Did you get a shiny new Ryzen 3000 CPU?

[Chomiq]

I have the original SX8200NP and a "last BIOS time" in Task Manager of around ~18 seconds. I previously thought it was a Ryzen 3000 + B450 problem.

The fast boot options don't make a difference for me; once Windows starts booting everything is hella fast. You should probably just move it up to the first slot, it's not like NVMe/PCIe isn't backwards compatible. You get to use the board's heatsink too.

I'm not even sure if it's NVMe that causes this delay at boot. It seems like the most plausible explanation, but in that case, how come just changing from 2133 JEDEC speeds to 3200 C16 XMP cut down my pre-Windows booting time to 1/3 of what it was on JEDEC? Literally nothing else had changed with my system when this happened.

I know it is, I'm just trying to save myself a hassle in case I'll ever get a 4.0 one.

 

[TheLostSwede]

I guess you guys boot 1 second faster than me, I'm on 19 seconds. Can't say it feels like it takes that long though...

 

[tabascosauz]

I know it is, I'm just trying to save myself a hassle in case I'll ever get a 4.0 one.

Bruh, you only need to undo 1 screw and reinstall it again I just received my replacement SATA daisy from Cablemod and was able to power up my MX300 and Blue3D again; working with 2.5" in confined spaces is such a bitch once you've experienced NVMe.

I guess you guys boot 1 second faster than me, I'm on 19 seconds. Can't say it feels like it takes that long though...

It's somewhere between 15-19 seconds on average. It's not consistent, if judging by the number of times the GPU fans "twitch" before the board LEDs come on. But the actual perceived time from pressing the power button to Windows loading is more like 4-5 seconds, so Task Manager's number feels a little BS-y to me.

 

[Chomiq]

I guess you guys boot 1 second faster than me, I'm on 19 seconds. Can't say it feels like it takes that long though...

20.3, so it's totally random. Realtek, OneDrive and Windows Security in startup. Funniest part is that my M6Pro on 3470 managed to boot in 7.3 sec.

 

[EarthDog]

All this boot talk.. what are each of you guys using to measure? I know task manager reads UEFI boot time, but that isnt Windows load...

 

[Chomiq]

Yeah the actual load is like 3 sec tops.

Anyhow. After coming back from sleep case fans are spinning at full speed, any ideas?

 

[Tomgang]

Man now that i have seen Intels comet lake line up, i am even more sure on my next upgrade will be Ryzen 3000. I cant wait to get a better gaming exsperience that a new CPU will give me over my old I7 980X cpu. AMD hurry up with that Ryzen 9 3950X. The wait has begin to become unbearable

And also i need new stuff to blow throw benchmark again.

 

[biffzinker]

Yeah the actual load is like 3 sec tops.

Anyhow. After coming back from sleep case fans are spinning at full speed, any ideas?

Doesn't even come out of sleep for me if you've switched the multiplier from auto. Change it back to Auto though, and no problem resuming from sleep.

 

[tabascosauz]

All this boot talk.. what are each of you guys using to measure? I know task manager reads UEFI boot time, but that isnt Windows load...

I'm not talking about Windows boot time. Windows boots fast.

I'm talking about the time it takes to get to the logo screen, and therefore the BIOS, and by extension the start of the Windows boot process. I don't think I've had a single rig that has a weird mandatory delay like this between pressing the power button and the start of BIOS / GPU fan activity / board lights / any hard drive activity whatsoever. Judging from that lack of HDD activity, I guess it's just an NVMe thing, then?

 

[biffzinker]

Just timed with a stopwatch how long after the power on switch is pushed it takes before Windows started to boot.

19.53 seconds

 

[EarthDog]

If you are timing the bios, use task manager.

 

[tabascosauz]

Just timed with a stopwatch how long after the power on switch is pushed it takes before Windows started to boot.

19.53 seconds

Wow, my perception of time is really skewed as hell nowadays. 15.90 timed with a stopwatch. Looks like Task Manager is pretty spot on.

15.90 until Windows 10 started to boot.

 

[biffzinker]

Wow, my perception of time is really skewed as hell nowadays. 15.90 timed with a stopwatch. Looks like Task Manager is pretty spot on.

That was up until it started loading Windows 10.

That was up until it started loading Windows 10.

How many more seconds on top of that when Windows finally loaded not sure. Ran out of time

 

[Chomiq]

So few questions:
- is boot time of 18 seconds normal? W10 on sx8200 Pro. With ultra-fast boot enabled. That being said it's from Task Manager, so who knows what it actually measures. Boot takes a bit long before Gigabyte logo is updated but I guess it's due to F3 bios I'm using which is stock
- temp fluctuations at idle - I assume that's normal (for Ryzen), at least from what I read in a thread over here
- ultra fast boot or regular? Ultra fast basically skips "del to enter bios"

I've mounted NVMe in second slot, since it's not a PCIe 4.0 drive. Should I keep it there, or mount it in first NVMe slot?
View attachment 129973

Specs are listed, all on X570 Aorus Elite

Switch over NVMe to the first slot:

 

[TheLostSwede]

Doesn't even come out of sleep for me if you've switched the multiplier from auto. Change it back to Auto though, and no problem resuming from sleep.

That sounds like a UEFI bug to me.

I'm not talking about Windows boot time. Windows boots fast.

I'm talking about the time it takes to get to the logo screen, and therefore the BIOS, and by extension the start of the Windows boot process. I don't think I've had a single rig that has a weird mandatory delay like this between pressing the power button and the start of BIOS / GPU fan activity / board lights / any hard drive activity whatsoever. Judging from that lack of HDD activity, I guess it's just an NVMe thing, then?

Keep in mind that you're graphics card can affect this as well, as the graphics card BIOS/UEFI is the first thing to start, you generally don't see that though, as the screen is blank on most boards with UEFI during that time. Then the UEFI kicks in. That said, my previous Asus board was just the same as this Gigabyte board, regardless of the OS drive being SATA or NVMe.

 

[Chomiq]

Tried to adjust ram freq and timing with DRAM calc. 3600 just wasn't stable with built in membench test - it threw errors on SAFE and FAST. 3466 is stable at both with XMP High Freq Support on lvl 2 I believe. Gonna DL memtest and validate.

 

[biffzinker]

Looks like Task Manager is pretty spot on.

I checked what Task Manager is reporting.

 

[tabascosauz]

So...the Gigabyte BIOS has this option called Core Performance Boost, with possible Auto and Disabled modes.

• With it on, temps + voltages + PWM speeds are a wild freaking ride, even at idle. SMT on, SMT off, Ryzen Balanced, Windows Balanced......none of it makes a difference.
• With it off, the system actually behaves like it's sane. Idling between 37-40C, no more random fan spikes, Vcore not peaking past 1.0V. Mirrors the behaviour out of setting 99% Maximum Processor State, but even more stable.

Problem is, Vcore doesn't peak past 1.0V. Since Ryzen 3000 intertwines voltage with clockspeeds, this means that the 3700X never boosts past its base speed (3.6GHz). Yes, this also means that at full load, it barely breaks 50C and the fans don't even change from idle, but it also means that it never reaches any boost speed, ever.

I'm not sure what to do with this, honestly. The rewards of 7nm are really obvious with CPB off; at almost the exact same load clocks, the 3700X is nearly 25C cooler than the 4790K, sips roughly 80% the voltage, and is faster while bringing 2x the cores and 4x the threads to bear (I had HT off on security recommendation). With CPB off and staying at base clocks, the efficiency of the 3700X is nothing short of mindblowing. Fan speeds don't change with load. At all.

But at the end of the day, this isn't what's advertised. I was sold a CPU that boosts from 3.6 to 4.4GHz, and while I can forgive ~30MHz of missing frequency to the advertised 4.4GHz, I cannot forgive missing out on 800MHz. Generations and generations of Intel and AMD CPUs have worked just fine with a few of the cores at maximum boost speed, without 1.5V Vcore, 55C idle temperature spikes, and fan speed spikes that make my rig sound like a GE90. the firmware and software that control the 3700X are such hot garbage that I wouldn't even call this a finished product.

I'm not really demanding much from my system at the moment, so I might just leave it like this with CPB off for now and wait for a F43 BIOS with 1.0.0.4 AGESA. Quite honestly, I'll much rather lose a few frames in CPU-dependent games than keep on having a jet engine on my desk and running 1.50V through my new 3700X. Chances are, nothing will change.

Between the complete inability of my B450 board to run 32GB properly at respectable speeds and timings, and now this 3700X's all-or-nothing approach to thermals, I'm just a teeny bit regretful I didn't just get a 9700K instead. At least I can enjoy the everyday breakneck speeds of the SX8200.

@TheLostSwede thanks for the heads up, it's getting late and my patience is wearing thin with AMD.

 

[TheLostSwede]

@tabascosauz you said on in both cases, so you might want to edit your post for clarity...

Also, what are you using to measure the CPU clock boosts with? HWInfo seems to be the most reliable thing.

 

[tabascosauz]

@tabascosauz you said on in both cases, so you might want to edit your post for clarity...

Also, what are you using to measure the CPU clock boosts with? HWInfo seems to be the most reliable thing.

HWInfo is generally reliable, with the one exception of idle clocks below 2.2GHz. For that, I have to visit Ryzen Master.

The difference in clockspeeds, voltages, temperatures and PWM are plain to see with HWInfo monitoring in the background. HWInfo, Ryzen Master, CPU-Z and OCCT all report the same thing: 100% load incites no more than 3.6GHz and 1.01V.

Also starting to think that Ryzen is going way too far with taking Vcore into its own hands. BIOS reports 1.225V. CPB on, 0.2/0.9/1.5V idle and 1.36V load. CPB off, 0.2/0.9V idle and 1.0V load. What's the point of still having a tweakable Vcore if it's just going to do whatever it likes?

 

[TheLostSwede]

HWInfo is generally reliable, with the one exception of idle clocks below 2.2GHz. For that, I have to visit Ryzen Master.

The difference in clockspeeds, voltages, temperatures and PWM are plain to see with HWInfo monitoring in the background. HWInfo, Ryzen Master, CPU-Z and OCCT all report the same thing: 100% load incites no more than 3.6GHz and 1.01V.

Also starting to think that Ryzen is going way too far with taking Vcore into its own hands. BIOS reports 1.225V. CPB on, 0.2/0.9/1.5V idle and 1.36V load. CPB off, 0.2/0.9V idle and 1.0V load. What's the point of still having a tweakable Vcore if it's just going to do whatever it likes?

Yeah, agreed, it doesn't seem to see the real idle clocks, in fact, I can't see anything below 3,075.25MHz...
That said, I just tried to enable all the overclocking features and what not to see if it made any difference on the latest UEFI for the X570 Master and it doesn't do anything compared to running the board at standard settings.

That said, I don't have the issues you're having with the high Voltages at all times, but it was an issue on earlier UEFI versions, so maybe they'll get that fixed on your board in the next update as well.

You're clearly not getting the right boosts due to too low Voltage with your current settings, but that seems like a bug. That said, I've always had CPB set to Auto, but I can have a go at turning off to see what happens.

As for your last question, that's something you're going to have to ask AMD, as I don't have an answer for you.

This is mostly idle at the desktop with a few background apps running as well as Chrome, but with all auto OC settings enabled and maxed.

And this is with all the auto OC settings manually disabled in the UEFI.

Seeing the same boost clocks in both cases, i.e. up to 4,500.2MHz, so no odd behaviour like in your case.

 

[Chomiq]

(...)

Set system fans to fixed rpm, set temp update interval to 5 sec on CPU fan in gigabyte software.

 

[TheLostSwede]

So...the Gigabyte BIOS has this option called Core Performance Boost, with possible Auto and Disabled modes.

Aha! That was hidden somewhere else in the UEFI, not under the overclocking settings.
And you're indeed correct, it limits my CPU clocks to 3.9GHz, i.e. base frequency of the CPU.
That's a very odd option to have indeed.
Doesn't do squat in terms of thermals at idle for me though, but under full load the CPU obviously doesn't get as hot.

 

[tabascosauz]

@TheLostSwede thanks. Yeah, the massive spread of idle voltage from 0.2V to 1.5V is a major concern for me. When you stress test with CPB off, does it actually hit 4.5GHz or does it just say that and actually top out at 3.9GHz in HWInfo?

Just another issue with saving money on a board not X570, I guess. Gigabyte does seem to release a lot of BIOSes for my board to keep it up to date with AGESA, but this problem clearly demonstrates that something's amiss.

Set system fans to fixed rpm, set temp update interval to 5 sec on CPU fan in gigabyte software.

Unfortunately, that's a bit like burying my head in the sand. With CPB on, it's still feeding a 1.4V spike into the CPU regardless of what the fan setting is. Right now the fan reaction time is delayed but still enough to keep the temps from reaching 70C. Reducing the polling rate would just allow the CPU to roast. The fans would still kick in on longer spikes anyways.

CPB off clearly demonstrates that it does know how to behave itself, it just chooses not too on CPB.

 

[TheLostSwede]

@TheLostSwede thanks. Yeah, the massive spread of idle voltage from 0.2V to 1.5V is a major concern for me. When you stress test with CPB off, does it actually hit 4.5GHz or does it just say that and actually top out at 3.9GHz in HWInfo?

Just another issue with saving money on a board not X570, I guess. Gigabyte does seem to release a lot of BIOSes for my board to keep it up to date with AGESA, but this problem clearly demonstrates that something's amiss.

Why is the spread an issue? Some of the cores are in deep sleep, so they go down to 0.2V, this happens normally for me now on the latest UEFI on my board.
It was an issue on early UEFI's where these ultra low Voltages weren't reported properly, but AMD fixed that in AGESA 1.0.0.3ABB and the new drivers.

And no, I can't boost at all with CPB disabled.