails at 250... this sucks... no progress...

 

I did a double CMOS clear with power to the board, even did it while holding F1...then I reflashed the BIOS too.... Neo says this can help stability sometimes... no luck though...

 

I reseated my RAM and tried in slots 2 and 3 instead or 1 and 3.... interested... in slots 2 and 3 when I go into memtest and I'm at 250 or above I start getting lots of errors in memtest #2.... test #2 !!!! and nothing fixes it unless I back down the FSB.... I go to 245 it's OK...

NOW that's just like windows.... 245 OK... but 250 and 255 no way!!!! so in slots 2 and 3 that makes sense... memtest and windows based programs fail at the same or very close FSB's....

 

However, when I put the RAM back into slots 1 and 3.... memtest does NOT show any errors when in test #2... and passes memtest reasonably at 255 as previously described...

 

So... maybe using slots 1 and 3 is just masking the errors the RAM is producing above 245... cuz they are defineately there!!! and using slots 2 and 3 reports the errors more accurately... my best guess anyways...

 

any thoughts on this...

 

all I know is I pretty F**King pissed at the RAM..... Bull$hit! BH-5 my butt... can't beat 245 1T.. pfft... sucks

[kitfit1]

Like you i found that slots 2 and 3 allways work better than 1 and 3, even if Memtest passes in both configs it would fall over in Prime or S+M when using 1 and 3. I must admit i'm a bit surprised at your results with the BH5. Although i am surprised, in another way i'm not. If you look at the dating when peeps were running BH5 in large numbers, it was just before the NF4's came on the market. Both of us have got fairly new mobos and i'm thinking that maybe the newer LPB's arn't as good as the older ones(just a thought). As a side note, got my 4800 plats today and running at 265x8 with an XP1700 on board. That seems to be the max they will do in the LPB. I put them in the NF4 a couple of hours ago and there running at 306x9 1:1, with a bit more to come i think. So you can see what i ment when i said about the newer LPB's.

[Guest Darien]

Like you i found that slots 2 and 3 allways work better than 1 and 3, even if Memtest passes in both configs it would fall over in Prime or S+M when using 1 and 3. I must admit i'm a bit surprised at your results with the BH5. Although i am surprised, in another way i'm not. If you look at the dating when peeps were running BH5 in large numbers, it was just before the NF4's came on the market. Both of us have got fairly new mobos and i'm thinking that maybe the newer LPB's arn't as good as the older ones(just a thought). As a side note, got my 4800 plats today and running at 265x8 with an XP1700 on board. That seems to be the max they will do in the LPB. I put them in the NF4 a couple of hours ago and there running at 306x9 1:1, with a bit more to come i think. So you can see what i ment when i said about the newer LPB's.

Well... again... all these problems I had with the BH5, Mushkin Blue and TwinMos SP, I did NOT have with the OCZ 4000VX... which leads me to believe its the RAM...

 

furthermore... the RAM didn't run better in slots 2 and 3.... just showed errors where 1 and 3 didn't and more closely matched the results of windows for a particular FSB...

 

I still perfer slots 1 and 3... as good or better for me and better cooling.

 

as for 265... that's where both my TCC5 and my BH5 crap out at 2t... thats the boards max....

 

I put the BH5 back in at 265 x 10 2T 2-2-2-8, 12-15, 3-3-4-3-2-2-3.... in memtest #5

at 3.6v 1.9vdd... 1000's of errors per pass

at 3.6v 2.0vdd... 100's of errors per pass

at 3.6v 2.1vdd... 0 errors per pass

 

so... I have to boost Vdd to get 265, and that's as much as it's memory controller can handle...

 

btw... what the best opty 170 stepping? I heard it's CCB1E 0550 VPMW... that true.. which are the good ones that can hit 3.0+ GHz ?

[NEOAethyr]

265 is more likely the max potential of the bios you're using when in dual chan mode.

 

The hellfire 3eg, it's most likely this one's real max is 258, and if you managed to get more, it would be an extreme case, one that I would double check at diff times of the days and etc.

I got it to 258, wen I ended up making my 1st fenrir bios(aggressive long ago), ended up being allmost identical with hellfire's bios of all things.

 

I allways distanced my self from the other bios's because they allways made my old 280 unstable bigtime, the single chan 280(probaly was'nt even stable but as far as I was concered it was stable for anythign I did back in the day).

Whay I did'nt notice...

 

 

The F1 key lol...

You need a special cmos backup in that bank in order to full use that trick .

It's meant to load up the cmos reloaded backups, F1-F4.

 

Flashing the bios and clearing the cmos can help from time to time.

For the avg thing though, just sticking through it works out at the end.

That's what I have been doing for a litl while now, if it's not stable, I've been say screw it and keep trying wihtout doing the run around thing.

Normally if you are truly stable, 100% hardcore, it should be stable even if you got a bad cmos.

There's those cases were you're on the edge and a cleaed cmos and such is what is needed, but those edges, I'd rather have them slighty further away these days.

 

 

You knwo what really sucks?

I just woke up a little while ago, freaking hot as heck in my room, my rig was oof the whole time anyways.

I checked out my stability on s&m, in win2k3(just finally got it installd, took me 7 installs last night to get my shiz figued out for that os again), @ 266.

With tighter timings, ones I broke in the other day, maybe it was yesterday.

 

10-3-2-3, 9-13, cas3 cmd2, 0-0-2-2-0-0-3, 1.7v vdd, 2.8v vmem, 8x multi @ 1.45v vcore.

Blah.

Anyways unstable .

Damn I mean, this is so low level, and I fail it when my crap gets super hot.

I defently knwo it's the psu to blame, though on the other hand, these timings are quite diffrent then I'm used to so perhaps in a few I'll try somehting more normal.

 

Matter of fact, later on today or tonight(as long as I let everything heatup with the window closed).

I will test this config:

3-3-2-3 9-13, cas3 cmd2, 0-0-2-2-0-6-3, vmem 2.9v and vdd of 1.8v.

 

I know I can pass it even if my rig is super hot, I have gotten it to doit long long ago, and it was that way forever until the old pop spill thing.

 

 

About my psu's.

The enermax, even though it's not liek it used to be setup, I can still get a understnding of how it acted.

The +12v was low, real low, around 11.75v.

The +5v is super hi, 5.25v I guess, this may of been lower when I had it setup good(filtering, load etc).

The +3.3v I dn, don't have the wires no more for the moment.

 

Anyways on load, and when it's heated up to who knows what, that enermax does not fluctuate it's rails from when it's cold to hot, not one bit^^.

This ocz on the other hand, it's rails jump up, up maybe 2-3v when it gets hot.

 

I allready know that if the rails are diff then the norm, you have to re-breakin.

I get stable at the lower, optimal voltages I setup when it's cool.

Then it rises and screws me over, and I have to breakin again.

I have no udea, but I hope that once I breakin both, it's all good for th most part(maybe won't get max clocks whe it's hot, but I shoudl be getting decent ones).

But if it throws it out of wack everytime and never ends up breakin in, that sorta thing, man...

 

 

That ocz 520w is'nt as good as you think it is Darien.

We have to many people that are pushing for the enthueist theme here.

To much so, people are pushing the bs for that cash.

Just think about it, we need this, we need that.

This is out, why would you want that?, etc.

I'm only nor realizing this.

 

Hence those ocz's, it's that image.

People think they are getting great parts.

The truth of the matter is this psu really does sorta suck.

I still can get it to work the way I need it even with 2 psu's.

Make me wish I never cut off my atx off the4 enemrax.

Then gain, the adjustable rails got me this far, but the fluctuation is keeping me from keeping it stable when the going gets tough, it's really sickens me.

I want to burn some damned dvd's off my hd for once, I'm 2 months behind, I got both my 120 and my 160 totally filled up, I can't even capture video.

I'll have to wait till night time just to start burning :.

 

 

I was reccomended this psu by AG, on irc.

I figuered that guys reccomends enough psu's and such, and since he gets enough parts to work with, he should know what sort of psu does great and all for those newer platforms.

I kow the guy doe'snt do much for oc'ing and such, but that psu thing I figuered he'd know pretty well.

Of course he reccomended me the ocz, over the enermax's btw.

So I got one.

.....

It's either for that sell, or..., you get the idea.

No offence intended at all.

I was just hoping the guy knew psu's.

 

Now I'm stuck with this $200 psu, a christmas present, I wouuld of been better off with my enermax, 6600gt(which I could load clock either one or the other, better then it is now).

Buying new memory for christmas would of been better.

Blah, anyways these ocz's, check your rails at diff times of day dude.

[kitfit1]

btw... what the best opty 170 stepping? I heard it's CCB1E 0550 VPMW... that true.. which are the good ones that can hit 3.0+ GHz ?

 

Yes i've heard much the same. What i will say is have a look at some of the 06 steppings, some peeps at Overclockers UK forums are saying good things about them. As peeps have said in your NF4 thread, it's pretty much a crap shoot and unless you can either specify or choose your stepping from the retailer, it's pot luck anyway. I must admit i don't really think there is a bad stepping, most go to 2.6 on stock voltage, after that it's down to luck.