Brand new 5700x underperforming in mulitcore

[Mazer]

I switched to a 5700x, and it seems I'm not getting the performance I should be, as I see on forums and various tech sites. I updated my chipset drivers, all settings in bios are default my R23 bench is really low only 12810 and not the score I see on here. Furthermore, I checked my temps, and it's low too in the low 60s on a thermaltake 120 assassin which cooled both my 5600x and 5800x (rip) fine underload so no problems there...
Ram is xmp enabled 3600mhz
Motherboard is up to date b550m motar wifi
All core boost is 3.9ghz for multicore shouldn't this be a little higher?

 

[ViperXZ]

Uninstall AMD chipset drivers / install it again. If nothing helps, make new Windows installation, that would usually fix the problem. Be sure CPU is properly cooled with good TIM, but I suspect it’s a software problem.

 

[Mazer]

Anything unusual here? I did see it boost in games I'm confused

 

[yfn_ratchet]

This looks pretty standard for stock performance. Do you have PBO enabled? It's likely best to look at it in Ryzen Master to see what you're being limited by if not thermals/voltages. As for game boost clocks, games are very single-threaded and will reach those boost clocks more reliably since it's per-core boost behavior and not all-core.

 

[Kodehawa]

It's not underperforming, the 5700X stock PPT is hilariously low. Try setting PBO to something like 115/85/90 if you want more performance out of it.

 

[Mazer]

This looks pretty standard for stock performance. Do you have PBO enabled? It's likely best to look at it in Ryzen Master to see what you're being limited by if not thermals/voltages. As for game boost clocks, games are very single-threaded and will reach those boost clocks more reliably since it's per-core boost behavior and not all-core.

It's on auto I'll need to check it again in game to see what is limiting me. Edit: Yep PPT is maxing out during game loads.

 

[tabascosauz]

It's on auto I'll need to check it again in game to see what is limiting me. Edit: Yep PPT is maxing out during game loads.

78W is pretty much stock (76W stock). Seems pretty normal for out of the box settings.

If you want to go beyond that, then use PBO. You can leave power limits in the board's hands (can work but not usually optimal), or set it to Advanced and work on your power limits yourself.

For Cinebench that really is pretty much all that matters since you won't be limited by Fmax (4650MHz) until you are pushing a serious amount of power.

If you care more about game performance that actually matters, then set boost override to its max +200MHz and see what your CPU can do on 1 or 2 threads, if it can attain 4850.

 

[R0H1T]

Anything unusual here? I did see it boost in games I'm confused

This is me with nearly stock settings for 5800x except a slight undervolt, I have 64GB (4x16) RAM though so it limits some options for me as it's not too stable at faster mem/IF speeds.

 

[Zach_01]

I switched to a 5700x, and it seems I'm not getting the performance I should be, as I see on forums and various tech sites. I updated my chipset drivers, all settings in bios are default my R23 bench is really low only 12810 and not the score I see on here. Furthermore, I checked my temps, and it's low too in the low 60s on a thermaltake 120 assassin which cooled both my 5600x and 5800x (rip) fine underload so no problems there...
Ram is xmp enabled 3600mhz
Motherboard is up to date b550m motar wifi
All core boost is 3.9ghz for multicore shouldn't this be a little higher?

View attachment 359863View attachment 359864View attachment 359865

Please do CB R23 again but make visible all values (current, min, max, average) from the sensors.

First
On main settings (right click tray icon) of HWiNFO enable the "Snapshot CPU Polling"



I would like to see all values from the sensors visible below while the benchmark is running for a few sec/mins.
Best way to do this by starting the rendering and then reset values of sensors window with clock down right.
Take the screenshot right before first pass finishes.

Example of HWiNFO (this is not on load)

 

[Mazer]

Please do CB R23 again but make visible all values (current, min, max, average) from the sensors.

First
On main settings (right click tray icon) of HWiNFO enable the "Snapshot CPU Polling"

View attachment 359907

I would like to see all values from the sensors visible below while the benchmark is running for a few sec/mins.
Best way to do this by starting the rendering and then reset values of sensors window with clock down right.
Take the screenshot right before first pass finishes.

Example of HWiNFO (this is not on load)

View attachment 359908

Had snapshot polling enabled.

 

[Zach_01]

Had snapshot polling enabled.

View attachment 359924


View attachment 359925

Ok I do not see exactly everything, like PPT, PPT limit, TDC limit, EDC limit… On the second column that stuff at bottom about the drive and GPU is unnecessary, the top end is more important.

But…
If CPU package power sensor value is the same as PPT (usually the have small deviation) then the true power of the CPU during that R23 is:

77.8W divided by 1.105 = 70.4W

That “Power Reporting Deviation” (110.5%) is the key

The true power draw of the CPU is 70.4W
And if “CPU PPT” reports slightly lower than “CPU Package Power” for example 76W then this is more accurate calculation:

76 / 1.105 = 68.7W

So you are missing about 10% of CPU power because the board over-states CPU power back to it, as feedback by about 10% (hence the 110.5%). So the CPU thinks it reached its power limit (76W) and regulates it self accordingly.

 

[Mazer]

Oh sorry about that @Zach_01, I will post a better one here

 

[Marcus L]

This should be easily doable, 115ppt, 90tdc, 110edc +200mhz core boost, negative 30 CO (ymmv) boosts to 4.6-4.65 all core in CB R23

Incidentally I have the opposite problem as you and my single core won't boost higher than 4.1Ghz for love nor money but MT will happily hit 4.6-4.7Ghz sustained!

 

[Zach_01]

Oh sorry about that @Zach_01, I will post a better one here
View attachment 359954

Thank you,
I think it doesn't change much of what I said earlier

Its 76 / 110.5 = 68.8W

Your CPU is using 68.8W while it "thinks" that uses 76W
So you are losing some performance

If your max temp is good you can open a little the CPU limits by using PBO advanced.
You have a gigabyte board so lets see if this can help...

Be back in ~5min

 

[Mazer]

You have a gigabyte board so lets see if this can help...

No, it's a MSI b550m motar wifi

 

[Zach_01]

Oh sorry...

Then you have to find the equivalent settings in your BIOS

You can start conservative (I would) by setting

PPT: 90W
TDC: 70A
EDC: 90A

Keep scalar X1 (=default) (read the description in screenshot)

You can add (positive) some clock too on the "Max CPU Boost Clock Override" but dont go to max at once with +200MHz. Go with +50MHz
In Curve Optimizer you can set all cores to negative steps and start with 10 (-10). Thats 30-50mV undervolt that in reality saves some power so the CPU can boost more under the same voltage until it hits again PPT limit.



Whatever you change run CB R23 again and take a screenshot of HWiNFO as we said previously while the render still runs.

Remember that CurveOptimizer benefits mostly multicore loads and max boost override mostly single core loads

There is a high chance that you will not be able to max both of these settings
CO = -30 steps (90-150mV undervolt)
boost override = +200MHz

Unless your CPU is qood silicon quality

 

[freeagent]

Give it the beans.. 200/140/180 +150 -25

That is what I run my 5600X at lol.. doesn't use it all.. but she boosts.

 

[Kodehawa]

Y'all got lucky with CO, my CPU gets unstable even at -5 lmao

Give it the beans.. 200/140/180 +150 -25

That is what I run my 5600X at lol.. doesn't use it all.. but she boosts.

In my experience giving it unlimited power ends up with worse performance as voltage goes way up

 

[iNinja9K]

Using a more aggressive curve optimizer setting, like -25 or -30, may help improve performance.

 

[DaemonForce]

This looks pretty standard for stock performance.

For a 5700X? That Passmark CPU bench sucks. My 3600 was kinda doggin it on v10 but I kind of go crazy with voltage.

That shouldn't be scaling so clean to a newer generation 8-core and newer version of the bench tool. Scammed?

 

[Kodehawa]

For a 5700X? That Passmark CPU bench sucks. My 3600 was kinda doggin it on v10 but I kind of go crazy with voltage.
View attachment 360222

That shouldn't be scaling so clean to a newer generation 8-core and newer version of the bench tool. Scammed?

AMD Ryzen 7 5700X Benchmark

He got pretty much an average.

You can get a lot more out of a 5700X if you're willing to push PPT higher, but at the same time it stops paying off after a bit. Also passmark is a lil' odd.

newer version of the bench tool

scores aren't comparable between versions

 

[Mazer]

Don't mean to bump this thread but I realized I didn't post back with my results.. sometimes life will do that

So, I went with 115 PPT, 80 TDC and left EDC alone at 90 stock. I did -20 all cores (best ones are core 2 and 3 according to ryzen master) with +100mhz with scalar set to 1x. Please tell me if anything is too much voltage on anything since I don't want to kill this one like I did with the 5800x on this board (b550m motar wifi) I really don't have the money to blow on more CPUs down the line right now or else I would've bought a 5700x3d haha..

 

[Zach_01]

As long as the CPU is on auto boost and PBO scalar is on default or X1 then all the silicon protections of the CPU are at play.

Can you show a full HWiNFO sensors window (no hidden sensors) while running CB-R23/24 all core?

 

[Mazer]

Can you show a full HWiNFO sensors window (no hidden sensors) while running CB-R23/24 all core?

 

[Zach_01]

View attachment 363267

Unfortunately that screenshot was not taken when CB-R23 all core was still running.

To be honest, even so... I'm not seeing any issues.
CPU temp peaks at around 75C which is safely away from the 90C peak limit.
Actually is what I'm getting on my 5900X temp wise.