Upgrade to 5900X

[tabascosauz]

@DebenPoison

1. SVI2 Vcore only measures properly at load. When you see 1.4V+ it's either at idle (where it's incapable of measuring actual Vcore behaviour), or high single core load. If you ran 1.4V during an all-core load you would have thermal shutdown a long time ago, Zen 3 is designed for a 1.2-1.25V all core Vcore at stock.

2. Auto PBO boost override is +0. If you want more you need to specify it.

3. All the boost override does is change your global limit (refer to your HWInfo sensor list). Stock is 4950MHz, goes up to 5150. If you're lucky you'll only ever bump into the global limit on 1 or 2 core load, and you might get some more ST performance out of a 5150 limit, but all-core will be unaffected by what you set because it will always clock much lower.

4. What exactly have you changed in BIOS? I've never seen CPU-Z drop only CCD2 clocks like that even in the Stress option. 4.55 is relatively high for CPU-Z MT but the score is low for that clock. Run some other benchmarks, I'd run CPU-Z for thermal testing and power deviation, not for the score.


5. I don't see VSOC being low or a problem, nor how it is relevant to cores performance.

 

[phill]

I'd say, if you'd like to upgrade then upgrade. As I think if your gaming the 5900X is a decent CPU from the 3700X. Of course as its been mentioned, if you don't really need the cores and just game, then a 5800X3D might be a better CPU for you overall but value for money, I think the 5900X would get the nodd from me I'll be moving from a Ryzen 2700 and a 2700X to one at some point for my girls rigs

 

[Zach_01]

Talk to me friends ~ what should I be enabling, disabling or manually adjusting in my BIOS to achieve better PBO results on this 5900X?

Currently all I have is everything default other than DOCP @ 3200mhz for memory ; and 3 x PBO enabled settings in BIOS. Enabled seems to grab an extra ~100mhz vs Auto. I did see an increase max cpu mhz PBO related setting with a limit of up to 200mhz but that is set to auto at present.

Running the Asus TUF X570 Wifi motherboard with a fairly current BIOS (2021), but there is newer available. Also running a Cooler Master ML240L V2 AIO which I'll be replacing soon with a 360mm Artic Liquid Freezer II A-RGB in the next couple weeks. Current thermal paste is MX-2.

I have set Thermal limit in bios to auto perhaps this should be set to 80'C or more? Also I note the voltage looks lower than it could be 1.288V CPU Core VID and SoC Voltage seems low compared to the above screenshots. Are those being set manually or do we typically leave them auto ?

View attachment 262196

I have a thermal limit at 75C and Curve Optimizer negative with magnitude from 8 (best cores) to 18 (worst cores)

"and 3 x PBO enabled settings in BIOS"
What does this mean?

Probably he means scalar...

@DebenPoison
4. What exactly have you changed in BIOS? I've never seen CPU-Z drop only CCD2 clocks like that even in the Stress option. 4.55 is relatively high for CPU-Z MT but the score is low for that clock. Run some other benchmarks, I'd run CPU-Z for thermal testing and power deviation, not for the score.

Never test it my self on CPU-Z but without CO second CCD is lower clocking on other MT tests when all cores in there are 7-12 on perf order.

 

[lexluthermiester]

Small FFT doesn't max out PPT anymore on stock Ryzen. Boost algorithm for Vermeer and newer CPUs basically recognizes Prime95 as a power virus and automatically throttles down to exactly base clock (3.7GHz), only draws about 130W.

That said, power deviation metric still appears to be accurate in Small FFT.

I did not know that. WTH AMD? That's some dumbass BS.

 

[tabascosauz]

Never test it my self on CPU-Z but without CO second CCD is lower clocking on other MT tests when all cores in there are 7-12 on perf order.

You observed CCD2 dropping 350MHz effective below CCD1 while having the same core clocks?

I did not know that. WTH AMD? That's some dumbass BS.

If boost algo treated Small FFT like regular AVX the CPUs would go up in flames, they just don't do all-core the same way as Intel. 3.7GHz still does almost 100W core power - there's no way not throttling would still stay within 142W PPT. The generally lower stock Vcore and more aggressive boost algo account for all of Zen 3's lower temperatures compared to Zen 2 - crank up the Vcore to Matisse levels (1.28-1.35V) and they will behave the same.

 

[lexluthermiester]

If boost algo treated Small FFT like regular AVX the CPUs would go up in flames, they just don't do all-core the same way as Intel. 3.7GHz still does almost 100W core power - there's no way not throttling would still stay within 142W PPT. The generally lower stock Vcore and more aggressive boost algo account for all of Zen 3's lower temperatures compared to Zen 2 - crank up the Vcore to Matisse levels (1.28-1.35V) and they will behave the same.

I hate to say this, but it's 6 years and I'm STILL trying to get a grip on how AMD does power delivery on Ryzen CPU's.. It's a complicated, convoluted mess. I get that it's a necessity, but still..

 

[Zach_01]

You observed CCD2 dropping 350MHz effective below CCD1 while having the same core clocks?

I will do tests w/ and w/o CO later when I get back home.
CB R23, CPU-Z…
Any other workload?

I hate to say this, but it's 6 years and I'm STILL trying to get a grip on how AMD does power delivery on Ryzen CPU's.. It's a complicated, convoluted mess. I get that it's a necessity, but still..

What I also find odd is temp of second CCD. Even on same discrete and effective clock the 2 CCDs have way different temp. I will also try to demonstrate this by showing the C-States of each core-CCD during the MT tests

 

[freeagent]

I open the taps for all the power the socket will give me, and adjust my curve to that. Runs a bit warm at full load, but it’s manageable, and within safe numbers so all good.

 

[tabascosauz]

I will do tests w/ and w/o CO later when I get back home.
CB R23, CPU-Z…
Any other workload?


What I also find odd is temp of second CCD. Even on same discrete and effective clock the 2 CCDs have way different temp. I will also try to demonstrate this by showing the C-States of each core-CCD during the MT tests

No, as in, effective clock 350MHz below core clock for 1 CCD is not normal in any MT benchmark lol, hence why I asked exactly what settings OP currently messed with

CPU-Z is not a good test for score comparison, but I know how Zen3 behaves during Stress test better than any other benchmark. It occasionally drops score suddenly around the 10-15min mark then spends a long time slowly building back up, but clocks/temps/power are always consistent.

Every CPU has different core quality and IHS contact. I don't even look at the CCD1 and CCD2 average anymore, useless after HWInfo added per-core temps. Mine looks wildly different to yours, CCD2 about 3 degrees warmer at iso-clock/Vcore/power, 13-15C deltas between some CCD2 cores - it doesn't mean anything if the cores are performing up to par and drawing reasonable power. Just production variation. Why would C-states ever matter if all cores should be at full tilt during a heavy, consistent all-core test?

 

[Zach_01]

No, as in, effective clock 350MHz below core clock for 1 CCD is not normal in any MT benchmark lol, hence why I asked exactly what settings OP currently messed with

CPU-Z is not a good test for score comparison, but I know how Zen3 behaves during Stress test better than any other benchmark. It occasionally drops score suddenly around the 10-15min mark then spends a long time slowly building back up, but clocks/temps/power are always consistent.

Every CPU has different core quality and IHS contact. I don't even look at the CCD1 and CCD2 average anymore, useless after HWInfo added per-core temps. Mine looks wildly different to yours, CCD2 about 3 degrees warmer at iso-clock/Vcore/power, 13-15C deltas between some CCD2 cores - it doesn't mean anything if the cores are performing up to par and drawing reasonable power. Just production variation. Why would C-states ever matter if all cores should be at full tilt during a heavy, consistent all-core test?

You are right of course about C-States...
But I noticed now that my second CCD (full MT load) has slightly higher effective clock but has lower power consumption and of course temp.
The eff. clock diff is not so distinct on CB_R23 as on other tests. The individual core power(s) is following exactly the "perf #" order. And this is true on both cases, w/ CO and w/o CO.

Settings:




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

CPU-Z stress test w/ CO


CPU-Z stress test w/o CO


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

FurMark CPU w/ CO


FurMark CPU w/o CO


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

P95 (small FFT, 128KB~128KB) w/ CO


P95 (small FFT, 128KB~128KB) w/o CO


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

Blender (barbershop) w/ CO


Blender (barbershop) w/o CO


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

CB R23 (MT) w/ CO


CB R23 (MT) w/o CO

 

[DebenPoison]

@DebenPoison

1. SVI2 Vcore only measures properly at load. When you see 1.4V+ it's either at idle (where it's incapable of measuring actual Vcore behaviour), or high single core load. If you ran 1.4V during an all-core load you would have thermal shutdown a long time ago, Zen 3 is designed for a 1.2-1.25V all core Vcore at stock.

2. Auto PBO boost override is +0. If you want more you need to specify it.

3. All the boost override does is change your global limit (refer to your HWInfo sensor list). Stock is 4950MHz, goes up to 5150. If you're lucky you'll only ever bump into the global limit on 1 or 2 core load, and you might get some more ST performance out of a 5150 limit, but all-core will be unaffected by what you set because it will always clock much lower.

4. What exactly have you changed in BIOS? I've never seen CPU-Z drop only CCD2 clocks like that even in the Stress option. 4.55 is relatively high for CPU-Z MT but the score is low for that clock. Run some other benchmarks, I'd run CPU-Z for thermal testing and power deviation, not for the score.


5. I don't see VSOC being low or a problem, nor how it is relevant to cores performance.

PBO Boost override - What is an optimal value to set or is auto preferred here? I occassionally see two cores hit 5Ghz, wonder if setting 200mhz is worthwhile then?

PS Was only running CPU-Z to measure the clocks under load, not score.

Probably he means scalar...


Never test it my self on CPU-Z but without CO second CCD is lower clocking on other MT tests when all cores in there are 7-12 on perf order.

The Asus X570 TUF is confusing at best when enabling PBO. It has PBO settings in multiple places and I didn't have the names on hand for each setting.

The first two screenshots seem to indicate the same PBO setting but weirdly they can have different settings configured, so they don't appear to be in sync.

The first screenshot offers a few additional settings too that the second does not, but it does allow you to select advanced which provides additional PBO settings including Scalar and others. The third screenshot references the advanced setting.

First SS -> AI Tweaker
Second SS -> Advanced Menu
Third SS -> Advanced Menu -> Advanced PBO setting unlocking additional PBO settings

Setting the 3 I have enabled to AUTO, provide a slightly worse clock on all cores under full load (about 100mhz), therefore I've got them set to Enabled.

"and 3 x PBO enabled settings in BIOS"
What does this mean?


The Tuf x570 is one of their weaker boards, performance is fine but they cut corners. Watch your chipset temps because if it's like my x570-F, the thermal pad they used will have dried up and gone to shit long ago


CPU-Z is not a good benchmark or stress test on AMD, cinebench is heavier but a more reliable test - if you're getting 4.55GHz all core, you're doing better than most of us.



HWinfo results dont tell us much if it's been running for under 2 minutes btw, you need to start it before the benchmarks for the idle results, then run the test programs so we have something to compare to and see how high temps get over time

See above post for reference of the PBO settings thanks!

I haven't seen an issue on the chipset temp ; what are chipsets designed to reach now days? 60'C max?

HWInfo was reset once the CPU was under full load ; then captured still under full load. I merely wanted to present what the clocks looked like under load. The idle results are pretty good, 34'C and below across most cores. I can reference this if there would be further value in seeing it

In game I'm not seeing above 64'C or 65'C but stress test with prime95 is murder, see below with idle temps. P95 test is Small FFTs for 5 minutes

 

[Zach_01]

PBO Boost override - What is an optimal value to set or is auto preferred here? I occassionally see two cores hit 5Ghz, wonder if setting 200mhz is worthwhile then?

PS Was only running CPU-Z to measure the clocks under load, not score.



The Asus X570 TUF is confusing at best when enabling PBO. It has PBO settings in multiple places and I didn't have the names on hand for each setting.

The first two screenshots seem to indicate the same PBO setting but weirdly they can have different settings configured, so they don't appear to be in sync.

The first screenshot offers a few additional settings too that the second does not, but it does allow you to select advanced which provides additional PBO settings including Scalar and others. The third screenshot references the advanced setting.

First SS -> AI Tweaker
Second SS -> Advanced Menu
Third SS -> Advanced Menu -> Advanced PBO setting unlocking additional PBO settings

Setting the 3 I have enabled to AUTO, provide a slightly worse clock on all cores under full load (about 100mhz), therefore I've got them set to Enabled.


View attachment 262295

View attachment 262296
View attachment 262297

Typically PBO on Auto is disabled.
Meaning PBO limits are on default settings.
For the 5900X they are:
PPT: 142W
TDC: 90A
EDC: 140A

When you turn PBO to Enabled, those limits are set by the board.
Typically boards are setting the limits to skyrocket values that completely throw the CPU out of optimal operation.

Observe your last P95 SS that you're hitting PPT = 200Watt. Hence the near limit (90C) temperature.

I'm not familiar with ASUS boards and BIOS, but I do know that that "PBO Fmax Enhancer" complicates things even more than already are.
I would keep PBO Fmax Enhancer Disabled, PBO on Advanced and set everything on manual to your preference (within reason).

First you have to know what you want more...
Best gaming performance or best 100% (MT) load performance. You cant have both. You can have a bit of both but not best of both at the same time.

By default the max allowed clock of the 5900X is 4950MHz. With boost override you can have additional 200MHz (with steps of +25MHz).
Curve Optimizer on the other hand is a way to alter the voltage/frequency (v/f) curve. With negative steps (1~30) you can set the magnitude of the curve shift (per core or entire CPU) to tell it to run higher boost (f) on the same voltage (v). This affects the entire clock range of the CPU.
With reasonable PPT/EDC limits you can have slightly higher clocks (within the same power envelope W) with each negative step.
Curve Optimizer affects mostly the 100% CPU load and less the single/middle threaded loads like gaming.

The thing is that you cant set +200MHz on boost override and go all the way (20~30 steps) on curve optimizer too, because the clock stretching will be too high and some cores will start loosing stability.
Both settings (boost override / curve optimizer) are stretching boost clock in a different way.

Like I said before you have to decide what you want/need more for your daily tasks (and not for benchmark scores, like too many users do).

 

[tabascosauz]

You are right of course about C-States...
But I noticed now that my second CCD (full MT load) has slightly higher effective clock but has lower power consumption and of course temp.
The eff. clock diff is not so distinct on CB_R23 as on other tests. The individual core power(s) is following exactly the "perf #" order. And this is true on both cases, w/ CO and w/o CO.

Wow! Seems like you really lucked out on that second CCD. Core deltas looking good. Higher clocks on that CCD maybe because it's 8C cooler, idk if that's a feature because I don't usually see CCD1-CCD2 gap so big. Small clock differences though.

PBO Boost override - What is an optimal value to set or is auto preferred here? I occassionally see two cores hit 5Ghz, wonder if setting 200mhz is worthwhile then?

PS Was only running CPU-Z to measure the clocks under load, not score.

The Asus X570 TUF is confusing at best when enabling PBO. It has PBO settings in multiple places and I didn't have the names on hand for each setting.

The first two screenshots seem to indicate the same PBO setting but weirdly they can have different settings configured, so they don't appear to be in sync.

The first screenshot offers a few additional settings too that the second does not, but it does allow you to select advanced which provides additional PBO settings including Scalar and others. The third screenshot references the advanced setting.

First SS -> AI Tweaker
Second SS -> Advanced Menu
Third SS -> Advanced Menu -> Advanced PBO setting unlocking additional PBO settings

In game I'm not seeing above 64'C or 65'C but stress test with prime95 is murder, see below with idle temps. P95 test is Small FFTs for 5 minutes

Enable snapshot polling so Core Clock metrics actually make sense.

You can set +200 if you want, it shouldn't hurt anything or change overall boost behaviour. If you have cores capable of scaling past 4950, they will make themselves known after you set +200. More aggressive CO offsets on those cores can take them further, to a point.

In Asus BIOS, you should be doing all your PBO settings in AI Tweaker. Except Curve Optimizer - last I checked AGESA 1207 still hasn't changed it and you need to go to Advanced>AMD OC to do your Curve Optimizer settings. There is a CO menu in Tweaker but it ranges from wonky to straight up doesn't work. Rest of the settings, do them in Tweaker not AMD OC.

The aggressive boost throttling only works at stock power limits. When you enable PBO (even if you don't touch anything and leave the settings at auto) most vendors treat it as a free "go fast" button. Boost algo is not nearly as conservative then.

Still not sure why you have a 400MHz gap between CCD1 and CCD2. Core clock looks fine but the bench scores so far look more like effective clock is telling the truth.

The 4201 BIOS works okay on my Impact, but even if they mostly copy+paste the same update between boards, sometimes vendors fuck up implementation for one particular board. Might be worth trying some other recent BIOSes to see if that behaviour continues.

 

[DebenPoison]

Some improvements have been made... still playing with various settings and once I get optimal results from CPU, will begin with memory (guess this thread should be moved to the OC channel )

 

[DebenPoison]

Any issues with these temps? They're higher than expected with an Artic Freezer ii 360mm AIO. The mounting really sucks on this AIO with the Asus X570 TUF mobo. First mount it wouldn't the PC wouldn't boot, probably shorting out even though there was just the slightest gap between the boards caps and the mounting bracket. I re-mounted with less pressure but I think I'm seeing this in higher than I'd like temps. Am I ok to run like this daily? In game I can see up to 68-72'C at peek