Ryzen 5800x 87c in Cinebench...

[Vogon]

5800X is running hotter than other 3 in most reviews. This is normal

It has all of 105W (or 140W) power budget at hand to feed into single small chiplet. 5900X and 5950X spread this power over two chiplets, 5600X has only 65W for same area.
This allows 5800X to have all-core boost some 100MHz higher than other chips (at 16-thread load),

Indeed it seems normal, but less than perfect in my case as I went for it over the 12 core part as I had assumed less cores=less heat.
I don't need 12 cores but would have happily paid the extra had I known the 5900 was cooler !
But if I had waited for reviews I wouldn't have anything, so there is that

 

[alec]

It is less heat, as in less power consumption - fewer watts dumped into case.
But silicon itself is running hotter.

There are rumors of more advanced undervolting being released in bios updates. Maybe it comes without performance reduction attached.

 

[Vogon]

Those rumours, if true, could make it better indeed.
Playing with RM I can all core @4.6 with 1.25v and appears stable (intelburntest maximum settings 4 passes, reputedly more stressful than prime95) but I won't run that as I want it downclocking when browsing etc.
Left to its stock settings as I am doing, running cinebench it only boosts to around 4500 all cores but it puts 1.4+ vcore in.

 

[RCLockwood]

Those rumours, if true, could make it better indeed.
Playing with RM I can all core @4.6 with 1.25v and appears stable (intelburntest maximum settings 4 passes, reputedly more stressful than prime95) but I won't run that as I want it downclocking when browsing etc.
Left to its stock settings as I am doing, running cinebench it only boosts to around 4500 all cores but it puts 1.4+ vcore in.

try running the system at default bios settings, dont adjust ram speed up and see if you get about 1.2v instead of 1.4v, for me as soon as i up the dram clock the voltage jumps up by .2 it feels like some sort of bug

 

[UK_SPAWN]

hmm seems like there may be some bugs from what i keep hearing around the net and things like this... i bet there will be a AGESA update within the next week or so

 

[Vogon]

try running the system at default bios settings, dont adjust ram speed up and see if you get about 1.2v instead of 1.4v, for me as soon as i up the dram clock the voltage jumps up by .2 it feels like some sort of bug

Thanks for the pointer, but it seems I don't have that issue, dropped the ram to non xmp and still have the same voltages.

I may have another bug... not sure as I am not fully conversant with AMD cpus, but when I try a negative voltage offset in bios it seems to diasable any auto boost and cpu runs at base clocks without boosting.

 

[Nick1881]

Indeed it seems normal, but less than perfect in my case as I went for it over the 12 core part as I had assumed less cores=less heat.
I don't need 12 cores but would have happily paid the extra had I known the 5900 was cooler !
But if I had waited for reviews I wouldn't have anything, so there is that

I also went 5800X assuming it would run cooler, bit gutted I didn't go for the 5900X. I might cancel as I'm not in a rush.

 

[amd7674]

Maybe upcoming new AGESA will help with thermals. According to AMD Director of Technical Marketing – Robert Hallock , there will be a new undervolt functionality for zen3.

But Robert, where is undervolting? Answer: IT’S GONE. j/k. Temporarily gone for 5000 series. It’ll be back in an upcoming AGESA with new functionality. It’s gonna be HYPE. Y’aint never seen undervolting quite like this.

“CAN YOU TELL ME MORE?” No. ¯_(ツ)_/¯

I have a feeling it will help all zen3 with minimal performance impact.

 

[Griff7520]

anyone else noticed a 500 point drop on cinebench r20 after undervolting, I did a 4.7 all core with 1.25v today and it was very stable and about 10c cooler but I lost 500 points on the bench marks, not sussed why, I think other people have seen the same, 5800x does not like you messing with voltage for some reason

 

[UK_SPAWN]

seems common for lower voltages to yield lower scores. the boost behaviour is changed with voltage so even though in HWINFO the boost may look the same remember that can only update every 100ms so the CPU can actually have changed clocks 100times in that amount of time (can change clocks every 1ms)

EDIT: I see your running all core OC but i think the voltage can still effect things in the sense that if there isnt enough voltage it will cause problems in the cpu trying to maintain clocks you have set. I cant really explain what that means because i dont know myself but its not unusual and why you should always validate the overclock just beacuase the clockspeed is higher sometimes performance can regress due to *reasons*

I would wait for AGESA updates to see if anything improves but it seems so far (other youtubers also report the same) that AMD has got the voltage pretty much exactly where it needs to be for optimal performance.

I would keep testing a few different increments and negative offsets like -0.075v to -0.05v. also try taking note of the maximum voltage reached at idle and what voltage while doing a all core CB20 run when all auto voltage. then try setting a manual voltage around there with a LLC setting that allows it to droop more/less and see how that effects scores.

EG: Everything Auto -
Idle Voltage = 1.45v
CB Load voltage = 1.35v

Try: Manual voltage 1.45v
LLC setting that allows it to droop to 1.33 / 1.30 / 1.28 or whatever


Anyone played with the PBO scalar settings and board power limits like TDC EDC PPT?

 

[adren]

I think some of it appears to be related to how fast the CPU fans can react to really fast temperature spikes because of the boosting. Looking closely at temp graphs and over long periods of high usage I'm pretty comfortable around 70c, but there's a few very small spikes that seem to suggest this behaviour. Could probably tweak the fan profile to bring the whole curve up, the singular max temperature value doesn't seem to hold as much importance as what I'm used to with old intel stuff, as long as it's not too crazy!

 

[Vogon]

TBH I am still having a hard time coming to terms with these temps, having the previous cpu never break 72 deg on a max intelburntest stress test and mid to low 40s in games !​

I have done some testing, I would be interested to hear opinions on the results (primarily temps and clocks, not so much CB scores).

Methodology was: fresh boot of windows between each test, measured with HWinfo which was loaded first then CB loaded, I then waited for cpu to settle to idle temps (32/33 deg atm for me) and reset HWinfo readings and hit CB run button. After the run I took the max readings from HWinfo.
Cooling is 360 AIO set to reach 80% fan speed at 60 deg 100% at 85 deg, but TBH during my earlier testing running a constant load for upwards of 20 mins (intelburntest maximum settings) fan speed above 80% does not improve temps at all.

Setting: Max temp deg c: Max Vcore: Max clock (average of all cores): Score:
Stock: 84 1.337 4575mhz 5926
Bios setting PBO 95w Eco: 76 1.28 4510mhz 5769
Bios setting PBO 65w Eco: 57 1.32 4375mhz 5351
RM all core clock 4600mhz/1.25v: 70 1.25 4600mhz 6040

 

[DanF]

There's some good replies in this thread
Some info about bad MSI bios atm. Also the 5900x has 2 CCX, one seems to be power efficient, the other power hungry and can go to higher single core boost clocks like the 5800x. The 5900x also draws a lot of W per core if only 6 cores are tested at the same time.

 

[UK_SPAWN]

im wondering if people are doing a blob of paste in the middle and its not spreading out over the die in the 5800x there in only one die. maybe thats why they get high temps and the cooler isnt heating up...

I havnt installed mine yet still gota get the mobo but its something iv been thinking about.

 

[Griff7520]

I actually put a blob and then spread it out across the whole surface with a spreader, dont know if thats the way to do it on these chips but if there is a better way please say

 

[UK_SPAWN]

I actually put a blob and then spread it out across the whole surface with a spreader, dont know if thats the way to do it on these chips but if there is a better way please say

no thats fine way to do it.

I just put a X combined with a + and then let it squidge out the sides

 

[PinkFloyd]

Fit my 5800X today along with a new set of RAM. Like others, I'm finding things toasty!

Ryzen 7 5800x at stock settings
X570 I AORUS PRO WIFI
2x16GB Crucial Ballistix 3600MHz CL16
Noctua NH-L12S with full size 120mm fan on top of the cooler
Coolermaster NR200P with glass panel

Idle temps are a pretty poor 50 degrees, rising up to 90 degrees when under a full all core load from Cinebench R20. I'd expected high temps considering the cooler is marginal, but with many others seeing comparable results with 280mm AIOs (or similar) and the fact that the heatsink is barely getting warm I'm starting to think it's just something we might need to get used too!

Performance stomps my old Ryzen 2600, but it's below what it should be based on posts here and on reddit. Cinebench R20 scores are at 582 single core, 5364 multi core. CPUz benchmarks give me 645 and 6463 points for single and multi core.

Is there a tool that will produce a nice solid graph of clockspeeds, voltages and temperatures? I'm not sure what to believe right now. I'm at a loss just trying to track clockspeed on it's own; CPUz jumps around between 3GHz and 5GHz faster than I can read the numbers, Task Manager has a mind of it's own, HWMonitor doesn't recognise the CPU and Ryzen Master also jumps around (albeit not as bad as CPUz). This is making it hard to judge what sort of clocks I'm boosting at.

As for thermal paste, I've gone with the central pea-sized blob approach. But will try something different if there's a strong consensus one way or another!

 

[UK_SPAWN]

Fit my 5800X today along with a new set of RAM. Like others, I'm finding things toasty!

Ryzen 7 5800x at stock settings
X570 I AORUS PRO WIFI
2x16GB Crucial Ballistix 3600MHz CL16
Noctua NH-L12S with full size 120mm fan on top of the cooler
Coolermaster NR200P with glass panel

Idle temps are a pretty poor 50 degrees, rising up to 90 degrees when under a full all core load from Cinebench R20. I'd expected high temps considering the cooler is marginal, but with many others seeing comparable results with 280mm AIOs (or similar) and the fact that the heatsink is barely getting warm I'm starting to think it's just something we might need to get used too!

Performance stomps my old Ryzen 2600, but it's below what it should be based on posts here and on reddit. Cinebench R20 scores are at 582 single core, 5364 multi core. CPUz benchmarks give me 645 and 6463 points for single and multi core.

Is there a tool that will produce a nice solid graph of clockspeeds, voltages and temperatures? I'm not sure what to believe right now. I'm at a loss just trying to track clockspeed on it's own; CPUz jumps around between 3GHz and 5GHz faster than I can read the numbers, Task Manager has a mind of it's own, HWMonitor doesn't recognise the CPU and Ryzen Master also jumps around (albeit not as bad as CPUz). This is making it hard to judge what sort of clocks I'm boosting at.

As for thermal paste, I've gone with the central pea-sized blob approach. But will try something different if there's a strong consensus one way or another!

As the single die is in the top left corner of the CPU making sure the paste is properly going to the corner would be worthwile. if you have a blob in the middle that isnt extending out when squashed down to the very corners i can see that being an issue.

When people do a pea size blob it sometimes doesnt cover the whole IHS.

Its better to have too much and squash it out the sides than not enough. also make sure u tighten down the cooler pretty tight. most cooler have a point where the srews stop and they should be tight to that point or atleast very finger tight.

 

[Vogon]

I knew the layout of the die/ccx, but not where in relation to orientation. So I put a rice grain and a half and used a plastic trim removal tool (only thing I could find that was flat and rigid apart from my CC !) to spread all over to within 1mm of the edges to be sure of coverage.
Idling dependant on ambient, but 30-34 consistently.

 

[rn2]

This is a very helpful and interesting thread. I'm planning to get the 5800x and an assassin 3 air cooler. I really don't want to pay for a 5900x to be honest. I'm going to wait a month and see how it goes.