upgrading from 2700X to 5900X on a MSI B450M Mortar

[Kodehawa]

Thanks. At the moment I'm still testing running stock. Based on the 5700X (and 7700X) reviews, I get decent Single Core results on Cinebench - 1524, but I can't reach 13000 on the 10min multicore test which seems odd because temps are max ~60º.

View attachment 286537
On the ol' 2700X I managed to get 9902 Multicore and 1079 Singlecore.

I get about 12600 without PBO with temps in the mid 50C. Depends on motherboard I guess.

Enabling PBO pushes it closer to 15k.

 

[tvshacker]

I get about 12600 without PBO with temps in the mid 50C. Depends on motherboard I guess.

That's about the maximum score I managed so far also. I expected to be closer to @W1zzard numbers since I managed to get within ~95% of his test scores on the others runs.
Also, the 2700X used more power, so I think I can scratch temps and power delivery as the "issue".

 

[Kodehawa]

That's about the maximum score I managed so far also. I expected to be closer to @W1zzard numbers since I managed to get within ~95% of his test scores on the others runs.
Also, the 2700X used more power, so I think I can scratch temps and power delivery as the "issue".

On two different motherboards I haven't really been able to touch 13k stock. The stock power limit is 77W (!) which is hilariously low and my CPU on all core workloads under non-PBO conditions runs at 1.0V. PBO tweaking can "fix" this, but I'd also argue the default configuration is the most efficient. Sure, I can push 4.45GHz at 130W, but at almost double the wattage of stock (3.9 all core for me)

Take your poison there, basically. If you tweak PBO (personally I have 120/80/125) for maximum performance or leave stock for maximum efficiency.

 

[Mussels]

Welcome to the 5700X club. If you have enough cooling enabling PBO leaves it close to a 5800X. Efficiency does go out of the window, though (+60W for 300 or 400MHz more). IMO the most efficient one is 90W PPT, you get about 4.1GHz, at 120 or 130W you get 4.4, but that's a lot more power.

the 90-110W is the happy range for all the 8 core zen3 CPU's

I haven't played with the higher core count ones enough to know for sure what they can run upto while remaining cool (mostly since the people i know who have them, put them on terrible boards where the VRM's couldnt go higher wattage anwyay)

Thanks. At the moment I'm still testing running stock. Based on the 5700X (and 7700X) reviews, I get decent Single Core results on Cinebench - 1524, but I can't reach 13000 on the 10min multicore test which seems odd because temps are max ~60º.

View attachment 286537
On the ol' 2700X I managed to get 9902 Multicore and 1079 Singlecore.

You're probably limited by the VRMs, set R23 to a 30 minute run and post a screenshot of HWinfo's sensors page in fullscreen, showing all the readings for the CPU

We need to see CPU Clocks, effective clocks, PPT, EDC, TDC values and %, and the motherboards temperature readings - odds are the boards throttling before the the CPU is, and a lower PPT setting will keep it more stable. People pushing boards to the limit find they get stuttering in games, since the GPUs heat makes the VRM's hotter and throttle earlier despite the CPU being cold itself

Enabling PBO without manually entering limits, unlocks it to effectively unlimited settings and the CPU will try for 142W - and budget boards just cant sustain that

 

[Kodehawa]

You're probably limited by the VRMs, set R23 to a 30 minute run and post a screenshot of HWinfo's sensors page in fullscreen, showing all the readings for the CPU

We need to see CPU Clocks, effective clocks, PPT, EDC, TDC values and %, and the motherboards temperature readings - odds are the boards throttling before the the CPU is, and a lower PPT setting will keep it more stable. People pushing boards to the limit find they get stuttering in games, since the GPUs heat makes the VRM's hotter and throttle earlier despite the CPU being cold itself

Stock it just reaches PPT of 77W, making clocks dip to 3.8GHz on my side. I *can* PBO it away to make it reach 4.35GHz or so, but I imagine that's about the normal range.

It does reach 13k, though (had to make cb23 priority high though, at "below normal" -what it defaults to- it does 12.8k)! Definitely not 13660, though.
As a random fact, this uses LESS power than my Ryzen 5 2600. The 2600 would max out at 89W, and it was cooled fine by a Wraith Stealth (at ~86C max). Also the CPU never reaches 60C on CB23, with the fans on the cooler spinning at 810RPM lol. I imagine it's because of the extremely low operating voltage (1.05V)

Few more notes: 90/80/95 PBO reaches 4.15GHz, scoring ~14250, while 130/80/125 PBO reaches ~4.33GHz, scoring ~14700. Up to the person to see if those 40W extra are worth the 450 points

 

[tvshacker]

It does reach 13k, though (had to make cb23 priority high though, at "below normal" -what it defaults to- it does 12.8k)! Definitely not 13660, though.

Tried this, but didn't improve much

We need to see CPU Clocks, effective clocks, PPT, EDC, TDC values and %, and the motherboards temperature readings - odds are the boards throttling before the the CPU is, and a lower PPT setting will keep it more stable. People pushing boards to the limit find they get stuttering in games, since the GPUs heat makes the VRM's hotter and throttle earlier despite the CPU being cold itself

Here are 2 screenshots: one at pass 23, and another after it had finished. I hope that the sensors names in Portuguese won't be an issue (I only noticed halfway through the 30min run):

All the help will be greatly appreciated!

 

[tabascosauz]

Tried this, but didn't improve much

Here are 2 screenshots: one at pass 23, and another after it had finished. I hope that the sensors names in Portuguese won't be an issue (I only noticed halfway through the 30min run):

All the help will be greatly appreciated!

Things look fine. Generally a little on the warm side, but I also don't know your ambient temp. Your case looks a bit restrictive as well so unless you're paying special heed to airflow, probably explains the slightly warm-ish Motherboard temp.

As for VRM......I have no idea where all this VRM obsession as of late came from, but the board is fine. The Mortar is fine, it's a reasonably heatsinked 4-phase with double 4C024N, miles and miles away from the worst AM4 boards. Judging from the readings from its own thermistor in the VRM, 59C for 77W is fine, runs slightly cooler than the 4-phase 50A IR3556 on one of my old boards (B450I Aorus Pro Wifi), which is a pretty good result for discrete parts.

ST scores seem reasonable - 1510 @ 4.65GHz, no clock stretching......my 5700G did 1560 @ about 4.8GHz, but the 16MB Zen 3 core is a little slower. My 5800X3D is 1495 @ 4.53GHz, my 5900X was about 1640 @ 4.93GHz. There's a ton of things that can affect ST score due to how long the test is. If you want to cut all those things out of the equation, just run the bench in safemode.

As for MT, there's not a whole lot you can do about any particular BIOS being weak/conservative for core performance. If you are on an AGESA 1207 or 1208 based BIOS, it's pretty normal to expect lower clocks, blame AMD. You can shop around to see how different BIOSes behave with your hardware. If you want to up MT clock though (it is a little conservative at 4.05GHz @ 77W), you will have to open up PPT and do some curve optimizer. Can just start at 88W, an easy number since 88/60/90 is the standard 65W power profile for CPUs that are not 5600/5600G/5600X/5700X.

 

[Kodehawa]

Tried this, but didn't improve much


Here are 2 screenshots: one at pass 23, and another after it had finished. I hope that the sensors names in Portuguese won't be an issue (I only noticed halfway through the 30min run):
View attachment 288829

View attachment 288824

All the help will be greatly appreciated!

Just hitting PPT limit there. Enable PBO if you want higher scores. 90W PPT should get you pretty much the best efficiency while giving you more performance.

 

[tvshacker]

Just hitting PPT limit there. Enable PBO if you want higher scores. 90W PPT should get you pretty much the best efficiency while giving you more performance.

I'm regretting not having taken similar print screens with the ol' 2700X to see the VRM temps and other sensors with it pulling 105W

Generally a little on the warm side, but I also don't know your ambient temp. Your case looks a bit restrictive as well so unless you're paying special heed to airflow, probably explains the slightly warm-ish Motherboard temp.

ambient temp should be below 20º
It's not very "airflow friendly" as it's supposed to be a silent type case. I currently have the "main" side panel off, but the open side is just 1cm away from a wall.

 

[Mussels]

side panels off can lead to worse temperatures, since the case is designed to guide airflow over things

A rear exhaust fan pulls air over the VRM location for example, but with the side panel off it grabs the air from outside to push it outside again, instead of pulling that hot air towards it

 

[bug]

side panels off can lead to worse temperatures, since the case is designed to guide airflow over things

A rear exhaust fan pulls air over the VRM location for example, but with the side panel off it grabs the air from outside to push it outside again, instead of pulling that hot air towards it

That's assuming hot air goes in circles somehow. It doesn't, it goes upwards.

 

[Mussels]

That's assuming hot air goes in circles somehow. It doesn't, it goes upwards.

It goes up extremely slowly, and fans nearby are far stronger than that effect
No idea how you got that air goes in circles, air moves in the direction of high pressure to low pressure?

An AIO top mounted with a side panel off is going to pull air from outside the case into the AIO and blow it out, following the path of least resistance - leaving motherboard components nearby with open air, but zero air flow over them

 

[tabascosauz]

ambient temp should be below 20º
It's not very "airflow friendly" as it's supposed to be a silent type case. I currently have the "main" side panel off, but the open side is just 1cm away from a wall.

I don't think having it 1cm from the wall counts as extra airflow lol

generally side panel off greatly benefits downdraft coolers but towers don't benefit significantly. Still, if fromt intake performance is that bad, side panel off should still be generally a positive.