Jump to content
Corsair Community

ABIT AV8, Athlon 64 FX-53, and TWINX1024-3200XLPRO


Zach

Recommended Posts

Okay, I'm just going to throw this out there, not out of deperation, but out of fatigue. (I'm just growing weary of running memtest86 and Prime95 day after day and night after night to test new settings. :(:)

 

System:

Abit AV8 (Bios v.1.3 -- just released yesterday)

(I may switch back to v.1.2 because regardless of Abit's claims, these seem to be less stable than the originals.)

Althon 64 FX-53

1 GB RAM (TWINX1024-3200XLPRO)

 

 

My primary interest is achieving rock solid stability, but at the same time, I want to maximize my performance as much as possible. :cool: (And because of this, please, no one suggest that I change my command rate to 2T. I simply refuse to go that route if at all possible; that's just way more memory bandwidth than I'm willing to lose.)

 

Okay, I can't pass memtest at the rated 2-2-2-5 speeds, at least not at 1T anyway. (I concede that no where does Corsair state that these timings were achieved at 1T, but others seem to manage it, so hopefully I can, too.)

 

As far as voltages, I only have two suitable choices: 2.75 V and 2.85 V. (The BIOS settings are for 2.7 and 2.8, respectively, but the voltage monitor always reads either 2.75 V or 2.85 V.)

 

 

Okay, now to the nitty-gritty "please give me some advice" part. :biggrin:

 

I seem to be stable at 2.5-2-2-5-1T (DDR400) @ 2.75 V. Memtest passed all its tests (standard + extensive), but I have only run Prime95 for about 30 minutes, so I don't really know for sure.

 

I have seen RamGuy repeatedly recommend using 2-3-2-6-1T timings for the 3200XL series (in conjuction with Abit and Asus brand new Socket 939 boards), so I guess I'll try that, but before I waste yet another 12-20 hours of my life running stress tests, what if I want to OC my processor/memory just a little bit? (Just so I can have that warm fuzzy feeling, you know? :)) I'm not looking for anything grand. Maybe setting my FSB to 210 maybe (or heck, even a marginal 205 would be nice). That would have me running at DDR420 (or DDR410 @ 205 MHz). Do those timing recommendations still hold true, or would I need to loosen them up even more?

 

Also, performance-wise, would I just be better off leaving my FSB at 200 MHz and going with the 2.5-2-2-5-1T timings? (Again, I haven't run Prime95 for the minimum 6 hours yet, so I'm not 100% confident that this settings even as stable as I think they are.)

 

What about the other memory settings? (row cycle time, row refresh cycle time, etc.) Some guy in another thread claimed to achieve stabilty by increasing the row cycle time. Is there any merit in that?

 

Thanks in advance for any and all help and advice!

Link to comment
Share on other sites

  • Corsair Employee

Zach

You should take things with a grain of salt and consider a few things!

1. The MB or MB's you are asking about are new and it's not unusual for new MB's to take a few bios revisions for them to work any minor issues out!

2. The modules you have are tested in another platform at the timings listed Cass 2-2-2-5! Please see XMS Qualification and Testing and look up your modules as all of our XMS modules are listed along with the make and model of MB we use to validate them!

3. No 2 MB's will test exactly the same and in this case the memory controller is in the CPU, so the CPU will play a bigger role with memory timings and the over clock achievable.

4. Memory is nothing more than an electronic switch and will either work or it will not!

Stability issues are often times hard to diagnose and many times are not caused by memory, but often times can be caused by a memory related setting! Either one that is changeable by you and or one that is not changeable but set by the MB maker! Thus getting the latest bios and resetting the bios and starting fresh is not a bad idea!

 

The goal should be with any system to find a starting point where the system is stable and work your way up to the point you want to attain with over clocking, knowing there is always a safe point you can go back too! And you should consider that when you over clock you may have to lower the timings or set them to a slower speed! In other words;

1. Find the safe starting point! I think you had Cass 2.5-2-2-5

2. Then set the settings to a more relaxed setting like Cass 3-4-4-8 and set the Dim Voltage to 2.85 Volts and see how high you can get with the Freq.

3. Once you know these 2 then you can find a happy medium of both to get the most out of your system.

Over clocking is not an easy thing sometimes and may take weeks to complete a battery of tests to find the best settings for a given system. I would not spend a lot of time doing long term testing until I am fairly confident in that a specific setting or range will run, then you would run an extended test and tweak the settings to make the system stable long term. Sometimes that is just a long walk in the park, so to speak and sometimes it is impossible to get a given system stable.

Link to comment
Share on other sites

Thanks for the quick reply!

 

And before I say anything else, I would like to quickly add that I am in no way criticizing these awesome memory modules or doing that average Joe "why don't these sticks run at their rated timings" dance. :p:

 

Also, I know intellectually (and professionally for that matter) that buying the fastest processor by definition does not give me much room (if any) to grow speed-wise. (They all come from the same wafer; after that point they just get rated and tagged. *thinking* Well, perhaps not the FX-53 yet, but you know what I mean.)

 

Like I said originally, though, getting that extra megahertz or two just gives one that warm fuzzy feeling.

 

Anyway, yes, I know I need find stability first, then proceed to overclocking tweaks, but this is the first memory/motherboard combo I have ever had that didn't at least run at stock speeds, so I am just getting a little frustrated (and impatient) with all the stress tests I'm running day and night and was looking to possibility try to kill two birds with one stone. :cool:

 

Okay, I'll take your suggestions and go off and do more testing. Since you didn't comment about the other timings (row cycle time, etc.), I assume that you are indirectly recommending that need not bother with those in my quest for stability, right?

Link to comment
Share on other sites

  • Corsair Employee
The other settings you asked about should not need to be changed, as they are usually set by the MB maker, but will sometimes base the setting on a SPD value. In any case I have never had to change these settings only the Cass latency. And if you like we can try replacing your modules, but if it were me, one way or the other I think I would try and test the modules on another platform that I knew would run these to be sure I am not fighting a failing or weak module! If you want to try and get these replaced, please follow the link in my signature “I think I have a bad part!” and we will be happy to replace them or it!
Link to comment
Share on other sites

Judging from all the posts I have been reading here the past few days, you seem to have a stellar attitude about replacing parts! :cool: Don't think I or we don't appreciate that! It's so refreshing considering the general level of customer support these days.

 

No, at this point, I am holding AMD and its onchip memory controller to blame. Maybe getting a couple of sticks of RAM with "slightly" better tolerances would help overcome AMD's deficiencies, but I am notoriously unlucky in that regard, so it would most likely be a waste of time anyway. :[pouts:

 

Of course, I suppose if I end having to use those 3-4-4-8 timings on a permanent basis to get stability, then I might click your link. ;):

 

Anyway, as an update, I can get my CPU up to 212 MHz x 12, pass memtest's standard tests, and run Prime95 for at least it's first test. So then next I reduced it to 205 MHz and changed the memory timings to your 2-3-2-6-1T values, and though that passes memtest, Prime95 fails within the five minutes. :(:

 

It looks like I'm locked into those 2.5-2-2-5 timings (at stock 200 MHz speed) I mentioned earlier, but only a full night's run of Prime95 will say for sure.

 

Hypothetical: Suppose both 2.5-2-2-5 and 2-3-2-6 gives me a stable system. Which set of numbers, in your (or anybody's) opinion would be better performance-wise? Okay, "perfomance" is a very broad term, so let me ask it another way. If this were your personal system, which set would you use on your system?

Link to comment
Share on other sites

I had an Abit AV8 and it wouldnt even post unless I set timings to 2.5-3-3-8 2T lol. As far as I'm concerned the AV8 is a huge hunk of crap and I will never buy abit again. Im using a MSI Neo2 and I like it much better. Also I've noticed that alot of these new 939 boards will only work with cas 2.5. I myself cannot run cas 2 but I can run Prime95 for 8 hours at 2.5-3-10-2 1T with no errors. I auto fail the first test on prime if I try cas 2.
Link to comment
Share on other sites

As far as I'm concerned the AV8 is a huge hunk of crap and I will never buy abit again.

 

I'm beginning to agree with you. :p:

 

For years now, I have flip-flopped between Asus and Abit. I generally prefer Asus, and I was actually going to buy an Asus A8V, but I read a lot of bad user reviews and problem reports (in tech support forums), so I hastily went with Abit's AV8 instead. (Had I researched Abit's board a bit more, I would have found just as many problem reports. :roll:)

 

Anyway, like I said above, my gut feeling is still telling me that the root of problem is in AMD's memory controller, not any one MB manufacturer. I am just hoping Abit's S/W guys can figure out a way to compensate because I am getting very, very frustrated here.

 

 

Okay, well, Ram Guy, never mind my hypothetical. Prime95 failed in 12 minutes at 200 MHz, 2-3-2-6-1T @ 2.85 V. Next I'll try my 2.5-2-2-5-1T, and if that fails, I'll, as a last resort, try 2.5-3-2-6-1T. If that doesn't work, then I don't know what I'll do--probably take the board out, place it gently on the floor, and then proceed to jump up and down on it with full force. :eek:

 

(Yes, I know, my true last resort is to switch to 2T, but I refuse. Seriously, if I can't even get 2.5-3-2-6-1T to work, then I'll probably just trash this $1000 MB/CPU and go spend another $1000 [*groan*] on a good old reliable Pentium 4 EE.)

Link to comment
Share on other sites

2.5-2-2-7-1T @ DDR400, 2.75 V : Failed Prime95 in 8 hours and 15 minutes.

 

Hmmm... I wonder if there is a pattern here.

 

Okay, I'm off to try 2.5-2-2-8-1T @ DDR400, 2.75 V ...

 

 

Edit: No, I think I'll just leave the voltage at the chip's rated 2.75 V.

Link to comment
Share on other sites

  • Corsair Employee

Zach

I would test the modules one at a time and see if you can identify one of the modules may be failing! However, I have no problem getting them replaced if you would like to try that!

Please follow the link in my signature “I think I have a bad part!” and we will be happy to replace them or it!

Link to comment
Share on other sites

Yeah, I know, and that is very kind of you, but I truly do not believe the physical memory is the problem. Like I just posted in that other thread, I can get your memory to pass memtest at amazing speeds. It's just when I get the processor more thoroughly involved in the process (i.e., like with Prime95 for instance) that things go awry.

 

For instance, as an extreme, I was able to pass memtest at DDR490 speeds using 3-4-4-8-1T timings!

 

You're the expert, I know, but I believe that alone proves beyond a doubt that the memory I have is very good not at all defective. <shrug>

Link to comment
Share on other sites

  • Corsair Employee
LOL! Well I would have to agree with you! Can a friend maybe help you test the CPU and memory in another MB? That might help to isolate the MB! In other words if you still see the same type of response with the CPU and memory in another system, I would RMA the CPU, but if the issue is not as apparent or non apparent I would RMA the MB!
Link to comment
Share on other sites

No, unfortunately, I'm the current trendsetter in my circle of friends and colleagues. :cool: By that I mean no one else I know is willing to use a Socket 939 board until 1.) it matures a bit more, or 2.) I tell them to. :p: (Obviously, I'm not going to tell them to at this stage.)

 

Besides, I'm not willing to separate the heatsink from the CPU now that it has bonded, and you can't take out the CPU without taking off the heatsink first. For an Intel, I might consider it, but AMD's already get too hot as it is. ;):

 

So for me, I wasn't kidding above. If I can't get stability, then I'm just going to trash the whole thing and get a P4EE instead. (No sense in gambling twice on the same processor!)

Link to comment
Share on other sites

you should have no problems running 2-2-2-5-1T on the k8n neo2.

 

Coincidentally, I was going to ask you about that in another thread, but I wasn't logged on, so I couldn't.

 

People's definition of "stability" varies greatly. Now being a senior design engineer who used to design embedded systems for a living, I am fairly intolerant I am afraid to say. For instance, my 8 hour Prime95 run above would be good enough for most people since statistically speaking, I'll get struck by lightning and die before my system ever crashes due to memory timing problems. :p:

 

In that other thread, you mentioned a lot of things, but you never actually mentioned ever running Prime95. Have you? And if so, how long did you let it run?

Link to comment
Share on other sites

*update*

 

2-2-2-6-2T @ DDR400, 2.75 V : Failed Prime95 in 10 hours and 18 minutes.

 

(Note the "2T" DRAM command rate.)

 

 

I give up. :[pouts: I don't know if it's just an AV8/Corsair compatibility issue, simply a MB issue, or worse-case scenario, a CPU problem.

 

I've read numerous success stories about how changing out the memory--the "brand" is not really a factor!--fixed all the problems. In other words, some are claiming that these 939 boards are extremely picky about the memory characteristics, and the board and the memory need to be perfectly matched. And, of course, it's all a gamble because you will never know if you have a matched set until you buy the memory and test it in the motherboard in question.

 

I'm not sure I buy that that theory. Besides, as usual, I have found that most people use the phrase "it works perfectly!" or "it's completely stable!" a little too loosely. Heck, one guy even stated outright, "...I ran Prime95 for nearly half an hour with no problems, so now I'm positive it's completely stable. These <insert memory brand name> are awesome!" :roll:

 

Yesterday, I was thinking (and hoping) that it was simply a heat issue, but when it failed just now, my CPU was only reading 30º C (86º F). (I don't know what temperature the memory was at since there is no monitor for that. :):)

Link to comment
Share on other sites

I read a lot about the AV8, Memory and other issues...

 

Some people claim to run 280MHZ FSB 1:1 with the ram primestable. Others cannot even run 220 FSB 1:1 stable with the same ram and MB(Abit AV8 in this case). They both had a A64 3500+. At the moment i am happy to run 229 FSB 1:1 at 2,5-3-3-10-1(The 10 are faster than 5. In a test at 200FSB 2-2-2-10-1 was faster(or as fast as) 2-2-2-5-1 with Sandra).

 

But the Row Cycle Time was critical, because ""Formula-wise, the row cycle time (tRC) = minimum row active time (tRAS) + row precharge time (tRP)""

 

As the tRC is at 12 on default it was to low for my "10+3". This caused my problem in the other threat you showed.

 

So my suggestion is that the internal memory controller cannot always handle dual-channel with high FSB(1:1) and/or aggressive timings.

 

Another thing is, that there are two revisions of the AV8. Rev 1.0 and 1.1. What the differences are, i do not know, but i have Rev 1.1.

Link to comment
Share on other sites

Thanks for the advice and the tips.

 

Unfortunately, I am not sure if it applies or not to my specific, apparently quirky hardware.

 

(1) Running at 2.5-2-2-7-1T-200 MHz, Prime95 runs for 8 hours and 15 minutes.

(2) Running at 2.5-3-3-10-1T-200 MHz, Prime95 runs for 8 hours and 3 minutes.

 

For both cases, my BIOS seems to automatically set tRC to 13 clocks regardless of what I tell it to use.

 

http://img26.exs.cx/img26/6459/cpu-z-memory.jpg

 

Yet, in BIOS, tRC is manually set to 11. Go figure. And yeah, I have a Rev. 1.1 board, too, at least according to the sticker on the the board. And I have done extensive testing with both version of BIOS (1.2 and 1.3). I have pages and pages of emperical evidence showing that the "new" 1.3 bios is even more unstable than the original 1.2, so now I'm back on 1.2 now.

Link to comment
Share on other sites

Hey, Ram Guy, in a last ditch effort to get a stable FX-53 system (before I throw in the towel and switch back to Intel and their more "stable" motherboards), I just purchased an MSI K8T Neo2-FIR motherboard. (It should arrive in 2-3 days I hope.)

 

Unfortuntely, the only "tested" Corsair module they cite close to mine is the CMX512-3500C2PT which apparently you guys discontinued because I can't find any PC3500 chips on your site. ;):

 

Have you had a chance to test your oft recommended TWINX1024-3200XLPRO's on this board yet? If yes (or even if no), do you have any recommended settings I should use other than the default 2-2-2-5-1T? (I suppose your answer will probably be 2-3-2-6-1T. :):) Also, do you recommend setting the voltage to 2.8 - 2.85 V from the get go, or only as a last resort? (I worry about heat issues, hence why I ask.)

 

Thanks!

Link to comment
Share on other sites

Er... when I said "...Intel and their more "stable" motherboards," I meant a Pentium 4 and a Socket 478 motherboard. (I didn't mean a literal "Intel" branded board--probably more like a Asus or something.)
Link to comment
Share on other sites

  • Corsair Employee
of them, that could not be addressed with a bios update. So any of these MB’s should be able to run these modules. However, with AMD64 the memory controller is in the CPU, so the MB will not have as much as an effect on the memory performance! I would suggest that you try and relax the timings if you are having trouble to Cass 2-3-3-6 and see if that helps! If it’s the memory it will help if it’s the CPU it will make little difference usually!
Link to comment
Share on other sites

However, with AMD64 the memory controller is in the CPU, so the MB will not have as much as an effect on the memory performance!

 

I beg to differ, especially when you take into account the trace layout and design (from memory slots to processor in the case of the AMD64) and its effect on latencies and worse, signal reflections due to loading. ;):

 

At any rate, I hear what you saying. :):

 

Actually, it is kind of depressing. On the MSI forums, there are entire, moderated threads dedicated to the subject of avoiding Corsair memory (and Kingston as well) for the Neo board. (I will be using a Neo2, so maybe I'll have better luck.) Same goes for the Asus forums and their A8V boards. The common theme I have deduced is that Corsair memory is very hit-or-miss with the A64's: either it works outstandingly, or it doesn't work at all. There seems to be no in between. :(:

 

I seem to be running stable now at 3-4-4-10-1T, but I'm sorry. Even though you seem to think nothing of it, having to clock down a 2-2-2-5 chip to 3-4-4-10-1T indicates that something is very, very wrong somewhere. :sigh!:

 

Oh well, anyway, thanks for all your help and patience with me. I greatly appreciated it, and it was comforting to know that I was not alone in my plight; however, I think I'll just admit defeat and shut up now and stop bothering you. :sunglasse

Link to comment
Share on other sites

  • Corsair Employee

That is a great point Zach, and you are correct! However, the design spec should not differ that much as far as trace lengths! But I would have too agree with you to some extent!

 

Well as far as this type of bashing, you really have to consider the source of these types of posts, and then if someone is having a problem what have they done to try and remedy the problem. In addition, this chipset is very new and it is not uncommon to see compatibility issues with a new platform when they are first released. That being said I would have to assume that the # of users having trouble are way less that the ones who are not based on the volume of modules we sell!

So we do watch these kinds of reports and we do testing on a Dailey basis to try and duplicate these types of problems; but I have to say that most I have been unable to duplicate! That would suggest to me there is likely some other problem or weakness in general that more than likely will be resolved one way or the other with time and or bios!

 

Bottom line if our modules will not run at the rated speed we will be happy to replace them, even if we do not think the modules are the problem! But there is no way I or anyone else can guarantee a given MB will over clock or be able to run with tight timings, all we can do is tell what we did in our testing and do our best to help you should you have a problem!

Link to comment
Share on other sites

  • 3 weeks later...

this thread is, to say the least, scaring me! :eek:

 

I am waiting for the delivery of my new:

 

Athlon 64 3500+ (sckt 939)

Abit Av8 3rd eye

TwinX 1024-3200XLPRO

 

I am no fan of overclocking and I am not planning on doing any on my new system. Will I be able to pull everything out of the box, assemble and run the system at the speeds advertised on the various components? :confused:

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...