Jump to content
Corsair Community

2 pairs of CMX4GX3M2A1600C9 with Asus M4A78T-E and AMD Phenom II X4 965


HighlyDubious

Recommended Posts

At the time I bought all the components for this system, I had purchased other memory (it was also DDR3 1600). Unfortunately, two of the sticks turned out to be defective, and I had to return them.

 

I had planned to return only the two defective sticks and replace them with the same product. (The other memory was not Corsair, but at this point I don't even recall what other brand/product.) But when I returned them, the store did not have any more of the other memory in stock. So, on the advice of the sales person, I purchased 8GB of CMX4GX3M2A1600C9 and returned all of the other memory.

 

When I got the home, I saw that the CMX4GX3M2A1600C9 packaging mentions compatibility with i5/i7, but figured that was just added to the package by the weenies in the Corsair Marketing Dept to try to bump sales. It seemed logical to me that DDR3 is DDR3, and there should not be any problems using DDR3 "for i5/i7' on my AMD Phenom II 965.

 

While building the machine I upgraded the BIOS to the latest version (v2105). I then installed Windows XP, and Windows 7 (both are x64), and had absolutely no problems. The system is rock solid and totally stable. I've run memtest86+ v4.0 through 3 full cycles, with no errors. And also run Prime95 for 30 minutes under both WinXP and Win7 with no errors. So I know the system is very stable.

 

However, every time I try to install Ubuntu 9.04 (x64), or Ubuntu 9.10 Beta (x64), I get random hangs/freezes, and general weirdness. I've tried installing from CD, and from a USB flash drive. After a dozen attempts, I did finally get Ubuntu 9.04 (x64) to install. But afterward it acted very oddly. Programs and windows would randomly crash/close. And even the X server would randomly crash and dump me back to the log-in screen. These errors would occur even if the machine had not been touched for quite a while.

 

I did some research and found a recommendation to manually set the DRAM Voltage in the BIOS to 1.5v, which I have done. But beyond that, I've left all other settings at their default values.

 

I was able to install the Linux distro of PCLOS (x32), and after a few hours of use it seems completely stable. But, given that I spent the money for 8GB, and all my other OS's (on this machine and my other machines) are x64, I really don't see an x32 OS as being acceptable.

 

I did find this today (link), so I'm thinking my next option may be to reduce the speed of the memory. The thing is though, ASUS clearly advertises this board as being fully ready to OC to 1600Mhz. On the other hand, it does not mention any Corsair memory as being compatible with 1600Mhz. I did not worry too much about this, because I felt it could easily be explained by the mobo manual being printed before your product was released.

 

Bottom line: I need this system to be stable with Linux in 64-bit mode, and want to be able to access all of my 8GB. I'd prefer to not cut my speed to 1066Mhz, because the board clearly claims to support it. But if that is my *only* option, then I guess I'll do it.

 

The reason I'm writing here is to ask if there's anything I've missed, or if you could perhaps present me with other options?

 

(BTW, on a side note, I also have an ASUS Rampage II Gene with i7 920, using 6GB of your TR3X6G1600C8. That machine works *perfectly* with WinXP, Win7, and Ubuntu 9.04 (all are x64). So I know your other memory products are very good.)

 

 

Thanks so much for your time and rapid response.

Link to comment
Share on other sites

Corsair can't guarantee 2 packs will run together since they weren't tested together, they may be using different ICs, and the memory controller (which is out of their control) may not like maxing out the slots. Also, the motherboard manual probably says 1600, 8 GB, but nto 8 GB AT 1600.

 

Drop the speed a notch to 1333, see how it goes.

Link to comment
Share on other sites

Corsair can't guarantee 2 packs will run together since they weren't tested together, they may be using different ICs, and the memory controller (which is out of their control) may not like maxing out the slots.

 

Would that be true even if all 4 sticks are the same model? Because that is what I have. I have 4 sticks of Corsair CMX4GX3M2A1600C9.

 

I don't mean for this to come out sounding the wrong way, but if 2 packs of the same model from the same company can't be expected to work together, then just how the heck is anyone supposed to get 1600Mhz to work above 4GB? :confused:

 

I mean, so far as I am aware, I can't go out and buy 4 sticks of 2GB that have been tested together.

 

I assure you I'm not trying to flame on this, but something seems fundamentally wrong with the entire computer and memory market if you can't reasonably expect to use all 4 slots at the advertised speed just because the sticks of the same brand and model weren't tested together.

 

That would mean nobody could reasonably expect to get 1600Mhz with 8GB on any dual channel mobo, unless a) it came as 2x4GB sticks or b) all 4x2GB sticks were tested together -- which, so far as I know, does not happen.

 

Please understand, I'm *not* saying that you are incorrect. I'm just saying that even if what you say is the real truth, it seems fundamentally wrong. :bigeyes:

 

I know this next example is a very poor analogy, but that would be like saying I could buy a car with a V8 motor, but it has the following limitations 1) if I want to go 100mph, then I have to run on only 4 cylinders, or 2) if I want to run on all 8 cylinders, then I could only go 65mph. :eek: Those kinds of limitations would be fundamentally absurd, and what you said strikes me the same way.

 

I'm not arguing that your facts are wrong, I'm just astonished if your facts are right.

 

Also, the motherboard manual probably says 1600, 8 GB, but nto 8 GB AT 1600.

 

You are correct. The manual does not mention any mfg with 8GB at 1600. *BUT* it does mention two different (non-Corsair) mfg with 6GB at 1600, and adds that those cases are "Kit of 3". In other words, at the very least, one of the channels as 2 sticks. So it's not a stretch to imagine 1600 with all 4 slots filled.

 

Drop the speed a notch to 1333, see how it goes.

 

Regardless of if I am shocked by your assertion above, yes, I agree with you I will try things out with the speed reduced and see if my results improve.

 

And I apologize if I seems like I'm flaming above. As I said, I'm just astonished at the idea that it's impossible to get 8GB of dual channel at 1600mhz with 4x2GB sticks. :eek:

 

Even so, I *DO* thank you for taking the time to reply, and for opening my eyes to something utterly mind boggling. :bigeyes:

Link to comment
Share on other sites

Would that be true even if all 4 sticks are the same model?
There's something like 9 different IC manufacturers that have ICs that can be used to make that part, hence 9 different version numbers per part (generally speaking).

 

 

I don't mean for this to come out sounding the wrong way, but if 2 packs of the same model from the same company can't be expected to work together, then just how the heck is anyone supposed to get 1600Mhz to work above 4GB? :confused:

 

I mean, so far as I am aware, I can't go out and buy 4 sticks of 2GB that have been tested together.

There's currently 2 DDR3 8 GB packs (4 x 2 GB) that will hit 1600 Mhz:

CMX8GX3M4A1600C9 (XMS heat spreader @ CAS 9)

CMD8GX3M4A1600C8 (Dominator heat spreader @ CAS 8)

 

 

Please understand, I'm *not* saying that you are incorrect. I'm just saying that even if what you say is the real truth, it seems fundamentally wrong. :bigeyes:
Yeah, it sucks, but it's pretty much the combination of the limits of technology and many factors at play.

 

Think of it this way, if they WERE to mix ICs and test it, there's over 6000 possible MANUFACTURER combinations if you're testing combos per slot assuming 4 slots (per pair would bring it down to 72 manufacturer combos), and then multiply that by the number of memory controllers out there. Then multiply that if the IC manufacturer has versions or varying batches. ICs from the same batch don't always play well with each other. That's silicon for you. No company is going to do all of that testing. Cost would be literally astronomical.

 

 

You are correct. The manual does not mention any mfg with 8GB at 1600. *BUT* it does mention two different (non-Corsair) mfg with 6GB at 1600, and adds that those cases are "Kit of 3". In other words, at the very least, one of the channels as 2 sticks. So it's not a stretch to imagine 1600 with all 4 slots filled.
The manual doesn't actually say 3 sticks at the same time, they just list a 3 stick pack part number. Probably because for at least the first half of the year there was a lot of memory made for the new Intel Core i7 9xx CPU platform which has a triple channel memory controller. Some of the most sought after memory (e.g. the Dominator GT line) was only coming out in 3/6 stick packs. Because of that, some people were buying 3/6 stick packs so they could use 2/4 sticks on their dual channel systems.

 

So it looks like Asus listed triple stick packs because you can always use 1 or 2 out of the 3 sticks. One thing though, some of the 3 stick packs are listed at lower speeds in your manual (e.g. the 2nd kit listed below is really a 1600 MHz kit). Here's the breakdown of the 3 stick packs listed:

 

6 GB 1600 = A

6 GB 1066 = B

3 GB 1600 = AB

6 GB 1333 = AC

3 GB 1333 = ABC

6 GB 1333 = ABC

3 GB 1066 = ABC

 

A = 1 stick in a single channel config

B = 2 sticks in a dual channel config (same colored slots)

C = 4 sticks in a dual channel config

 

 

I know this next example is a very poor analogy, but that would be like saying I could buy a car with a V8 motor, but it has the following limitations 1) if I want to go 100mph, then I have to run on only 4 cylinders, or 2) if I want to run on all 8 cylinders, then I could only go 65mph. :eek: Those kinds of limitations would be fundamentally absurd, and what you said strikes me the same way.
Think of the memory controller as a cop directing traffic. It's easy to direct a decent flow of traffic going 40 MPH, but double that traffic and he'll probably have to slow down traffic to keep it manageable.

 

 

And I apologize if I seems like I'm flaming above. As I said, I'm just astonished at the idea that it's impossible to get 8GB of dual channel at 1600mhz with 4x2GB sticks. :eek:
No worries, your posts are perfectly fine. Note that it's improbable and simply not tested, not impossible. That's what makes it fun :) This issue has been around since the days of DDR1. Generally loosening timings and slowing down memory helps a lot, but some ICs just don't play well with others, even certain batches of themselves.
Link to comment
Share on other sites

There's something like 9 different IC manufacturers that have ICs that can be used to make that part, hence 9 different version numbers per part (generally speaking).

 

Hey, thanks for such a long and informative reply! :)

 

I've had time to cool down a bit, and I have to admit that I've been playing in the computer game for going on 25 years now. So, yah, I know that the same basic part might have IC's from different mfgs. And, although I almost never thought about it, I agree it does make sense that the different IC's can have differing characteristics. I knew and understood why memory mfgs test their sticks together, but I didn't think it would make that big a difference when using one set in one channel, and the other set in the 2nd channel.

 

I mean, part of the memory mfg's job is to ensure that every stick (or pair of sticks) meets certain clearly defined operating criteria. And so, although one shouldn't mix the two sets, I would have thought that using 2 sets on 2 channels would be "good enough" because both sets must fall within certain boundaries defined by the model line itself.

 

Of course, that doesn't mean I'm right, it's simply what told myself when I bought 2 sets of 4GB with plans to plug them into my machine.

 

 

There's currently 2 DDR3 8 GB packs (4 x 2 GB) that will hit 1600 Mhz:

CMX8GX3M4A1600C9 (XMS heat spreader @ CAS 9)

CMD8GX3M4A1600C8 (Dominator heat spreader @ CAS 8)

I was unaware of that. Next time I build a new machine I'll be MUCH more careful about what memory I buy.

 

(As an aside, it makes me sad to think about this, because I've been thinking about upgrading the 6GB of 1600 on my i7 920 machine up to 12GB. But with what you're telling me, I may end up shooting myself in the foot if I just toss in another set of TR3X6G1600C8 :sigh!:)

 

The manual doesn't actually say 3 sticks at the same time, they just list a 3 stick pack part number.

 

True, but that's down right tacky. I mean, ASUS advertises this mobo as being able to accept up to 16GB of RAM. And it proudly boasts of being able to handle 1600Mhz. But it almost borders on deceitful to not clearly state that the only way to get high speed AND gobs of ram, is to buy 4 matched sticks.

 

To reuse the same lame car analogy I used above, its kind if like if the car mfg just casually neglected to mention the cylinders/speed limitations. Then when the buyer complained, they replied with "Oh, but we didn't *SAY* you could use all 8 cylinders to go 100mph. You just assumed it, and we aren't responsible for your assumptions"

 

I've not contacted ASUS about this topic (yet), so I'm not saying that is their real attitude. But the way you define the situation, it could very well be.

 

Probably because for at least the first half of the year there was a lot of memory made for the new Intel Core i7 9xx CPU platform which has a triple channel memory controller. Some of the most sought after memory (e.g. the Dominator GT line) was only coming out in 3/6 stick packs. Because of that, some people were buying 3/6 stick packs so they could use 2/4 sticks on their dual channel systems.

 

So it looks like Asus listed triple stick packs because you can always use 1 or 2 out of the 3 sticks.

 

Yes, that does sound perfectly logical.

 

One thing though, some of the 3 stick packs are listed at lower speeds in your manual (e.g. the 2nd kit listed below is really a 1600 MHz kit). Here's the breakdown of the 3 stick packs listed:

 

6 GB 1600 = A

6 GB 1066 = B

3 GB 1600 = AB

6 GB 1333 = AC

3 GB 1333 = ABC

6 GB 1333 = ABC

3 GB 1066 = ABC

 

A = 1 stick in a single channel config

B = 2 sticks in a dual channel config (same colored slots)

C = 4 sticks in a dual channel config

 

Yes, it does seem they are following your logic about letting people know they can (if they choose), use just 1, 2, or 4 sticks out of a 3/6 stick pack.

 

Think of the memory controller as a cop directing traffic. It's easy to direct a decent flow of traffic going 40 MPH, but double that traffic and he'll probably have to slow down traffic to keep it manageable.

 

Yes, your analogy paints a much clearer picture than my own. Even so, it still seems less than honest to neglect to clearly explain this to potential board buyers.

 

No worries, your posts are perfectly fine. Note that it's improbable and simply not tested, not impossible. That's what makes it fun :) This issue has been around since the days of DDR1. Generally loosening timings and slowing down memory helps a lot, but some ICs just don't play well with others, even certain batches of themselves.

 

Well, for what it's worth, after getting off my soapbox above, I did follow your advice to reduce the speed.

 

In the process of which, I was embarrassed to discover that, in fact, I had *NOT* previously upgraded my BIOS to the latest version. All I can say is that I built my i7 920 machine and AMD 965 machine only a few weeks apart, and I must have gotten confused as to which one I upgraded while building the system.

 

But regardless of that, I went ahead and updated the BIOS to the latest version, and made sure the memory was set to 1333Mhz.

 

At which point, Ubuntu 9.04 (x64) installed and ran PERFECTLY. Not a single hiccup or glitch. I've since run a full update of all packages, and even added the Kubuntu-desktop modules, and its still humming like a perfectly oiled machine.

 

At this minute I've not tried to crank the memory speed back up. So I don't know if the lower speed, or the BIOS update (or both) solved the problem. But for now I'm happy, and can finally rest easier about this board. (I was seriously beginning to believe I'd have to dump the board and buy something different for my AMD 965 because I need Linux on this machine, and I wasn't willing to accept only 4GB)

 

Thank you so much for your informative responses. I've learned alot. Now I suppose you can mark this thread as "Solved", or whatever you guys do in this forum. ;):

 

Have a good one, and Thanks Again!

Link to comment
Share on other sites

Archived

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

×
×
  • Create New...