Abnormalia

1 hour ago, mariushm said:

Look into how much the extra cache in 5800X3D affects the tools and stuff you use with that system.  Could be those extra 64 MB would benefit you a lot, more than the extra couple of cores a 5900 would give you. 

 

On one hand I want to say wait a couple months to see what processors they announce for socket AM5, see if they're gonna have more processors with that extra cache BUT on the other hand for the first 3-6 months or so after release/launch it's gonna be lots of bios updates, maybe memory compatibility issues etc etc.. 

yeah but it overprices ATM and hard to get. So it would be better to get 2 x 32 kit ATM ? I've read multiple times that its better for X570 to have 2 vs 4 mem. Can you guys recommend field tested kits ?

24 minutes ago, ZetZet said:

Why not play around, underclock your RAM first? If you can spend 0 money and make it stable I think that's worth a negligible performance loss.

I tried different BIOS versions, tune memory manually, reading tons of forums. My CPU was rated Bronze. Now mem runs 1-1 1800 XMP. It stable but occasionally there is USB or WHEA errors. This is why I think either getting new CPU or CPU + Mem swap could be good idea. I want to know which kits are recommended and field tested for 64GB scenarios.

So I got build for work ( VR games, video rendering, modeling etc. ) and is was mostly flawless with Asus X570 and 4X (G.Skill Trident Z Neo Series DDR4 3600 CL16-19-19-39). Yes I need 64GB from time to time when I make builds daily. I had issues with WHEA errors occasionally. Most stable BIOS I found is 3801. So is it 4X (16GB) memory that is causing instability ? What do you recommend go for 2 x 32 GB ( which kits would you recommend in this case?) or replace CPU with 5900X or 5950X ?

Thanks guys, "Operations repaste" is officially over. Results went from 95C to 65C at max load on both GPU and CPU. Performance is back as it was 2.5 years ago, when I bought it. I've used Dremel 4000 (Outdoors, w mask and eye protection just in case) this header on first image.

Have opened Asus Scar III as its thermals got worsen recently. This is what I see now.

I cleaned CPU and GPu easily. But IHS have it oxidized and now they stuck really hard to surface, alcohol and tissue have almost no effect.

How can I clean IHS effectively to reapply paste ?

I bought this custom build laptop In Aug, 2019

Custom Built ASUS ROG Strix Scar III G531GW-KB71 - 15.6" FHD 240Hz - i7-9750H - RTX 2070

Also I've chosen this option: 

Thermal Interface Materials
Thermal Grizzly Conductonaut on CPU + GPU, and Fujipoly Extreme Thermal Pads on heat sensitive surfaces – BEST PERFORMANCE

 

Below is 3D Mark chart after I received laptop and it was scoring approx. 7400-7500

 

Recently I noted while working on VR games and playing CP 2077 performance was terrible. 

I logged temps and noted that when PC was waking up it was hitting 95C briefly on silent mode frequently.  

I launched 3D Mark and scored average 4400-4500 with tons of thermal trotting on CPU, Before it was 4100 Ghz during stress tests (see attached images). 

 

So It obvious I need to open and reapply paste. Is it normal for LM to degrade like that in 2.5+ year ?

Do you recommend to reapply same or more traditional paste this time ?

Update 1 week later : tested every possible way, work as usual not an single accident on idle and on high loads

It seems sparks originated form that area.

10 minutes ago, Mel0nMan said:

That's definitely not good. A lot of things could have caused that: improperly seated memory modules causing a short, PSU issue, or motherboard issue to name the most likely. You said the mem LEDs were turned on? It could be faulty ram sticks or not properly seated, take them out, blow them off with a can of compressed air, stick em back in and see if the sparks happen again. If no sparks, fixed. If sparks, motherboard issue.

Will try that soon, thanks !

7 minutes ago, Chris Pratt said:

What PSU?

Corsair AX 850

I got this build more than 1.5 year and its working mostly stable except few bad BIOS versions. X3900 + Asus Hero VIII + 64 GB memory + 1080 Ti.

So I went out of city for weekend and when returned and tried to turn PC, nothing no reaction, only memory LEDs were turned on, not a single Fan or other reaction.

Tried to turn off/on PSU, nothing.  Opened side panel, turned off PSU, waited 5 sec and pushed start button on MB, same only memory leds turned on.

Repeated and this time, sparks somewhere near memory slots and computer turned on. It booted and seems ok.

Inspected MB no burned marks, what may cause this ? What should I check or test? Is MB dying ?

I got AMD based Desk 3900/64BG/M2 with 5000+ read speed and when I open same project on desk it takes approx 2.5 min.

Also got Laptop 9750H/32GB/M2 3000+ read speed and same project takes approx 26-30 seconds.

Tested many times. Its not much time just what may cause such behavior ?  

So last 2 weeks noted this weird behavior. 

Windows 10 2004. All updates, BIOS, Drivers etc. Power plan - AMD Balanced

BIOS settings: 3900X, DOHC 3600, vCore undervolted -0.1v, rest settings are mostly default, PBO auto. 

So It works fine 2-3 days no crashes etc. If I shutdown and turn on in the morning, as soon as windows logged in it restarts suddenly (no BSOD) in 2-3 sec, and after that works fine.

If I restart it after that np. Eventviewer shows nothing except unexpected shutdown.

Temps are fine, 47-50 idle and 75-77C on max synthetic test and load boosting all cores 4200-4300.

Was part of CPU missing part of the paste before ? Or too much pressure caused this?

Ok I build this system 8 month ago and decided to clean and reapply thermal compound. So this is how it looked when I removed cooler. It seems it was not evenly spread on surface right ?

Reapplied, all defaults, voltage vcore -0.1v, PBO on. I tested with Prime95. Last time I tried it temps were getting to high. Now I accidentally left it for 40 min and when noticed 100% CPU in task manager, turn it off. CPUID highest temp was 80C. Boosting 4300 all cores.

10 minutes ago, boggy77 said:

that's a bit much, I think -0.05 would be more stable.

still better on auto for regular use. only some applications would benefit from all core fixed 4.2ghz.

no, the mb won't degrade you cpu over time. the higher voltages are in short bursts, only when needed.

Just booted with -0.1v, PBO is off. Performance is 1-2% better. vCore 1.39-1.4v on Idle 1.25-1.27v on load and max temp 75C on air cooler.

1 minute ago, DJ46 said:

I'd probably save the OC you have as a profile, reset it to factory defaults and see what clocks and voltages it runs at. Then look into all the automatic overclock settings and set a negative voltage offset.

A voltage offset won't do much with a fixed clock OC. But with Ryzen, a fixed all core OC is already sub optimal.

I believe default is 1.45 on Auto. Will try to set negative once more. But if this runs stable fixed 1.2v and would not harm system in long term besides loosing some perf, I do not mind. I was just worried f 1.4+ Ryzen Master was showing was true.

1 minute ago, DJ46 said:

Personally I prefer setting a negative voltage offset.

You've no doubt noticed that on Auto the voltage dips under an all core workload. Ryzen runs at a higher voltage in single core tasks to reach higher frequencies and at a lower voltage in all core tasks to lower temperatures.

With a voltage offset your voltage will be lower in both scenarious. With a static voltage (and a set all core clock which it looks like you have too) you loose some singlethreaded performance and it runs at a higher voltage under a full load.

I tried set -0.1v negative in BIOS and cannot post, have to clear CMOS.

3 minutes ago, TofuHaroto said:

why did you set it to a static voltage? unless you need it for something specific just dont use a static voltage.

the reading is 1.2.

It works stable at this 24/7 all core 4200 so still better to have auto ? Does MB will try to push higher values which degrade CPU over time?

1 minute ago, SavageNeo said:

well on hwinfo it says that you got 1.2V...

Ryzen Master showed 1.4+ I thought it was most accurate. HWInfo also shows 2 values. That's why I get confused.

In bios I set fixed 1.2v, So seeing this in different programs. I run them all together or separately same results. What is actual vcore  voltage CPU is getting?

R20 7356 -   3900X@4200 all cores while vcore fixed 1.22v

Attached some screens of system state before and after R20 run:

Just updated to latest bios and saved profile gets lost. I built this system almost an year ago and do not remember what were exact bios settings. I found so many guides but as I suspect some are outdated and some even contradicting themselves.

RIG:

MB: ASUS ROG Crosshair VIII Hero X570

CPU: 3900X

MEM: G.Skill Trident Z Neo Series 4x16 3600

Cooler : Noctua NH-U12A

My current settings in BIOS are:

VCore 1.22v

Multiplier fixed at 42

XMP - 3600

Fabric - 1800

PPT - 300

TDC - 230

Windows: 

All latest chipset drivers

Power plan - Ryzen Balance

Cinebench R20 - 7380

So with hot summer ambient at 30-32C it idles 47-50 and reaches 80C at stress test with all cores at 4200.

p.s.

Ryzen Master shows incorrect vcore sitting fixed at 1.1 all the time.

So are those correct settings or there is 2020 updated guide for approved settings. My goal to have stable "creative mode" machine for development all cores working stable at 4200-ish. Thanks all in advance.

42 minutes ago, Mira Yurizaki said:

The either:

• Set the UEFI boot mode to Compatibility Mode (usually shorthanded to CSM)
• Disconnect any other drives you have in the computer and try again.

Tried with on and off. Maybe I need drivers ?