• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Ryzen 7 7800X3D Issues

I got the Trident f5-6000J3038F16GX2-TZ5NR
Might be interesting to post a ZenTimings screenshot of your memory.

If the motherboard and memory combination is the culprit, we might see some odd voltages in that screenshot.

Generally speaking, the old kit would have been perfectly fine, if you know how to tune memory manually on Ryzen 7000. If you don't know, don't buy G.Skill kits since their XMP/EXPO profiles assume that your CPU is actually able to run that speed which isn't a given for DDR5-6400 and might still need some manually tweaking on DDR5-6000 in case you got really unlucky with your CPU.
If you want "safe" XMP/EXPO profiles, buy Kingston Renegade or Fury Beast kits with the lowest tCL available, usually CL 30 or 32. Those kits will typically come with two speed bins. If the higher one doesn't work, they lower one usually will.
 
Generally speaking, the old kit would have been perfectly fine, if you know how to tune memory manually on Ryzen 7000. If you don't know, don't buy G.Skill kits since their XMP/EXPO profiles assume that your CPU is actually able to run that speed which isn't a given for DDR5-6400 and might still need some manually tweaking on DDR5-6000 in case you got really unlucky with your CPU.
If you want "safe" XMP/EXPO profiles, buy Kingston Renegade or Fury Beast kits with the lowest tCL available, usually CL 30 or 32. Those kits will typically come with two speed bins. If the higher one doesn't work, they lower one usually will.
What nonsense is this?
 
Check if you have problems with stock settings. So ram to 4800 without the EXPO.
And if there arent any issues. its the RAM.
Also you could slowly step the ramspeeds up, untill you know where it issues lies with in the speeds.
 
What nonsense is this?
You are aware that nobody is forced to use an XMP/EXPO profile, if they want to set their memory timings manually.
Not sure what you are referring to, but AMD only guarantees DDR5-5200 speed on JEDEC timings with their Ryzen 7000 CPUs. Just because DDR5-6000 is referred to as sweet-spot, it doesn't mean it will work on every CPU.

Also, most higher end G.Skill kits only come with one single XMP profile, unlike Kingston kits that typically include two.
 
Might be interesting to post a ZenTimings screenshot of your memory.

If the motherboard and memory combination is the culprit, we might see some odd voltages in that screenshot.

Generally speaking, the old kit would have been perfectly fine, if you know how to tune memory manually on Ryzen 7000. If you don't know, don't buy G.Skill kits since their XMP/EXPO profiles assume that your CPU is actually able to run that speed which isn't a given for DDR5-6400 and might still need some manually tweaking on DDR5-6000 in case you got really unlucky with your CPU.
If you want "safe" XMP/EXPO profiles, buy Kingston Renegade or Fury Beast kits with the lowest tCL available, usually CL 30 or 32. Those kits will typically come with two speed bins. If the higher one doesn't work, they lower one usually will.

I have never had a G.Skill kit refuse to run at the sweet spot frequency. G.Skill bins are almost always better than other vendors as well, either through subtimings or lower voltage. Corsair for example often needs more voltage than G.Skill to achieve the same frequency and worse timings. DDDR5 6000 will work on every Ryzen processor regardless of whether you win or loose the silicon lottery, especially since Hynix A die started replacing M die. With M die there was a 0.01% tiny chance that a DDR5 6000 kit from any vendor (not just G.Skill) might not work properly but current timings used for A die are extremely relaxed compared to what it's capable of so even a CPU with a horrid memory controller is going to be fine.

Let's not discuss brands, that will just derail the conversation and detract from helping the OP. The OP doesn't need to buy yet another RAM kit, they need to test what they already have.

Check if you have problems with stock settings. So ram to 4800 without the EXPO.
And if there arent any issues. its the RAM.
Also you could slowly step the ramspeeds up, untill you know where it issues lies with in the speeds.

Which reminds me, OP should try swapping RAM slots as well. In terms of diagnosing the memory, it's possible (but not likely) that memory instability be caused by the motherboard and can be "fixed" by reducing memory frequency. Sometimes this can be caused by a defective motherboard or over-tightened CPU cooler mounting screws and it may affect just a single RAM slot or all of them. Just another thing to knock off the list of course, it isn't a very likely scenario.
 
Usermenkbarch is a fraud, and soon to be a paywalled fraud. If you asked them the reason your machine blows is because you have an AMD processor, I kid you not.

Update BIOS to get AGESA up to par, install Windows and your games on the SSD (reminder to certify that all your drivers are correctly installed, including chipset, audio, etc. - I see a LOT of people just using the Microsoft audio driver obliviously because "some sound comes out" and they couldn't fathom the fact they aren't even using chipset drivers to have their motherboard operate correctly) - use HDD only and exclusively for files that aren't frequently accessed or audio and video files exclusively. No HDD models, including high-RPM or large cache models are performant enough to run any modern video game off them.

You are aware that nobody is forced to use an XMP/EXPO profile, if they want to set their memory timings manually.
Not sure what you are referring to, but AMD only guarantees DDR5-5200 speed on JEDEC timings with their Ryzen 7000 CPUs. Just because DDR5-6000 is referred to as sweet-spot, it doesn't mean it will work on every CPU.

Also, most higher end G.Skill kits only come with one single XMP profile, unlike Kingston kits that typically include two.

I consider myself a pretty well versed fellow, but I take cow's advice without thinking twice when it comes to memory ;)

But anyway update AGESA and you're golden, 6000 is easy peasy to achieve regardless of setup. That memory kit should run its XMP or EXPO in a 7800X3D without any difficulty whatsoever. OP's issues lie elsewhere, I reckon
 
If you don't know, don't buy G.Skill kits since their XMP/EXPO profiles assume that your CPU is actually able to run that speed which isn't a given for DDR5-6400 and might still need some manually tweaking on DDR5-6000 in case you got really unlucky with your CPU.
Again, what is this is nonsense? How does G.SKILL assumes your CPI have a given limit? They have QLVs you know.

If you want "safe" XMP/EXPO profiles, buy Kingston Renegade or Fury Beast kits with the lowest tCL available, usually CL 30 or 32. Those kits will typically come with two speed bins. If the higher one doesn't work, they lower one usually will.
How does Kingston Renegade or Fury provide "safe" EXPO / XMP if you run at the same speed to G.SKill

Edit: After reading into this more, I can translate. G.SKILL doesn't have multiple XMP/EXPO profiles in their kits so you either get the sold bin or JEDEC to pick from. Kingston, PNY, Patriot, Neo Forza and a few others have a extra lower bin profile.

Maybe next time take the time to explain what your saying. Otherwise it is just nonsense.
 
With M die there was a 0.01% tiny chance that a DDR5 6000 kit from any vendor (not just G.Skill) might not work properly but current timings used for A die are extremely relaxed compared to what it's capable of so even a CPU with a horrid memory controller is going to be fine.
Going from Hynix M-die to A-die didn't have much to do with memory compatibility.
AMD releasing AGESA 1.0.0.7c and updating their IMC was one of the two reasons that allowed almost every CPU running DDR5-6000. Luckily the OP has a BIOS that includes those updates. That said, AMD still doesn't guarantee anything above DDR5-5200 JEDEC according to their own website.
The other reason why most CPUs can actually run most XMP/EXPO profiles nowadays are BIOS updates that included preloaded settings for many, many memory kits and manufacturers in the last months. Especially voltages are still very tricky, if the motherboard doesn't preload them. According to ASRock's website there are two beta BIOSes available for the board that the OP uses. Might be worth checking out.
Let's not discuss brands, that will just derail the conversation and detract from helping the OP. The OP doesn't need to buy yet another RAM kit, they need to test what they already have.
Well, I didn't suggest that the OP should blindly replace his memory kit. That might be a discussion if anything substantial hints at memory being the culprit. That's the reason I asked for that ZenTimings screenshot in the first place.
Again, what is this is nonsense? How does G.SKILL assumes your CPI have a given limit? They have QLVs you know.
And none of those two kits that the OP bought are supported by G.Skill's own QVL on his motherboard. That means that he has to hope that ASRock preloads the proper BIOS settings for those kits.
 
And none of those two kits that the OP bought are supported by G.Skill's own QVL on his motherboard. That means that he has to hope that ASRock preloads the proper BIOS settings for those kits.

That isn't how it works at all, the motherboard will simply load the values stored on the memory's XMP/EXPO data banks. And QVL doesn't always mean full compatibility: my kit is the highest end one on my MSI Z690 ACE's QVL and it flat out doesn't run at 6800, even with a i9-13900KS processor. I was able to run tightened 6400 which in all fairness, works just as well, if not better.

The critical issue with AMD and memory compatibility is and has always been AGESA, but we're in a tangential here: this more than likely doesn't have anything to do with OP's stuttering problems.
 
That isn't how it works at all, the motherboard will simply load the values stored on the memory's XMP/EXPO data banks.
Not necessarily, only if the BIOS doesn't come with settings to preload.
Especially for a lot of AM5 boards there have been many BIOS updates that specifically mention "improved memory compatibility" by pretty much all major manufacturers which usually translate into some sort of preloaded setting(s), mostly voltages it seems.

From my personal experience, one 96GB G.Skill kit that I own comes with 1.4V VDD & VDDIO, 1.35V VDDQ. If I use it on an older Gigabyte BIOS on either my B650E Aorus Master or X670E Aorus Pro X it uses those settings. However, on later BIOSes both boards preload 1.25V VDDIO & VDDQ since that seems to be the preferred signalling voltage by Gigabyte.
EVGA did something similar with their older DDR4 boards and outright preloaded Samsung B-die settings regardless of the memory sticks in use.

Just because you set an XMP or EXPO profile, doesn't mean the board will necessarily have to use all of those settings. It's also possible to activate additional settings if the manufacturer wants them to. E.g., Gigabyte changes some memory training parameters if you select XMP or EXPO on AM5. IIRC, ASRock even adjusts 2:1 mode at a certain speed threshold.

Keep in mind that modern UEFI acts like its own OS to a certain degree and can do a lot of things the BIOSes of the old days couldn't do. Injecting driver installers into a fresh Windows install is only a minor thing that is possible nowadays.

this more than likely doesn't have anything to do with OP's stuttering problems.
I doubt that as well, but it would be nice to see the OP's memory voltages (and timings) to be sure about that.
 
Not necessarily, only if the BIOS doesn't come with settings to preload.
Especially for a lot of AM5 boards there have been many BIOS updates that specifically mention "improved memory compatibility" by pretty much all major manufacturers which usually translate into some sort of preloaded setting(s), mostly voltages it seems.

From my personal experience, one 96GB G.Skill kit that I own comes with 1.4V VDD & VDDIO, 1.35V VDDQ. If I use it on an older Gigabyte BIOS on either my B650E Aorus Master or X670E Aorus Pro X it uses those settings. However, on later BIOSes both boards preload 1.25V VDDIO & VDDQ since that seems to be the preferred signalling voltage by Gigabyte.
EVGA did something similar with their older DDR4 boards and outright preloaded Samsung B-die settings regardless of the memory sticks in use.

Just because you set an XMP or EXPO profile, doesn't mean the board will necessarily have to use all of those settings. It's also possible to activate additional settings if the manufacturer wants them to. E.g., Gigabyte changes some memory training parameters if you select XMP or EXPO on AM5. IIRC, ASRock even adjusts 2:1 mode at a certain speed threshold.

Keep in mind that modern UEFI acts like its own OS to a certain degree and can do a lot of things the BIOSes of the old days couldn't do. Injecting driver installers into a fresh Windows install is only a minor thing that is possible nowadays.


I doubt that as well, but it would be nice to see the OP's memory voltages (and timings) to be sure about that.

Improve memory compatibility is largely due to AGESA on AMD and microcode on Intel, though. Mobo makers have a ton less input on this than one may initially think. It's part of the CPU initialization routine and takes precedence to even the UEFI :)
 
Improve memory compatibility is largely due to AGESA on AMD and microcode on Intel, though.
AGESA is manufacturer agnostic and is only supplied as binary blob by AMD as far as I know.

Do you have any source that explains how AGESA sets signalling voltages differently for boards of different manufacturers or why it sets different voltages for different PMICs used by different memory module manufacturers? I observed the behavior with different signalling voltages between a beta and release BIOS from Gigabyte on one of my boards. Both BIOS versions were supposed to have the same AGESA.
Something similar happens if I use another memory kit that does come with different VDDQ & VDDIO settings. Old BIOS uses 1.3V each, new BIOS defaults them to 1.25V like the G.Skill kit.

Also how does AGESA change the training parameters in the UEFI? Not trying to be dense here, but that's the first time I see someone basically claiming that it does. Then again, from my experience with Gigabyte boards, enabling XMP/EXPO for the first time definitely changes some of the visible/adjustable training parameters in the BIOS. Usually memory trains faster after that.
 
XMP/EXPO profile only stores a dozen values, EXPO has a few more. The rest come from motherboard tables either by using AMDs suggest values from the current AGESA code or the vendor BIOS team may tinker with it more, but that is usually reversed for the performance mode settings.
 
And none of those two kits that the OP bought are supported by G.Skill's own QVL on his motherboard.
Good point. :) If everything fails it might be worth a shot to get a kit that is on the supported list for testing:
AsRock B650M Pro RS WiFi: Memory QVL List

Especially for a lot of AM5 boards there have been many BIOS updates that specifically mention "improved memory compatibility" by pretty much all major manufacturers which usually translate into some sort of preloaded setting(s), mostly voltages it seems.
There where loads of BIOS updates for the board (inclusive for memory compatibility). OP is on v2.02 and there are 2 newer beta versions available.
AsRock B650M Pro RS WiFi: BIOS Updates
AsRock B650M Pro RS WiFi: Driver & Software Downloads

The games are installed on the SSD, the HDD is for my videos / movies I have downloaded. I recently upgraded from a i7 - 9700k to my new Ryzen along with a new motherboard and new ram. Regardless though, the freezing is happening outside of the games so I don't think the drives are the problem.
I would try what happens with the HDD not connected since your HDD performance numbers from your two benchmarks are quite random. The stuttering can happen from a HDD that has bad sectors and/or is on it's way out or has a bad Molex connection. Might run a HD Tune benchmark & a SMART scan (not the "Quick" one) on the drive.
 
Going from Hynix M-die to A-die didn't have much to do with memory compatibility.

It certainly does, A die can have tighter sub-timings than M-die and at lower voltages. By extension this means there's more leeway for the Motherboard on deciding how aggressive they want to set memory sub-timings(excluding those that are set by EXPO of course). ASUS and MSI in particular like to be more aggressive with their out of the box settings. For a CPU that might be riding the edge due to a low quality IMC, having A die over M die with a motherboard that sets aggressive sub-timings might be the difference between having a stable system or crashing.

AMD releasing AGESA 1.0.0.7c and updating their IMC was one of the two reasons that allowed almost every CPU running DDR5-6000. Luckily the OP has a BIOS that includes those updates. That said, AMD still doesn't guarantee anything above DDR5-5200 JEDEC according to their own website.

No, most AM5 CPUs could do that anyways prior to the update. AGESA 1.0.0. 7c mostly improved the max obtainable frequency and fringe stability issues. I don't really see the point of mentioning AMD doesn't guarantee hitting the sweet spot frequency, it's common knowledge that neither Intel nor AMD make that guarantee.

The other reason why most CPUs can actually run most XMP/EXPO profiles nowadays are BIOS updates that included preloaded settings for many, many memory kits and manufacturers in the last months. Especially voltages are still very tricky, if the motherboard doesn't preload them. According to ASRock's website there are two beta BIOSes available for the board that the OP uses. Might be worth checking out.

Motherboards absolutely do not preload settings on a per kit basis. The motherboard has a set of memory related CPU voltages (like SOC, VDDR_IO, ect) for the default, one for if XMP / EXPO is enabled, and maybe one or two more if you enable memory OC / select a more aggressive timing mode. SOC voltage isn't going to be changed on a per kit basis, that doesn't make any sense. It makes far more sense to have the motherboard set things it knows better and the Memory set things it knows better (which is what currently happens).

The critical issue with AMD and memory compatibility is and has always been AGESA, but we're in a tangential here: this more than likely doesn't have anything to do with OP's stuttering problems.

Agreed, irregardless of whether it does turn out to be a memory issue this side discussion is starting to derail the thread.
 
Low quality post by Space Lynx
hi guys, below are my specs. My pc is stuttering hard every minute or so and benchmarking in the bottom 20% on UserBenchmark. (I benchmarked multiple times and this was the best it ever was https://www.userbenchmark.com/UserRun/67161985) Now I know people are not fans of this site so i also downloaded cinebench and everything benched fine.

View attachment 332368
with a 2060 Super.

Also I have 32 gigs of ram, i just took one out for troubleshooting when i took this screenshot

What do??
Did you figure out the problem? It looks like the last page of posts went off the rails.
 
No, most AM5 CPUs could do that anyways prior to the update. AGESA 1.0.0. 7c mostly improved the max obtainable frequency and fringe stability issues. I don't really see the point of mentioning AMD doesn't guarantee hitting the sweet spot frequency, it's common knowledge that neither Intel nor AMD make that guarantee.
Before 1.0.0.7 Gear 2 didn't work. Now more motherboards can run unfavorable memory configurations. I rather just see it not* work for performance reasons, but the average person doesn't know the difference. Computer turns on and that's all that matters.
 
Last edited:
Hey guys, long time no see. School got to me and I've been crazy busy. Here are some dumps from cs2 and overwatch. In CS2 I suffer from high frame times and in OW im experiencing full game stutters that last about a second. This also happens in cs sometimes but didnt while i was recording info.

A little extra info: Whenever i change my graphics settings in cs2, i get consistent 270 fps for abt a minute then it drops back again, even on max settings. I play on 1280 x 960 res

Might be interesting to post a ZenTimings screenshot of your memory.

If the motherboard and memory combination is the culprit, we might see some odd voltages in that screenshot.

Generally speaking, the old kit would have been perfectly fine, if you know how to tune memory manually on Ryzen 7000. If you don't know, don't buy G.Skill kits since their XMP/EXPO profiles assume that your CPU is actually able to run that speed which isn't a given for DDR5-6400 and might still need some manually tweaking on DDR5-6000 in case you got really unlucky with your CPU.
If you want "safe" XMP/EXPO profiles, buy Kingston Renegade or Fury Beast kits with the lowest tCL available, usually CL 30 or 32. Those kits will typically come with two speed bins. If the higher one doesn't work, they lower one usually will.
 

Attachments

  • info.zip
    info.zip
    4.2 MB · Views: 57
  • Capture.PNG
    Capture.PNG
    31.9 KB · Views: 119
Last edited:
1. I got stutters in fortnite dx12 first time using EXPO 6000. I lowered it to 4800mhz by turning off expo, the stutters were gone during game. After that I rebooted again with EXPO 6000 and no more stutters.
2. you could also remove your GPU, install AMD drivers for iGPU and try the game in low-setting to reproduce the issue.
3. If issue persist, then try updating motherboard BIOS.
 
Hey guys, long time no see. School got to me and I've been crazy busy. Here are some dumps from cs2 and overwatch. In CS2 I suffer from high frame times and in OW im experiencing full game stutters that last about a second. This also happens in cs sometimes but didnt while i was recording info.

A little extra info: Whenever i change my graphics settings in cs2, i get consistent 270 fps for abt a minute then it drops back again, even on max settings. I play on 1280 x 960 res

This looks like a bog standard G.skill Hynix XMP, except the complete lack of voltages and cadbus/setup is concerning

Do you have chipset drivers installed? Are you sure the 2.02 BIOS flash is good and not actually corrupted? Did you TM5 memtest to make sure this XMP setup is stable?

I'm also on B650E ASrock and have access to values in zentimings except the usual VDDG culprits.
 
This looks like a bog standard G.skill Hynix XMP, except the complete lack of voltages and cadbus/setup is concerning

Do you have chipset drivers installed? Are you sure the 2.02 BIOS flash is good and not actually corrupted? Did you TM5 memtest to make sure this XMP setup is stable?

I'm also on B650E ASrock and have access to values in zentimings except the usual VDDG culprits.
View attachment IMG_3743.webp

The image is of the TM5 memtest. I believe I installed the chipset drivers. And not sure if the flash was corrupted. How would I check?
 
View attachment 337984

The image is of the TM5 memtest.

No, it isn't, idk what you just ran but it's not TM5/Karhu/HCI


Chipset driver should be fairly obvious if you see missing devices with yellow triangle in Device Manager. But you can always get the latest off AMD website if you search for B650 drivers.

BIOS only way to know for sure is to reflash or try a different BIOS. I'm also on 2.02 which is an AGESA 1100 for mine B650E PG-ITX, but can't check anything until monday

There should not be a single board in existence that doesn't at least show you VSOC, and UCLK/FCLK/MCLK, so a reflash might be in order.
 
Back
Top