• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Long AM5 POST times

I recently changed my RAM default timings from 40-40-40-77 to 38-38-38-77 via the MB BIOS while keeping the default 4800MHz frequency. The boot times have not been affected by this at all and it takes around 20 secs to boot into windows log in screen. I haven't tried loading XMP with the latest BIOS 1.21 on my Asrock MB but the earlier BIOS versions did have long boot times with anything other than the default JEDEC speeds.
 
Considering the efforts to improve boot times, this does seem ridiculous, if DDR5 needs this to be stable it probably shouldnt have been considered ready.

I wake my PC from sleep, and desktop is showing in 2-3 seconds, would this even slow down waking up from sleep?

I have decided when I upgrade will be intel on DDR4, DDR5 the benefits dont seem to outweigh the hassles right now. Save myself money on ram also although I know ram is cheap right now.
 
Blame Microsoft (and all the 'throwaway' PC manufacturers).

If standby/sleep/hybrid sleep/etc.etc. hadn't become 'normal' in consumer PCs (and OEMs hadn't resisted SSDs), AMD (and partners) couldn't 'get away' with this.

Up until these last few generations, it seemed like faster POST+boot was an improving a feature each new generation.

I can remember booting my Ivy Bridge build for the first time, and it was considerably faster than my Phenom II to boot.
-My Phenom II was 'average' vs. the multiple C2Ds+mobos I had before it.
-My C2Ds were faster to boot than my Athlon 64 x2 and Northwood P4 before it.
-My P4 was faster to boot than my Talutin Celeron.
-My Celeron was faster to boot than my Slot1 500mhz P3

Now,
My Ryzen 5 5600 boots slightly slower than my R5 3600 (same mobo.), and both booted slower than my Ivy Bridge builds.

Why the reverse trend?
The excuse that there's so much going on at POST and boot, doesn't make good sense.
5+ generations of systems, reaching back to the Win9x era, and they (almost) all improved boot time from the previous generations, (new features and complications included)

Fast cold booting just isn't a priority anymore, clearly.
 
@jaxzor42 I'm working on the X670 AX Elite review right now. 2x16GB at DDR5-6000 is fine, no double or triple boot for me on the lasted BIOS. How much ram do you have and configuration?

Are you sure it's not a PBO overclocking fault error instead of the ram?
2x16GB 5600MHz
Ryzen 7 7800X3D base clock...
RX 7900 XTX also base clock... haven't done any OC...just activated the XMP / EXPO 2 (the one that boost the RAM to 5200MHz) cause EXPO 1 (5600MHz) is not stable...
the system works just fine, only that 10s DRAM + 10s VGA POST timer is worrying me... the display has no signal until POST sequence is over and windows loads...
 
@jaxzor42 idk about X3Ds ( don't have one), but 5200 is really low for 2x16 on none X3D. Something is wrong. CPU, Ram or MB
 
m8 i have the same MB... and my POST sequence starts with CPU > DRAM > VGA > BOOT...
CPU and BOOT takes 1 second, but DRAM and VGA takes 10s each... i don't rly know what the issue is... if i try to use the max speed my RAM can do (5600MHz) after windows loads my screen turns off and on twice and it's followed by AMD Adrenaline reporting a bug...if i use EXPO 2 and setup the RAM at 5200MHz the POST timers are the same but i no longer get the screen turning off and on and the Adrenaline bug report also doesn't popup... I really don't know what the issue is since gaming performance is not compromised at all...and I also believe it started to be like this after latest F8c BIOS update...
let me know if you fixed the POST long timers!
cheers!
That's just the DRAM being too fast for your IMC (memory controller)
Normally we'd advise raising the SoC voltage, but with the current issues with AM5 CPU's burning up due to motherboard faults i'd advise keeping that 5200MT/s speed and locking the SoC voltage to something safe like 1.2v


IR_cow might have better manual settings to suggest since he's got the direct experience, my experience with AM5 is that SoC voltages on auto can be wonky as heck, DRAM works better paird in the slots further from the CPU and not closer, and sometimes an over-tightened CPU cooler can cause issues with long boots (but that shouldnt be your issue, since you boot fine at lower speeds)

One good point is that you say "2x16" but nothing else - that could be samsung, hynix, single or double rank - you guys could be comparing very VERY different RAM here.

My DDR4 system has uncommon Hynix MJR chips and behaves totally different to the common DDR4 overclocking advice, as an example.

Blame Microsoft (and all the 'throwaway' PC manufacturers).

If standby/sleep/hybrid sleep/etc.etc. hadn't become 'normal' in consumer PCs (and OEMs hadn't resisted SSDs), AMD (and partners) couldn't 'get away' with this.

Up until these last few generations, it seemed like faster POST+boot was an improving a feature each new generation.

I can remember booting my Ivy Bridge build for the first time, and it was considerably faster than my Phenom II to boot.
-My Phenom II was 'average' vs. the multiple C2Ds+mobos I had before it.
-My C2Ds were faster to boot than my Athlon 64 x2 and Northwood P4 before it.
-My P4 was faster to boot than my Talutin Celeron.
-My Celeron was faster to boot than my Slot1 500mhz P3

Now,
My Ryzen 5 5600 boots slightly slower than my R5 3600 (same mobo.), and both booted slower than my Ivy Bridge builds.

Why the reverse trend?
The excuse that there's so much going on at POST and boot, doesn't make good sense.
5+ generations of systems, reaching back to the Win9x era, and they (almost) all improved boot time from the previous generations, (new features and complications included)

Fast cold booting just isn't a priority anymore, clearly.
Generally, the more out of spec the system is the slower the boots. High RAM speeds have been a big part of slowing boot times on ryzen since day one.

I disable unwanted devices in my BIOS, lock down the boot order (CSM disabled, one device only in the list) and it speeds things up considerably.

My 4th gen intel laptop can boot instantly, but it only does so because it's so cut down it doesnt need to take time to get the DRAM from stock 2133 to 3800, scan SATA ports, NVME ports, negotiate PCI-E generations and so on.
 
Last edited:
I can confirm: "Memory Context Restore" finally works on the latest MSi BIOS. The 28 second waiting time to boot is gone.
 
I can confirm: "Memory Context Restore" finally works on the latest MSi BIOS. The 28 second waiting time to boot is gone.
That's awesome, I wonder if its related to the SoC voltage fixes somehow - that mechanism being screwy (raising voltages every attempt?) could be what was going wrong
 
That's awesome, I wonder if its related to the SoC voltage fixes somehow - that mechanism being screwy (raising voltages every attempt?) could be what was going wrong
Could be, although my SoC voltage was fine even before the fix, yet MCR didn't work.
 
64 gigs ddr4 in and tested, (also did several hours stressapptest).

It did work on XMP timings, but dropped to 3000mhz to match the QLC. (the QLC had 3000mhz of the ram but I brought 3200 as was cheaper).

At first it had long reboots, but after a power cycle its as fast as before.

Hopefully AM5 gets fixed with a newer update on the post times.

This box is used for proxmox VM's including my NAS hence the high amount of ram. This windows I keep on a partition just for testing stuff barebones.
 

Attachments

  • karhu64gigs3000hynix.png
    karhu64gigs3000hynix.png
    666.9 KB · Views: 142
I updated my bios to f9b to see if it fixed sleep issues I was having with f6b (which it did) and noticed it boots faster by default now with xmp enabled. I cleared cmos and then enabled xmp and nothing else. In the past I would have had the long POST times if I didn't manually enable memory context restore.
 
I can now confirm that on my motherboard with BIOS 1.24 (latest stable) the BOOT times even with XMP profile has come down to normal (5-10 secs) as opposed to 30-40 secs on every boot with early versions. I do however have Memory Context Restore turned ON. So yeah patience is the only way forward. It has its demerits of being early adopters but that's the price one has to pay to have the latest tech in hand. Also waking up from sleep takes about 5-10 secs which is normal.
 
Hi, I’ve the same problem with my pc running 7900x and ddr5 from gskill at 6000mhz cl36 (trident z neo rgb) officially supported in the memory list.

Sometimes don’t even post, I have to shutdown and reboot, but when I simple reboot it doesn’t do anything with two leds on the mb flashing . I enabled memory context, but this have not changed the issue. This is the third memory kit I change with no luck.

The mb is a x670 gigabyte gaming ax with latest bios. Also, to update the bios was a pain in the ass, as the bios ask to reboot and the reboot failed every time
 
Hi, I’ve the same problem with my pc running 7900x and ddr5 from gskill at 6000mhz cl36 (trident z neo rgb) officially supported in the memory list.

Sometimes don’t even post, I have to shutdown and reboot, but when I simple reboot it doesn’t do anything with two leds on the mb flashing . I enabled memory context, but this have not changed the issue. This is the third memory kit I change with no luck.

The mb is a x670 gigabyte gaming ax with latest bios. Also, to update the bios was a pain in the ass, as the bios ask to reboot and the reboot failed every time
What is your SoC voltage?
 
IMG_2323.jpeg
(thanks for the reply)

@AusWolf i also tried to
Set the vcore Soc from auto to 1.115 but nothing change
 
Trying with 1.2v vcore soc and memory context on enable, and is the second time posting, but after 60/90 seconds.

It’s already something …

p.s. I’m on bios f9.d for my mb
 
Trying with 1.2v vcore soc and memory context on enable, and is the second time posting, but after 60/90 seconds.

It’s already something …

p.s. I’m on bios f9.d for my mb
Try 1.25 and memory context disabled.
 
Memory context is bad news. It just causes problems. What is does is saves the training part of the memory boot and applies it next time. However AMD didn't design it to work that way so those same settings might be unstable the next boot.
 
Hi, I’ve the same problem with my pc running 7900x and ddr5 from gskill at 6000mhz cl36 (trident z neo rgb) officially supported in the memory list.

Sometimes don’t even post, I have to shutdown and reboot, but when I simple reboot it doesn’t do anything with two leds on the mb flashing . I enabled memory context, but this have not changed the issue. This is the third memory kit I change with no luck.

The mb is a x670 gigabyte gaming ax with latest bios. Also, to update the bios was a pain in the ass, as the bios ask to reboot and the reboot failed every time
SoC voltage?

@AusWolf i also tried to
Set the vcore Soc from auto to 1.115 but nothing change
That's really low, which is why it didnt help. These companies were setting 1.30v to get 6000 working

(No, we don't recommend that high. 1.20 to 1.25v is needed sometimes however)

Memory context is bad news. It just causes problems. What is does is saves the training part of the memory boot and applies it next time. However AMD didn't design it to work that way so those same settings might be unstable the next boot.
From what i've read those changes are voltage related, the memory training is a resistance test and each boot pushes voltages up til it posts - manual settings tend to alleviate that.

Throw in the auto settings also messing with Uclk/Fclk like how MSI were lowering it to 850 or whatever (This was patched, but it did happen early on with AM5 on at least one board), and the longer boots get explained.


As a made up example of what i've understood to be happening:
Warm boots could work just because voltages, timings etc are still in use from the prior boot - like when the bios safe modes kick in and force you to jedec ram speeds, the incorrect settings are still chosen in the BIOS, but ignored.

1. The system tries to boot with internally saved 'cold boot' settings, fails.
2. It then tries to boot with higher voltages (Some boards just kept going infinitely here, and thats what killed CPUs)
3. Then it tries to downclock the related dividers - 6000MT/s could be 6000/2000/2000, and then it'll try a safer setting (varies by brand and bios) and try 6000/1800/850, or whatever they feel like.

Then if all else fails, it should kick in to the 'failed boot' mode and run safe settings - even if they're just JEDEC defaults.
Usually the fails were at 3 boots in the past, so I can imagine 2-3 attempts at each setting is why things take so long.



My AX370 Gaming 5 was like this with 3200 on the DRAM, it would work fine in everythiing but then have cold boot issues.
It'd fail from every single shutdown, so it wasn't power or temperature related - just that the settings the BIOS used for 3200 were not stable.
You could then load the saved 3200 profile on that first boot, and it'd work problem free until the next shutdown, much like the old 'memory hole' issues on 1156 where you could use various FSB settings and memory dividers to get the exact same clocks working one way, but not another

That system could only be fixed by lowering the RAM speed/IF clock, 3133 works perfectly with zero issues and boots were almost instant - by the time the screen powered on, i was seeing the windows desktop.


TL;DR: If you get these long boots, manually set your DRAM voltage, SoC voltage, MCLK UCLK and FCLK. Keep them in sync and slowly lower them til the issue is gone, and leave it there until new BIOSes come along.
 
Last edited:
Using a vcore soc voltage of 1.25v and disabling memory context resulted in about 2 minutes to boot (135 seconds to be precise)

I didn’t change the settings of the voltage for the dram, everything is controlled by expo. I’ll try with manual voltage for the dram, but what settings I have to use ? Those from expo ? and what about Mclk uclk and fclk?

thansk again for your help guys
 
Using a vcore soc voltage of 1.25v and disabling memory context resulted in about 2 minutes to boot (135 seconds to be precise)

I didn’t change the settings of the voltage for the dram, everything is controlled by expo. I’ll try with manual voltage for the dram, but what settings I have to use ? Those from expo ? and what about Mclk uclk and fclk?

thansk again for your help guys
DRAM voltage is not the problem. If the RAM is designed to run at X Volts, then it's fine.

Your 2-minute boot should only be the first one. Try booting again and see if it's still that slow.
 
vcore soc
Vcore and VSoC are NOT the same thing, do not mix them up!

Vcore is your CPU voltage, Vsoc is SoC voltage. DRAM voltage is something else again.
 
Memory context is bad news. It just causes problems. What is does is saves the training part of the memory boot and applies it next time. However AMD didn't design it to work that way so those same settings might be unstable the next boot.
The thing that bugs me about MCR is if its saved to a bios profile, then load another bios profile (with MCR on when it was saved) it still takes a couple minutes to boot with the chosen bios profile despite the fact that the bios profile is already saved (with the MSR on) . So shouldn't the motherboard be remembering the precise settings for the RAM from the bios profile in the first place & that of course includes MCR being on? I've seen this problem on an Asrock, MSI & even a Gigabyte AM5 boards I have here despite agesa updates for several months already. There is still work to be done to fix these bios problems with Zen 4.
 
Back
Top