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

AMD Ryzen 9000 Zen 5 "Granite Ridge" Desktop Processors Launch Late-July

Joined
Sep 23, 2022
Messages
1,302 (1.51/day)
Isn't this only for memory training on the first boot or when frequency/timings are changed in the BIOS and then every boot is normal speed rather than several minutes.?
Unfortunately it's not. In later AGESA, AMD offers an option called Memory Context Restore that when enabled will attempt to boot/reboot with the last trained settings. It's not 100% failsafe though. It's also been a cause of BSOD's and instability for some.

In my case, with a GB X670 and 7950X3D, the only way I'm able to boot and reboot successfully every time is by turning off MCR. When enabled, it's about a 12 second BIOS time reported by Windows. Disabled, in my case with 64GB of 6000/CL30, it's an 86 second BIOS time, each and every time. This is running either straight EXPO/XMP, or manually entering in all timings.

It's not the end of the world until you need to be doing something that requires multiple reboots in a row.
 
Joined
Feb 12, 2021
Messages
220 (0.15/day)
Unfortunately it's not. In later AGESA, AMD offers an option called Memory Context Restore that when enabled will attempt to boot/reboot with the last trained settings. It's not 100% failsafe though. It's also been a cause of BSOD's and instability for some.

In my case, with a GB X670 and 7950X3D, the only way I'm able to boot and reboot successfully every time is by turning off MCR. When enabled, it's about a 12 second BIOS time reported by Windows. Disabled, in my case with 64GB of 6000/CL30, it's an 86 second BIOS time, each and every time. This is running either straight EXPO/XMP, or manually entering in all timings.

It's not the end of the world until you need to be doing something that requires multiple reboots in a row.
Thanks for the info, I was clearly unaware of this.!
 
Joined
May 22, 2024
Messages
418 (1.62/day)
System Name Kuro
Processor AMD Ryzen 7 7800X3D@65W
Motherboard MSI MAG B650 Tomahawk WiFi
Cooling Thermalright Phantom Spirit 120 EVO
Memory Corsair DDR5 6000C30 2x48GB (Hynix M)@6000 30-36-36-76 1.36V
Video Card(s) PNY XLR8 RTX 4070 Ti SUPER 16G@200W
Storage Crucial T500 2TB + WD Blue 8TB
Case Lian Li LANCOOL 216
Power Supply MSI MPG A850G
Software Ubuntu 24.04 LTS + Windows 10 Home Build 19045
Benchmark Scores 17761 C23 Multi@65W
Unfortunately it's not. In later AGESA, AMD offers an option called Memory Context Restore that when enabled will attempt to boot/reboot with the last trained settings. It's not 100% failsafe though. It's also been a cause of BSOD's and instability for some.

In my case, with a GB X670 and 7950X3D, the only way I'm able to boot and reboot successfully every time is by turning off MCR. When enabled, it's about a 12 second BIOS time reported by Windows. Disabled, in my case with 64GB of 6000/CL30, it's an 86 second BIOS time, each and every time. This is running either straight EXPO/XMP, or manually entering in all timings.

It's not the end of the world until you need to be doing something that requires multiple reboots in a row.
I'm under the impression that if Memory Context Restore does not work, something along the IMC-MB-RAM axis is probably being pushed too hard, or maybe it's the luck of the draw. Stock EXPO/A-XMP settings were overvolting too much and too loose in secondary timings on mine, and non-MCR boot would take more than two minutes. Right now it seemed stable.

Might worth some tuning to find out which. Hopefully, this next generation would do better.
 
Joined
Sep 23, 2022
Messages
1,302 (1.51/day)
I'm under the impression that if Memory Context Restore does not work, something along the IMC-MB-RAM axis is probably being pushed too hard, or maybe it's the luck of the draw. Stock EXPO/A-XMP settings were overvolting too much and too loose in secondary timings on mine, and non-MCR boot would take more than two minutes. Right now it seemed stable.

Might worth some tuning to find out which. Hopefully, this next generation would do better.
I'm sure it could have been a combination of any or all of those. I did a good bit of trial and error and it was boot/reboot was mostly stable except for a failure to reboot now and then. Unfortunately, the last time it happened was during a BIOS update reboot (initiated from the BIOS) that left the board essentially dead until performing a USB BIOS flashback. At that point I went full manual timings and MCR disabled and it is now rock solid, but with slow post times. The frustrating part is that no matter if there were boot problems or not, the system passed every type of possible stress tests memory tests, and never once had any stability issues. It's currently been running over a week straight at 100% CPU load without a hitch.

I agree though, hoping next gen is better. It's a little amateurish when it's either a choice of 1.5 minute+ post times or faster unreliable boots that require way too much trial and error. My Intel machine posts 48GB @ 8200 in about 8 seconds with zero effort needed.
 

ARF

Joined
Jan 28, 2020
Messages
4,670 (2.55/day)
Location
Ex-usa | slava the trolls
Not because of the weird design though, that's just the thickness ;)

It can't be just the thickness, which obviously adds a substantial amount of thermal resistance, but also the placement of the hot chiplets close to the edge/periphery, which means that there is a direct thermal resistance by the surrounding air.
 
Joined
Feb 12, 2021
Messages
220 (0.15/day)
It can't be just the thickness, which obviously adds a substantial amount of thermal resistance, but also the placement of the hot chiplets close to the edge/periphery, which means that there is a direct thermal resistance by the surrounding air.
Gamers Nexus did a video inside AMD and the design, thickness etc were discussed with one of the engineers that developed it.
 
Joined
Jun 18, 2021
Messages
2,645 (1.99/day)
It can't be just the thickness, which obviously adds a substantial amount of thermal resistance, but also the placement of the hot chiplets close to the edge/periphery, which means that there is a direct thermal resistance by the surrounding air.

The chiplet placement was already an "issue" with AM4 - and when I say issue I mean not ideal because with AM4 didn't really make much of a difference. With AM5 it's more relevant - and thus offset mounts became more common even if still pretty rare - because of the thickness disadvantage. The surrounding air makes no difference, it's not like older chips were vacuum sealed, the heat transfer resistance of air is so much bigger compared to the piece of metal on top it doesn't really contribute to anything.
 
Joined
Jan 11, 2022
Messages
1,070 (0.96/day)
That was true in the DDR4 days where most common stick size was 8GB. With DDR5 most common stick size has doubled to 16GB and most people buy kits of two meaning 32GB. Unless you're into production work etc (in which case you buy higher capacity 48GB or 96GB kits from the get go) then 32GB will be enough for most people during the lifetime of AM5. Also like others have said it's not that simple to just add two more sticks without compatibility issues or speed decrease.

Exactly. Two DIMM boards help with maintaining those speeds and with 96GB now and hopefully 128GB kits of two available soon i just dont see the reason for four DIMM board like i did in the past.
memory use is going to balloon in the next couple of years thanks to copilot and what ever Google is going to jam into chrome

and adding a bit more ram to that 2 year old computer is going to be just as attractive then as it is now
 
Low quality post by NoneRain
Top