Infinity Fabric trouble?

[Sadtiric]

So, I just splurged to upgrade my RAM to hit "the sweet spot" for my 5900X. I picked up a 64GB CL16 3600MHz kit in hopes to get that 1800 all around, but when I opened up ZenTimings (1.2.5) I noticed that I'm getting
MCLK 1800.00

FCLK  1600.00

UCLK   900.00

As far as I understood it, having the DOCP/XMP profile set to 3600MHz, which I can confirm it is via Task Manager, I should have been getting 1800s across the board for that 1:1:1, did I do something wrong? I've never really messed with this kind of thing before, but my old kit of 32GB CL 16 3200MHz was hitting 1600 for all 3. My BIOS is up to date on my ASRock B550 Taichi at P2.20. Thanks for whatever help!

[RONOTHAN##]

It happens sometimes where you need to manually set the FCLK in the BIOS to 1800MHz. Just jump into the BIOS and change FCLK to 1800 and set UCLK mode to UCLK == MCLK. They're usually right next to the memory settings in the BIOS, though there are a couple BIOS revisions out there where you have to set those in the AMD Overclocking menu. 

[Sadtiric]

2 hours ago, RONOTHAN## said:

It happens sometimes where you need to manually set the FCLK in the BIOS to 1800MHz. Just jump into the BIOS and change FCLK to 1800 and set UCLK mode to UCLK == MCLK. They're usually right next to the memory settings in the BIOS, though there are a couple BIOS revisions out there where you have to set those in the AMD Overclocking menu. 

Sorry for the late response, I didn't expect such a quick reply. 

So, I went into my BIOS and tried to set that up twice and it seems to have just set me back to 2133MHz both times. I'm going to try a CMOS clear and see if that changes anything. Is there specific terminology I should be looking for? I didn't see anything specific about the FCLK or UCLK, but what I DID find was a secondary option in my BIOS for MHz that had a lower range, and I made sure that everything else was set to "Auto," which I assumed would keep everything synced. I'll keep looking, and will take a picture of my BIOS options if the CMOS doesn't help. 

 

Update: I've gotten 3 different blue screens after tweaking this, which is a shame, because right before the first crash, it was all 1800s. I posted the pics in order of how they happened, starting from the BIOS change.

 

Note: I do see the mention off to the right now.

[Sadtiric]

46 minutes ago, Sadtiric said:

Sorry for the late response, I didn't expect such a quick reply. 

So, I went into my BIOS and tried to set that up twice and it seems to have just set me back to 2133MHz both times. I'm going to try a CMOS clear and see if that changes anything. Is there specific terminology I should be looking for? I didn't see anything specific about the FCLK or UCLK, but what I DID find was a secondary option in my BIOS for MHz that had a lower range, and I made sure that everything else was set to "Auto," which I assumed would keep everything synced. I'll keep looking, and will take a picture of my BIOS options if the CMOS doesn't help. 

Uhh... So PC won't boot anymore. No BIOS, no nothin'. It turns on, lights and fans are all on, but the screen is black. Did another CMOS clear after the BSs and then tried a full power drain after that with no luck. Dr. Debug is giving me an A6 code, which seems to imply some SATA issues, though my Windows install(s) are both on NVMe and I'm not even getting to BIOS.. Hmmm..

 

Update: Update: got it to boot after powering off my PSU and holding my power button for a minute after another CMOS clear. 

[RONOTHAN##]

58 minutes ago, Sadtiric said:

Sorry for the late response, I didn't expect such a quick reply. 

So, I went into my BIOS and tried to set that up twice and it seems to have just set me back to 2133MHz both times. I'm going to try a CMOS clear and see if that changes anything. Is there specific terminology I should be looking for? I didn't see anything specific about the FCLK or UCLK, but what I DID find was a secondary option in my BIOS for MHz that had a lower range, and I made sure that everything else was set to "Auto," which I assumed would keep everything synced. I'll keep looking, and will take a picture of my BIOS options if the CMOS doesn't help. 

 

Update: I've gotten 3 different blue screens after tweaking this, which is a shame, because right before the first crash, it was all 1800s. I posted the pics in order of how they happened, starting from the BIOS change.

 

Note: I do see the mention off to the right now.

So the first one could be memory controller voltages being set too low, try manually setting SOC voltage to 1.15V and see if that helps. 

 

Those second two error codes aren't memory related though. The first one is the OS being super bugged out, so that very well might be the OS being on its last legs. 

 

If I still had my Taichi (X570, not B550, but the BIOS layout should be identical) I'd experiment a bit along side you, though I just sold it like 2 weeks ago. 

[Sadtiric]

49 minutes ago, RONOTHAN## said:

So the first one could be memory controller voltages being set too low, try manually setting SOC voltage to 1.15V and see if that helps. 

 

Those second two error codes aren't memory related though. The first one is the OS being super bugged out, so that very well might be the OS being on its last legs. 

 

If I still had my Taichi (X570, not B550, but the BIOS layout should be identical) I'd experiment a bit along side you, though I just sold it like 2 weeks ago. 

Yeah, I noticed that about the OS. It probably has to do with some dual boot experiments I was doing with a new NVMe I got. I thought I got away with having Win 10 on one drive and 11 on the other with the same key and no consequences. The 10 OS I've been on for the past year or two has been amazingly stable. Today was my first BS, so it seems odd. Even when I was messing with it last week I had no issues. I wouldn't think the OS had much to do with RAM OC or the BIOS not booting, but like I said, I've never really messed with any of this before, not to this degree.

 

That's a shame about the timing of your sale. I have a friend who has the X570 Taichi, but he doesn't mess with this stuff. 

 

As far as SOC voltage goes, where do I look? While browsing other forums I saw something about a tRC value being represented as 85 when its supposed to be 58. I tried messing with it manually, but ended up back with my PC not booting at all. So I think I might just cut my losses, assuming I can get this thing to boot again. The 5900X is only a few months old and the RAM is brand new, so I'm a little confused/bummed, but I can't afford to lose my system and it seems I will if I keep tweaking it.

 

Update: I found the SoC Voltage options, tried one last tweak with no luck. I may mess with this again later, but for now, I'm not feeling lucky. If you have any more suggestions, I'm all ears. I think I may do some more research with your input and then attempt this again at a later time. 

[RONOTHAN##]

6 minutes ago, Sadtiric said:

As far as SOC voltage goes, where do I look? While browsing other forums I saw something about a tRC value being represented as 85 when its supposed to be 58. I tried messing with it manually, but ended up back with my PC not booting at all. So I think I might just cut my losses, assuming I can get this thing to boot again. The 5900X is only a few months old and the RAM is brand new, so I'm a little confused/bummed, but I can't afford to lose my system and it seems I will if I keep tweaking it.

The SOC voltage will be on the main screen, set the "SoC/Uncore OC Mode" to "Fixed" then it should give you the option to manually set the voltage to whatever you want. Most Ryzen 3000/5000 chips sweet spot somewhere between 1.1 and 1.2V, my 5900X personally seems to sweet spot at 1.175V. 

 

tRC is an odd setting. Ideally you want to set tRP, then adjust tRAS and tRC at the same time so tRC = tRP + tRAS. It won't affect stability having tRC being higher than tRAS + tRP, but tRP and tRAS don't affect performance, only tRC does, tRAS and tRP only affect stability. 

 

It might be worth messing with the BIOS revision a bit though. There are a number of different revisions on my X570 Master where the memory and FCLK overclocking are completely broken, so it's possible that the latest one is broken (the way AMD memory overclocking works, AMD releases something called the AGESA which along with other things contains the memory training algorithms, something every board vendor uses, thus why I'm comparing the two X570 boards). I've never seen it where 1800MHz FCLK doesn't work, even on the revisions where it's completely busted, but anything is possible I guess. 

 

It shouldn't break or anything by tweaking everything, as long as you don't throw 2 volts into the SOC or memory sticks you shouldn't have any issues. You might just have to be clearing the CMOS a lot. 

[Sadtiric]

1 hour ago, RONOTHAN## said:

The SOC voltage will be on the main screen, set the "SoC/Uncore OC Mode" to "Fixed" then it should give you the option to manually set the voltage to whatever you want. Most Ryzen 3000/5000 chips sweet spot somewhere between 1.1 and 1.2V, my 5900X personally seems to sweet spot at 1.175V. 

 

tRC is an odd setting. Ideally you want to set tRP, then adjust tRAS and tRC at the same time so tRC = tRP + tRAS. It won't affect stability having tRC being higher than tRAS + tRP, but tRP and tRAS don't affect performance, only tRC does, tRAS and tRP only affect stability. 

 

It might be worth messing with the BIOS revision a bit though. There are a number of different revisions on my X570 Master where the memory and FCLK overclocking are completely broken, so it's possible that the latest one is broken (the way AMD memory overclocking works, AMD releases something called the AGESA which along with other things contains the memory training algorithms, something every board vendor uses, thus why I'm comparing the two X570 boards). I've never seen it where 1800MHz FCLK doesn't work, even on the revisions where it's completely busted, but anything is possible I guess. 

 

It shouldn't break or anything by tweaking everything, as long as you don't throw 2 volts into the SOC or memory sticks you shouldn't have any issues. You might just have to be clearing the CMOS a lot. 

That's reassuring, and yeah, it looks like I'm going to have to Flashback to 2.10 and give that a shot. It is still odd to me that this isn't as "plug and play" as all the YouTubers seem to make it out to be. I figured this far into Ryzen it would have been a bit more automatic for the general user. As far as repeatedly clearing the CMOS goes, what's the quickest method to get it to boot back into the BIOS? What I've been doing hasn't been consistent and the time between boots is somewhere around 5 mins with at least one false BIOS boot (just a black screen) and one boot straight to the OS before making it to the BIOS, which seems wrong. What keeps happening is not what I'm used to with BIOS tweaks. Generally if it failed previously it would try to boot, click once or twice and then boot me back into the BIOS. Tonight it has been clicking once or twice, booting to the OS and then crashing on loop with 1-3 various BSOD error codes back to back, but as I said, only after a "bad" BIOS tweak. That doesn't seem typical. If I just hit D.O.C.P./XMP settings and boot, everything works as normal.

 

And that's very interesting about the tRC. I noticed that after BIOS clears the value for the tRC had changed a few times. I saw 58, 85 and then 51 and now I seem to be back at 85 which is what had me looking into it. Though, my BIOS is also reporting 85(tRC), which is definitely not 19(tRP) + 39(tRAS). 

 

So, with your SoC at 1.175V and you suggesting 1.15V for me to try, I'm assuming it (1.15V) would be higher than whatever "Auto" has me at and that the 1.15V is the starting point; I didn't really see any way to monitor what my current SoC voltage was. I've always been hesitant to mess with voltages manually outside of any "Auto OC" settings. That being said, would it help or hurt to OC my 5900X? I would imagine it could give more headroom, but that it also might increase a frequency more than that 1800MHz multiplier and somehow make things worse. 

 

I'm also going to attach a timings screenshot to see if there's something I SHOULD be reporting that I'm not.
Thanks for giving such detailed explanations, it's much appreciated. 
 

[RONOTHAN##]

1 minute ago, Sadtiric said:

As far as repeatedly clearing the CMOS goes, what's the quickest method to get it to boot back into the BIOS?

If I need to clear the CMOS, I flip the switch on the PSU and hold the button for 10 seconds, hit the power button, then spam the delete key while staring at the POST code to make sure it's going correctly (I've overclocked enough memory on AM4 systems to know how the POST codes are supposed to run). 

 

4 minutes ago, Sadtiric said:

So, with your SoC at 1.175V and you suggesting 1.15V for me to try, I'm assuming it (1.15V) would be higher than whatever "Auto" has me at and that the 1.15V is the starting point; I didn't really see any way to monitor what my current SoC voltage was. I've always been hesitant to mess with voltages manually outside of any "Auto OC" settings. That being said, would it help or hurt to OC my 5900X? I would imagine it could give more headroom, but that it also might increase a frequency more than that 1800MHz multiplier and somehow make things worse. 

It depends on the board for what auto is. I've seen some where it defaults to 1 volt, I've seen them where it defaults to 1.2V. 1.15V is just almost optimal for 98% of CPUs, mine can only really do an extra tick one of the subtimings going to 1.175 compared to 1.15V, so that's why I said put it at that voltage. I wouldn't go above 1.2V for daily usage though, you can degrade the SOC if you're using 1.3V+ for an extended period of time, plus it doesn't really help anyway. I wouldn't really try to OC the 5900X, I've tested it with PBO and dialed in limits compared to a daily stable 4.55GHz all core OC, you lose 1-2% performance in multi core AVX workloads and gain 5-10% in single thread workloads, and 2-3% in lighter multi core workloads like Cinebench. You're better off enabling PBO and setting the limits PPT 300, TDC 177, EDC 188 (what I found was optimal on my chip and board) then calling it a day. Besides, OCing your CPU wouldn't really affect memory support at all, and in all reality might make it ever so slightly worse. 

 

8 minutes ago, Sadtiric said:

I'm also going to attach a timings screenshot to see if there's something I SHOULD be reporting that I'm not.
Thanks for giving such detailed explanations, it's much appreciated. 
 

OK, so this might be the culprit. You're running 4 DIMMs of dual rank memory, effectively quad rank. The Ryzen memory controller hates memory ranks, the single rank max memory speed you can expect is 4800MHz+ with a good motherboard, the dual rank max speed is about 4000MHz with a good board, and it trails off from there. You need a really good CPU to have a shot at running quad rank at 3600MHz CL16, just because the memory controller can't keep up. Odds are it's not the infinity fabric that can't keep up, it's the memory controller (UCLK) that can't keep up. If upping the SOC voltage (this might help a fair bit, I forget where ASRock sets their auto SOC voltage) doesn't get it stable at 1:1:1, try lowering the maximum frequency to 3533MHz or 3466MHz to see if that will help performance, and run something that's heavy on the memory controller like Linpack Xtreme to stress test to make sure that it's stable. 

[Sadtiric]

10 hours ago, RONOTHAN## said:

If I need to clear the CMOS, I flip the switch on the PSU and hold the button for 10 seconds, hit the power button, then spam the delete key while staring at the POST code to make sure it's going correctly (I've overclocked enough memory on AM4 systems to know how the POST codes are supposed to run). 

 

It depends on the board for what auto is. I've seen some where it defaults to 1 volt, I've seen them where it defaults to 1.2V. 1.15V is just almost optimal for 98% of CPUs, mine can only really do an extra tick one of the subtimings going to 1.175 compared to 1.15V, so that's why I said put it at that voltage. I wouldn't go above 1.2V for daily usage though, you can degrade the SOC if you're using 1.3V+ for an extended period of time, plus it doesn't really help anyway. I wouldn't really try to OC the 5900X, I've tested it with PBO and dialed in limits compared to a daily stable 4.55GHz all core OC, you lose 1-2% performance in multi core AVX workloads and gain 5-10% in single thread workloads, and 2-3% in lighter multi core workloads like Cinebench. You're better off enabling PBO and setting the limits PPT 300, TDC 177, EDC 188 (what I found was optimal on my chip and board) then calling it a day. Besides, OCing your CPU wouldn't really affect memory support at all, and in all reality might make it ever so slightly worse. 

 

OK, so this might be the culprit. You're running 4 DIMMs of dual rank memory, effectively quad rank. The Ryzen memory controller hates memory ranks, the single rank max memory speed you can expect is 4800MHz+ with a good motherboard, the dual rank max speed is about 4000MHz with a good board, and it trails off from there. You need a really good CPU to have a shot at running quad rank at 3600MHz CL16, just because the memory controller can't keep up. Odds are it's not the infinity fabric that can't keep up, it's the memory controller (UCLK) that can't keep up. If upping the SOC voltage (this might help a fair bit, I forget where ASRock sets their auto SOC voltage) doesn't get it stable at 1:1:1, try lowering the maximum frequency to 3533MHz or 3466MHz to see if that will help performance, and run something that's heavy on the memory controller like Linpack Xtreme to stress test to make sure that it's stable. 

Thanks a ton for the clarification and detail! I will try messing with it again tonight after I work and report the results. 

 

I specifically went with a 4 stick kit vs 2 because of some videos that showed that for gaming, having the 4 sticks was most optimal for FPS. As far as the capacity, I was just wanting the 64GB for rendering videos and dealing with larger files a bit more quickly. I was capping out my 32GB kit pretty bad. I do know that 2 sticks are ideal for stability and such, but again, I thought it would all be more stable in different configs by now and I could cherrypick my pros and cons and still get that 1800.

 

So, then if I were to drop to 3533MHz or 3466MHz should I just try "auto?" If not, do the multipliers for MFLK, FCLK and UCLK round?

 

Thanks again for all the help. I'll update when I get a chance to mess with it later tonight.

[RONOTHAN##]

7 minutes ago, Sadtiric said:

I specifically went with a 4 stick kit vs 2 because of some videos that showed that for gaming, having the 4 sticks was most optimal for FPS.

The reasoning behind that is a lot more nuanced. The real reason why 4x8GB configs usually do better than 2x16GB configs is because you're comparing dual rank vs. single rank. Dual rank is effectively adding 300-400MHz worth of performance for the same exact settings. The problem is that rank interleaving (the reason why dual rank performs better than single rank) scales logarithmically, 2 is much better than 1, but 3 is barely better than 2 and the difference between 3 and 4 is unnoticeable, while weighing down the memory controller a ton and limiting the maximum frequency you can hit. If you're gonna be doing an Intel rig and willing to overclock the RAM, you will likely want to use single rank memory since you can overcome the advantages of dual rank by just clocking it so much higher (it's a little more nuanced than this, 11th and 12th gen DDR4 systems for example you still want to stick to dual rank since the memory controller doesn't get above 4200MHz in gear 1, and outside of a couple benchmarks you want to stick to Gear 1 for the best performance). On an AMD rig where you're limited in the max frequency you an run by the FCLK and can't just clock memory to the moon or you just want to enable an XMP profile on Intel and expect it to work, you want to run dual rank if possible since it's like getting 300-400MHz of more RAM performance while maintaining the same speeds. 

 

As for 4 DIMM vs. 2 DIMM configs, 8GB DDR4 is pretty much exclusively single rank until you start looking at really early DDR4, but since you're running 2 DIMMs per channel, you're effectively getting dual rank memory. 16GB DIMMs are also, for the most part, single rank since late 2018-2019. If you buy one earlier though, or go for an aggressive memory bin, you can still get dual rank DDR4 16GB DIMMs as you have done. 32GB DIMMs though, what you were trying to avoid, are actually impossible to get in single rank unbuffered configs and always come in dual rank (technically you could get server memory that's single rank 32GB DIMMs, I just don't think I've ever seen them). Personally, if this is an option, I would actually return your current config and get a 2x32GB kit instead, it will be easier to work with and you won't lose any performance by doing so. 

 

26 minutes ago, Sadtiric said:

So, then if I were to drop to 3533MHz or 3466MHz should I just try "auto?" If not, do the multipliers for MFLK, FCLK and UCLK round?

I'd still manually set the FCLK and UCLK mode. Enable XMP, but before you reboot change the memory frequency to 3533MHz, the FCLK to 1766MHz, and the UCLK mode (if available) to 1:1. That should get it to train and you won't lose that much performance by doing so, then run some stability tests (mainly ones that focus on the memory controller, Memtest5 and Linpack Xtreme are pretty intensive on the memory controller, Linpack especially). If the stability tests fail, drop the memory frequency to 3466MHz and the FCLK to 1733MHz respectively. 

[Sadtiric]

On 8/8/2022 at 1:36 PM, RONOTHAN## said:

The reasoning behind that is a lot more nuanced. The real reason why 4x8GB configs usually do better than 2x16GB configs is because you're comparing dual rank vs. single rank. Dual rank is effectively adding 300-400MHz worth of performance for the same exact settings. The problem is that rank interleaving (the reason why dual rank performs better than single rank) scales logarithmically, 2 is much better than 1, but 3 is barely better than 2 and the difference between 3 and 4 is unnoticeable, while weighing down the memory controller a ton and limiting the maximum frequency you can hit. If you're gonna be doing an Intel rig and willing to overclock the RAM, you will likely want to use single rank memory since you can overcome the advantages of dual rank by just clocking it so much higher (it's a little more nuanced than this, 11th and 12th gen DDR4 systems for example you still want to stick to dual rank since the memory controller doesn't get above 4200MHz in gear 1, and outside of a couple benchmarks you want to stick to Gear 1 for the best performance). On an AMD rig where you're limited in the max frequency you an run by the FCLK and can't just clock memory to the moon or you just want to enable an XMP profile on Intel and expect it to work, you want to run dual rank if possible since it's like getting 300-400MHz of more RAM performance while maintaining the same speeds. 

 

As for 4 DIMM vs. 2 DIMM configs, 8GB DDR4 is pretty much exclusively single rank until you start looking at really early DDR4, but since you're running 2 DIMMs per channel, you're effectively getting dual rank memory. 16GB DIMMs are also, for the most part, single rank since late 2018-2019. If you buy one earlier though, or go for an aggressive memory bin, you can still get dual rank DDR4 16GB DIMMs as you have done. 32GB DIMMs though, what you were trying to avoid, are actually impossible to get in single rank unbuffered configs and always come in dual rank (technically you could get server memory that's single rank 32GB DIMMs, I just don't think I've ever seen them). Personally, if this is an option, I would actually return your current config and get a 2x32GB kit instead, it will be easier to work with and you won't lose any performance by doing so. 

 

I'd still manually set the FCLK and UCLK mode. Enable XMP, but before you reboot change the memory frequency to 3533MHz, the FCLK to 1766MHz, and the UCLK mode (if available) to 1:1. That should get it to train and you won't lose that much performance by doing so, then run some stability tests (mainly ones that focus on the memory controller, Memtest5 and Linpack Xtreme are pretty intensive on the memory controller, Linpack especially). If the stability tests fail, drop the memory frequency to 3466MHz and the FCLK to 1733MHz respectively. 

Sorry for the delays, I've become quite busy all of a sudden. Okay, so all of what you said makes sense to me. As soon as I get a moment, I'm going to go in and mess with it all and see if I can find a sweet spot, if not, then I'm still within my return window for the 4 sticks and will trade out for a pair of 32GB sticks instead. 

I'm not sure I can get around to it until the weekend, but I will update you as it seems you genuinely want me to succeed here, and it is much appreciated. Least I can do is keep you posted. Thank you so much for all your help with this so far!

[Sadtiric]

On 8/8/2022 at 1:36 PM, RONOTHAN## said:

The reasoning behind that is a lot more nuanced. The real reason why 4x8GB configs usually do better than 2x16GB configs is because you're comparing dual rank vs. single rank. Dual rank is effectively adding 300-400MHz worth of performance for the same exact settings. The problem is that rank interleaving (the reason why dual rank performs better than single rank) scales logarithmically, 2 is much better than 1, but 3 is barely better than 2 and the difference between 3 and 4 is unnoticeable, while weighing down the memory controller a ton and limiting the maximum frequency you can hit. If you're gonna be doing an Intel rig and willing to overclock the RAM, you will likely want to use single rank memory since you can overcome the advantages of dual rank by just clocking it so much higher (it's a little more nuanced than this, 11th and 12th gen DDR4 systems for example you still want to stick to dual rank since the memory controller doesn't get above 4200MHz in gear 1, and outside of a couple benchmarks you want to stick to Gear 1 for the best performance). On an AMD rig where you're limited in the max frequency you an run by the FCLK and can't just clock memory to the moon or you just want to enable an XMP profile on Intel and expect it to work, you want to run dual rank if possible since it's like getting 300-400MHz of more RAM performance while maintaining the same speeds. 

 

As for 4 DIMM vs. 2 DIMM configs, 8GB DDR4 is pretty much exclusively single rank until you start looking at really early DDR4, but since you're running 2 DIMMs per channel, you're effectively getting dual rank memory. 16GB DIMMs are also, for the most part, single rank since late 2018-2019. If you buy one earlier though, or go for an aggressive memory bin, you can still get dual rank DDR4 16GB DIMMs as you have done. 32GB DIMMs though, what you were trying to avoid, are actually impossible to get in single rank unbuffered configs and always come in dual rank (technically you could get server memory that's single rank 32GB DIMMs, I just don't think I've ever seen them). Personally, if this is an option, I would actually return your current config and get a 2x32GB kit instead, it will be easier to work with and you won't lose any performance by doing so. 

 

I'd still manually set the FCLK and UCLK mode. Enable XMP, but before you reboot change the memory frequency to 3533MHz, the FCLK to 1766MHz, and the UCLK mode (if available) to 1:1. That should get it to train and you won't lose that much performance by doing so, then run some stability tests (mainly ones that focus on the memory controller, Memtest5 and Linpack Xtreme are pretty intensive on the memory controller, Linpack especially). If the stability tests fail, drop the memory frequency to 3466MHz and the FCLK to 1733MHz respectively. 

So, I finally got around to swapping the new RAM that came in late, and I have the exact same issue 1800, 1600, 900. So I'm going to flashback to BIOS 2.10 and see if that helps. 

 

Update: well shoot, that did the trick! I'm at 1800 all around. Thanks so much for all your help and for the lesson! I'm going to do some practical testing to see if I can get the 4 sticks running at 1800 1:1 and which is faster for gaming in my rig. 

[Sadtiric]

2 hours ago, Sadtiric said:

So, I finally got around to swapping the new RAM that came in late, and I have the exact same issue 1800, 1600, 900. So I'm going to flashback to BIOS 2.10 and see if that helps. 

 

Update: well shoot, that did the trick! I'm at 1800 all around. Thanks so much for all your help and for the lesson! I'm going to do some practical testing to see if I can get the 4 sticks running at 1800 1:1 and which is faster for gaming in my rig. 

I got 5FPS higher on average with 4 sticks vs 2 with them both at 1:1. Thanks again for all your help! @RONOTHAN##