7 modules report @ 2400; 1 module reports @ 2133

[jhayat2]

Hi all. I have an email into Crucial and am waiting for a reply back, but so far, they have been unable to help.

MB - ASUS TUF X299 Mark 1 (supports 128 GB RAM)
BIOS vers - 1503
AM Megatrends: 2171246
CPU: Intel i9 7980x (supports 128 GB RAM)
RAM: 8 16 GB BALLISTIX modules (one order) that, according to the Crucial site, are for my MB

With the understanding that I cut power to the MB before swapping RAM modules...

I ordered 8 Ballistix modules from Crucial. Seven of them report @ 2400 in the BIOS; one reports @ 2133. This module is in the B1 slot. I swap the B1 & B2 modules, and now the module in the B2 slot reports @ 2133, while the module in the B1 slot now reports @ 2400.

I swap the A1 & B2 modules, and now the module in the A1 slot reports @ 2133, while the module in the B2 slot now reports @ 2400.

I call Crucial tech support and reported my findings, and the tech agreed that it seemed like a bad module. We did an RMA, and I received the new module today. This new module reports @ 2133. I did the same testing as above with the same results. Then, I removed all of the modules, and took one of the seven, and placed it in the C1 slot (the slot that the manual says to use if you only have one module). That module reports @ 2400. I swapped that module for another one of the seven, and that too reports @ 2400. I then swapped that module for the one I received today, and that reports @ 2133.

This is whether or not I set the DRAM freq to AUTO, or to 2400.

I spoke with a tech today, and he suggested I set that module's freq to 2400. How do I do that? He is not sure. I call ASUS, and they tell me that there is no way to set individual freqs; it is global only.

Again, I am waiting to hear back from Crucial via email, but was wondering if anyone has seen anything like this, and if there are any other thoughts.

Thanks in advance.

 

[Darkbreeze]

This makes no sense anyhow. Modules cannot run at different speeds. All modules are reduced automatically to the speed of the lowest speed module.

This sounds like a buggy bios. I would check your motherboard product page to verify whether or not you have the newest bios version installed. If not, update it to the latest version. Make sure you have XMP enabled.

As far as the module itself is concerned, the whole scenario is fubared from the start. Sending you a single module is BS. Memory should ALWAYS be purchased, used and replaced in FULL SETS. If you need two modules, buy a kit, WITH two modules.

If you need four, buy a kit, WITH four modules. Etc., etc.

Then, if one of them is determined to be bad, they need to be replaced AS a set, not send you one module. Any module that did not come together as part of a set stands a VERY good chance of not playing nice with all the other modules that came in the set and this stands true whether during an RMA process OR when you are initially buying them. Don't buy two sets of 2 modules. Buy ONE set of four modules. Or ONE set of 8 modules. Otherwise, you are rolling the dice that problems exactly like the one you are dealing with are not going to happen, when in reality, they are extremely common. Probably even getting to be MORE common than not.

 

[jhayat2]

Thank you.

I updated the BIOS to the latest version. I also cleared the CMOS. Sorry, forgot to mention that earlier.

I get what you are saying about kits and pairs.

I bought one set of 8 modules. They all shipped in one pkg. One module showed @ 2133. Crucial replaced that one. That new one also shows @ 2133 - with other modules, and by itself.

It makes no sense, other than bad module(s)?

 

[Darkbreeze]

They should have replaced the entire set, not one module. There is no way for them to know or verify that the one module they sent works or plays well with the other modules from the original set, EVEN IF IT IS THE EXACT SAME MODEL NUMBER with identical specifications. From production run to production run they often use different memory chips, rank the chips differently and a number of other possible changes.

See here for an example:

http://www.tomshardware.com/answers/id-3610013/amd-ram-compatibility.html#20562100

I would contact them and address that concern. However, that aside, if the new module works, but is still only working at 2133mhz, regardless of where it is installed, I would first try increasing the DRAM voltage by .005v, then .010v, etc. to see if all the modules will at some point run correctly together. With that many modules it's common for issues to happen at the stock voltage. Do not exceed 1.45v IF you cannot get them stable and at the correct speed at a lower voltage than that and do not just jump to that voltage. Try to increase in .005v increments until you have success or determine that it isn't going to work.

Also, be sure to save the bios settings between voltage adjustments. I would also make sure that the memory fast boot setting is disabled. Do not confuse the memory fast boot setting with the regular storage device fast boot setting.

 

[jhayat2]

Thanks again.

So, Crucial is telling me that XMP should be enabled. Is that true? To me, that makes no sense. If 7 modules report 2400 and one reports 2133 with XMP off, why should I need to turn on XMP to "fix" the issue?

So, do I just turn it on? Or do I turn it on AND set something else?

 

[Darkbreeze]

XMP always needs to be enabled unless you manually configure all the memory settings yourself, manually. Leaving everything on Auto usually defaults the whole works to the minimum default speed and timings.

Yes, enable XMP. That's one of the first things you should usually do unless you are going to configure an overclock on the CPU, and then that should usually be done FIRST, and then memory changes get addressed.

Like I said, with that many modules, you might also have to bump the DRAM (Memory) voltage a bit as well if it won't POST at the XMP settings.

 

[Karadjgne]

Sounds like you got ram from 2 different batches. 7 from 1 batch that defaults to 2400, and 1 stick that defaults to 2133, same as the replacement. They just happen to be compatible tested at the factory and passed.

If XMP and a small bump to dram voltage (helps with memory controller compatability and stability) doesn't fix the issue, RMA all the sticks, tell Crucial to send 8 more instead of the band-aid one, since there's now a possibility of incompatability due to mix and match kits.

 

[jhayat2]

Thanks to both of you ^

Sounds like you got ram from 2 different batches

Yep - I def did. The markings are different. (Which I just noticed, BTW)

First, so we are clear, I am not OCing the CPU.

Secondly, I have now enabled XMP.

Thirdly, I have not yet bumped the DRAM voltage as suggested; I am still talking to Crucial, and I wanted to give them specs w/o introducing any variables.

Here are three shots from CPU-Z in W7, w/one of the 2400 modules, and the 2133 module

http://www.jeffreyhayat.com/temp/CPUz-1.jpg

http://www.jeffreyhayat.com/temp/CPUz-2.jpg

http://www.jeffreyhayat.com/temp/CPUz-3.jpg

And here is the answer I got from Crucial, tho it might not be complete:


Thanks for the CPUz info. Yea, you can actually see the difference between both parts pretty clearly. Both have correct XMP which is 3200 (DDR) or 1600 (SDR). With the 2133 parts you can see the JEDEC timings are reported as 2133 (1066 SDR). The odd thing is this doesn't appear to be effecting the memory performance at all, since the memory will run at full XMP or lower speeds to match the other memory just fine. This leads me to believe the numbers we're seeing are actually a lie, and the memory isn't running at 2133, that's simply what the memory controller on the board is telling us. With the first picture you attached you can see your memory is actually running at 2700 MHz, and not at 2133Mhz. If you run full XMP speed you'll see 3200Mhz instead.

We've seen strange SPD info bugs like this in the past, where the computer says the memory is one thing when in reality it's another. Sometimes a future BIOS update to the motherboard may fix this, but since it doesn't actually effect the performance of the memory, or cause system stability issues, it could be look at simply as UI bug and nothing more.

I'll let you know when we get a more solid answer from Engineering.

 

[Darkbreeze]

I've NEVER ever heard an answer that was so full of BS from a major company before. Well, maybe, but that one's up there with it. I'd tell them I'm not buying it. You bought a kit with what were supposed to all be matched modules, and clearly they were not, and are still not since they sent you an unmatched stick to throw into the mix, and that you want what you paid for which is a kit with all memory modules that are matched and tested to operate together in the way they were intended.

 

[Karadjgne]

From the first pic, it clearly also states dual channel. With 8 slots populated, you should most definitely be running in quad channel if all that ram is the same. So it's not all the same and you are running 4x dual channel instead.
It's a Skylake-X cpu, it's going to run the ram at 2666MHz default if it can. That's the quip about the NB being 2.7Gb. With XMP enabled, that ram, according to its model, should be running at 3200MHz, not 2666MHz, so obviously something is wrong.

Wrong all around.

If you ordered that much ram, at that cost, (phew!) then it should be perfectly matched, not just a "oh well, there's a bug on your end, it looks like maybe it's working, sort of..."

Send the whole thing back, every last stick (save any receipts, emails, texts or other correspondence, take good pics of differences between the sticks) and tell Crucial you either want satisfaction for what you paid for, or a refund so you can purchase from Corsair instead.

 

[Darkbreeze]

Exactly. That's exactly what I'm saying. I wouldn't take no for an answer and I'd ask to kick it up the food chain if anybody wants to give you grief. You should get exactly what you paid for, not just oh well settle for this.

 

[jhayat2]

From the first pic, it clearly also states dual channel. With 8 slots populated, you should most definitely be running in quad channel if all that ram is the same.

So, that pic was with two modules installed. I just threw in the others, and with all 8, CPU-z reports QUAD in that box.

See the week/year in the pics? Is that when they were manu'd? The seven all say 15/18 - 3rd wk of April, 2018. The one says 2/17 - 2nd wk of Jan, 2017. Hmmmm...

If you ordered that much ram, at that cost, (phew!)

$2,118.47

then it should be perfectly matched, not just a "oh well, there's a bug on your end, it looks like maybe it's working, sort of..."

You should get exactly what you paid for, not just oh well settle for this.

Can't agree more. Let's see what they say when they come back tomorrow.

 

[Darkbreeze]

If I might ask, what are you DOING that you need that much RAM? Running a bunch of VM's or serving off that machine?

 

[jhayat2]

I need to add more stuff!

I write music for a living, primarily using sample libraries. The audio files of which are (partially) loaded into RAM. I have a HUGE template, and my current machine, which has 64GB of RAM, is almost maxed out. I need to add more libraries (.wav files) to my template/RAM. I have a second machine, which, through an app called VEPro, enables me to have more stuff loaded on that machine, and have the samples come back to my main machine over LAN, but that 2nd machine is maxed out @ 24 GB. So, I will use my new 128GB machine as my main machine, and this one as my secondary, and take what is now my secondary OOS.

Cheers.

 

[jhayat2]

So here's what I get back a few moments ago:

Hello Jeff,

We are currently still waiting to hear back from our engineers on this matter. We have been told that the memory is running at the correct speeds and it looks like the SPD is just reading information incorrectly and is showing different timings then what it is actually running at.

We have no further updates beyond that and will need to wait until we hear back from our engineers on this. I do not have an ETA as to when we will hear back, but as soon as we get any updated information, we will be in contact with you.

Ok, now I am crying BS.

I told them that "it looks like the SPD is just reading information incorrectly and is showing different timings then what it is actually running at" is unacceptable, and demanded a proper answer by tomorrow.

 

[Darkbreeze]

Yes, complete BS. I'd tell them if they are going to do you like this, you are going to take your documented situation to someplace like HardOCP who doesn't mind running stories like this when companies turn crappy.

In the future, my recommendation would be (And usually is anyhow) to stick with G.Skill, Corsair and Mushkin when it comes to memory purchases. And probably in that exact order.

 

[jhayat2]

So, while I wait for Crucial.... I just realized something.

If we look back at those pics, #2 shows a "good" module, with JEDECs of 1200, and an XMP of 1600 (which is going to be half of 3200, b/c DDR4 is double data rate, I think?)

#3 shows the "bad" module, with JEDECs of 1066 (a concern for me), and an XMP of 1600 as well. Does this mean that this module is running at the same freq as the other?

 

[Darkbreeze]

Just because the profile is there doesn't mean that it's running at that speed.

I would suggest installing ONLY the questionable module, and enable it's XMP profile, then boot into windows and see what HWinfo tells you. Another item you might want to check out is Thaiphoon burner which will also give you detailed reports on the memory.

http://www.softnology.biz/files.html

That is interesting and kind of neat, the paid version will even allow you to burn new profiles to the modules or overwrite old ones, but truthfully the only thing you need to really do is download HWinfo, install it, run "Sensors only" option and then if you would, take a screenshot of the memory information in HWinfo and post that here.


*How to post images in Tom's hardware forums



Run HWinfo and look at system voltages and other sensor readings.

Monitoring temperatures, core speeds, voltages, clock ratios and other reported sensor data can often help to pick out an issue right off the bat. HWinfo is a good way to get that data and in my experience tends to be more accurate than some of the other utilities available. CPU-Z, GPU-Z and Core Temp all have their uses but HWinfo tends to have it all laid out in a more convenient fashion so you can usually see what one sensor is reporting while looking at another instead of having to flip through various tabs that have specific groupings.

After installation, run the utility and when asked, choose "sensors only". The other window options have some use but in most cases everything you need will be located in the sensors window. If you're taking screenshots to post for troubleshooting, it will most likely require taking three screenshots and scrolling down the sensors window between screenshots in order to capture them all.

It is most helpful if you can take a series of HWinfo screenshots at idle, after a cold boot to the desktop. Open HWinfo and wait for all of the Windows startup processes to complete. Usually about four or five minutes should be plenty. Take screenshots of all the HWinfo sensors.

Next, run something demanding like Prime95 version 26.6 or Heaven benchmark. Take another set of screenshots while either of those is running so we can see what the hardware is doing while under a load.

*Download HWinfo

 

[jhayat2]

Due to the amount of images, I skipped over the CPU and GPU, as I guess they are unnecessary here?

First three are the 2133 module by itself in the C1 slot, the next three are from the 2400 module.

The mem timings seem to be the same?

Thanks.

 

[Darkbreeze]

Based on those screenshots, I'd say they were right and that they are all running at the same timings at 3200mhz frequency. Weird that they are not the same in other reporting modules, but based on that HWinfo data it all looks right to me. Now install all the memory back in and take another screenshot of the memory settings to see if they retain the same speed and timings.

 

[Karadjgne]

I guess I'm a little picky. OP spent over $2k on ram that's supposed to be identical. Sure it all runs equal at 3200, but only when xmp is set. To me that'd be like buying 4 tires and 1 has a different tread pattern, even if same model and brand and size. I'm paying for 4 identical tires, not an oddball, even if it is close enough and does the job.

 

[jhayat2]

Thats with all 8 sticks.

So, I guess HWInfo doesn't show Individ modules? Still says 1600 (3200), while CPU-z still shows the same discrepancy. I am still concerned - if this app says 3200, why does the BIOs say 2400?

 

[Karadjgne]

If you assume the email is correct, and there's an issue with how the SPD is being read, then the real question is which is the correct reading? The software or the bios? If the bios is correct, the ram is funked and not working as it is supposed to. If the software is correct, the bios is wrong and there's nothing to worry about. Except the fact you can never be sure just what is happening and if the ram is really running at 3200 and not 2400.

 

[Darkbreeze]

I still agree that you SHOULD have a right to all identical modules. I won't argue that. I think I'd trust the HWinfo data over the bios though. BIOS could be buggy, HWinfo probably isn't. It's weird all the way around really.

I'd still insist that I bought a MATCHED set, and I WANT a matched set. If one module is different, it is NOT a matched set.

BUT, based on the screenshots it WOULD seem to be running right. Like I said before, download Thaiphoon burner and see what it says for EACH of the modules.

 

[jhayat2]

If you assume the email is correct, and there's an issue with how the SPD is being read, then the real question is which is the correct reading? The software or the bios? If the bios is correct, the ram is funked and not working as it is supposed to. If the software is correct, the bios is wrong and there's nothing to worry about. Except the fact you can never be sure just what is happening and if the ram is really running at 3200 and not 2400.

Could not have said it any better.

This is thaiburner on my current machine when I dbl-click the exe, or r-click and 'run as admin':

I then click close and have a gui.

This is thaiburner on my new machine when I dbl-click the exe, or r-click and 'run as admin':

I then click close and have no gui.

Not too sure what to make of that.... :/