Question Failed stress test after enabling XMP

[Orgazia]

Hello there,

So recently I've upgraded my CPU to 5800x3d from 3600. I've always had an unstable system with enabled XMP. But I've never really needed the extra ram boost so I've never tried to figure the issue out, so I ran it at base frequency which js 2133mhz.

After the upgrade I want to run some of the latest games (cyberpunk and hogwarts legacy) so I want to push my ram a bit. So I enabled the xmp profile 1 and ran an OCCT test and my pc froze. I lowered the frequency to 3466mhz and pc didn't post. So I reset the CMOS and got it to post.

I am very new to OC. So I didn't want to mess around too much in the bios, blindly.

Here are my specs:

CPU - AMD Ryzen 7 5800x3d
Mobo- MSI B450 Gaming Plus Max - latest bios
Ram- G.Skill Ripjaws 32gb (4x8gb)
Gpu- MSI 6750xt 12gb oc
Psu- Corsair RM750 Gold, 750W.
Operating drive- WD Blue Sn550, m.2, 500gb.

Forgot to add. Stress test on default settings was a pass. I also got Kombo Strike set at 3.

What should i try to up my RAM frequency?

Thank you in advance.

 

[Roland Of Gilead]

Hey there,

There's a few things to look at here.

So recently I've upgraded my CPU to 5800x3d from 3600. I've always had an unstable system with enabled XMP. But I've never really needed the extra ram boost so I've never tried to figure the issue out, so I ran it at base frequency which js 2133mhz.

So the difference in ram at 2133mhz and 3600mhz, is enormous, roughly 40% increase in gaming FPS and also for apps in general. It's hobbling your system in terms of performance, and the same could be said of the R 5 3600 too. Lets aim to get the ram running XMP.

So I enabled the xmp profile 1 and ran an OCCT test and my pc froze. I lowered the frequency to 3466mhz and pc didn't post. So I reset the CMOS and got it to post.

So, you mentioned you have 4 x 8gb 3600mhz, right? Were the DIMMs all part of one 4 pack, or two separate 2x kits? If so, using 4 x DIMMs often means a trade off in max mem speed because all four slots are occupied. Typically it may limit max speed to 3200mhz or 2666mhz (Jedec default). Not only that, ideally you needed to get a 2x16gb kit which would be optimal, as opposed to the 4 x 8gb.

If they are two separate packs, then try only 2 x kit matched and XMP. See how it works. More stable? Still issues? etc

You can also test both kits out with memtest86+ outside of windows to detect faulty DIMMs. Try each pack, then each individual DIMM and rule out errors. If you have any errors on memtest after 4 passes, your ram is faulty.

Let us know how that goes.

 

[Orgazia]

Hey there,

There's a few things to look at here.
How

So the difference in ram at 2133mhz and 3600mhz, is enormous, roughly 40% increase in gaming FPS and also for apps in general. It's hobbling your system in terms of performance, and the same could be said of the R 5 3600 too. Lets aim to get the ram running XMP.

So, you mentioned you have 4 x 8gb 3600mhz, right? Were the DIMMs all part of one 4 pack, or two separate 2x kits? If so, using 4 x DIMMs often means a trade off in max mem speed because all for slots are occupied. Typically it may limit max speed to 3209, mhz or 2666mhz (Jedec default). Not only that, ideally you needed to get a 2x16gb kit which would be optimal, as opposed to the 4 x 8gb.

If they are two separate packs, then try only 2 x kit matched and XMP. See how it works. More stable? Still issues? etc

You can also test both kits out with memtest86+ outside of windows to detect faulty DIMMs. Try each pack, then each individual DIMM and rule out errors. If you have any errors on memtest after 4 passes, your ram is faulty.

Let us know how that goes.

Thank you. I'm gonna try it all out and update. I got 2 different kits for the RAM. So I'm going to down grade to 2 identicals and try and sell the other 2.

Should I use OCCT to stress test or stick with memtest86+?

Also, would you suggest disabling Kombo Strike before stabilizing the RAM?

 

[Roland Of Gilead]

Thank you. I'm gonna try it all out and update. I got 2 different kits for the RAM. So I'm going to down grade to 2 identicals and try and sell the other 2.

Should I use OCCT to stress test or stick with memtest86+?

Also, would you suggest disabling Kombo Strike before stabilizing the RAM?

Using OCCT with just the mem test if fine, however, memtest will test outside of windows, thus reducing the chances some windows bug or something.

I've not used Kombo Strike, but yes, I would disable that.

 

[Orgazia]

Using OCCT with just the mem test if fine, however, memtest will test outside of windows, thus reducing the chances some windows bug or something.

I've not used Kombo Strike, but yes, I would disable that.

So I've finally was able to run a few memtest tests on 2 sticks. On default settings it was a pass. Xmp enabled with frequency set to 3600mhz and 3200mhz was a major fail. Tested the other pair, same thing.

I read somewhere that an issue might be not enough voltage, could that be an issue? Or should I try something else?

Ram is installed in the slots, as per manual.

 

[kira-faye]

Bump VDIMM up 0.1v and see how that goes. If it fails, RMA the RAM - everything from G.Skill is lifetime warranty, assuming you have proof of purchase. If you bought it used, probably out of luck and time to get a new kit.

 

[Orgazia]

Bump VDIMM up 0.1v and see how that goes. If it fails, RMA the RAM - everything from G.Skill is lifetime warranty, assuming you have proof of purchase. If you bought it used, probably out of luck and time to get a new kit.

Would you recommend just enabling the xmp or setting up frequency/latency /voltage manually? Gonna try it later today. Really hope it works.

 

[kira-faye]

Would you recommend just enabling the xmp or setting up frequency/latency /voltage manually? Gonna try it later today. Really hope it works.

Might as well set it manually. XMP is one of those things that "should" work and is like 50/50 in practice, from my experience.

 

[Roland Of Gilead]

Would you recommend just enabling the xmp or setting up frequency/latency /voltage manually? Gonna try it later today. Really hope it works.

This is also an option. Most of the timings required to change will be on the label on the DIMM. You can also read them with CPU-z.

Just to be clear, when you enable XMP on the ram, it fails memtest?

 

[Orgazia]

This is also an option. Most of the timings required to change will be on the label on the DIMM. You can also read them with CPU-z.

Just to be clear, when you enable XMP on the ram, it fails memtest?

Yea, it fails with xmp enabled, few seconds in, errors galore.🙁

 

[Roland Of Gilead]

Yea, it fails with xmp enabled, few seconds in, errors galore.🙁

Okay, simples, RMA them. Forget trying to troubleshoot.

 

[Orgazia]

Okay, simples, RMA them. Forget trying to troubleshoot.

Will do! Thank you all for all the help. I'm still going to try the voltage just for good measure lol

 

[Roland Of Gilead]

Will do! Thank you all for all the help. I'm still going to try the voltage just for good measure lol

You're welcome! Good luck. Let us know if you get it sorted

 

[Orgazia]

You're welcome! Good luck. Let us know if you get it sorted

So the voltage didn't work either. I had a pair of corsair sticks that I bought because it was a really good price (lol) that I tried, xmp enabled @3200mhz and it was a pass. So I'm gonna use them for a bit. Gonna try the other pair of g.skill at some point.

Also, started rma on the initial pair. Never done it before, so we'll see how she goes.

What a process though. Tried about 10 different combinations of sorts. Luckily a lot of them failed at the beginning of tests lol.

 

[PiersPlowman]

Memory is only a bottleneck where memory is the bottleneck. Sounds dumb, but it needs stating that (and contrary to popular opinion) faster memory does not automatically translate into MoAr FPS. In many cases you won't know any difference outside of AIDA.

What faster memory does do is eat into the CPU's power budget (reducing clocks) and make idle power draw higher. Noticeably so, in some cases.

 

[Orgazia]

Memory is only a bottleneck where memory is the bottleneck. Sounds dumb, but it needs stating that (and contrary to popular opinion) faster memory does not automatically translate into MoAr FPS. In many cases you won't know any difference outside of AIDA.

What faster memory does do is eat into the CPU's power budget (reducing clocks) and make idle power draw higher. Noticeably so, in some cases.

Care to elaborate? In terms of settings, what can I try?

 

[SkyNetRising]

Care to elaborate? In terms of settings, what can I try?

Can you show screenshots from CPU-Z - memory and spd sections?
(upload to imgur.com and post link)

 

[Orgazia]

Can you show screenshots from CPU-Z - memory and spd sections?
(upload to imgur.com and post link)

Why? What am I looking for?

 

[PiersPlowman]

Care to elaborate? In terms of settings, what can I try?

Each Ryzen processor is allocated a power budget. I'm the case of your old 3600 that was 88 watts. With my 5600 it is 77 watts. Your 5800x 3D will be more, but I cannot quote you a figure from memory.

This power budget is shared between the CPU cores and the SoC - a separate chip on the CPU die that controls memory and input/output. This SoC will typically draw around 20 watts (varies slightly depending on CPU and RAM), but enabling XMP / DOCP jacks up a bunch of voltages, which, together with the higher clocks, results in increased power consumption. Now the SoC burns juice at a constant rate no matter what you are doing with the PC, and this includes doing nothing.

Remember that power budget I mentioned? Well, on a 77 watt processor, instead of 20 of those watts being eaten up by the SoC, you now have 30 (or more). As a result, your CPU cores have less power to work with and your boost clocks will go down.

 

[SkyNetRising]

Why? What am I looking for?

You're looking for information contained in CPU-Z screenshots.
Currently used ram settings, spd and xmp table entries.
This information is necessary for manual ram overclocking, since automatic overclock is failing.