F5F Stay Refreshed Hardware Desktop AM5 DDR5 6000 EXPO Challenges - Not warming up after reboot

AM5 DDR5 6000 EXPO Challenges - Not warming up after reboot

AM5 DDR5 6000 EXPO Challenges - Not warming up after reboot

T
techiseasy
Senior Member
688
11-18-2023, 04:13 AM
#1
I bought an AM5 configuration some time back and faced RAM problems. After a bit over a minute (post-initial memory training) the system froze cold, and later I noticed it couldn’t warm up at all. I thought the RAM issues were early signs of platform instability, similar to what others reported. I kept updating the firmware regularly, just like with the X370, hoping for a fix. Recently I read about AGESA offering much quicker RAM support than what I had. I contacted ASUS, suspecting a board-level problem. They tested single sticks and confirmed the issue affected all slots with both sticks. They went back and forth, even challenging me to try another RAM module before they would open an RMA. I did that, purchasing another identical kit. The problems persisted—this was expected since the kit had passed numerous memory tests at high speeds. They proceeded with an RMA, which I wasn’t sure would resolve my situation. For temporary use, I bought the cheapest x670 board and listed my RAM on the QVL. Right after setting it up for return, I transferred the components to a new board (still using an older BIOS), but the cold boot issue returned immediately. The same pattern appeared with the Creator board: it would start up cold, then hang during reboot. Eventually, I concluded the memory controller on my chip was faulty. I reached out to AMD for an RMA and am curious about the outcome. The suggestion to run Infinity Fabric at optimal speed really swayed my decision. I wondered if boosting the SoC voltage could help, but I hoped for a smoother out-of-the-box experience with such a premium component. I managed to run a previous-gen processor at the recommended speed using (3900x) with the same RAM, so stability seemed possible. Still, I’d love feedback from the community.

Original Components:
- Ryzen 9 7950X3D
- GSkill Trident Z5 Neo 6000
- CL30 (F5-6000J3040G32GX2-TZ5NR)
- ProArt X670E-Creator
- SuperFlower Platinum 1000W
- MSI Pro X670-P
- EVGA Platinum 750W

Interim Upgrades:
- MSI Pro X670-P (same as Creator)
- EVGA Platinum 750W

Notes:
- I’m open to suggestions and ready for more questions.
- Updated August 5, 2023 – grammar and spelling corrections applied.
T
techiseasy
11-18-2023, 04:13 AM #1

I bought an AM5 configuration some time back and faced RAM problems. After a bit over a minute (post-initial memory training) the system froze cold, and later I noticed it couldn’t warm up at all. I thought the RAM issues were early signs of platform instability, similar to what others reported. I kept updating the firmware regularly, just like with the X370, hoping for a fix. Recently I read about AGESA offering much quicker RAM support than what I had. I contacted ASUS, suspecting a board-level problem. They tested single sticks and confirmed the issue affected all slots with both sticks. They went back and forth, even challenging me to try another RAM module before they would open an RMA. I did that, purchasing another identical kit. The problems persisted—this was expected since the kit had passed numerous memory tests at high speeds. They proceeded with an RMA, which I wasn’t sure would resolve my situation. For temporary use, I bought the cheapest x670 board and listed my RAM on the QVL. Right after setting it up for return, I transferred the components to a new board (still using an older BIOS), but the cold boot issue returned immediately. The same pattern appeared with the Creator board: it would start up cold, then hang during reboot. Eventually, I concluded the memory controller on my chip was faulty. I reached out to AMD for an RMA and am curious about the outcome. The suggestion to run Infinity Fabric at optimal speed really swayed my decision. I wondered if boosting the SoC voltage could help, but I hoped for a smoother out-of-the-box experience with such a premium component. I managed to run a previous-gen processor at the recommended speed using (3900x) with the same RAM, so stability seemed possible. Still, I’d love feedback from the community.

Original Components:
- Ryzen 9 7950X3D
- GSkill Trident Z5 Neo 6000
- CL30 (F5-6000J3040G32GX2-TZ5NR)
- ProArt X670E-Creator
- SuperFlower Platinum 1000W
- MSI Pro X670-P
- EVGA Platinum 750W

Interim Upgrades:
- MSI Pro X670-P (same as Creator)
- EVGA Platinum 750W

Notes:
- I’m open to suggestions and ready for more questions.
- Updated August 5, 2023 – grammar and spelling corrections applied.

M
masonight
Member
175
11-19-2023, 06:28 PM
#2
Initially I'd assume Memory Context Restore isn't functioning properly with your CPU. Most boards should include this feature by default, but until the latest AGESA update to version 1.0.0.7b it wasn't fully dependable across many configurations. It's still not ideal, though it works for many users. To clarify, this behavior doesn't apply to Ryzen 7000 processors. They permanently reset the FCLK, meaning regardless of whether you're using DDR5 5200, 6400, or DDR5 8000 on the newest AGESA, the FCLK will default to 2000MHz unless adjusted manually. The choice of 6000 CL30 was based on being the highest frequency that generally supported 99% of CPUs up until recently—it wasn't really a sweet spot, but now it's closer to one because it reliably runs in single mode for most systems.
M
masonight
11-19-2023, 06:28 PM #2

Initially I'd assume Memory Context Restore isn't functioning properly with your CPU. Most boards should include this feature by default, but until the latest AGESA update to version 1.0.0.7b it wasn't fully dependable across many configurations. It's still not ideal, though it works for many users. To clarify, this behavior doesn't apply to Ryzen 7000 processors. They permanently reset the FCLK, meaning regardless of whether you're using DDR5 5200, 6400, or DDR5 8000 on the newest AGESA, the FCLK will default to 2000MHz unless adjusted manually. The choice of 6000 CL30 was based on being the highest frequency that generally supported 99% of CPUs up until recently—it wasn't really a sweet spot, but now it's closer to one because it reliably runs in single mode for most systems.

V
valentijn10000
Junior Member
49
11-19-2023, 10:38 PM
#3
I switched MCR before seeing any improvement, but the Creator board stays active. The main board still uses the default setting. I’m testing the 1.0.0.7b (1516) BIOS on another board tomorrow and will review it more closely. I suspect the 2GHz/6000 setting was chosen for a matching ratio. It’s disappointing my chip clocked at the very bottom of the controller selection list.
V
valentijn10000
11-19-2023, 10:38 PM #3

I switched MCR before seeing any improvement, but the Creator board stays active. The main board still uses the default setting. I’m testing the 1.0.0.7b (1516) BIOS on another board tomorrow and will review it more closely. I suspect the 2GHz/6000 setting was chosen for a matching ratio. It’s disappointing my chip clocked at the very bottom of the controller selection list.

P
132
11-19-2023, 11:43 PM
#4
2GHz was chosen as the standard for compatibility across all processors. A speed of 6000MT/s paired with a 2200MHz clock will outperform 6000MT/s at 2000MHz clock under identical conditions. Similarly, 6200MT/s with a 2000MHz clock will surpass 6000MT/s when using the same 2000MHz clock speed. The results simply aligned as expected.
P
PremierTrotsky
11-19-2023, 11:43 PM #4

2GHz was chosen as the standard for compatibility across all processors. A speed of 6000MT/s paired with a 2200MHz clock will outperform 6000MT/s at 2000MHz clock under identical conditions. Similarly, 6200MT/s with a 2000MHz clock will surpass 6000MT/s when using the same 2000MHz clock speed. The results simply aligned as expected.

M
Mickey_Games
Junior Member
3
12-02-2023, 02:38 AM
#5
AMD processed a return request for my CPU, but the replacement part didn't contain the problem.
M
Mickey_Games
12-02-2023, 02:38 AM #5

AMD processed a return request for my CPU, but the replacement part didn't contain the problem.