New 5800x.

[Speedyblupi]

I only hope, My liquid cooling for the CPU will be able to keep up with the 5800x. It does well with my current 2700x.

It should do. The Ryzen 7 5800X uses less power than the 2700X, so should be easier to cool.

 

[Zach_01]

It should do. The Ryzen 7 5800X uses less power than the 2700X, so should be easier to cool.

On all core workloads they are using about the same power.
7nm chiplets tend to work warmer because of heat density as others said already. 5800X is the warmest of all 5000series. Not by much but still.
It’s the only single 7nm CCD CPU with 142W max stock power. I guess they compensate this with better binned chiplets for the 5800X but still it’s not enough. Anyway it’s not all that bad.

EDIT: typo

 

[toilet pepper]

That's what I said too. I came from a 3600 to a 5800X. After some tweaking here's what I got.

Here's Cyberpunk:

 

[ratirt]

That's what I said too. I came from a 3600 to a 5800X. After some tweaking here's what I got.

View attachment 201844

Here's Cyberpunk:
View attachment 201846

1.48v core? That's quite a bit though. It's more than what my 2700X needs. Obviously your is clocking way higher but still. Can you try the single core and see how it goes and post it here please?
I hope that the CPU will get here today and if I'm not busy and all is working as intended, I will get some benchmarks done and post those here for reference.

 

[tabascosauz]

1.48v core? That's quite a bit though. It's more than what my 2700X needs. Obviously your is clocking way higher but still. Can you try the single core and see how it goes and post it here please?
I hope that the CPU will get here today and if I'm not busy and all is working as intended, I will get some benchmarks done and post those here for reference.

I forgot for a sec that you're still on a 2700X I was like where have you been the past 2 years lol

it's just how boost works since Ryzen 3000, nothing to worry about. The Vcore visible in open source software is pretty useless until your CPU is under heavy (MT) loads, at lighter loads you can just check the Vcore in Ryzen Master.



And I got it wrong, your board does have BIOS flashback (Q-flash), it's just not on the rear I/O. There's a white button in the bottom left of the board.

 

[ratirt]

I forgot for a sec that you're still on a 2700X I was like where have you been the past 2 years lol

it's just how boost works since Ryzen 3000, nothing to worry about. The Vcore visible in open source software is pretty useless until your CPU is under heavy (MT) loads, at lighter loads you can just check the Vcore in Ryzen Master.

View attachment 201914

And I got it wrong, your board does have BIOS flashback (Q-flash), it's just not on the rear I/O. There's a white button in the bottom left of the board.

Well You know. 2700X has served me well but it is not its day to die. It will serve another home where it will make a user very happy.
With all the shortages and such I had to wait with an upgrade but the time has come and voila. 5800X it is.

I know the Vcore goes higher when the CPU is boosting but I didn't realize the Vcore goes to almost 1.5v for the 5000 series CPUs. What you are saying is the software shows inaccurate values? If that is the case by how much?

 

[tabascosauz]

Well You know. 2700X has served me well but it is not its day to die. It will serve another home where it will make a user very happy.
With all the shortages and such I had to wait with an upgrade but the time has come and voila. 5800X it is.

I know the Vcore goes higher when the CPU is boosting but I didn't realize the Vcore goes to almost 1.5v for the 5000 series CPUs. What you are saying is the software shows inaccurate values? If that is the case by how much?

When you're at idle, open source software will show 1.4V+ on the SVI2 TFN sensor. It's not actually accurate, Ryzen idles at sub-1.0V, which you can see in Ryzen Master. It's not "inaccurate", just unable to read the true (rapidly fluctuating) Vcore at idle; the heavier (and more cores) the load, the closer RM and HWInfo readings get until they read the same Vcore at full all-core load.

At lightly threaded loads PB2 allows up to 1.5V. It's alright, current is low, that's why it's allowed. As more cores come online (more current) Vcore progressively decreases until about 1.1-1.25V @ 100% all-core depending on which Ryzen 5000 SKU you have. It's been this way since Ryzen 3000, but Ryzen 3000 drew higher all-core Vcore (1.25-1.35V). Same "1.4-1.5V" at idle.

Picture:

 

[ratirt]

When you're at idle, open source software will show 1.4V+ on the SVI2 TFN sensor. It's not actually accurate, Ryzen idles at sub-1.0V, which you can see in Ryzen Master. It's not "inaccurate", just unable to read the true (rapidly fluctuating) Vcore at idle; the heavier (and more cores) the load, the closer RM and HWInfo readings get until they read the same Vcore at full all-core load.

At lightly threaded loads PB2 allows up to 1.5V. It's alright, current is low, that's why it's allowed. As more cores come online (more current) Vcore progressively decreases until about 1.1-1.25V @ 100% all-core depending on which Ryzen 5000 SKU you have. It's been this way since Ryzen 3000, but Ryzen 3000 drew higher all-core Vcore (1.25-1.35V). Same "1.4-1.5V" at idle.

Picture:

View attachment 201917

Man, I got to test the sumbitch when it is here I'm really curious.
I tried the ryzen master with the 2700x and the Vcore (if I remember correctly) stayed at 1.45 max when fully loaded or when you switch the PBO on.

 

[Zach_01]

I hope you don't freak out with the voltages you'll see. As @tabascosauz already said this is the mechanism of boosting for ZEN2/3. High voltage for single core and progressive voltage decrease as threads and load goes up (=more current/EDC). You can see all that in HWiNFO64 sensors.
According to HWiNFO64 author the "CPU Core Voltage SVI2 TFN" is the core voltage as accurate as it can be. VIDs is not the supplied voltage but just a per core request. The board may also provide VRM voltage sensors (VR VOUT) for cores and SoC individually but stick with SVI 2 TFN sensors for both.

Depending on the power plan you'll use or your background tasks, you'll see from 0.8~0.9V up to 1.5V.
For CB R20/23 expect a lot more under 1.3V and for heavy AVX under 1.2V. Always at stock boost.

Remember that ZEN2/3 CPUs can alter their speed and voltage every 1ms. No software can catch that even at the lowest possible polling period. HWiNFO is by default polling every 2000ms. You can set it to 1000ms but going under it will keep more cores from entering low power stages (C1=halt and C6=sleep/power down) and keep them more at active state (C0) with higher avg. speed/voltage than the system really needs.

 

[ratirt]

I hope you don't freak out with the voltages you'll see. As @tabascosauz already said this is the mechanism of boosting for ZEN2/3. High voltage for single core and progressive voltage decrease as threads and load goes up (=more current/EDC). You can see all that in HWiNFO64 sensors.
According to HWiNFO64 author the "CPU Core Voltage SVI2 TFN" is the core voltage as accurate as it can be. VIDs is not the supplied voltage but just a per core request. The board may also provide VRM voltage sensors (VR VOUT) for cores and SoC individually but stick with SVI 2 TFN sensors for both.

Depending on the power plan you'll use or your background tasks, you'll see from 0.8~0.9V up to 1.5V.
For CB R20/23 expect a lot more under 1.3V and for heavy AVX under 1.2V. Always at stock boost.

Remember that ZEN2/3 CPUs can alter their speed and voltage every 1ms. No software can catch that even at the lowest possible polling period. HWiNFO is by default polling every 2000ms. You can set it to 1000ms but going under it will keep more cores from entering low power stages (C1=halt and C6=sleep/power down) and keep them more at active state (C0) with higher avg. speed/voltage than the system really needs.

I wont freak out. It's not like I don't know the voltage rises when boost kicks in. It's just hard to believe (when this conversation started) that hitting a 90c on a CPU is inevitable.
I will check all of those when I get the parts.

 

[lemoncarbonate]

Can't say about Elite series, but I'm running an older AX370 Gaming 5 here, for over 4 years now. I love this board, a little bit expensive when I got it but it's a solid board, with 3y warranty.

Although, it bricked itself twice. Apparently from some forum this was a common issue with GB board. The solution is as simple as removing CMOS battery, wait 5min and put it back in.

My complains about GB board are mostly related to their RGB.
1. Their RGB Fusion software sucks. I use the very old version (late 2017) because version released from 2018 and recent, the software was hogging the CPU so much in idle. Regardless, the software is slow and painful to use. Although still better than the newest version of MSI Mystic.
2. The RGB LED also sucks, inconsistent color across the board.

 

[Speedyblupi]

On all core workloads they are using about the same power.
7nm chiplets tend to work warmer because of heat density as others said already. 5800X is the warmest of all 5000series. Not by much but still.
It’s the only single 7nm CCD CPU with 142W max stock power. I guess they compensate this with better binned chiplets for the 5800X but still it’s not enough. Anyway it’s not all that bad.

EDIT: typo

True, but in more common workloads like gaming a 5800X would usually use less power because it would usually not be running at full speed on all cores as it's more likely to be bottlenecked by the GPU... though I thought it also used less power at full multicore load because the 5600X uses less than the 2600X, but after checking I can see that while the 5600X uses less than the 2600X, the 2700X and 5800X use about the same, so thanks for the correction.

The higher density does also make them tend to run hotter than older 8-core Ryzen CPUs. This usually doesn't cause any problems as TSMC 7nm can work well at relatively high temperatures (though it clocks higher if it can be kept cooler), and the overall amount of power/heat that needs to be dissipated is about the same.

 

[Zach_01]

I wont freak out. It's not like I don't know the voltage rises when boost kicks in. It's just hard to believe (when this conversation started) that hitting a 90c on a CPU is inevitable.
I will check all of those when I get the parts.

You have H115i so I would worry much about it. Of course depends on the system setup, use case scenario, and/or ambient temp also.
I'm using H110i 280mm with my little 3600 and it manages to keep it around 60~62°C with ~1400 fan rpm and low pump speed, when it draws 90+W on CB R20 (Ambient 22C).
If you're not running stress tests all day I believe it won't go too far. On gaming ZEN2/3 are very comfortable, temp wise.

Can't say about Elite series, but I'm running an older AX370 Gaming 5 here, for over 4 years now. I love this board, a little bit expensive when I got it but it's a solid board, with 3y warranty.

Although, it bricked itself twice. Apparently from some forum this was a common issue with GB board. The solution is as simple as removing CMOS battery, wait 5min and put it back in.

My complains about GB board are mostly related to their RGB.
1. Their RGB Fusion software sucks. I use the very old version (late 2017) because version released from 2018 and recent, the software was hogging the CPU so much in idle. Regardless, the software is slow and painful to use. Although still better than the newest version of MSI Mystic.
2. The RGB LED also sucks, inconsistent color across the board.

I uninstalled RGB Fusion ASAP from the start of getting X570 AorusPro. Since, I keep RGB off from BIOS. No need for blinkers...

By the way... @ratirt I assume the board won't come with the initial X570 BIOS (F3~4), but it would be nice to update it from the start. Pre F10~12 versions cause a lot issues, and some times could brick the board even with just messing around with DRAM settings. Thankfully BIOS Q-Flash Plus saved the day(s).

 

[ratirt]

You have H115i so I would worry much about it. Of course depends on the system setup, use case scenario, and/or ambient temp also.
I'm using H110i 280mm with my little 3600 and it manages to keep it around 60~62°C with ~1400 fan rpm and low pump speed, when it draws 90+W on CB R20 (Ambient 22C).
If you're not running stress tests all day I believe it won't go too far. On gaming ZEN2/3 are very comfortable, temp wise.

That's weird. I got my 2700x and it has never reach 50c while gaming. never reached 70c while benching all threads till reached stable temp. I know these are different CPUs but come on.
What I don't understand here is,
The reviews have been done by TPU for instance, Guru3d (They used a 280mm LC cooler) and others with a temp never crossing 80c yet here everyone is telling me a Corsair h115i pro rgb will not be enough to cool a 5800x CPU? No I don't believe it but I will definitely monitor the temp and get back to you.
Why would I stress test my CPU whole damn day? If you bench for all day long non-stop you better make sure your case airflow is flawless otherwise you will burn the god damn thing down.

Guys I did it. It's been a hustle today though. It's late already.
Here's some things I wanted to share with you.
This is just a first bench.
Simple stock CPU. Mem 3200 CL14.
And the temps.
I ran C20 single core and here it is you can see the temp. It's pretty fair since it takes a while to finish it. (no adjustments to the corsair software and the RPM/temp curve)

for what I know, It can go better.
Here is the multicore score (8 runs)

I know It can go higher it's just, I didn't have much time.
Suggestions? any settings I should try please share.
I'd be grateful for that

 

[Zach_01]

Mem is at 2400MHz and not 3200MHz. And I can’t see FCLK/UCLK speeds. Although RAM performance is not so significant for CB scores.

 

[Jack1n]

They get hotter than other Ryzen CPUs but I doubt it will reach 90c

Mine goes over 90C

 

[ratirt]

Mine goes over 90C

It can't go over 90. that's a max

Mem is at 2400MHz and not 3200MHz. And I can’t see FCLK/UCLK speeds. Although RAM performance is not so significant for CB scores.

Damn. how the hell.... ?
brb I'm changing it :/

 

[freeagent]

Both my 3600XT and 5600X have kissed a bit over 100 once.. whoops.

They are ok still.. I think

Both times were pretty much 20 minutes out of the box and somehow with 1.6v.. whoops.. bad math.

 

[ratirt]

Here is with 3200 @Zach_01 Thx for the catch. I will try getting 3600 and/or more if possible

What is weird, the single core score has dropped a bit but that is a problem for another day. Added the H115I Pro settings
Any suggestions let me know

 

[tabascosauz]

Both my 3600XT and 5600X have kissed a bit over 100 once.. whoops.

They are ok still.. I think

Both times were pretty much 20 minutes out of the box and somehow with 1.6v.. whoops.. bad math.

Kinda sketchy but as long as you didn't run any benches or loads greater than idle current (~10-20A(?)) should still be okay. Once you run say >50A current through the CPU at those temps and Vcore, you create the unholy trifecta you'll lose a couple hundred MHz within half an hour or a few benchmark runs.

Here is with 3200 @Zach_01 Thx for the catch. I will try getting 3600 and/or more if possible


What is weird, the single core score has dropped a bit but that is a problem for another day. Added the H115I Pro settings
Any suggestions let me know

A reliable ST score is really hard to achieve in Cinebench R20 or R23. The test takes too long, so any background processes will affect your score quite a bit. For example HWinfo running usually doesn't have too much impact on most benches, but in R20/23 ST it definitely is noticeable. Windows updates, indexing service, Windows Defender, etc. all potentially interfering if unlucky. If you max out your fan speed on air coolers, you can even boost your ST score a bit.

Once you get your baseline it's time to work with Curve Optimizer

 

[freeagent]

No benching, except on the XT.. that was really an oops. the 5600 only made it to desktop and when I launched hwinfo I saw the vcore and just shut her down and cleared the cmos. No harm no foul.. that was before I posted any kind of benching with it so I think she's ok. Well. wont bench @ 5ghz.. that's the only problem

I wouldn't say its ok to do it.. but then I said that with Intel too and have been to 100 more than once

Lol ok.. I know that sounds bad.. but trust me it was just briefly.. it reboots at 107c.. the Intel's that is

I've been nice to my AMD's

 

[GamerGuy]

I have the X570 Aorus Xtreme and it's been great for me, BT/WiFi 6 work as they should. I'd recently gotten a WiFi 6 router to fully utilize the improved speed (my Note 10+ and Fold 2 are also WiFi 6 capable). No issue with BIOS updates, got SAM to work right off the bat.

 

[Zach_01]

A reliable ST score is really hard to achieve in Cinebench R20 or R23. The test takes too long, so any background processes will affect your score quite a bit. For example HWinfo running usually doesn't have too much impact on most benches, but in R20/23 ST it definitely is noticeable. Windows updates, indexing service, Windows Defender, etc. all potentially interfering if unlucky.

That^^

It’s pretty difficult to reproduce same scores in a row. Also, 1-2C temp fluctuation between runs can result in different score. 622–>617 is less than 1%.

The best way to capture with HWiNFO these runs (any benchmark/test CPU related) is to start first the load and after 2-4sec reset the values in HWiNFO with clock down right and few seconds before it ends you take the screenshot.
This way you capture values only during the test and you can study them later.

Better to have HWiNFO full screen with multiple columns so you can see all CPU related values visible.

From settings you can enable “Snapsot CPU Polling” as its more accurate and minimizes the observer effect.

Tool tip is also nice to have when hovering mouse pointer over sensors. But this affects CPU idle states unless you place the pointer off of them.

Collapsible sensor sections are nice too, keeping unwanted sensors invincible but it’s disabled if you re-arrange their order.

 

[ratirt]

A quick 3dmark timespy test

There's still a bit I can improve but that is for later.

A quick 3dmark timespy benchmark.

There is still some stuff I can improve. I shall try with the curve optimizer but some other time.
What I noticed was the 5800x during the benchmark was pulling almost a 100Watts during the CPU test. That was surprising.

I've tested the CPU with the Euro Truck Simulator. Damn the GPU is being maxed out all the time. With my older CPU 2700X it was really hard to achieve.

 

[Zach_01]

You were expecting more or less than 100W?
It is after all a ~140W CPU