The Corsair User Forums  

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

Reply
 
Thread Tools Search this Thread Rate Thread Display Modes
  #31  
Old 10-08-2004, 07:41 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 # = 147867
RAM GUY Reputation: 10
Default

I would get them both replaced!
__________________
Support accounts and tickets can be created at https://support.corsair.com.
Reply With Quote


  #32  
Old 10-11-2004, 01:10 AM
cimerol cimerol is offline
Registered User
 
Join Date: Jul 2004
Posts: 42
POST ID # = 148105
cimerol Reputation: 10
Default

I would RMA the set and get a brand new TWINX set. This board runs 2-2-2-5-1T @ 2.75v easily, so you must have a bad module.
__________________
MSI K8N NEO2 Platinum-s939
AMD FX-53
ThermalTake Pipe 101 HS w/VAntec Tornado Fan
Artic Silver 5 Thermal Paste
TwinX1024-3200XLPRO 2-2-2-5-1T @ 2.75v
PNY 6800 Ultra
Creative Audigy 2 ZS
2 WD 74 Raptors SATA RAID0
Enermax 600W PSU
ThermalTake XaserIII Super Series Case
VERY STABLE!
Reply With Quote


  #33  
Old 10-11-2004, 12:53 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 # = 148181
RAM GUY Reputation: 10
Default

I would tend too agree, or maybe some other issue, but more it does sound like you just have a weak module!
__________________
Support accounts and tickets can be created at https://support.corsair.com.
Reply With Quote


  #34  
Old 10-12-2004, 11:49 AM
herman's Avatar
herman herman is offline
Registered User
 
Join Date: Dec 2003
Posts: 60
POST ID # = 148421
herman Reputation: 10
Default

Keep in mind that they may not test the ram on as fast a board as the neo2 or even a AMD 64 at all, which can make a huge difference, especially for the tests that really push the memory like test 5.

I can tell you that I have twin 512 3200XL as well and they fail on my neo2 3500+, but pass on my asus A7N8X Deluxe with a 2500+ clocked at 2.0 and 400. I also have Twin 512 3200LL which pass on both boards.

It would be really easy for ram to pass on the slower A7N8X and thus pass Corsair QA, but still fail for us.

Lastly, the slots on the neo2 are really close together and these XL sticks get much hotter thean the LL. They dont feel absurdly hot, but it these memories are weak anyway, it probably wouldn't take much. For me the ram never fails the first time if mem is cold (yes RG, I have a modern well ventilated case and tried your stated voltages explicitly).

In my case, they only fail in dual but I know its one of the two sticks, because if I swap channels the errors are basically the same offset by 2MB. When warming up (second pass if run only test 5 and mem cold) I get failure on about 20, 40, 560, 6XX, and after heat plains out they fail at 560, 6XX only. Swap them and the errors are offset by about 2MB.

-----

Fun info:

I just bought mine like less than a week ago from new egg. I'm kind of pissed becuase the same thing happened when I bought the LL, and took me a month to figure out what was wrong with the system. I'm much more educated now. I would blame myself for everything, but it was just the memory was bad. Ever since I have had the new pair of LL (I had to RMA too), they have never failed, which was what I thought I was buying with all the expense in the first place. Once you get good sticks they rule.

When I bought the LL, the official DDR 400 nforce2 hadn't really been out a couple of months. After two mother boards (both A7N8X official DDR 400, second was the next rev asus) and discovering memtest86, I finally realized it was the memory. Months later and after the RMA I got good sticks. I guess I never thought in a million years it would be the Corsair memory.

Either packaging or shipping is ESDing the sticks and killing them, or they are just testing them in what they think is most popular and hedging their bets. This as opposed to requireing all sticks meet the specs of the harshest environments which would cut down on the number of sellable sticks.

Just some speculation, but I have to say after the second time I'm pretty disenchanted too.


It is important to remember that these are out of spec parts to begin with. This is one of the original ideas of XMS, knowing that there are many modules (chips) that can run better that the spec. In this case these are really like 500s or 550s that are programmed as clocked down 400s to get lower latencies. This is an age old trick and not necessarily a bad thing, but you have to re-test if you sell them that way (as opposed to me taking a part sold as say 400 and using it to get CAS2 on a 333 bus myself judging by the individual chip specs). The problem is that by the nature of a qualitative test only certain cercumstances are guarenteed. In my case XL on my 2500+ A7N8X clocked at 2.0 and 400 dual will get ~1423 MS/s (from memtest86) but the 3500+ Athon 64 939 neo2 will get ~2399 dual. Thats a *HUGE FRIGGIN DIFFERENCE* and more than enough to have the sticks work on one but not the other, as the timing for the back to back R or W couldn't possibly be the same, even if the other specs / timings are followed to a "T".

I would almost guarentee for both of us that *if* these were tested on AMD, they were tested on nforce2 400, and most likely a A7N8X just like my XL pass flawlessly on, not a 939.

I remember when Corsair made some of the first solid pipeline burst cache modules (back in the 486 days when I worked for comp manuf.) and they have ruled ever since as far as build quality and consistancy when dealing with fast and progressive parts for the masses. Maybe I am just unlucky but the XMS thing is startig to hurt. If there are this many problems with XL i'm tempted to just get LL or some other band. The XL on our boards (if not over clocking, saying you could) only give about 66 MB/s more, like about 2333 vs 2399 MB/s on a 3500+.

Last edited by herman; 10-12-2004 at 01:05 PM.
Reply With Quote


  #35  
Old 10-12-2004, 12:24 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 # = 148431
RAM GUY Reputation: 10
Default

I would RMA the pair, you did get a Twinx set right?
__________________
Support accounts and tickets can be created at https://support.corsair.com.
Reply With Quote


  #36  
Old 10-12-2004, 12:51 PM
herman's Avatar
herman herman is offline
Registered User
 
Join Date: Dec 2003
Posts: 60
POST ID # = 148440
herman Reputation: 10
Default

Quote:
Originally Posted by RAM GUY
I would RMA the pair, you did get a Twinx set right?
Ya, TwinX 3200XL PT.

Man, I went to town on that post. I guess I was frustrated. At least I can use single channel LL with both machines for a while this time, unlike last time where I had to borrow half of my then girlfriends memory who I had just given my old comp to, or go pay $300 again for a crap shoot. That just sucked.

This is a PIA because I work from home necessiarily sometimes where I have to run alot of enterprise software (DB, WS, AS, etc...) for development that easily takes 1GB.

------

Is this soimething that I can always expect with XMS in the future?

Am I better off just getting LL instead, or is XL just not happy yet? Will it ever be happier?

Last edited by herman; 10-12-2004 at 12:57 PM.
Reply With Quote


  #37  
Old 10-12-2004, 12:59 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 # = 148443
RAM GUY Reputation: 10
Default

Herman, if the modules will pass on one MB and not another the first conclusion should be the MB. If a MB is as you say more aggressive with memory than another 1/2 of the memory on the market would not run in that MB. That would suggest that maybe the MB and or bios needs to mature to me. However, in general there is not as much margin in the chipset and CPU's as there was in the past, so that will add to this as well. In other words, on one MB the modules fail and pass on another or several others that would suggest either MB does not have as much margin in the memory buss as other MB's; not necessarily that you have weak memory. In addition you are running the timings at an aggressive setting, which is over clocking the timings to achieve more memory performance rather than over clocking the CPU Frequency! The fact that you changed the modules and that seems to have solve the problem, does not always suggest that you had a failing module, rather that maybe the modules tolerance with the specific MB did not run as well in another system. This is the reason we have the RMA procedure and have a fairly open RMA procedure. In addition by the shear volume we sell you are a small percentage of the unfortunate one to get a module that you had to RMA. The return rate on this part is less than .01%, so I have too assume that either you are trying to bee too aggressive with your settings or I am sorry but just unlucky.
__________________
Support accounts and tickets can be created at https://support.corsair.com.
Reply With Quote


  #38  
Old 10-12-2004, 01:54 PM
herman's Avatar
herman herman is offline
Registered User
 
Join Date: Dec 2003
Posts: 60
POST ID # = 148473
herman Reputation: 10
Default

In my case I have a fairly popular board for Athon 64 dual channel, the MSI neo2 393, and probably the most popular Athon 64 dual channel chip, the 3500+.

I agree with your assertions and other generally accepted wisdom, and it's true the board or proc could be at fault or faulty. Or are you saying that the 3500+ or neo2 don't have the tolerances in general to support XL memory? Are you saying that you would not support the Athon 64 dual channel?

I ran the memory at spec. All timing as volatges by SPD, and even manually set the voltage 2.7, 2.75, and 2.8 just in case. With "agressive" bios option off, where I started, it failed. With it on it failed. With spread spectrum off and on it failed (off makes tighter clock signal, at expense of possible RFI). And like many of us who come here as a last resort, I tried every combination I could think of.

Corsair testing memory on an A7N8X @ 400 DDR 2225 which gets 1423 MS/s with XL, with agressive "on" (which still has LL spec timings to the memory), and the 3500+ (which has the actual mem controller) @ 400 DDR 2225 with agressive "off" still gets ~2158. You cant possibly tell me that a difference like that with both having the same signal rate of 400 DDR and same 2225 mem settings, would not affect the apparent stability memory module.

The reason that the Athon 64 is faster is that it has a faster memory controller among other things and lower latency to get to the point of accessing the memory. Once accessing the memory it is followng the same rules as the nforce2 controller, but from the memory's point of view there are just fewer missed but still entirely appropriate opporunities for getting used.

Is at least possible that you are passing memory through QA with an A7N8X @ 400 that just cant keep up with 393, or gets too hot being accessed that often (same thing usually for chips).


--------

All that asside, one of the two modules is definatly weaker than the other as the errors seem to fall on (and follow) one modules half of the interleaved addresses.

One thing I am not going to do is set up a specail fan for the XL, when my LL works perfectly fine with the same setting except 2326 (and at default MB mem voltage 2.6 !!, though I tried 2.7, 2.75 just to be sure that didnt cause the issue on the CPU or MB when the LL was in with that voltage).

If you say go get another 3500+ and another MSI neo2, I suppose I might try that. But when I have seen so many others have issues untill they get a good RMA, and yet others have no problem out of the box, I feel compelled to encourage you to start tesing then the newer lower latency controllers like on the 3500+ if you are not already doing so.

Last edited by herman; 10-12-2004 at 02:05 PM.
Reply With Quote


  #39  
Old 10-12-2004, 02:04 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 # = 148480
RAM GUY Reputation: 10
Default

I would as we suggested try replacing the modules and then if you still have a problem, I would do more testing. But we also use the ASUS P4C800-E and DFI Infinity MB's as well, so in general while some chipsets or platforms seem to have more margin than others if you have a bad module or failing IC on a given module it will fail no matter what MB you run it on. For anything else it can get really hard to speculate, but we have tested this MB in our lab and found it too be one of the best MB's for this platform, so if you do continue to have problems I would suggest we take a look at your configuration and try to make the best choice on how to proceed in solving the issue should you have one. I think what I am trying too say here is it's best too keep an open mind when problem solving and try and verify what ever you feel or find with other testing. By no means am I saying that you do not or did not have a weak or failing module only some prudence may be best to solve the problem long term. And we will with out a doubt stand behind our modules and I will do what ever I can to help you or anyone solve a problem!
__________________
Support accounts and tickets can be created at https://support.corsair.com.
Reply With Quote


  #40  
Old 10-12-2004, 02:33 PM
herman's Avatar
herman herman is offline
Registered User
 
Join Date: Dec 2003
Posts: 60
POST ID # = 148491
herman Reputation: 10
Default

I understand and thanks for your time. If I had two of everyting I would certainly test till I had more certainty about what is wrong. I will post something to the other thread for RMA and see what happens.

Sorry if I was being irate. I was in support one time too. :-)

-----

On what was intended to be a separate issue:

I would still have to respectifully disagree (reguarless of what my issue turns out to be) that a dual channel low latency controller, might very well reveal deficiencies in some modules currently being sold. The Athon 64 939 has the lowest latency dual controller normal people can buy right now, by a good margin.

I know that, relatively, not alot of people have these 939s yet, but I suspect this change very very soon. 90nm, PCI-E, price-cuts, etc... are comming. It will be the new Socket-A. You guys might consider it.

And I agree the toleraces migh just be different reguarless. Which is probaby an even better reason to start now. Give you somethign to yell at AMD about.

Thanks again.

Last edited by herman; 10-12-2004 at 02:47 PM.
Reply With Quote


  #41  
Old 10-12-2004, 02:48 PM
herman's Avatar
herman herman is offline
Registered User
 
Join Date: Dec 2003
Posts: 60
POST ID # = 148497
herman Reputation: 10
Default

What should these be run at 1t or 2t? which is faster?
Reply With Quote


  #42  
Old 10-12-2004, 03:38 PM
herman's Avatar
herman herman is offline
Registered User
 
Join Date: Dec 2003
Posts: 60
POST ID # = 148505
herman Reputation: 10
Default

ok

I got one to fail in single channel and the other has no errors as far as I can tell.

The one that fails has errors in memory location that are about half of the same places that fail in dual.

some row or column cant do the deed on one chip, as the failing bit is always 00080000 or somehting. same failing bit each time.

I think its getting worse, or I am letting the test run longer and letting it get to its normal heat. Once it starts failing around second or third pass, it just keeps failing every pass. This happens in both green slots, though slot one seems a bit more stable by a hair, which I guess is to be expected.
Reply With Quote


  #43  
Old 10-12-2004, 05:26 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 # = 148526
RAM GUY Reputation: 10
Default

Can you tell me the exact settings you have set in your bios?
__________________
Support accounts and tickets can be created at https://support.corsair.com.
Reply With Quote


  #44  
Old 10-27-2004, 02:40 PM
Zoltan Zoltan is offline
Registered User
 
Join Date: Oct 2004
Posts: 5
POST ID # = 151034
Zoltan Reputation: 10
Default same

I have the MSI neo2 plat mb and dual 512mb 3200XL memory.

I have the same problem like the rest of the ppl here when running at 2-2-2-5 settings (1,8v). Memtest86+ finds 9 errors each loop.
I tried everything mentioned on the forums and on the links to other forums.

The best stable value I got (so far, test still running) is 2-3-2-5 1T. 2.8v Agp 1.6v

My question is now this:

How much performace will I lose by useing 2-3-2-5 instead of 2-2-2-5 ?
I paid almost twice as much as for a cheaper memory to get the speed and now I cant use it fully. Should I return the memory and get new once ?

Last edited by Zoltan; 10-27-2004 at 03:02 PM.
Reply With Quote


  #45  
Old 10-27-2004, 02:51 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 # = 151036
RAM GUY Reputation: 10
Default

Please check your Voltage, 1.8 Volts would be the CPU, the Dim or DDR voltage should be 2.8 Volts. And I would test the modules one at a time and make sure that the bios you have supports the CPU you have and or just get the latest version of bios!
__________________
Support accounts and tickets can be created at https://support.corsair.com.
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 06:19 AM.


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