Thursday, September 1st 2022
ASRock X670E Steel Legend Motherboard Needs Hundreds of Seconds at First Boot or Clear CMOS to Train Memory
At this point, we don't know if this is a limitation at AMD's level or ASRock's, but someone with access to a retail ASRock X670E Steel Legend motherboard, with all its packaged paraphernalia in place, spotted an interesting sticker covering the board's four DDR5 DIMM slots. The sticker has some info on the ideal DIMM slot selection for dual-channel memory (4x sub-channels); but what catches our eye is a table which states just how long the motherboard will take to train the memory the first time it's booted up, or after a clear-CMOS operation (where your BIOS settings are erased).
The table says that a typical setup with two 16 GB modules (read: two single-rank modules in a 1 DIMM per channel/1DPC configuration), takes 100 seconds to train (or until first boot). Two 32 GB modules (typically a pair of dual-rank modules in 1DPC configuration) take 200 seconds, as do four 16 GB modules (four single-rank modules in a 2DPC configuration). The least optimal config, four dual-rank modules in a 2DPC configuration, takes a whopping 400 seconds (almost 7 minutes) to train. That's 100 to 400 seconds of a black screen, or no display signal, enough to unnerve anyone and assume something is DOA.
Update Sep 2nd: The source behind this story confirmed that this is an ASRock-level issue, and that it's been "fixed" with the latest BIOS.
Update Sep 8th: This has been fixed according to ASRock.
Here's the kicker—since UEFI BIOS updates typically clear CMOS, you'll have yourselves some nerve-racking hundred(s) of seconds until the display lights up, letting you know that the BIOS update went through. Interestingly, we haven't yet seen anything to suggest that memory overclock (which involves dozens of reboots and re-training of memory), takes hundreds of seconds—not unless you clear CMOS for some reason.
Source:
HXL (Twitter)
The table says that a typical setup with two 16 GB modules (read: two single-rank modules in a 1 DIMM per channel/1DPC configuration), takes 100 seconds to train (or until first boot). Two 32 GB modules (typically a pair of dual-rank modules in 1DPC configuration) take 200 seconds, as do four 16 GB modules (four single-rank modules in a 2DPC configuration). The least optimal config, four dual-rank modules in a 2DPC configuration, takes a whopping 400 seconds (almost 7 minutes) to train. That's 100 to 400 seconds of a black screen, or no display signal, enough to unnerve anyone and assume something is DOA.
Update Sep 8th: This has been fixed according to ASRock.
Here's the kicker—since UEFI BIOS updates typically clear CMOS, you'll have yourselves some nerve-racking hundred(s) of seconds until the display lights up, letting you know that the BIOS update went through. Interestingly, we haven't yet seen anything to suggest that memory overclock (which involves dozens of reboots and re-training of memory), takes hundreds of seconds—not unless you clear CMOS for some reason.
89 Comments on ASRock X670E Steel Legend Motherboard Needs Hundreds of Seconds at First Boot or Clear CMOS to Train Memory
This clearly ASRock's fault, they've been notoriously inflating their number and extremely finicky with memory timings. Heck even 1usmus's DRAM Calculator for Ryzen doesn't work at all.
I'm using ASRock B550M Steel Legend and back then, GSkill Flare X 3200MHz CL14 had trouble booting even at stock speed, let alone overclocked say at 3466Mhz and thus, I choose to live with loosely timing Team Dark α .
It's not even their high end, Taichi or OC Fomula for this matter, this is Steel Legends targeting casual user. If you want memory overclocking I suggest you look no further than MSI or Asus, as Gigabyte seem to fall in same ballpark as ASRock.
edit :Who would've thunk :rolleyes:
Then check the timings numbers under the SPD tab (which is the next one on your CPU-Z screenshot), and go to BIOS in order to make the timings as shown in the SPD tab!
No way 82 clocks for the tRC Bank Cycle Time!
AMD release AGESA PI 1.2.0.7 back in June and until now ASrock still stuck in Beta phase, so no update (yet).
RAM overclocking is heavily dependant on divider, clock multiplier, etc it's boiled down to motherboard, why updating BIOS will make any differences? Just grab Asus or MSI for RAM overclocking.
I don't see what kind of errors would it give you which wouldn't be present otherwise (quick boot = ON).I think you got it wrong. It's 200 sec as per the first image in the article. So about 3 minutes or a third of what you think it is.Just in time till the 3D-V cache models appear. Ironed out MB, BIOS and platform issues with the new cpus. Perfect.
.. it didn't overclock, at all
(and now here's a board that takes three minutes to POST after a CMOS clear from an explosive memory config hee hee, jab jab, hopefully a fluke rather than a mainstay)
Has everyone already forgotten their AMD history? Am I becoming the old man now?:toast:
Overclocking modern AMD hardware does make sense, though not in the sense of raising voltage and clocks;
more in cutting lard bad for system responsiveness, dialing in the right voltage combinations to minimize fabric error corrections, making the chips stop pointlessly gargling watts ...
In terms of "total throughput", outside of memory overclocking increasing it when memory-performance limited, since Zen2 it has been mostly in the realm of +0-4%
In terms of responsiveness and performance stability, then Zen3 (Vermeer) in particular was incredibly annoying to work with.
It's just not like intel where you can throw 400W at it for minimal extra gains
5800x is advertised as 3.8GHz to 4.7GHz
Yet stock settings will run all-core load at 4.4GHz to 4.85GHz boost, with PBO (which is overclocking) you can get 4.6GHz-5.05GHz
You could also run static OC's which are more limited, but my present 4.6GHz all core is hardly an issue - i can still easily game at 1440p 165Hz and be GPU limited with a 3090
So when they clock higher than advertised, PBO overclocking goes higher again, and then all core overclocking can boost sustained MT at the cost of some time limited ST - you call it a disaster?
4.4GHz - 4.85GHz stock to 4.6GHz to 5.05GHz with PBO is a 200MHz overclock, or around 4%. Hitting those numbers at all is not common on the 5800X, which is an incredibly hot CPU.
I see you're running a custom loop for your 4.6GHz, which is lower than the all-core load frequency you listed as stock above ^^ Even if you pump that up to 5GHz, you're talking about a custom loop for a 2-10% overclock, whereas I was comparing to 30%+ overclocks on air on older CPUs.
Nobody is contesting whether it's a capable CPU or not, the matter being discussed is OC headroom, of which there is very little. AM4 also covers Ryzen 1000, 2000(G), 3000(G) and 4000G, all of which pretty terrible overclockers with the odd exception here and there.
If you look at my post,
- Athlon Thunderbird, highest SKU was the 1.4GHz, architecture topped out around 1.7GHz on air cooling (21%). The AXIA 1000 could do 40% with air cooling, with some doing 50% or more.
- Athlon XP Thoroughbred, highest SKU was 2.2GHz, architecture topped out around 2.5GHz on air cooling (13%). The lower end 1700+ could come pretty damn close to maxing out the architecture on air (around 64% on air cooling for a good number of them)
- Athlon64 San Diego, highest SKU was 2.8 GHz, architecture topped out around 3GHz on air cooling (7%). Take the lower end 3700+ which could do the same +/- 3GHz, and you're talking 36% on air.
Compare that to what you're running, which is (even by your definition) slap bang what you should expect from stock ignoring PBO. To get the overclocking headroom I was talking about, you'd need to run that somewhere around the 6.5 to 7.5GHz range on air to match headroom of old platforms - hence overclocking is significantly worse on AM4 than prior platforms.Even if we take OC headroom as every MHz above the base speed and ignore boost/PBO/etc, 3.8GHz to 4.6GHz is 21%, and you'd need to be running in the 4.75 to 6GHz range on air to match.
I'm not sure why you brought Intel into it, as Intel as never part of the debate. The debate was old AMD vs new AMD OC headroom.
You've taken this far off topic, though, as my post was originally replying to "AMD has historically never been an overclocker's dream..."
As well as various AGESA versions breaking various things about CPUs running non-stock, RAM overclocking suddenly requiring different SOC/CPU 1.8v voltage to run [X Settings], with no warning.
The "non-deterministic boost clock" algorithm is very impressive on paper, but ultimately serves as a way to increase throughput at the expense of latency.
So yes, a Zen3 CPU can hit 5.1 GHz with non-deterministic boosting - but it's going to be less responsive and barely ""perform better"" (potentially, worse!) than having EVERY SEMI-POINTLESS BELL AND WHISTLE turned off running a static clock at 4.6-4.75 Ghz.
very nice for long, non-gaming workloads though! so is ddr5 ...
There's a very high likelyhood that Zen4 is no different in this regard, but at least the ceiling is likely to be much higher now.
And I hope, I really hope that you can adjust SOC/IOD/CCD voltage in-OS now, without having to Reboot/POST/BIOS/Adjust/Reboot/POST/Boot/Test/[Repeat] to test each +0.01v step with 90 permutations to go through to eradicate fabric error corrections at high FCLK.