***The Official 5900X \5950X owners thread***

[Arron17]

I won't claim to be an expert but I will state two things:

1. The developer himself states that even 'one' error is cause for concern:
Starting at how many errors do you consider a computer unstable ?
Seriously : One. - https://www.ocbase.com/#_faq

2. I used this test against 1 core at a time for 1 hour each and was able to pass every core without a single error. That said, what I thought was previously a stable curve optimisation was proven otherwise by this test. I would usually have to knock the curve down about 5 increments on every core before the OCCT test stopped reporting errors. So, I can attest that it was helpful in finding 'unseen' errors and there was a level of consistency to when it stopped reporting errors (it wasn't random chance).

EDIT: Just to add - if you are experiencing errors at stock, do not panic. It could very well be a BIOS problem. Update to the latest revision if you haven't already and test again.

I've never actually had any problems with the machine. And I only get the errors when running 1 thread with the affinity set to one core and it's thread. If I just set occt to 1 or 2 threads manually it's fine. BIOS is up to date, although is a beta.

It's just basically impossible to stability test anything in the way you described as it will just error out regardless.

 

[VortX]

I've never actually had any problems with the machine. And I only get the errors when running 1 thread with the affinity set to one core and it's thread. If I just set occt to 1 or 2 threads manually it's fine. BIOS is up to date, although is a beta.

It's just basically impossible to stability test anything in the way you described as it will just error out regardless.

Have you tweaked curve optimiser yet? Do you have access to it on your motherboard? If yes and you haven't yet tweaked it, maybe try changing it to a positive curve of +5 on all cores to see if it makes any difference? It could simply be that you have a dodgy core somewhere.

 

[Razor Time]

When you say you're hitting 75-77C, what are you using? Also, what else is on the loop?

To add some context, with my 5950X I am currently running 2x triple radiators on a custom loop each with a single bank of fans. I am cooling the CPU and MOFSETs using this: https://www.ekwb.com/shop/ek-quantum-momentum-aorus-x570-master-d-rgb-plexi. I haven't yet added my GPU to this loop.

With PBO enabled + a small OC I can still hit 90C running a small FFT stress test for a couple of hours. I do have the fans at a relatively quiet curve, but still. General gaming I am seeing anywhere between 50C and 70C. I too reapplied paste two times to make sure.

It just seems these CPUs run HOT. I am not looking forward to adding a GPU to the loop

Ok so sounds fairly similar to what I am seeing.

My specs are:

AMD 5900X + HeatKiller Pro IV with an Asus X570 Crosshair Dark Hero and an EVGA 3080 + EK Block in the single loop.

Cooling is:

Alphacool XT45 420mm & 480mm/ / VPP755 pump/ Heatkiller D5 Top/ Aquaero 5 XT & LT Slave/ Bequiet Silent wings 3 on all rads.

What threw me off was seeing the same temps on gaming which was only using 30% cpu load to running cinebench multicore at 100%, water delta was fairly similar.

 

[VortX]

Ok so sounds fairly similar to what I am seeing.

My specs are:

AMD 5900X + HeatKiller Pro IV with an Asus X570 Crosshair Dark Hero and an EVGA 3080 + EK Block in the single loop.

Cooling is:

Alphacool XT45 420mm & 480mm/ / VPP755 pump/ Heatkiller D5 Top/ Aquaero 5 XT & LT Slave/ Bequiet Silent wings 3 on all rads.

What threw me off was seeing the same temps on gaming which was only using 30% cpu load to running cinebench multicore at 100%, water delta was fairly similar.

You'll probably find that is just due to the difference in behaviour between the benchmarks. I mean, Cinebench multicore will stress all cores simultaneously with medium load. A game may stress individual cores sporadically at high load. The end result is similar temperatures. That said, run Cinebench for long enough and the temperatures will continue to rise well beyond gaming temperatures.

 

[Razor Time]

You'll probably find that is just due to the difference in behaviour between the benchmarks. I mean, Cinebench multicore will stress all cores simultaneously with medium load. A game may stress individual cores sporadically at high load. The end result is similar temperatures. That said, run Cinebench for long enough and the temperatures will continue to rise well beyond gaming temperatures.

When I ran cinebench I was doing 10 mins loops, not just a single pass.

 

[Raiden85]

I'd expect gaming to be warmer, on my 5950x it's alway warmer than putting 100% load on 32 threads. All cores fully loaded at least when stock will run at a lower frequency with less voltage but when at 30 to 50% load when gaming the voltage can go much higher but with much higher clocks when spikes occasionally to 4.8ghz/4.9ghz so I'd expect higher temps.

Fully stock with no pbo in cinebench R23 my temps will be around 50 to 55c max but games such as the division 2 will get my temps to around 68c to 71c.

 

[Arron17]

Have you tweaked curve optimiser yet? Do you have access to it on your motherboard? If yes and you haven't yet tweaked it, maybe try changing it to a positive curve of +5 on all cores to see if it makes any difference? It could simply be that you have a dodgy core somewhere.

I just ran the same test as yesterday and no errors this time.
I did install the new chipset drivers last night, so maybe that sorted it.

Edit: nevermind it happened again. Only way I seem to be able to get it not to error is put +10 on the core in curve optimiser.

 

[Razor Time]

I'd expect gaming to be warmer, on my 5950x it's alway warmer than putting 100% load on 32 threads. All cores fully loaded at least when stock will run at a lower frequency with less voltage but when at 30 to 50% load when gaming the voltage can go much higher but with much higher clocks when when spikes occasionally to 4.8ghz/4.9ghz so I'd expect higher temps.

Fully stock with no pbo in cinebench R23 my temps will be around 50 to 55c max but games such as the division 2 will get my temps to around 68c to 71c.

That makes sense, I will have a look at the clock speed and vcore next time I fire up frycry new dawn which was the game I was playing. I was just playing cyberpunk and CPU temps were in the low 60s so a fair bit lower than farcry.

 

[Tuvoc]

I've had my 5900x for just over a week now, incredible machine. It was built to order from a spec sheet, the only change I made was to downgrade the graphics card. I have no intention of overclocking or fiddling, it is running perfectly as is. It is on an MSI MAG B550M Mortar Wifi board with Corsair RAM running at XMP 3200. I see now that a new BIOS is out, which says "Update to ComboAM4PIV2 1.2.0.0". Is there any benefit of upgrading ? I am worried I would lose memory settings etc after the flash - is that the case or should it just boot up as before ? I've always updated to latest BIOSs in the past, but in this case I'm inclined to think, "if it ain't broke don't fix it" ! What are your views ? Bizarrely the current BIOS is 1.52 according to CPU-Z and the new BIOS is 1.5, but the existing BIOS according to CPU-Z is only 1.1.0.0 and dated 13 Nov 2020 so clearly older.

 

[Arron17]

Wondering if anyone has any ideas about this I'm trying to work out if my CPU is defective or not.

I've been doing stability testing before trying out the curve optimiser and one of my cores the "best" core is failing stability tests.

Both OCCT small set SSE 1 thread which it fails just before 10 minutes and Prime95 small fft none avx.

None of the other cores fail anything on stock. I've tried multiple BIOS and am at factory default settings.

The main difference I see with this core and all the others is that it runs the tests at 4.881ghz, where the next best core will be at 4.8ghz.

Wondering if it's just some strange issue with the CPU pushing a little too hard in a specific scenario or if there's something wrong. It seems strange that only the best core would have issues.

One thing to note is of I run the problem core at +10 on curve optimiser it no longer fails the tests. But runs a little slower.

 

[MikeS3000]

Wondering if anyone has any ideas about this I'm trying to work out if my CPU is defective or not.

I've been doing stability testing before trying out the curve optimiser and one of my cores the "best" core is failing stability tests.

Both OCCT small set SSE 1 thread which it fails just before 10 minutes and Prime95 small fft none avx.

None of the other cores fail anything on stock. I've tried multiple BIOS and am at factory default settings.

The main difference I see with this core and all the others is that it runs the tests at 4.881ghz, where the next best core will be at 4.8ghz.

Wondering if it's just some strange issue with the CPU pushing a little too hard in a specific scenario or if there's something wrong. It seems strange that only the best core would have issues.

One thing to note is of I run the problem core at +10 on curve optimiser it no longer fails the tests. But runs a little slower.

Same exact issue on 5900x. Fails single thread on #1 core in occt and prime non-avx. AMD just approved RMA for that issue. You have a defective CPU. I would RMA. Cpu is shipping out tomorrow overnight to AMD. hopefully won't take too long to ship out replacement. Cpu is 2047PGS for reference.

 

[Tuvoc]

A question on voltage on the 5900x. I have everything on mine at stock. When I'm loading it with 20+ threads the voltage is at 1.2v or even less, and the CPU runs between 4.3 and 4.5Ghz. When I have it loaded on 1 core only, it boost to a little over 4.9Ghz, but also boost the voltage to as much as 1.48v according to Dragon Centre and Core Temp. Is that normal ?? Seems incredibly high.

 

[denigor777]

I've had my 5900x for just over a week now, incredible machine. It was built to order from a spec sheet, the only change I made was to downgrade the graphics card. I have no intention of overclocking or fiddling, it is running perfectly as is. It is on an MSI MAG B550M Mortar Wifi board with Corsair RAM running at XMP 3200. I see now that a new BIOS is out, which says "Update to ComboAM4PIV2 1.2.0.0". Is there any benefit of upgrading ? I am worried I would lose memory settings etc after the flash - is that the case or should it just boot up as before ? I've always updated to latest BIOSs in the past, but in this case I'm inclined to think, "if it ain't broke don't fix it" ! What are your views ? Bizarrely the current BIOS is 1.52 according to CPU-Z and the new BIOS is 1.5, but the existing BIOS according to CPU-Z is only 1.1.0.0 and dated 13 Nov 2020 so clearly older.

If it's working ok for you and you are happy with ir, then just leave it as-is. If you get pleasure fr the engineering and tweaking side of things, then here's yoir chance!

 

[Nitefly]

My core clock boosts and ram timings seem a bit poopy! Running at stock... can’t hit FCLK 1900 which is disappointing.

Is this normal or am I missing something? I’m a complete noob so all input welcome!

 

[Naff8000]

A question on voltage on the 5900x. I have everything on mine at stock. When I'm loading it with 20+ threads the voltage is at 1.2v or even less, and the CPU runs between 4.3 and 4.5Ghz. When I have it loaded on 1 core only, it boost to a little over 4.9Ghz, but also boost the voltage to as much as 1.48v according to Dragon Centre and Core Temp. Is that normal ?? Seems incredibly high.

Yes that's normal and it's pretty much what mine does.

 

[Tuvoc]

Yes that's normal and it's pretty much what mine does.

Thanks for confirming.

 

[Tuvoc]

If it's working ok for you and you are happy with ir, then just leave it as-is. If you get pleasure fr the engineering and tweaking side of things, then here's yoir chance!

I think I'll leave it as-is for a while. Support for Zen3 is probably still in the process of maturing BIOS-wise, so perhaps I'll wait another couple of BIOSs for things to settle down.

 

[MultiDoc]

Has anyone tried CTR 2.0 yet ? What’s your experience so far ?
I’m eagerly waiting for v2.1 that will allow for CO profile too

 

[grumpybeard]

Hey all. Been long long time since done any benchmarking. Having recently upgraded to setup in sig from 5820K, would like to check how the system runs.

Grateful for guidance on the recommended benchmark tool/test suit?

I note in hwinfo64 temp sensors, for cpu, there's CPU Package and CPU (the former tending to run a few deg higher than latter). Which one should I pay attention to when benching?

Edit - Just seen in hwinfo64 there's a load more readings associated with CPU, for e.g. CPU Die, CCD1 etc. Which one do you normally report on?

Thanks!

 

[Tuvoc]

I'd expect gaming to be warmer, on my 5950x it's alway warmer than putting 100% load on 32 threads. All cores fully loaded at least when stock will run at a lower frequency with less voltage but when at 30 to 50% load when gaming the voltage can go much higher but with much higher clocks when spikes occasionally to 4.8ghz/4.9ghz so I'd expect higher temps.

Fully stock with no pbo in cinebench R23 my temps will be around 50 to 55c max but games such as the division 2 will get my temps to around 68c to 71c.

When I load my 5900x up on 20 cores, the temps are actually lower than when it is under much lesser load as you say, because the board bumps up the voltage very significantly under lower load to achieve higher clocks. Quite bizzarre, that is a new experience for me. Coming from much older generartion CPU's the more you load them up the hotter they run, even when fans spin up to try and compensate, very predictable.

Anyway, my cooling is currently being tested as it is sitting in a room with no aircon, and temps outside of 30C. Core temps topping out at around 75C with headroom still remaining on fan rpm. (Silverstone PF360W-ARGB)