PDA

View Full Version : Need advice with my Mushkin BH5



ic3m4n2005
06-16-2005, 05:32 AM
OK Here we go :)

First of all my System: DFI NF4 SLI-D, Venice 3200+ (0516 GPAW), 2x512 MB Mushkin PC3500 LVL2 (old BH5), Windows 2k3

I`ve tested my cpu and it seems to be stable @ 2760 Mhz with 1,56 Volt (read out in Windows), tested with 166 Divider. After this i tested the rams, put the Voltage to 3,3 V (running off the 5V Line) and going further by 5 mhz. 255 seems to be cool, so i decided to go for 260, the first one or two times test 5 throws no errors, but then it startet, 3rd round 2 errors, 4th 20-30, then hundreds. Cranking up the voltage to 3,5 V solves this problem until 265 Mhz. First and second run of test #5 no errors, then crapping out starts again :rolleyes: . Really wired is that loosening the timings even makes it worse, e.g. Drive Strenght for bh mem should be 5 or above, mine seems to be best @ 3 or 4 or lowering RW Bypass doesnt help at all. After trying many things i got completly :confused: . Don´t know what to try next. Atm i am using bios 310, tried bios 510-2 (both, yellow an orange slots) seems to be same problem. Thought the problem is heat but theres a 120mm fan over the rams and the Heatspreader isn´t really hot. My goal is to get my sys 1:1 with FSB, so 270 or more would be let me say VERY NICE :) so i hope u guys can help me out. Would be best if we start from scratch, because after trying different settings my knwledge about it becomes this
:banana: :banana: :banana:

attachment: Timings i used last session (tref is auto, 7.8 us in memtest)

Jaco
06-16-2005, 05:47 AM
well , 264 Mhz is probably the max of your memory. maybe a few mhz more with some tweaking.
for 2*512MB sticks that is VERY good !! :toast:

you can try dram drive strength 6 or 8 . that's usually a good setting for BH-5.

ic3m4n2005
06-16-2005, 05:53 AM
I´ve tried high drive strenghts already, e.g. with 5 i got 2k errors in memtest #5, with 3 no error in the first pass of #5. Best would a kind of guide like kakarotos TCCD guide.

Aleman
06-16-2005, 06:23 AM
I got this settings for BH-5´s which were the best that worked for me.

2-2-8-2-7-14-0-2-1-2-3120-1

Auto-0-L8-L1-5.0ns-7.0ns-16-E-16x-7x-Disabled

ic3m4n2005
06-16-2005, 08:38 AM
Flashed to Bios 510-2, now its not stable with 250 @ 3,3 Volts :(
Don`t know whats going on here, passed 4 times test #5 then fifth time 32 errors, next pass 4 errors.

Lord_RTKK
06-16-2005, 03:12 PM
It seems to me a heat problem...
Are you cooling your vdimm mosfet?? Mine gets REALLY HOT when feeding it with 5v line...
I have measured the temp with an external temp probe and it have readed temps of 102ºC and so... so could it best as you can, to have stable volts for your ram.
If you are already cooling well the mosfet, what PSU are you using??

ic3m4n2005
06-17-2005, 02:57 AM
Thats another confusing thing, the 120 mm Fan blowing onto the rams is cooling everything, even the mosfets and the northbridge. Now the wired point is tha the mosfet in the upper right corner is warm to touch. PWMIC readout in Windows says around40 °C.
PSU is OCZ Powerstreram 600 W.

Lord_RTKK
06-17-2005, 03:16 AM
My PWMIC reads around 40 too, but putting a thermal probe between heatsink and mobo, touching the side of the mosfet, gave me temps of 100+ºC, when feeding the mosfet with 5v line, so PWMIC reads a "little bit bad"... :D :D
When I feed the ram with 3.3v, the mosfet stands in 30 ºC or so... ;)
The PSU that you got is a good one, so it is not the problem. I am running with a Enermax Noisetaker 600W and runs fine too...

ic3m4n2005
06-17-2005, 05:05 AM
Does ur Heatsink get warm ? Because mine seems to be like the 40 degrees itsays :)

Anyone something else it could try ? Or how i could proceed in testing timings etc...

krampak
06-18-2005, 02:44 AM
I have the same memory as you, I can't pass 260Mhz stable but at 260 I got about 200 errors at every loop since I change my 8cm fan (about 40cfm) for my vantec tornado 9cm at 110cfm, then errors went to 0, but 261 still not stable.
And another thing! I get less errors at cas 1'5 than at cas 2 :confused: now I need to play with dram strenght and that timings, maybe those are the most important ones to have it stable.


PD: I get thousands of errors with yellow slots and 0 errors with orange slots at 250 2-2-2 3'35v. (260 needs 3'55v)
I've properly cooled the mosfets and of course, the memory. using bios 510-2.

By now this timings works the best for me (in 1MB superpi):
http://img149.echo.cx/img149/9687/timings2cr.th.jpg (http://img149.echo.cx/my.php?image=timings2cr.jpg)

ic3m4n2005
06-18-2005, 03:05 AM
THX krampak, seems to be same problem i`ve got to fight with :) will try ur timings, what value has tref in bios since a64 tweaker shows different value.

Edit: your values for drive strenght and data drive strenght ?? 3 or 4 seems to work best for me, have u got similar values ?

coop
06-18-2005, 08:08 AM
Here are some settings I used with my winch. 3500 when I had the same rams as you. 267 and 270, http://www.xtremesystems.org/forums/showthread.php?p=772077#post772077

ic3m4n2005
06-22-2005, 07:13 AM
After being unhappy all the time with my Mushkins i decided to do a little test. First check is a few loops of test #5 with 3,3 volts @ 260 Mhz
=> pass(errors): 1(0), 2(14), 3(256), 4(660), 5(1400), abort by myself after this. Then i tried to boost voltage. Again test #5 with 3,6 volts @ 260
=> i´ve only run one test since it gave me 20000 errors.

The Mems became really warm, but i could touch them without harm.
So what would u guys say, heat problem or bad rams ?? Anything what i could try ? btw there`s a 120 mm blowing onto the rams. Below the settings i used during the test.

G H Z
06-22-2005, 09:52 AM
CPU cooling effects Ram clocks when you get to a certian point. I have the same Ram and past 267 I needed phase cooling to go further with them. First thing I would try is more voltage. Over 260 is 3.65v territory for me.

And another thing I found is that test 5 doesn't cut it, you need to loop 7 & 8 as well as 5 with this mem and board.

ic3m4n2005
06-22-2005, 10:36 AM
thx for info, for memtest i have to get rid of errors in test 5 first then i can check others :D . CPU is not getting too hot i think, max is 36°C @ 10x260 (stable till 2760). A little bit wired is that with more voltage more errors apear.

THunDA
06-22-2005, 11:26 AM
You can try Preamble @ 5.5ns sometimes that helps with stability without much of a performance drop..

edit..

Have you tried each stick seperatly to see if one is holding you back ?

coop
06-22-2005, 03:04 PM
Once you get to the point of more errors/volts, your just beating a dead horse. Your memory timings are not "proper" if you will. No one ever said it would run with those settings. For the rams sake, you should not need over 3.6v for only 260, especially for just memtest.Try setting your tras to 5 at least and 6 could be better, and the trrd to 2 and Trtw to 2 also. If you try forcing them to run under conditions they won't you can volt them up to 4.8 or whatever until they stop. Also Idle Cycle Limit plays a role in stability at the very top so put something in there.

ic3m4n2005
06-22-2005, 03:29 PM
OK thanx so far guys, i will try that tomorrow. Hope to get those babys stable soon :)

ic3m4n2005
06-24-2005, 05:17 PM
I found the time to gather some informations :)
tested only one stick so far, but here we go ...
Bios: 510-2 FIX
Settings: tcl 2, trcd 2, tras 6, trp2; bank interleave disabled, Max Async Latency 7, dram response normal, read preamble 5, idlecycle limit 128, R/W Queue Bypass 16x, Bypass max 7, 32 byte gran. disabled, rest automatic

The test: orange slot (far from CPU), single stick, short check of 10 passes of test #5; started with 3,3v dram, 1,2v ldt and 1,5v nb; MHZ: passnumber(number of errors in this pass, not added)

255: 10 Passes clean, going further.
260: 3(4), 6(4), 7(2), 9(4)
->set nb to 1,6v
260: 4(4), 5(3), 8(5), 9(4)
->set nb to 1,7v
260: 4(2), 5(6), 6(6), 7(4), 8(2), 9(2)
-> set nb to 1,8v
260: 4(2), 9(4)
->set nb to 1,8v & ldt to 1,3v
260: 5(2), 6(4), 7(4), 8(4), 9(4)
->set nb to 1,8v & ldt to 1,4v
260: 1(2), 4(4), 6(2), 7(2), 9(2)
->set nb to 1,8v & ldt to 1,5v
260: 7(4), 9(6)
->set nb to 1,5v & ldt to 1,5v
260: 4(1), 5(2), 8(6)
258: 18 passes without errors, then abort by me
->swithing to yellow slot (farest from cpu)
258: 5(4), 9(2), then abort by me, because :woot: (yellow should be better ?)

Hope this is enough information for you to give some advise. Post everything, sometimes even simple things help a lot. Tomorrow i am going to check timings of this one and the other stick. Thanks in advance

coop
06-24-2005, 05:31 PM
Orange on mine are better. 510-2 That is not bad 255= up to 260 with only 3.3v. Enable the =.03v under your vdimm and see how much that helps. Also see what changing your MAL from 7 to 8 does just for kicks. If you are going to run that fast though you most likely will nee 3.4+ the .03. Just put a fan on them. I was running SuperPi on mine last night @270 and found that the orange was better, but I needed to have 3.6 + the .03. Also my Mushkin 3500 needs to be in the top slot and the xms3500 needs to be in the lower one. They are my best rams so I use them together. ADDED** you will notice that in Windows you will need more voltage or lower clock to be A-ok.

ic3m4n2005
06-25-2005, 03:29 AM
My 120mm Fan is maybe 5-10 cm above the Rams, would it be better if the Fan is directly attached to the Ram or is it good enough ?? Ram really getting hot with 3,3v :)

ic3m4n2005
06-25-2005, 07:07 AM
Today i made a shot test of stick no. 2. For the test i used the same settings as yesterday. I only tested wether its stable @258 or not. 20 Passes of test #5 only 10 errors, so stick one is the better one. After this i pushed the "bad" stick in the orange slot near the cpu the other one in orange slot above, same setttings and 255 Mhz. But then something really :censored: happens. I got 43000 errors in the first pass :shock2: . I tried to enable Bank interleaving an got 48000 errors :shocked: .

So please save from going crazy with this. I had no Problems like this with my old corsair ram (2x256 bh5). The more i try even more errors i get.

coop
06-25-2005, 08:52 AM
Fan should be fine. The 510-2 bios may work better for you as well. Set Read Preamble to 5.5. Try different slots to see if it helps, but sounds like that one stick is not so good. Just work with the lesser stick for now. Enable the +.03v option. Try it in all slots. You want it in the best one. Also you don't want any errors. You may need to lower your speed you are running them at. You don't need to run them 1:1 (at200) so it is not the end of the world. Just get them working as good as possible for now. Running both sticks requires more effort on the part of the memory controller thus you need a little less speed or more power,- provided you still have no errors. You can run a 10x multi with your cpu and a 5/6 divider and do fine. I don't believe upping the ldt or chipset voltage more than .1v from stock is helpful. These speeds your working at could benifit from testing with 3.4v and the +.03v option enabled.

coop
06-25-2005, 09:23 AM
Instead of working from the top down, set your settings like discussed above and set your voltage, (3.4v +.03 opt.). Set ram speed to 240 and work up from there. Once you get an error keep settings the same and try different slots to see if it helps. Go back down in speed until the error is gone. Add the second "better stick to the other same color slot. This is about as good as it is, see how it does. If it gives you errors try going up .1v. If it don't help go back down in speed. Depending on your cooling a 5/6 divider will privide plenty room for oc'ing.

ic3m4n2005
06-25-2005, 02:30 PM
for divider its enough, i need only 255-260 with 9/10 divider from dfi board. But 1:1 is way faster i think, have compared scores from my old winchester (runs only 2580-2600) and with divider my new cpu isn`t really better even with 2700+ mhz.

coop
06-25-2005, 02:49 PM
I was doing divider testing the other night because I thought 1:1 was faster as well (even though theoretically it can't be). To my surprise running at 2700 I did some 1m Pi test a couple nights ago with 5/6 divider and found that 9x@300 was fastest, 30.128. Then 10x@270 was next at 30.492 and finally 11x@245 was at 31.087. I was surprised to find those results. I did get better results with my Win. as well with 1:1. That is why I did not bother to try different settings with this chip I have now. Why the diff? I don't know.

ic3m4n2005
06-25-2005, 03:17 PM
A few minutes ago i tried work on this mem again, but its going worse. MAL 8 & preamble 5,5, everything else at loosen timings, too. I know that Ram is stable at 250 with 3,3v (tested by the person i bought it from, on dfi Nforce2 and by myself, too), so i started at 250. First round of test #5 gave me 1800 errors. So what the f**k is going on here ??? After raising ldt & nb voltage 0,1v ive "only" 700 errors. I`am at the point of throwing the hole computer out of the window :(

Edit: this time a was using Hellfire Bios, any recommendations on bios version ?

Edit2: 250 seems to be stable (13 passes/10 minutes of test #5) trickwas drive strenght 5 data drive strenght 1. Apart from this hints or recommendations are welcome :)

ic3m4n2005
06-28-2005, 03:09 PM
Some news:

seems to be that lower drive strenght works better, like 3 > 4 > 5. Whats going on with my rams, thought higher numbers for that would be better for bh or ch ram.

Edit: And another problem apeared, 3,5v +0.3v option seems to be nearly fine for 268. I decided to put 3,6 on it and expected less errors, but i got thousands :mad: Should i try to burn them in ? I don´t know if they are already. Only thing i know is that the last owner was putting them on nf2 mobo with max 3,3v, don´t know if he has thrown 3,3v on rams all the time.