The Corsair User Forums  

Go Back   The Corsair User Forums > Corsair Product Discussion > Memory

Reply
 
Thread Tools Search this Thread Rate Thread Display Modes
  #61  
Old 10-28-2004, 07:49 PM
Zoltan Zoltan is offline
Registered User
 
Join Date: Oct 2004
Posts: 5
POST ID # = 151348
Zoltan Reputation: 10
Default

yeah, true.

I just got all the new stuff this monday, and have been trying to get it stable for the last few days with very little sleep. Its so crushing when you just bought new stuff and it dosnt work. Atleast I could install the OS now with the 2-3-2-5 setting and let out my aggression on some poor sools in Call Of Duty :)
Reply With Quote


  #62  
Old 10-28-2004, 09:41 PM
herman's Avatar
herman herman is offline
Registered User
 
Join Date: Dec 2003
Posts: 60
POST ID # = 151360
herman Reputation: 10
Default

Ya... I had to do the same thing.

I hate it when after you install the OS you discover the memory (or something affecting it) is bad.

You know you have to install the OS again fresh.

memtest86 first from now on.
Reply With Quote


  #63  
Old 10-29-2004, 08:39 AM
Walker Walker is offline
Registered User
 
Join Date: Oct 2004
Posts: 11
POST ID # = 151405
Walker Reputation: 10
Default

Hello I'm sorry to say I'm in the same boat as others here. I seem to have stability problems at 2-2-2-5 @200mhz If I lax latency or lower to 166mhz it seems to run stable I've run memtest86 V 3.1 With memory in many differnt configurations with with memory in slots 1/3 I would get 15000+ on the first pass on test 5 with memory in slots 1/2 I still get errors usually only 2-4 on the first pass but continued running only brings more fails per pass I have had the voltage at 1.75 even tried 1.80 with no help I can be lax on the timing and be ok but I clearly paid for 400mhz 2-2-2-5 My set up is as follows

Athlon64 3000+ Skt 939 90nm Winchester
MSI K8N Neo2 PT Bios 1.3
1GB XMS3200XL TWiNX 2x512
Radeon 9800Pro
2x WD Raptor 36Gb
Antec NeoPower 480
XP Pro SP2
Please feel free to contact me for detailed system specs and error reports
Reply With Quote


  #64  
Old 10-29-2004, 08:51 AM
Wired's Avatar
Wired Wired is offline
Administrator
Wired's PC Specs
 
Join Date: Aug 2000
Posts: 32,184
POST ID # = 151412
Wired Reputation: 32
Default

Quote:
Originally Posted by Walker
Hello I'm sorry to say I'm in the same boat as others here. I seem to have stability problems at 2-2-2-5 @200mhz If I lax latency or lower to 166mhz it seems to run stable I've run memtest86 V 3.1 With memory in many differnt configurations with with memory in slots 1/3 I would get 15000+ on the first pass on test 5 with memory in slots 1/2 I still get errors usually only 2-4 on the first pass but continued running only brings more fails per pass I have had the voltage at 1.75 even tried 1.80 with no help I can be lax on the timing and be ok but I clearly paid for 400mhz 2-2-2-5 My set up is as follows

Athlon64 3000+ Skt 939 90nm Winchester
MSI K8N Neo2 PT Bios 1.3
1GB XMS3200XL TWiNX 2x512
Radeon 9800Pro
2x WD Raptor 36Gb
Antec NeoPower 480
XP Pro SP2
Please feel free to contact me for detailed system specs and error reports
FYI, it MAY be because you have one of the new 90 nm cores, lots of people have had to update their BIOS to be fully compatible. May want to check with MSI to see if they have a beta BIOS for it.
__________________
No, I don't work for Corsair
...you need to flow like a river...
Forum Rules
Admin of this forum / webdesignforums.net / Petri's IT Forum / The Corsair Memory Steam Group
Founder/Creator/Admin of ZE SECRET PROJECT (Coming Soon)
Reply With Quote


  #65  
Old 10-29-2004, 05:59 PM
Walker Walker is offline
Registered User
 
Join Date: Oct 2004
Posts: 11
POST ID # = 151593
Walker Reputation: 10
Default

I dont see how it could be the cpu it passes memtest with flying colors at 333 but at 400mhz it has errors I'm willing to give this some time in regards to the memory and the mobo but if it could be a weak cpu I purchased it oem and only have 30 days to rma it how can i narrow this problem down?? If i get errors on test 5 at 400mhz does this point to a weak or failing stick? I paid for 400mhz at 2-2-2-5 and if it creating errors in memtest at this speed I would point to the memory not living up to its advertised speed.
Reply With Quote


  #66  
Old 10-29-2004, 07:10 PM
herman's Avatar
herman herman is offline
Registered User
 
Join Date: Dec 2003
Posts: 60
POST ID # = 151619
herman Reputation: 10
Default

There are a lot of reasons in theory that a new versions of parts like mem controller, a new core, new chipset, or even cards, could be a problem.

It is still up to the bios to basically "program" these components when the system starts, where there are all kinds of variables. A new mem controller might have new or changed options or otherwise need different settings to work correctly (intentional in the design or otherwise).

It kind of works basically like this: power on, cpu goes into 16bit mode, starts running bios code, bios then does about a billion things, bios runs small boot loader chain on drive, boot loaders start OS which changes to 32/64 bit mode, then happiness. Somewhere in the "billion things" part the bios has to do stuff like configure the memory controller (which is much more than the settings we normally see), map cards into memory, set aside memory for various things/reasons like agp, interrupt vectors, etc... and all of this has to be done with intimate knowedge of the memory, mem controller, and all the various components.

Timing is a serious problem that gets easily screwed. The first thing one of the Ram Guys might be inclined to think is that the tolerances on the 90nm parts are just different and the bios needs to account for things differently. This is a real possiblity. Its also just as possible that it is choosing to set things up so that memory is read or written at what has become a bad time considering new cache, AGP, HT, or interupt timings, or that the memory map (what goes where) has changed or was wrong before and just shows up on the new 90nm.

This stuff may just not show up at 333 or whatever, but at spec it would. Even the mem map, where memtest never collided with memory written on an interrupt service routine before but now does becuase the mem map is/was wrong.

----------

So blah blah blah....

One thing to notice is that memtest86 sets aside large areas to not even test. It does this to avoid some of what I was saying, areas that it shouldn't read or write.

This is one of the only real flaws/limitations of memtest86. That the most critical areas of memory in the lower regions are not ever tested. It sucks.

Last edited by herman; 10-29-2004 at 07:35 PM.
Reply With Quote


  #67  
Old 10-29-2004, 07:26 PM
RAM GUY's Avatar
RAM GUY RAM GUY is offline
Corsair Product Guru
 
Join Date: Apr 2001
Location: Fremont, CA
Posts: 118,269
POST ID # = 151624
RAM GUY Reputation: 10
Default

Well errors in test 5 might suggest some other problem. If you have a bad or failing module it will fail at any speed or setting you have it set too, normally on test 2-4-6 at a specific address. I would for the time being just run it at a slower setting like Cass 2-3-2-6 and set the Dim Voltage to 2.8 Volts to see if that does not solve the problem, but from what I have seen it looks like it may just take some time for the bios to mature a bit more!
If you just want to try replacing your modules, please follow the link in my signature “I think I have a bad part!” and we will be happy to replace them or it!
__________________
Support accounts and tickets can be created at https://support.corsair.com.
Reply With Quote


  #68  
Old 10-29-2004, 07:54 PM
herman's Avatar
herman herman is offline
Registered User
 
Join Date: Dec 2003
Posts: 60
POST ID # = 151641
herman Reputation: 10
Default

Ya. If it is a purely bad stick with totally bad board, bad addressing, bad cell/row/column, etc... then it should fail at any pretty much any speed.

Sticks can definatly not operate correctly at spec, but work at lower speeds. If you consider spec to be a part of it working correctly, then I would say it is a bad stick. If not, then I guess you could say it was just not spec'd correctly.

A cheesy example would be one capacitor at a certain temp on one chip may just not discharge fast enough, thus fails at a certain speed, and lowering the speed or raising the timing params (spacing) it starts working again. Its not that the capacitor is bad, or the module is bad, its just not that fast. This may never (or take a really long time) to show up on a A7N8X, but not very long on a neo2; it may take not very long to find at 50C but almost forever if ever to find at 45C.

The beauty of test five is that it hauls ***, i.e. is much more in line with the processors natural best case, and increases the chance of actually taking advantage of the low timings (spacings). This means more heat and more often really doing 2225 vs. 2325+. This is particularly true on A64's. Prime95 is even faster (but more random), why it finds stuff so fast sometimes.

I think more and more things seem like timing issues these days. The manufacturers and you guys are doing great at catching blatantly bad chips and modules. That is why I love test 5. Timing issues get very complicated, so many ways it can happen in a PC and look like bad memory wether is is or not.

I was just saying in the last post *why* it could be the bios might need to be updated, since he asked.

Last edited by herman; 10-29-2004 at 09:14 PM.
Reply With Quote


  #69  
Old 10-29-2004, 10:00 PM
Walker Walker is offline
Registered User
 
Join Date: Oct 2004
Posts: 11
POST ID # = 151647
Walker Reputation: 10
Default

Looking in to answer to this problem I found this quote on MSI forum:

"The memory controller of AMD64 does not officially support 400mhz speed on two doublesided sticks of RAM, or on 3 sticks of RAM (regardless of single/double side), however this board seems to be very tolerant concerning this, change your ram setting from auto to 200, and see if it helps - if not, try with other RAM, there's no guarantee it will be able to function at 400MHz.
Please note that this is a limitiation in the AMD 64 CPU's memory-controller, and not a limit in the motherboard!"


I have asked amd as to what the official specs are for the on board memory controller with no answer yet. I'm still not completely sure the cpu is to blame though I can run endless cpu tests and i get no errors if I run memtest I get memory errors if its true memtest does not test all the ram that scares me even more if it finds errors only testing limited amounts of ram there is only the possibilty that more errors are there if it was completly tested. This memory is supposed to be 2-2-2-5 This link shows that there is no * for the XMS3200LLPro for amd or intel platforms http://www.corsairmemory.com/corsair/xms.html
And here is the link for the specs of my ram
http://www.corsairmemory.com/corsair...-3200xlpro.pdf
One spec i dont see anywhere is, are these single or double sided? If its the cpu then I'll have live with that. But why wouldnt the memory pass the tests at its specd speeds?? I dont have another computer that can test these sticks at 400mhz 2-2-2-5 so its hard for me to know for sure Whats going on. You guys want to blame MSI, MSI wants to blame AMD, and its yet to be found what amd has to say. For some reason I'm guessing that amd will point me back at Corsair(wouldnt that complete the circle of blame the other guy) I know one thing for sure I should have did some more homework before I put this system together.
Reply With Quote


  #70  
Old 10-30-2004, 12:18 AM
herman's Avatar
herman herman is offline
Registered User
 
Join Date: Dec 2003
Posts: 60
POST ID # = 151657
herman Reputation: 10
Default

Ya...

I dont know if your mem is officially "double sided" either. There are chips on both sides, eight per side, but that may not equate to actually being organized as double sided. I honestly cant remember if a "bank" is the same as a "side". But I do know that during post the bios shows each stick taking two banks or whatever.

Also its not clear if they are talking about just a single channel or not. In my case I am using one stick per channel, not two on one channel. I am thinking that it would be silly to not be able to run at least one stick per channel at 400 DDR.

What I heard:

I had already heard that two "double sided" on one channel for the A64 would cause reduction of speed to 333 DDR to be officially supported. Maybe thats what they are talking about. I think its that the controller can only do two banks/sides total at 400 DDR per channel. This is illustrated somewhat in the MSI manual.

I also heard that if you get currently made 1 GB sticks it has the same effect, in that you have to lower to 333 and/or maybe use other loose timings. Maybe a 1 GB stick is really just like two 512, where is it realy organized as *4* sides/banks.

-----

Probably we both need to find out the real deal.

I don't think any of this is going away; it's the new world order as they wring out the last bits of speed from current paradigm. So this will be good to know.

Last edited by herman; 10-30-2004 at 12:23 AM.
Reply With Quote


  #71  
Old 10-30-2004, 12:44 AM
herman's Avatar
herman herman is offline
Registered User
 
Join Date: Dec 2003
Posts: 60
POST ID # = 151660
herman Reputation: 10
Default

********* MSI Forum *********

Bios 1.4 is out there somewhere.....

From poster in MSI forum (they have neo, single channel, not neo2).
They evidently ran two XL 512s on one channel @ 2225 1T 400 DDR.

-------------------------
Corsair TWINX1024-3200XLPRO is extremely flexible!!!

I use DDR slots 1 and 2 (overclocks better than 1 and 3)
BIOS 1.1 works best so far. (have tried 1.2xx, 1.3xx, 1.4xx)
BIOS 1.3 is second best. Hopefully the official 1.4 will replace my 1.1

UPDATE BIOS 1.4 beta 3 ROCKS!!!!! UPDATE

All configs below are 100% stable over 48 hours stress testing.

For tight timings it can handle:
1T, 2-2-2-5, 200MHz 1:1, 5xHTT (BIOS 1.1)
****************************************************

Another using neo2 where they ran LLs in dual, which work great for me too.

----------------------
Well so far, using Memtest86+ v 1.20 my 2 x 512 Corsair 3200 xmsLLPT cas2 in dimms 1&2 @dual 200 has been running without any errors for 8.5 hrs, so (touch wood) its looking good.

AMD 64 3800+ Zalman CNPS7000B-Cu
K8N Neo 2 Platinum
2 x 512 Corsair 3200 xmsLLPT cas2
**************************************************

Hey can I get LLs for my XLs like this guy did for his C2s? :-)
Just kidding.

-------------------------------------------------------
1GB Corsair TwinX1024-3200C2PT

That RAM had a bad chip, it was returned and replaced with Corsair TwinX PC3200LLPT running 2-3-2-11 at 2.75 volts. Stock speeds, I don't overclock.

Super stable, runs Prime95 all night long no problem.

Gets about 6,150 MB/s int and 6,100 MB/s float in Sandra SP2b memory bandwidth test.

MSI K8N Neo2
Athlon64 3800+
Corsair TwinX 1024MB PC3200LLPT 2-3-2-11 @ 2.75v
****************************************************

XL Pro and Neo2
running bios 1.3 like me.
----------------------------------------------------------

got it running in 400mhz dual channel at 2-2-2-5-1T @2.75v (as per Corsair recommendation)

My god this system screams! - Had allsorts of memory trouble with the Gigabyte Socket939 K8NSNXP - Never buy gigabyte again - it would only run my memory at 333mhz

Keep up the gud work MSI :>

AMDFX-53 Skt939 on MSI K8N Neo2 Platinum (Bios 1.3)
1Gb Corsair TwinX-3200XLPRO (2-2-2-5-1T)
****************************************************

Guy doesn't say what board, but has to be neo2 if dual channel, as this was AMD64 nForce section.
-----------------------------------------------------

Corsair TwinX - DDR3200XLPT @ 2,2,2,5 (stock) 1T in channels 1 and 2

No problems after (checks watch), hmmm...4.5 days.
****************************************************

This guys clams they are double sided and has two running running per SPD and 400 DDR on one channel fine.
-----------------------------------------------------

two 512 Mb sticks of DoubleSided Corsair PC3200 XL's no problem here
at stock speed.

have to try higher speeds though.

K8N Neo (non Platinum)
A64 3400 Newcastle with Zalman CNPS7000A Cu+ AS5
2*512 Corsair PC3200 XL's
****************************************************

Fun thread is the stuff that doesn't work section:

---------------------------------------------------------------
Neo2 here.....s939

Corsair TwinX1024-3200XLPRO absolutely hates this mobo.

215x11 1:1 5x 2-2-2-10 is MAX w/memory 1:1.......second Neo2 board as well......first one did the same thing.

Gonna try some Crucial Ballistix PC4000 Friday.
---------------------------------------------------------------

My Corsair 1GB TWINXMS3200XL cannot run at CAS 2.0-2-2-5 1T (Losk up every now and again in Windows) on my MSI K8N Neo2 Platinum

I have the memory running at CAS 3.0-4-4-11 2T and will work my way doen to CAS 2.0-2-2-5 1T and see where it bails?

Last edited by herman; 10-30-2004 at 01:12 AM.
Reply With Quote


  #72  
Old 10-30-2004, 01:11 AM
Walker Walker is offline
Registered User
 
Join Date: Oct 2004
Posts: 11
POST ID # = 151662
Walker Reputation: 10
Default

Here is a link for the AMD Athlon 64 Whitepapers
http://www.amd.com/us-en/assets/cont...docs/24659.PDF
It says "Low latency high bandwith" on board memory controller
it also says that it supports 200mhz. It dosnt say anything about single or double sided, and it dosnt say anything about latency. I've been running more tests @200Mhz 2-2-2-5 with voltages set as follows
CPU 1.4 Bios is set to 1.5 but cpuz says 1.4
Memory 2.75
AGP 1.6
With this setup I can get thru 1 pass error free of tests 1-7
With test 5 on constant loop I get 2-10 errors by pass 3 I had 40 errors by the 8th pass.
On test 8 I get multiple errors on pass 1 the error code looked like this
Test #8 Pass #1 Failing adress 0001a7f0db0 495.0 MB Good 00000000 Bad 00000100 Error bits 00000100 Count Chan 1
Reply With Quote


  #73  
Old 10-30-2004, 01:16 AM
herman's Avatar
herman herman is offline
Registered User
 
Join Date: Dec 2003
Posts: 60
POST ID # = 151663
herman Reputation: 10
Default

Ya thats what I see. At least as far as test five by itself getting it hot enough before it starts blowing. Once it gets hot enough, it more or less keeps failing worse and worse.

Let it cool down and starts over. More and more I think the A64 is just working it really hard.

Put an ice cube on it and see if it keeps passing. (Just kidding, dont do it.)

---back of board----
....stuff stuff stuff...
............CPU 1234T. .... T is termination.
..................GPGP... .... Green and Purple
...........power.........
---front of board----

The MSI stupid BTX layout attempt sux. They tried for meeting BTX layout and overclocking, which was rumored to be a big pain for A64 with the memory controller on the CPU. They alternated channel slots to keep each matching channel slot as close together as possible to keep the traces as as equal as possible to the CPU and to the termination, and close to the CPU as possible.

This causes us to fill 1&2 or 3&4, squeezed right next to each other, and placed away from the big fan flow. Where if you have Zalman 7000 the 1&2 is right next to the flower, or 3&4 where it is more out of the flow of the big fan. And aligned with the right edge makes them against the grain if you have a PSU fan pulling up.

Normally this shouldn't be a problem, but this mem gets really hot. That and being double sided puts more of a strain.
----------

I have a Fluke multimeter with a temprature probe. I'm tempted to see how hot this gets between the sticks. And I want Corsair to make a statement about max temp.

If its realy just temp then there is this.... The other problem are these heat spreaders. I almost guaentee that they hurt not help with heat and are more to keep grubby fingers from causing ESD. These are stamped and painted, which is horrible for dissipation. If they are going to do it, they need to be aluminum (better yet machined not stamped) and either anodized black or brushed (for texturing) for PT, which over many years are well proven strategies. Maybe time for Corsair to rethink spreader design; as they push it and A64 dual hits critical mass, this is just going to get worse, if it really is a heat problem for us.

Last edited by herman; 10-30-2004 at 02:02 AM.
Reply With Quote


  #74  
Old 10-30-2004, 01:54 AM
Walker Walker is offline
Registered User
 
Join Date: Oct 2004
Posts: 11
POST ID # = 151670
Walker Reputation: 10
Default

Well I've been tweaking voltages some more with the cpu set at 1.55 cpuz is still only showing 1.40v The specs say 1.5 volt cpu I'm not sure if this could be the culprit. with memory at 1.80v and agp at 1.60 I get thru 8 passes of test 5 with no error then on pass 9 I start getting errors 2+ at a time. Voltage seems to be the most sensitive setting in regards to errors and the motherboard just dont seem to be getting the cpu to 1.5v which leaves me wondering if the memory is suffering the same problem? How can i check Ram voltage?
Reply With Quote


  #75  
Old 10-30-2004, 02:19 AM
herman's Avatar
herman herman is offline
Registered User
 
Join Date: Dec 2003
Posts: 60
POST ID # = 151674
herman Reputation: 10
Default

You CANNOT trust the on board voltage readings.

You might just go get a nice $20-$30 multimeter from radio shack.

If you test the major 3 under load (try power torture on prime95), the you eliminate the PSU. I've never seen regulation of the board go out without the board just completly freaking out, so worry about that last. Real ram voltage is hard/scary to test. The rest is easy.


How:

The only ones that matter are 5 12 and 3.3. 5 and 12 can be tested at a molex power plug. Stick ground probe first in black wire hole, doesn't matter which. Then stick in ether red or yellow, yellow is 12v, red 5v.

For 3.3 you can use the ATX power plug, where i think orange is 3.3. Plenty of places on net to show how, but you basically look at pinout and stick the probes, always ground first, into the back of the plug until makes contact.


Why:

You CANNOT trust the on board voltage readings. They vary form board to board. Most of the time the readings are made *before* the power goes through any onboard regulation curcuitry, and are produced using cheap uncalibrated multifunction chips. They are awful. Winbond was very popular for a while, not high class stuff.

They capacitors and other components of the readers very wildly for consitancy in manufacturing. The normal solution is to calibrate them, but they can't calibrate each chip for each finction because it would be way to expensive. They should be used only to see if stuff is way way off.
Reply With Quote


Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT -4. The time now is 04:25 AM.


Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2019, vBulletin Solutions, Inc.