CPU 5800x3D

scatterforce

Master of Cramming
Original poster
May 21, 2018
408
325
Now that reviews are out, I have been trying to determine how hard this chip is to cool. While overclocking the chip is off the table, is undervolting an option? Is there an Eco Mode that runs at 65W? Can it be cooled on a sub 50mm cooler?

For the 5800x, the answer was no. Unless you heavily undervolted it or apply Eco Mode, the 5800x runs HOT. None of the major outlets that I've seen have addressed this. I've seen one benchmark pull 112W under load, but that only tells part of the story.
 

scatterforce

Master of Cramming
Original poster
May 21, 2018
408
325
OK, so on a 360mm radiator, the 5800x3D hit max 83* C after steady state. The 5800x hit 87* (source, Hardware unboxed)

So 4 degrees cooler, but these temps are still high for a 360mm rad. =(
 

REVOCCASES

Shrink Ray Wielder
REVOCCASES
Silver Supporter
Apr 2, 2020
2,057
3,331
www.revoccases.com
as far as I know the voltage is fixed and AMD only allows to play around with memclk, uclk, fclk

cooling seems to remain challenging but slightly better compared to the 5800X - this is a benchmark I found using a 280mm AIO, comparing 5800X and 5800X3D in Cinebench

Bild_2022-04-15_122051000.png
 
  • Like
Reactions: scatterforce

gffermari

SFF Lingo Aficionado
Jan 7, 2017
91
91
The 5800X3D would have been a great cpu for a SFF case, having lower clocks than the 5800X, fixed vcore 1.35V, no PBO. I expected to be way cooler but it's hot as hell.
 

scatterforce

Master of Cramming
Original poster
May 21, 2018
408
325
Hardware Canucks provided numbers with cooling on a Noctua NH-U12S. It reached 83-84 degrees under load on different motherboards. If this chip is getting this hot under the best cooling conditions, there is no way it will do well in a SFF case on air.

SO, I guess the best upgrade path is going to be the 5700X..... or a bigger case. I do really like the looks of the Dan C4-SFX.
 
  • Like
Reactions: Arboreal

pr0m3t

Chassis Packer
Jan 17, 2022
19
7
I imagine the 5800x3d would do fine as long as you have a 240rad to cool it. For example I think it'd be a great match for the A4 h2o
 
  • Like
Reactions: rfarmer

teamet

Chassis Packer
Mar 17, 2018
19
1
I dont get it .. is it not possible to _reduce_ power usage through the normal ways? I mean .. perhaps not an eco mode but custon PPT/TDC/EDC limits .. why would they ever remove that?

Does somebody have a connection with some of the testers, it should be any easy thing to confirm if you have the chip!
 

ikanos

What's an ITX?
New User
Oct 22, 2019
1
0
Do you guys think 5800x3D in a Ncase with C14s work? Any in sights would be appreciated.
 

scatterforce

Master of Cramming
Original poster
May 21, 2018
408
325
Do you guys think 5800x3D in a Ncase with C14s work? Any in sights would be appreciated.
Yes. A C14s with intake from the side will be plenty. Under full synthetic load, it might reach 90, but it will never get that hot while gaming.
 
  • Like
Reactions: ikanos

Skripka

Cat-Dog Perch Manager
May 18, 2020
443
543
It is an interesting CPU I'll give it that...other than 'because we can' I'm not sure of the point. This fall their new CPU platform will be out that will likely bury it in performance....only caveat being early-adopter BIOS/UEFI issues.

I'm much more interested in a B650motherboard and their new 65W chips, than their X3D chip.
 

scatterforce

Master of Cramming
Original poster
May 21, 2018
408
325

sneedster

Trash Compacter
Apr 22, 2022
40
25
Some redditor reported that the options to access edc/tdc/ppt and eco mode on his x570 strix disappeared after upgrading to x3d :/
 

Runamok81

Runner of Moks
Jul 27, 2015
445
621
troywitthoeft.com
Some redditor reported that the options to access edc/tdc/ppt and eco mode on his x570 strix disappeared after upgrading to x3d :/
That's not great. Also, does anyone know of any AM4 ITX mobos even have an external generator?

Update: Looks like MSI is gearing up to OC the 5800x3d. Still, anyone know if they have mITX boards that will accept this new OC firmware?
 
Last edited:

sneedster

Trash Compacter
Apr 22, 2022
40
25

Its very confusing how people report findings that completely conflict that wccftech article. According to people in the linked thread, all power management options are completely locked out in both bios and ryzen master.
 
Last edited:

sneedster

Trash Compacter
Apr 22, 2022
40
25
Again a completely different finding:


I wonder what sense to make of this. Undervolting the X3D Must be very particular to which motherboard you're on.

(sorry for double posting, can't edit my replies yet. )
 
Last edited:

Skripka

Cat-Dog Perch Manager
May 18, 2020
443
543
Again a completely different finding:
I wonder what sense to make of this. Undervolting the X3D Must be very particular to which motherboard you're on.

(sorry for double posting, can't edit my replies yet. )
This reminds me of the problem with Intel CPU/motherboard reviews.

Namely 'official guidance' of how the motherboard UEFI should be set up....and then mainboard makers not only turn on full-time MCE by default but they can even screw with BCLK to win benchmarks: the former is bad-practice while the later is utterly insane from a stability standpoint. Then people wonder why Intel chips run so hot, or can be unstable.

An analog on AMD some might remember from the Good Old Days....The AMD Phenom 2 CPUs that had 3 cores. Actually they had 4 cores, but one was disabled at the factory for reasons undisclosed. Maybe binning, maybe stability who knows. Well Biostar boards were famous for letting you flip a BIOS switch and getting that off-CPU-core back online, and it mostly just worked fine. AMD didn't officially care--but it was unsupported and luck of the draw.
 

Runamok81

Runner of Moks
Jul 27, 2015
445
621
troywitthoeft.com
Again a completely different finding:
I wonder what sense to make of this. Undervolting the X3D Must be very particular to which motherboard you're on.

(sorry for double posting, can't edit my replies yet. )

The mobo in that redddit post was a Asus b550 strix. OP stated that here.

UPDATE!
Overclocking looks to be on the table. News story here.
For undervolting there is some good/crunchy info on overclock.net forum thread here.

In that thread users Bloot and Blameless are undervolting. They are able to lower temps 15C. Bloot's undervolt is not achieved via BIOS, but by using PJVol's PBO2 Tuner in Windows here. Temps dropped from 83C to 68C with a negative 100mv vcore offset. Evidence here. So it looks like anyone can undervolt, but it just requires a tool in Windows that needs to be set each boot. With regards to making those vcore offsets sticky (automatic), it is up to the motherboard manufacturers to expose those controls. Which SFF boards have that? From that thread and this reddit post, it looks like Asus boards seems to have exposed the vcore negative offset in BIOS! MSI had those same vcore adjustment controls, but then mysteriously removed them when 5800X3D support was added. No information on Gigabyte boards.

Summary: For SFF boards, it looks like the ASUS ROG Strix B550-I Gaming AMD AM4 Mini-ITX SFF can undervolt and lower temps!
 
Last edited:

Draco25240

Efficiency Noob
Apr 26, 2022
7
14
Just to provide an update for this... Ended up grabbing a 5800X3D myself and have done a bit of testing.
Motherboard is Asus Strix B550-I running BIOS version 2603 (Agesa 1.2.0.6b), and the CPU is cooled by a 47mm Black Ridge with a Noctua NF-A9x14 fan swap in a Sliger SM550.
  • Setting Package Power Limit in BIOS, as described by the reddit comment I linked above, did not work for me personally. CPU did not acknowledge the limits, and hwinfo did not even see them.
  • Voltage offsets were available to me in BIOS, but I haven't touched them (yet?). May only be an Asus thing though, since every time I've seen it mentioned by others, it's been on Asus boards.
  • I have however been able to successfully set PPT, TDC and EDC under AMD CBS > NBIO Common Options > XFR Enhancement > Precision Boost Overdrive by setting PBO to manual. Upon stress testing in Cinebench, the 5800X3D strictly adhered to the limits I set, no matter what.
    I can't confirm it works on all boards, but I've seen comments saying it works on various Asus and Gigabyte boards at least, and maybe ASRock. Yet to see anyone confirm whether it's possible or not on MSI boards though.
  • I haven't tried Curve Optimizer yet, but I'm yet to see anyone say the PBO2 Tuner tool from the overclock.net forum didn't work for them, so that's promising.

I did a few quick multicore benchmarks in Cinebench R23 (which doesn't seem to benefit from extra cache?) yesterday to test how PBO limits impacted the performance and temperatures:
- Dialing it in to match a 5600X (88 PPT, 60 TDC, 90 EDC, so basically ECO mode) got me a max temperature of 74.6 C while hitting roughly 91% of the score of stock values reported online (13,603 vs 15,000-ish)
- Highest score I got without throttling was 14,388 (96% of stock) with 120 PPT, 75 TDC and 110 EDC, and capping at 88.5 C.
- Found my sweet spot at 120 PPT, 70 TDC and 105 EDC, with a max temp of 82-83 C and a score of 14,300 - 14,350 (96% of stock).
The actual, realistic performance hit with extra cache taken into account is probably less though, and in the single core benchmark I was within margin of error of stock scores no matter my values, while only hitting 64.0 C at max.

I'll see if I can get the chance to a bit mess around with Curve Optimizer via the PBO2 Tuner tool over the weekend, and what kind of improvements I might be able to get with it.
(Note, this is my first Ryzen chip, my first time doing any tweaking to a CPU and my first time benchmarking, so bear with me)



TL;DR: Setting PBO values on the 5800X3D works on my motherboard and improves thermals enough to work comfortably under my 47mm cooler without too big of a hit to performance. Also possible to manually replicate ECO mode. However your mileage may vary depending on your motherboard.