Solved! My 5800x + 3080 thoughts - Read if upgrading

[Zstream]

Hi Everyone,

I thought I would post this in the CPU section, because the majority of time spent tweaking has been on the CPU side of things. If you want this thread moved, feel free to move it.

Just to give everyone a quick overview of what I've been dealing with.

Old Setup:

• Asus TUF Gaming (wifi) x570 board
• AMD 3600
• 32gb 3200 (b-die) O/C to 3400 16/17/17/36
• Two Intel 512GB SSD's 545S Series
• Cooler Master 240 AIO
• Nvidia 2060 EVGA

New Setup:

• Asus TUF Gaming (wifi) x570 board
• AMD 5800X
• 32gb 3200 (b-die) O/C to 3400 16/17/17/36
• Two GAMMIX S11 PRO
• Cooler Master 360 AIO
• Nvidia 3080 EVGA

The journey started when I upgraded the BIOS to 2607 from ASUS the day before I picked up my 5800X. After I updated the BIOS, I unplugged the machine, took out the cooler, and CPU and proceeded to wait. After the extremely long wait at Microcenter, I picked up the chip, put it in and did all the usual with paste and what not for the new AIO. I did not swap out my SSD's as of yet, because I wanted to see the differences right away, and to be honest, installing Window 10 isn't exactly five minutes of work

Well the CPU was stuck on 1.1GHZ because that bios didn't support the version I was running. To be honest I either clicked the wrong download for the BIOS or did multiple downloads on the USB drive, because the chip wasn't really acting properly at all. I ended up download the latest version again, even though I thought I downloaded it before. Apparently not! I finished installing the latest bios that had support for the new chips (2802). At least Windows started up properly, and the speed was all good. I did notice however that the core being fed to this chip was around 1.4/1.49... I was shocked and frankly scared to proceed any further after I found that out.

I went into the Asus BIOS and struggled with getting this thing to accept the bios vcore setting of 1.25, or any other rational setting for that matter. I ended up running Asus O/C tuner which somehow managed to feed the CPU 1.09 while maintaining 4.2ghz. With Ryzen master reporting better temps, and core voltage I felt a little better as there is no way it needs that much to function.

After that cooled my emotions for a bit, I also noticed that ASUS had new chipset drivers. Noob Tip: Check for bios and chipset drivers at the same time!

Upgraded the latest chipset drivers and while installing the drivers the install screen went to **** and either the GPU panicked or CPU went crazy because it kept launching in a one inch by inch window.. I managed to install the drivers and all went well. Noticed that the Ryzen power setting went away, which if I'm being honest.. annoyed me.

After all said and done I ran some benchmarks and compared that to my 3600 + 3080 scores, and nothing really showed distinct increase in measurement. Now, maybe my benchmarks aren't a list of the latest games, but it was within a 10-15% range. Some FPS managed to stay higher longer, and the rapid change in FPS was much less. Outside of that, not much of a difference.

Once that was all done, I went ahead and went with a clean install of Windows with the new NVME's. Well if you're new to those, you realize that these things run HOT. I mean they both run at 44C, which is often 10 degrees warmer than the CPU. For reference the Intel SSD's run at 20-25C.

Over the past four hours I've moved to the 2812 bios, and again the vcore sky rocketed, and was forced to run the ASUS O/C which stepped in to change vcore setting and max clock. I can't for the life of me be able to set this manually..

TLDR - Update BIOS to the latest version, watch your vcore with Ryzen Master or just enter the bios! There are issues on Nvidia's side with drivers, and what seem to be bios' on motherboards. I'd personally wait a few weeks or maybe a month or two if I had to make this decision again.

 

[alcoholbob]

44C is pretty cool, heh mine idles at 48-50C, and when I'm installing a big game from Steam it will go up to the 70s. I have an ITX board so there's literally no airflow under the case where the NVME drives are.

 

[lightmanek]

It is expected that Ryzens can use up to 1.5V STOCK (for ST boost). Nothing broken with your BIOS or board, it is expected to see that voltage when idle. Go back to defaults and run any multi-core stress test to see that voltage will go down to 1.1-1.2V automatically.

 

[Zstream]

It is expected that Ryzens can use up to 1.5V STOCK (for ST boost). Nothing broken with your BIOS or board, it is expected to see that voltage when idle. Go back to defaults and run any multi-core stress test to see that voltage will go down to 1.1-1.2V automatically.

It doesn’t. It also increases core temp across the board. Additionally, in many games and stress tests, the scores are equal if not lower in some cases.

 

[coercitiv]

Noticed that the Ryzen power setting went away, which if I'm being honest.. annoyed me.

Robert Hallock from AMD had the following to say on the subject:

Ryzen 5000 Series doesn't need a power plan. Don't expect to see one.

But you CAN tweak Ryzen perf vs. power with the Win10 Power & Sleep sider AFTER you install the chipset driver.

He also said undervolting is not available temporarily, new AGESA will actually bring improved undervolting over previous gens. Now if I understand correctly, you're having trouble even selecting a static Vcore in the current BIOS, which is probably not related to undevolting per se, since it affects overclocking as well (and we know people have already overclocked with static voltage options enabled in BIOS).

My advice would be to reset BIOS settings and don't enable XMP if you have ON currently. (just write down your main timings and set mem clocks manually) The reason for this is XMP might be linked with some "enhancement" on the Asus side that pushes voltages because it doesn't know any better. IIRC Asus did this one some Intel boards too in a way, enabling XMP also turned on MCE.

 

[Zstream]

Robert Hallock from AMD had the following to say on the subject:

He also said undervolting is not available temporarily, new AGESA will actually bring improved undervolting over previous gens. Now if I understand correctly, you're having trouble even selecting a static Vcore in the current BIOS, which is probably not related to undevolting per se, since it affects overclocking as well (and we know people have already overclocked with static voltage options enabled in BIOS).

My advice would be to reset BIOS settings and don't enable XMP if you have ON currently. (just write down your main timings and set mem clocks manually) The reason for this is XMP might be linked with some "enhancement" on the Asus side that pushes voltages because it doesn't know any better. IIRC Asus did this one some Intel boards too in a way, enabling XMP also turned on MCE.

good idea. I’ll do that later today and see what happens. The only difference in what you stated and what happened is that I chose the profile first and than modified it after. I’ll disable the profile, and test. However I know for a fact (if it’s being reported properly) that this thing can go to 4.2 all core on 1.1 volts. I have no idea why default it would be at 1.4/.49 - yes, it will fluctuate down a bit, but anywhere near 1.09/1.1.

I personally think the bios’ are too new and not working as intended.

 

[Zstream]

Same result it seems. It appears you can't set the voltage, and the board wants to jump to 1.49 on too many occasions. I sure hope the chips can handle that amount... I'm more intrigued as to why that is. I can achieve an all core o/c of 4.4 with the Asus o/c tool, and it stays at 1.1. Is the bios and software not reading the voltage correctly in that state? It seems weird to me that it can get that high on 1.1 and need 1.4 to get a single core up to 4.8?

 

[curryloti]

1.50V is normal transient boost voltage for Zen 2/3 chips

 

[thesmokingman]

I did notice however that the core being fed to this chip was around 1.4/1.49... I was shocked and frankly scared to proceed any further after I found that out.

You really should have read up on how Ryzens works and that IS how they work. I would not be setting voltages manually w/o knowing or having any experience with how these chips use voltage.

Also the bios' just released are not full fledged bios unless they are AGERSA 1.1.0.0. The bios' still have a ways to go though so early adopter rules at play.

 

[Zstream]

You really should have read up on how Ryzens works and that IS how they work. I would not be setting voltages manually w/o knowing or having any experience with how these chips use voltage.

Also the bios' just released are not full fledged bios unless they are AGERSA 1.1.0.0. The bios' still have a ways to go though so early adopter rules at play.

yeah, no. I’ve been setting voltages for over twenty years. I’m not sure where your response is coming from, but saying near 1.5 is normal, or desirable is foolish and shows a temperament not yet mature.

 

[Zstream]

1.50V is normal transient boost voltage for Zen 2/3 chips

I’ve never seen my 1700, 3600 get anywhere close to that.

 

[IEC]

Where to even start with the (probably well-intentioned) misinformation in this thread.

You should not be manually setting voltages for Zen based chips. You can attempt to use an offset or limit PPT via PBO in order to fine-tune load power consumption, but Zen 2 and Zen 3 should really be run at stock settings unless you know what you are doing. Just because you see the chip request a voltage of 1.47V-1.50V transiently, does not mean it is running out of spec or that it is receiving anywhere near that voltage. That is normal behavior for the chip. By manually setting voltages or other limits you may significantly decrease your performance.

However, if you never see it running below those numbers while idle at desktop, it is likely that you have some application which is nearly-constantly using at least a single thread and preventing low power states. Examples of this that have been found are Corsair iCue, certain hardware monitoring tools, and some cloud-connected applications. You can only experiment by killing off applications until you find which is the one preventing your chip from going to sleep. If you install Ryzen Master and you have a true idle power state at desktop it should look like this:

YMMV depending on specific motherboard and BIOS version. Seems unlikely it is the BIOS given it is ASUS, but you can always re-flash latest version and try reloading default settings to rule that out.

 

[Kenmitch]

Maybe it would help with the diagnosis by posting images of the uEFI?

What happens if you enable ECO mode?

Seems unlikely it is the BIOS given it is ASUS, but you can always re-flash latest version and try reloading default settings to rule that out.

ASUS from time to time will release a whacked uEFI just like others. It could be something silly like AMD Cool and Quiet disable by default that's the culprit. There's also things like ASUS core enhancement, game mode/boost, etc

One thing for sure is it can be challenging at times to diagnose with hands on in the uEFI, but it's even more so doing it blindly.

 

[thesmokingman]

yeah, no. I’ve been setting voltages for over twenty years. I’m not sure where your response is coming from, but saying near 1.5 is normal, or desirable is foolish and shows a temperament not yet mature.

Ignorance is bliss. You still don't know wtf you are talking about regarding Ryzens and voltage. Consider this a warning about the voltage.

 

[Zstream]

Where to even start with the (probably well-intentioned) misinformation in this thread.

You should not be manually setting voltages for Zen based chips. You can attempt to use an offset or limit PPT via PBO in order to fine-tune load power consumption, but Zen 2 and Zen 3 should really be run at stock settings unless you know what you are doing. Just because you see the chip request a voltage of 1.47V-1.50V transiently, does not mean it is running out of spec or that it is receiving anywhere near that voltage. That is normal behavior for the chip. By manually setting voltages or other limits you may significantly decrease your performance.

However, if you never see it running below those numbers while idle at desktop, it is likely that you have some application which is nearly-constantly using at least a single thread and preventing low power states. Examples of this that have been found are Corsair iCue, certain hardware monitoring tools, and some cloud-connected applications. You can only experiment by killing off applications until you find which is the one preventing your chip from going to sleep. If you install Ryzen Master and you have a true idle power state at desktop it should look like this:

YMMV depending on specific motherboard and BIOS version. Seems unlikely it is the BIOS given it is ASUS, but you can always re-flash latest version and try reloading default settings to rule that out.

I’m not going to argue with you two but it’s annoying to have someone come in here and tell me I don’t know what I’m doing when I’ve been doing fit successfully for a very very long time. I have extremely high performance numbers and targets. To willfully just state that the hardware monitor prob is the cause is foolish, if not downright ignorant. I have been doing this on all zen models starting at 1700.

I thank you for your attempt at “something” but ok.

 

[Rigg]

Just so we are clear....

With stock default settings you are seeing 1.4v in a high current all core workload? What is the PPT during this load condition? Do you have multiple monitoring apps running at the same time? Can you post a screenshot(s) with HWI64 and a heavy all core workload running?

What BIOS version? Version 2812?

Have you cleared CMOS and double checked that all advanced AMD settings are set to default?

 

[Zstream]

Just so we are clear....

With stock default settings you are seeing 1.4v in a high current all core workload? What is the PPT during this load condition? Do you have multiple monitoring apps running at the same time? Can you post a screenshot(s) with HWI64 and a heavy all core workload running?

What BIOS version? Version 2812?

Have you cleared CMOS and double checked that all advanced AMD settings are set to default?

I'm seeing 1.4v without workload at peak.. Here is the default settings on the 2812 bios. This is with a clean install of windows, all drivers blah blah. I'll try negative offset again, and test.



Set the mem to usual settings, and received a less score in 3dmark in graphics and CPU tests. If I set the all core max to 4.4ghz, and use the Asus O/C settings, I receive a higher score and use less voltage and the temps are much less.

 

[Zstream]

The BETA 2812 bios does decrease some of the voltages in the day to day running of windows. It still often spikes to 1.3/1.4. I've tried the negative offset, and it works in this BIOS, with little downfall on performance. All in all, if the voltages of ASUS O/C are accurate, I'd leave the system at that. Here are the compares. The score of 16772 is on the 2812 beta bios with default CPU settings. The vcore bounces around a bit, reaching 1.4+ on that bios. The 16721 is with ASUS/OC set with the multiplier to get 4400mhz across the board, using less voltages throughout. I'm 100% certain that the chips do not need to be fed 1.4 (if asus o/c and ryzen master are accurate).

 

[tamz_msc]

I'm seeing 1.4v without workload at peak.. Here is the default settings on the 2812 bios. This is with a clean install of windows, all drivers blah blah. I'll try negative offset again, and test.

If you're seeing 1.4V in the BIOS then that's normal. If you are seeing 1.4V at idle in Windows, then it is possible that voltage monitoring on the motherboard is not accurate. I know that the Crosshair VIII Hero has proper voltage monitoring, and it shows <1V idle in CPU-Z from the various videos that I've perused.

I've seen 4.7GHz stable all core Cinebench R20 runs at 1.25V with the Crosshair VIII Hero and 5800X.

 

[.vodka]

yeah, no. I’ve been setting voltages for over twenty years. I’m not sure where your response is coming from, but saying near 1.5 is normal, or desirable is foolish and shows a temperament not yet mature.

So have many of us here for even longer than that, but Zen+/Zen2/Zen3 don't work that way. Zen1 (the 1700 you mentioned) was the only Zen generation where voltages could be set similar to what we've traditionally been doing and out of the box behavior was also similar to older generations and Intel platforms.

Have a look at what AMD's Robert Hallock has to say about the 5000 series, but it applies to the 3000 series as well. This is official AMD information, so straight from them:

https://twitter.com/x/status/1324561815104356352

This is how AMD designed these chips to operate. According to the guy the chip does up to 1000 measurements and corrections per second, monitoring software is a few orders of magnitude slower so we're not getting the full picture of what's happening under the hood.

You only have to worry about keeping them as cool as possible and using the best motherboard you can throw at your build (best VRM you can afford). Let the boost algorithm do its thing. Apply an offset undervolt when that becomes available in the next AGESA update just like we did on Zen2 and enjoy the lower temperatures and probably even higher boost.

I’m not going to argue with you two but it’s annoying to have someone come in here and tell me I don’t know what I’m doing when I’ve been doing fit successfully for a very very long time. I have extremely high performance numbers and targets. To willfully just state that the hardware monitor prob is the cause is foolish, if not downright ignorant. I have been doing this on all zen models starting at 1700.

I thank you for your attempt at “something” but ok.

He's right, you know.

The observer effect is real on Zen2 and Zen3. Just waking them up for a status check with the wrong monitoring software will show values that are far from what you'd expect on idle. Crappy manufacturer provided software (RGB control BS, etc) will also prevent it from entering sleep states and keep voltages up.

Use hwinfo, or ryzen master to check. Anything else just isn't up to par. A better way to see if these CPUs are truly idling is to watch the power consumption on each core directly from the CPU's internal telemetry (the SVI2 TFN values in hwinfo64). If truly idle it should be measuring ~0.00x W per core save the one or two that are awake and running the OS, you'll also see C6 residency per core in hwinfo at high percentages. Don't rely only on the voltage measurements on these chips.

If you want to see Windows monitoring software show low voltages, then go to the current power plan settings and check if for some reason the default minimum CPU state is 99% instead of 0%. It doesn't have an effect on idle states, idle power will be the same, but monitoring software will now report the low voltage idle states.

------------------------------------------------------------------------------

Having said that, unlike Zen2, doing a manual OC on Zen3 at a fixed voltage is viable and as you see you can get excellent results relative to the boost algorithm. Have fun with that, but don't dismiss what we're telling you about these chips and how they behave out of the box, even when taking into account possible BIOS bugs that have yet to be ironed out (which is the usual for a new hardware release)

Our >20 year experience doesn't quite apply to these things. They behave more like a modern GPU.

 

[Graphenewhen]

You can overclock and undervolt with Ryzen Master although the amount of headroom left appears to be pretty low. Undervolting my 3700x has resulted in significantly lower temps though (8-10°C under load) whilst ensuring a higher sustained all core boost clock of 4.3Ghz achieving 5129 in CBR20. Not to mention about 20w lower power consumption. This is all assuming Ryzen Master is accurate.

Does Ryzen Master work with Zen 3?

 

[Zstream]

You can overclock and undervolt with Ryzen Master although the amount of headroom left appears to be pretty low. Undervolting my 3700x has resulted in significantly lower temps though (8-10°C under load) whilst ensuring a higher sustained all core boost clock of 4.3Ghz achieving 5129 in CBR20. Not to mention about 20w lower power consumption. This is all assuming Ryzen Master is accurate.

Does Ryzen Master work with Zen 3?

Not sure, it does detect bios vcore, cpu temps and load on cores. I don't use it to O/C at all, so maybe?

 

[Zstream]

So have many of us here for even longer than that, but Zen+/Zen2/Zen3 don't work that way. Zen1 (the 1700 you mentioned) was the only Zen generation where voltages could be set similar to what we've traditionally been doing and out of the box behavior was also similar to older generations and Intel platforms.

Have a look at what AMD's Robert Hallock has to say about the 5000 series, but it applies to the 3000 series as well. This is official AMD information, so straight from them:

https://twitter.com/x/status/1324561815104356352

This is how AMD designed these chips to operate. According to the guy the chip does up to 1000 measurements and corrections per second, monitoring software is a few orders of magnitude slower so we're not getting the full picture of what's happening under the hood.

You only have to worry about keeping them as cool as possible and using the best motherboard you can throw at your build (best VRM you can afford). Let the boost algorithm do its thing. Apply an offset undervolt when that becomes available in the next AGESA update just like we did on Zen2 and enjoy the lower temperatures and probably even higher boost.



He's right, you know.

The observer effect is real on Zen2 and Zen3. Just waking them up for a status check with the wrong monitoring software will show values that are far from what you'd expect on idle. Crappy manufacturer provided software (RGB control BS, etc) will also prevent it from entering sleep states and keep voltages up.

Use hwinfo, or ryzen master to check. Anything else just isn't up to par. A better way to see if these CPUs are truly idling is to watch the power consumption on each core directly from the CPU's internal telemetry (the SVI2 TFN values in hwinfo64). If truly idle it should be measuring ~0.00x W per core save the one or two that are awake and running the OS, you'll also see C6 residency per core in hwinfo at high percentages. Don't rely only on the voltage measurements on these chips.

If you want to see Windows monitoring software show low voltages, then go to the current power plan settings and check if for some reason the default minimum CPU state is 99% instead of 0%. It doesn't have an effect on idle states, idle power will be the same, but monitoring software will now report the low voltage idle states.

------------------------------------------------------------------------------

Having said that, unlike Zen2, doing a manual OC on Zen3 at a fixed voltage is viable and as you see you can get excellent results relative to the boost algorithm. Have fun with that, but don't dismiss what we're telling you about these chips and how they behave out of the box, even when taking into account possible BIOS bugs that have yet to be ironed out (which is the usual for a new hardware release)

Our >20 year experience doesn't quite apply to these things. They behave more like a modern GPU.

Thank you for posting the voltages. I'm shocked that this is enabled by default, but I'll take their word for it. With that said, based off real world workloads, the voltages sent IMO is a form of abuse. The amount of temp changes that occur, and so far lack luster performance due to the added voltages seem silly.

I'm aware of some software causing issues with older ZEN platforms, but I never use anything that isn't outside the norm. The norm = Ryzen Master and bios modifications. This isn't the case, it's just as you have stated above, the voltages are just high for some reason. Keep in mind, Asus O/C sets the voltages much lower, and I get exact same performance with much less heat output. This could be due numerous factors, which haven't been outlined yet.

 

[Graphenewhen]

Not sure, it does detect bios vcore, cpu temps and load on cores. I don't use it to O/C at all, so maybe?

It will probably be your best bet at the moment, why not give it a go?

 

[moinmoin]

This reminds me of some of the discussions we had after Ryzen 3000 was launched. Weren't you around back then @Zstream?