How To RAM is running at 2133MHz instead of 3200MHz. Unable to fix this issue.

[ZenoSama]

Not sure where in the world you are, or how you purchased the RAM.

If I was in your shoes:
1) return as defective/not compatible since they can't even run at XMP speeds in a single stick installed configuration
2) look towards credit card return protection if seller won't take back
3) try to RMA to manufacturer
4) buy a new kit: RAM is pretty cheap. Sure, it stinks to have to shell out more, but it is not a lot more
5) try to sell ram you won't use off

I'm from India. Purchased from a known seller in local PC parts market.

 

[ZenoSama]

Hmm, that is a way to look at it, like a factory OC. Like how some custom cards have higher rated clocks then reference. So not technically an end user, warranty voiding OC, (as the RAM is rated to run at that by the manufacturer) but still a higher speed than standard.

As for the issue, @ZenoSama have you tried memtesting for stability, at both DDR4 2133 and the XMP settings? If you find errors on either setting, chances are the RAM is defective. I agree that you should try to return the RAM, or get a refund from CC.

Personally I would be looking at good Gksill kits, DDR4 3200 to 3600. You can also look into what actual ICs different RAM kits have, and try to buy one with Samsung Bdie, or similar that are known to work well with Ryzens, particularly Zen 2 and 3. For instance, you could look into Gskill Ripjaws, TridentZ or FlareX with 3200C14 ratings, those are almost all Bdie kits.

I have plans for memtesting, but I'm trying to see if I can just get these RAM sticks exchanged or replaced via my seller for a better compatible sticks (DDR4, 16GB-dual-channel, 2933/3200MHz, maybe C16 or C18).

Is there a posibility that memtest might damage the RAM sticks ? If not, I can maybe run it and find out if they are defective afterall, so that I can have a solid proof to get them exchanged for like a G-Skill kit.

Even with G-skill or any brand, I just want to make sure that if I end up having an option to get a kit as a replacement for the existing ones, I don't wanna end up buying a kit that will have similar issues. Any recommendations regarding this ? Asking this because, no matter how much I search, most recommendations are the same RAM sticks that I have with me right now. IDK if they are using a different model RAM stick (having the same name).

 

[Shmee]

Memtest shouldn't cause any damage, and if it is showing errors, that means there was already an issue. I agree with you wanting to return the 2 separate sticks for a good kit, and if they error in memtest, this could help prove they are defective.

 

[ZenoSama]

Memtest shouldn't cause any damage, and if it is showing errors, that means there was already an issue. I agree with you wanting to return the 2 separate sticks for a good kit, and if they error in memtest, this could help prove they are defective.

How long should I run memtest ? @Shmee

Anything over stock RAM speed is an overclock. XMP is overclocking.

I suspect the B450 board specs just say that 3200 is overclock because of the available processors at the time of release not officially supporting those faster speeds.

You could try: enable XMP (but don't save and exit the BIOS yet), then reduce the RAM multiplier so you are running at 3000 or 2933. You shouldn't really have to adjust the voltages or anything.

But the best solution would be to just get a kit of 3200 and try that instead.


Yeah, that's why you don't mix and match RAM. Suppliers will sometimes change underlying chips without changing model numbers, causing all sorts of issues. However, if even individual sticks aren't running at the rated speed, I'd opt for a return under the idea that something is defective.

@Brainonska511 I tried the part where you asked me to enable XMP but lower the RAM multiplier to 2933. It ACTUALLY booted to windows. Checked task manager. It shows RAM speed as 2933Mhz (DDR4-3200 16-20-20-38-58-1.35V) . In CPU-Z, the Uncore Frequency and DRAM frequency is showing around 1463MHz. But setting the RAM multiplier to anything higher than 2933 results in no boot.

 

[Shmee]

I would run download the latest version of memtest86, put it on a bootable USB, and run it in UEFI mode for at least 2 passes. Perhaps just run it for the maximum length of 4 passes. You can always restart and run again as well. Alternatively, if Memtest86 can't find any errors, you could try OCCT memtest in Windows, and even try out the other OCCT stress tests.

 

[igor_kavinski]

I tried the part where you asked me to enable XMP but lower the RAM multiplier to 2933. It ACTUALLY booted to windows.

I think you should be content with that. It's possible that your particular mobo does not support running DDR4-3200 with this CPU or maybe can't run DDR4-3200 with ANY CPU. If your current timings are 16-20-20-38-58, you could try tightening them a bit, something like 15-18-18-35-55. If you get a Windows boot and it seems stable and Memtest is successful, try 14-18-18-35-55 and then leave it at that if that works too.

 

[Brainonska511]

How long should I run memtest ? @Shmee

@Brainonska511 I tried the part where you asked me to enable XMP but lower the RAM multiplier to 2933. It ACTUALLY booted to windows. Checked task manager. It shows RAM speed as 2933Mhz (DDR4-3200 16-20-20-38-58-1.35V) . In CPU-Z, the Uncore Frequency and DRAM frequency is showing around 1463MHz. But setting the RAM multiplier to anything higher than 2933 results in no boot.

Depending on where you look in cpu-z, it only shows the half-rate, so it sounds fine.

I'd let memtest run for at least a few hours: just let it do the basic tests: if there are errors, it will usually show quickly. It won't damage the RAM.

I would think the board + CPU could do 3200, but the mismatched sticks are preventing that. 3200 vs 2933, but maybe you just got unlucky with this board and it's official support cutting off at 2933. You probably won't notice much of a difference between 2933 and 3200. You could also try playing with the timings as someone else noted. It could help make up for the small loss in frequency.

I would continue to watch for other stability indicators running at 2933 (random system crashed, etc). Successful booting is just the first step.

 

[ZenoSama]

I think you should be content with that. It's possible that your particular mobo does not support running DDR4-3200 with this CPU or maybe can't run DDR4-3200 with ANY CPU. If your current timings are 16-20-20-38-58, you could try tightening them a bit, something like 15-18-18-35-55. If you get a Windows boot and it seems stable and Memtest is successful, try 14-18-18-35-55 and then leave it at that if that works too.

Yeah, I actually came across Ryzen DRAM calculator tool as well as memtest. Currently, I'm running memtest86 for one stick at a time to see if the RAM sticks are the issue or not. If they are all good, will run a memtest with both of them in dual channel and see if that raises any errors (although I'm not sure if I should do this coz like the first stick took close to 2hrs for 4 pass test).

If they clear all memtest, will see if I can use the DRAM calculator tool and take one step at a time towards C14.

 

[ZenoSama]

Depending on where you look in cpu-z, it only shows the half-rate, so it sounds fine.

I'd let memtest run for at least a few hours: just let it do the basic tests: if there are errors, it will usually show quickly. It won't damage the RAM.

I would think the board + CPU could do 3200, but the mismatched sticks are preventing that. 3200 vs 2933, but maybe you just got unlucky with this board and it's official support cutting off at 2933. You probably won't notice much of a difference between 2933 and 3200. You could also try playing with the timings as someone else noted. It could help make up for the small loss in frequency.

I would continue to watch for other stability indicators running at 2933 (random system crashed, etc). Successful booting is just the first step.

I'm actually testing each ram in their actual dual channel designated slots. RAM in DDR4_1 is currently under test (already been 2hrs. I thought each stick would take 1hr, like mentioned in the docs and corsair memtest article). After that, will have the next stick in DDR4_2 slot. I want to make sure MoBo is not the actual issue here

 

[Brainonska511]

I'm actually testing each ram in their actual dual channel designated slots. RAM in DDR4_1 is currently under test (already been 2hrs. I thought each stick would take 1hr, like mentioned in the docs and corsair memtest article). After that, will have the next stick in DDR4_2 slot. I want to make sure MoBo is not the actual issue here

Make sure you also test them together in the final configuration: it's an important check. Just run it overnight of you're worried about time.

Not sure I'd bother testing individual motherboard slots beyond the single slot you'd occupy in single channel, single stick mode (as in, the mode you're using to test individual sticks). Unless you come across some problem later (like both sticks test fine individually, but then have errors together; though, for that you could try stepping back on the multiplier, if that happens)

 

[ZenoSama]

Make sure you also test them together in the final configuration: it's an important check. Just run it overnight of you're worried about time.

Not sure I'd bother testing individual motherboard slots beyond the single slot you'd occupy in single channel, single stick mode (as in, the mode you're using to test individual sticks). Unless you come across some problem later (like both sticks test fine individually, but then have errors together; though, for that you could try stepping back on the multiplier, if that happens)

Oh, I forgot to tell you that I reset BIOS before I started memtest as if was a recommended step. Enabling XMP and doing memtest would lead to false positive type errors is what some articles pointed out and they all said to reset BIOS. So, did that.

 

[Brainonska511]

Oh, I forgot to tell you that I reset BIOS before I started memtest as if was a recommended step. Enabling XMP and doing memtest would lead to false positive type errors is what some articles pointed out and they all said to reset BIOS. So, did that.

You should test with xmp on and the multiplier reduced to the level that it boots at (2933). Getting no errors at 2133 doesn't really tell you about what is going on at 2933. Easy to find stories of memory fine at JEDEC standards, and then throw tons of real errors and stability issues at XMP speed.

 

[ZenoSama]

You should test with xmp on and the multiplier reduced to the level that it boots at (2933). Getting no errors at 2133 doesn't really tell you about what is going on at 2933. Easy to find stories of memory fine at JEDEC standards, and then throw tons of real errors and stability issues at XMP speed.

I just started the 2nd stick test. Will let it complete this. Then run it with both sticks.
And then repeat all three with 2933 and check stability. Is that better or would you suggest something better ?

 

[Brainonska511]

I just started the 2nd stick test. Will let it complete this. Then run it with both sticks.
And then repeat all three with 2933 and check stability. Is that better or would you suggest something better ?

Seems like a fine approach.

And if you do get any errors at 2933, step the multiplier back one more.

I have a B450 motherboard in my wife's computer that couldn't do 3000 MHz (or really any xmp) with a Zen 2 2600 - it would just crash randomly on her, so for a long time I left it at stock: 2400). I didn't spend much time trying to diagnose it. However, XMP runs fine now with the same board and a 5600X. Sometimes, it seems like luck of the draw.

 

[Shmee]

Yeah, memtesting extra never hurts. But if you do get any errors, whether at rated XMP settings or below, that right there tells you there is a problem, and most likely it will be with the RAM.

Now this could be an issue with your Ryzen APU's memory controller, but I find this less likely, as Zen 3 is far newer and more robust from the Zen+ 2600 that is mentioned (keep in mind Ryzen 2000 was considered Zen+, while the 3000 series was Zen 2, and 5000 being Zen 3. Yeah, kind of weird.)

But yeah, if you get even one error at rated DDR4 or lower, that should be sufficient evidence to return the RAM as defective.

And just double checking, but you did get the newest BIOS for your motherboard right? Latest AGESA have a lot of the older memory issues worked out, so this shouldn't be a motherboard or BiOS issue.

 

[ZenoSama]

Yeah, memtesting extra never hurts. But if you do get any errors, whether at rated XMP settings or below, that right there tells you there is a problem, and most likely it will be with the RAM.

Now this could be an issue with your Ryzen APU's memory controller, but I find this less likely, as Zen 3 is far newer and more robust from the Zen+ 2600 that is mentioned (keep in mind Ryzen 2000 was considered Zen+, while the 3000 series was Zen 2, and 5000 being Zen 3. Yeah, kind of weird.)

But yeah, if you get even one error at rated DDR4 or lower, that should be sufficient evidence to return the RAM as defective.

And just double checking, but you did get the newest BIOS for your motherboard right? Latest AGESA have a lot of the older memory issues worked out, so this shouldn't be a motherboard or BiOS issue.

Yup. That's the first thing I did when my CPU landed the MoBo. I used the Q-Flash tool in bios and updated it to latest F64 version.

Individual sticks seem to be working fine at their base clocks of 2133MT/s. Both of their tests are looking green. Lets see what the results will look like, when put to test together.

Also, a weird thing. On Day 1, when I had Windows 10, without XMP, system would never POST with just one of the either sticks, which I actually found it weird. I installed W11 fresh yesterday. And today just before memtest, wanted to see if system POST and boot to windows. And it DID ! (actually weird). I feel like there's a lot of minute factors influencing stability here.

And then, I came across this :

https://www.reddit.com/r/overclocking/comments/mcfkan

Literally an indirect middle finger to memtest and ryzen dram calculator 😅

 

[kschendel]

I think there are a couple things going on here.

First, as others have suggested, ditch those sticks and buy a matched pair. Kingston is good if you can get them. Memory vendors actually buy the memory chips from one of a very few manufacturers (there are only two or three) and then put together and grade the sticks. I've not had any real trouble with Kingston, Team, G.Skill, Oloy. I doubt that Corsair is statistically worse, I haven't tried them.

The other point that hasn't been discussed is that slower memory shouldn't matter all that much. You might see a difference in benchmarks, and if you're using the integrated graphics, you'll probably see a gaming difference because the APU wants all the memory bandwidth it can get. But I just don't see slower memory causing significantly slower application loading; the effect is on the order of a few percentage points, and if you can subjectively detect that, you're doing a lot better than me. It's possible that Windows decided it couldn't use all of your memory so it was only running with 8GB, and that could matter.

So yes, resolve your memory problems and get a full kit of stable memory in there. Then you can figure out whether you still have an issue somewhere in the build or software.

 

[Mantrid-Drone]

Maybe I missed this earlier in the discussion and it may be irrelevant anyway but just for clarity:-​

When ZenoSama says he is using RAM sockets 1 & 2 does that mean the first two adjacent sockets of the four or the two pairs which are actually the dual channel enabled ones and, usually, (physical) sockets 1/3 and 2/4.​

 

[Shmee]

Yeah, I wouldn't bother with the DRAM calculator. Memtesting is important, but he is right that multiple tests are important for true stability and troubleshooting. The point being, I think, that you can't just rely on one tool. Also, if you are tweaking or troubleshooting, change one thing at a time.

 

[ZenoSama]

Maybe I missed this earlier in the discussion and it may be irrelevant anyway but just for clarity:-​

When ZenoSama says he is using RAM sockets 1 & 2 does that mean the first two adjacent sockets of the four or the two pairs which are actually the dual channel enabled ones and, usually, sockets 1/3 and 2/4.​

I actually meant by the color paired slots on the mother board, starting from the farthest slot from CPU (closer to the edge of the MoBo). I also referred the manual which said to use DDR4_1 and 2 for dual channel. I looked up the slot markings on the MoBo and installed accordingly.

 

[ZenoSama]

I think there are a couple things going on here.

First, as others have suggested, ditch those sticks and buy a matched pair. Kingston is good if you can get them. Memory vendors actually buy the memory chips from one of a very few manufacturers (there are only two or three) and then put together and grade the sticks. I've not had any real trouble with Kingston, Team, G.Skill, Oloy. I doubt that Corsair is statistically worse, I haven't tried them.

The other point that hasn't been discussed is that slower memory shouldn't matter all that much. You might see a difference in benchmarks, and if you're using the integrated graphics, you'll probably see a gaming difference because the APU wants all the memory bandwidth it can get. But I just don't see slower memory causing significantly slower application loading; the effect is on the order of a few percentage points, and if you can subjectively detect that, you're doing a lot better than me. It's possible that Windows decided it couldn't use all of your memory so it was only running with 8GB, and that could matter.

So yes, resolve your memory problems and get a full kit of stable memory in there. Then you can figure out whether you still have an issue somewhere in the build or software.

Same thoughts. And that's why I'm testing to see if any issues are present in the sticks. Sellers won't accept a return or even an exchange without good proof most times. I did try the fact that even though its rated to run at 3200MHz but failed to post when I did that in BIOS and the response was that either my MoBo or CPU have issues and not the RAM stick.

 

[ZenoSama]

RAM Configurator - G.SKILL International Enterprise Co., Ltd.

Find compatible memory kits for your motherboard. Just select the motherboard from the dropdown list and click on Search for a list of compatible G.SKILL memory kits.

www.gskill.com

I'm thinking of getting the ripjaws V, as they seem to fit my expenses at the moment (currently on a tight budget and can't get the trident Zs). But I don't want to fall back to the same issue where enabling XMP will fail to post. How can I make sure about this before purchase ? Like they too have similar looking models with model no.

F4-3200C16D-16GVRB, F4-3200C16D-16GVKB, F4-3200C16D-16GVGB​

Is it just color difference or is there an actual difference these ?



And also, I feel like I should have gone with B550 instead of B450 MoBo. Seems like it has better hardware for an extra INR2000( 25 dollars).

 

[ZenoSama]

I think you should be content with that. It's possible that your particular mobo does not support running DDR4-3200 with this CPU or maybe can't run DDR4-3200 with ANY CPU. If your current timings are 16-20-20-38-58, you could try tightening them a bit, something like 15-18-18-35-55. If you get a Windows boot and it seems stable and Memtest is successful, try 14-18-18-35-55 and then leave it at that if that works too.

How do I set these timings in BIOS ? Like which fields to change and how many steps per change ?

 

[Brainonska511]

How do I set these timings in BIOS ? Like which fields to change and how many steps per change ?

Perhaps it would be best to focus on the overall RAM multiplier issue. Tightening timings probably won't do much.

 

[igor_kavinski]

How do I set these timings in BIOS ? Like which fields to change and how many steps per change ?

So what have you decided? Are you going to get G.Skill and B550?

For memory timings, check the CPU-Z memory tab and you will know what each timing parameter is called and then locate it in BIOS and change it. Usually, it's better to change them one step at a time.