Long AM5 POST times

[OkieDan]

This is on a Gigabyte X670 Aorus Elite AX using latest BIOS and G.Skill DDR5 6000 CL30-40-40-96 (XMP kit, full part no in my system specs).

On every boot/reboot it takes 45 seconds to complete POST and the DRAM LED on the board is lit for the vast majority of the time. This only happens when the XMP profile is enabled, it only takes 12-15 seconds w/o XMP enabled. The system is stable using XMP profile and the primary timings from the XMP profile are reflected as in use in HWinfo, CPU-Z, and ZenTimings once in Windows.

In theory, would the delay most likely be caused due to the BIOS trying to determine safe timings? If so, would the most likely fix be to set all "auto" timings to manual and using the actual timings reflected in ZenTimings?

I did notice tRDWR on one stick is 21, while it's 22 on the other stick. Would it be worth only setting that manually to 22 in the BIOS first or should I just set all to manual?

 

[HammerON]

Read W1zzard's review as he discusses the long boot time issue with AM5, in specific the 7950X:

AMD Ryzen 9 7950X Review - Impressive 16-core Powerhouse

The Ryzen 9 7950X is a monster CPU. When paired with the right workload it will eat even the 12900K for breakfast. As our review shows, the performance uplifts can be massive: +30-50% gen-over-gen is totally possible. What makes things complicated though, is that keeping the beast cool is almost...

www.techpowerup.com

 

[Super Firm Tofu]

Your post times seem about right @ 45 seconds. From power button to the flashing cursor in the top left corner is 30 seconds for me each boot, but I only have 32GB of RAM - the more RAM the longer the post time. Mine is EXPO rather than XMP, but from what I've gathered across the forums, that shouldn't make a difference.

I can't remember where I read it, but there's been talk of providing an option in an upcoming AGESA to skip the memory test/training that goes on during that long pause, but it's not really a fix, just a bypass.

I haven't attempted to enter any of the memory setting manually as my setup has been 100% stable, and I've just made myself push the power button and then go get something to drink in the kitchen.

 

[ir_cow]

Every single time the MB boots, it does some memory training. The first time you enable XMP, its like 2-3 minutes, every time after that is 30~ seconds. I did notice a option to disable the memory extra memory training, but it did some wacky stuff to perf. Also I see you have dual-rank memory. Those take even longer to boot I've noticed. I spend a lot of time watching the codes haha.

I haven't tried this, but maybe typing in all the timings will speed up the boot sequence. It does for DDR4 and for Intel DDR5.

Edit: If you do set those timings, don't forget to manually set the voltages as well for VDDIO_MEM and SoC. These will change based on the timings when its set to auto every single time.. Also UCLCK DIVI and FLCK. That is part of the memory training process. Set those and the process speeds up quite a lot.

I was wrong for AM5 (for GB at least). nothing really helps the speed up unless you do the memory training restore function. I get all the motherboards mixed up now and the features in the BIOS. I do remeber BIOSTAR X670E being the worst. Like 2 minute boot every single time.

Time to BIOS From Power OFF state: Seconds
XMP + Auto Everything =33
Just FLCK and DVI set = 32
FLCK + DVI + Voltage = 32
FLCK + DVI + Voltage + Timings typed in =30
Context Restore = 17
FLCK + DVI + Voltage + Timings typed in + Context Restore =15

Gigabyte Context Restore:

 

[AusWolf]

Your post times seem about right @ 45 seconds. From power button to the flashing cursor in the top left corner is 30 seconds for me each boot, but I only have 32GB of RAM - the more RAM the longer the post time. Mine is EXPO rather than XMP, but from what I've gathered across the forums, that shouldn't make a difference.

I can't remember where I read it, but there's been talk of providing an option in an upcoming AGESA to skip the memory test/training that goes on during that long pause, but it's not really a fix, just a bypass.

I haven't attempted to enter any of the memory setting manually as my setup has been 100% stable, and I've just made myself push the power button and then go get something to drink in the kitchen.

This. Long boot times never bothered me much, to be honest. My head is still somewhat stuck in the Windows 98 + hard disk era.

OP: Yes, a 30-45 second POST time is normal.

Every single time the MB boots, it does some memory training. The first time you enable XMP, its like 2-3 minutes, every time after that is 30~ seconds.

It might be different with EXPO, as my motherboard booted in roughly 30 seconds even on the first try.

 

[ir_cow]

It might be different with EXPO, as my motherboard booted in roughly 30 seconds even on the first try.

Nope. I have many many EXPO and XMP kits. Boot time is roughly the same. Not enough for me to retest what I did above. Newer BIOS versions solved some of the initial long boot cycle. I guess the confusion here is when you first power on the computer (after a power loss) will be 2-3 minutes. If you never cut power again, its about 30 seconds. I think it just depends on the MB and how much it saves from the previous boot. But you clear CMOS, its that same 2-3 minutes.

 

[OkieDan]

Every single time the MB boots, it does some memory training. The first time you enable XMP, its like 2-3 minutes, every time after that is 30~ seconds. I did notice a option to disable the memory extra memory training, but it did some wacky stuff to perf. Also I see you have dual-rank memory. Those take even longer to boot I've noticed. I spend a lot of time watching the codes haha.


Time to BIOS From Power OFF state: Seconds
XMP + Auto Everything =33
Just FLCK and DVI set = 32
FLCK + DVI + Voltage = 32
FLCK + DVI + Voltage + Timings typed in =30
Context Restore = 17
FLCK + DVI + Voltage + Timings typed in + Context Restore =15

Gigabyte Context Restore:
View attachment 269147

My board apparently doesn't have this "Context Restore" option or I can't find it.

 

[ir_cow]

@OkieDan Its deep in the menu for some reason. I think a earlier BIOS had it next to everything else on the Tweaker tab.

Advanced BIOS (F2) > Settings Tab > AMD CBS > UMC Common Options > DDR Options > DDR Memory Features > Memory Context Restore

Press Insert KEY while highlighting DDR Memory Features to add it to the Favorites Tab (F11)

 

[mama]

It will get looked at in a future BIOS update no doubt.

 

[OkieDan]

@OkieDan Its deep in the menu for some reason. I think a earlier BIOS had it next to everything else on the Tweaker tab.

Advanced BIOS (F2) > Settings Tab > AMD CBS > UMC Common Options > DDR Options > DDR Memory Features > Memory Context Restore

Press Insert KEY while highlighting DDR Memory Features to add it to the Favorites Tab (F11)

Thanks, POST now takes 21 seconds instead of 45 to complete!

 

[Keullo-e]

Your post times seem about right @ 45 seconds. From power button to the flashing cursor in the top left corner is 30 seconds for me each boot, but I only have 32GB of RAM - the more RAM the longer the post time. Mine is EXPO rather than XMP, but from what I've gathered across the forums, that shouldn't make a difference.

I can't remember where I read it, but there's been talk of providing an option in an upcoming AGESA to skip the memory test/training that goes on during that long pause, but it's not really a fix, just a bypass.

I haven't attempted to enter any of the memory setting manually as my setup has been 100% stable, and I've just made myself push the power button and then go get something to drink in the kitchen.

Wait, what? Does modern hardware have longer POST times with more memory? I thought that it was a thing of the past.

 

[Calenhad]

Wait, what? Does modern hardware have longer POST times with more memory? I thought that it was a thing of the past.

Memory training takes longer with more memory

 

[Keullo-e]

Memory training takes longer with more memory

Hm, at least I learned something new today.

 

[Super Firm Tofu]

Wait, what? Does modern hardware have longer POST times with more memory? I thought that it was a thing of the past.

For AM5 it appears it does. The BIOS the boards initially shipped with were especially bad. Remember the AsRock memory slot stickers that made the news at launch?

 

[Keullo-e]

For AM5 it appears it does. The BIOS the boards initially shipped with were especially bad. Remember the AsRock memory slot stickers that made the news at launch?

View attachment 269263

Totally missed that one well damn.

 

[Super Firm Tofu]

Totally missed that one well damn.

Even the stickers themselves were a problem

ASRock's X670 Motherboards Have Numerous Issues... With DRAM Stickers

This one is likely to go down ASRock's internal history as a failure of sticking proportions. Namely, it seems that some ASRock motherboards in the newly-released AM5 X670 / X670E family carry stickers overlaid on the DDR5 slots. The idea was to provide users with a handy, visually informative...

www.techpowerup.com

 

[Keullo-e]

Even the stickers themselves were a problem

ASRock's X670 Motherboards Have Numerous Issues... With DRAM Stickers

This one is likely to go down ASRock's internal history as a failure of sticking proportions. Namely, it seems that some ASRock motherboards in the newly-released AM5 X670 / X670E family carry stickers overlaid on the DDR5 slots. The idea was to provide users with a handy, visually informative...

www.techpowerup.com

Ah, I did read that one quickly but I seem to have skipped the main part (the boot time).

 

[AusWolf]

Nope. I have many many EXPO and XMP kits. Boot time is roughly the same. Not enough for me to retest what I did above. Newer BIOS versions solved some of the initial long boot cycle. I guess the confusion here is when you first power on the computer (after a power loss) will be 2-3 minutes. If you never cut power again, its about 30 seconds. I think it just depends on the MB and how much it saves from the previous boot. But you clear CMOS, its that same 2-3 minutes.

Well, my board never needed 2-3 minutes for some reason. Not even on its first start.

 

[Hofnaerrchen]

This. Long boot times never bothered me much, to be honest. My head is still somewhat stuck in the Windows 98 + hard disk era.

If you consider how much influence you had on your OS back then I fully understand this. Win98 still was - more or less - a GUI for DOS and with DOS you had no bloatware OS extensions. Today an OS is large compared to the early 90s. Some of this comes from drivers (GPU for example) but appart from that: Staying in control of your computer is much complicated than back then and some of this is based on "quality of life improvements". Did OS become more secure? Not really in my opinion. When a user does not know what the OS is doing it's already a hazard.

If Linux was more user friendly - I am just not willing to put time into becoming a practical Linux user - it would be an alternative. But right now Linux - and maybe how people want it to be (ELITE) - is not even close to be an alternative to Windows. Even Android (based on Linux) is more appealing than Linux itself, while it also shares the bloat feature of "modern" OS.

 

[AusWolf]

If you consider how much influence you had on your OS back then I fully understand this. Win98 still was - more or less - a GUI for DOS and with DOS you had no bloatware OS extensions. Today an OS is large compared to the early 90s. Some of this comes from drivers (GPU for example) but appart from that: Staying in control of your computer is much complicated than back then and some of this is based on "quality of life improvements". Did OS become more secure? Not really in my opinion. When a user does not know what the OS is doing it's already a hazard.

If Linux was more user friendly - I am just not willing to put time into becoming a practical Linux user - it would be an alternative. But right now Linux - and maybe how people want it to be (ELITE) - is not even close to be an alternative to Windows. Even Android (based on Linux) is more appealing than Linux itself, while it also shares the bloat feature of "modern" OS.

I completely agree with you, although what I meant was that I can't give a rat's arse whether my PC boots up in 20, 30 or 45 seconds. As @Super Firm Tofu said, boot time is a good time to go grab a drink from the kitchen.

 

[Hofnaerrchen]

I completely agree with you, although what I meant was that I can't give a rat's arse whether my PC boots up in 20, 30 or 45 seconds. As @Super Firm Tofu said, boot time is a good time to go grab a drink from the kitchen.

That's what I do after getting up... turn on the PC and go to the kitchen (and maybe/probably some other place you might want to visit after getting up) ;-) to grab something (can't help to think of Sheldon from TBBT^^) warm to drink.

 

[AusWolf]

That's what I do after getting up... turn on the PC and go to the kitchen (and maybe/probably some other place you might want to visit after getting up) ;-) to grab something (can't help to think of Sheldon from TBBT^^) warm to drink.

And that's what I do after work (I work night shifts).

 

[OkieDan]

I can't remember where I read it, but there's been talk of providing an option in an upcoming AGESA to skip the memory test/training that goes on during that long pause, but it's not really a fix, just a bypass.

I really hate how they handled this. I don't see a reason they couldn't create a hash from the RAM part/serial# + freq + timings and store a count of successful posts under that hash in NVRAM. Once the count of successful POSTs reaches a certain threshold they could bypass training on subsequent boots until the RAM or settings are changed.

 

[ir_cow]

Email the engineering team at AMD

 

[Tech_fanatic]

Email the engineering team at AMD

I guess they already know about the issue but haven't been able/ not cared to issue a fix till date as is the case with their graphics driver bugs (they will eventually fix it but only GOD knows when).