RainingTacco

Due to how AMD driver is made, the draw calls use a single process.There is only so much a CPU can do, before the thread is overloaded, so you will feel very little improvement between let say ryzen 3700x and i10900k in this regards, maybe 15-20%. What can excarberate the issue is If the same thread used by draw calls is also used for some other calculations like sound, AI etc. Draw calls is usually a huge load for a thread taking 50-75% or more of thread performance, depending on the game. So you can clearly see why multithreading draw calls is that important. This is what nvidia driver do with DX11 applications. If you want to test this easily run 3DMARK api draw call benchmark with AMD gpu and Nvidia GPU. Nvidia performance is roughly double of AMD. Better CPU won't help much, since single thread performance hasn't changed that much, and games develop to be more geometry/shading intensive. Sometimes the draw call load is so high, it overloads the thread and you get truly disgusting stutter. All while other threads are relatively lazying with 8 or more core CPUs.

Show frametimes graphs and compare that to nvidia equivalent of that time or vs current nvidia gpu. Different architectures are affected differently. I found that RX580 was less affected than 5700XT, maybe because it had less frames, but AFAIK i was using FPS standarized tests. Something terrible happened with RDNA and onward. Seems that RDNA2 has the same problems. Anyway, API overhead in DX11 affects all AMD gpus, some more some to lesser extent. I know that i won't buy another AMD product after RDNA fiasco. And its sad that they haven't fixed their drivers ever since. DX11 was left in dust bin for AMD driver team

AMD is completely rubbish in high CPU usage, DX11 games. See this: https://community.amd.com/t5/general-discussions/amd-when-will-you-fix-your-terrible-driver-overhead-for-dx11/td-p/298059 Odyssey is high cpu usage[lots of actors, open world] DX11 game = perfect recipe for disaster. One of threads is pegged by API draw calls, and frame render stalls until it resolve all draw calls. That's why its more stuttery. If you plan to play DX11 games i strongly advise to sell AMD gpu and buy Nvidia. I've researched this issue for a year, spare yourself troubleshooting and pulling hairs -it's AMD fault. They said they will NOT fix this issue EVER. So if you want to play odyssey with much smoother frametimes it's time to splurge on that RTX 3070.

Memtest doesn't heat CPU much, you have either bad cpu or ram. Does it crash/stop worker when running prime95?

My stock ryzen 3600 in prime95 smallest FFT has 70-71*C with kryonaut using a cheap Scythe Mugen 5 rev.B, with MX-2 it gets 77*C. I would reseat AIO, or check the flow if the aio provide the data.

Have they lowered/stabilized over time? This is horribly wrong temps for stock ryzen 3600 and AIO. Something is wrong.

Run prime 95 smallest fft for at least and hour, and run memtest for few loops at least. Only paid version of memtest give you more loops sadly.

Ok fair point. @jonnyguru are there PSUs that use fuses and current shunts in case something goes awry?

Am i wrong assuming that this is your work PC? Because otherwise, why put so much money in an absolutely stellar reliability under any circumstances? Unless you really love PSUs or money isnt an issue at all

1. A cooler might be designed to work best with intel AHS, and you have AMD ihs 2. Ryzens have terrible heat density 3. cpu manufacturers lie about TDP, a 65W Ryzen 3600 pulls 90W in full core tasks

It's a shame that biostar is rarely available in my country. They make cheap and good motherboards, and cheap motherboard is all i need.

It doesn't matter beside the question -are you willing to dust off your PC once in a while? If you don't mind, then you can have more exhausts. It will just pull more dust from outside.

Maybe the update again tinkered with gpu clocks? Check those, if they are higher than before update. Lower them by 100mhz and check if that helps. Switch display to lower refresh rate for now and check. AMD might have bugged something again, or reintroduced some bug

If you want ray tracing then the choice is simple.

A highly OCed 3090 sli will pull 1KW with worst case scenario like furmark. Still your PSU is such an overkill for that system[kinda stupid question when myself i paired ryzen 3600 with 5700xt but reading the horror storied about AMD gpus peaking at 2 times their wattage, i wanted an overkill and i plan to use this PSU for two builds, so its best if it runs at below its 50% capacity] Why?

I have rmx 850[the 2018 revision] and can vouch as excellent choice. Rock solid, but if you can AVOID psus with inline caps, they are nightmare to build with.

Oh so it's EU defence minister meetings? EU doesn't have an army or any military jurisdiction and power over that matters. That's pretty nonrelevant, nothing of value could be gained by hacking such conference -just a chit chat with big words like they do in EU parliament. NATO on the other hand...

PSU gurus are like any other specialist in their fields -they take edge cases and blow it out of proportions. Like if you asked a virologist what he recommends right now, he/she would say that for best safety you should go with full hazmat all the time, but the truth is -life is taking risks and assessing those risks and cases where they can appear. No one in right mind would wear full hazmat everyday 24h, just like i won't put this crappy PSU under a artificially manufactured case scenario with extreme transient load, switching every few seconds for hours. I know scientists and engineers love to put things to test and see how well they've made their titantium full proof psu with capacitors on every damn cable[fuck inline caps] but hey, it will only make a nice flat graph to feel you better Sorry for the semi rant, but sometimes people just take things too seriously, like an average user was buying a PSU to run a bank vault security system or his livelihood depends on the psu becasue he make a render that will render for two weeks straight and needs that AXi 1600W PSU[who the hell needs a 1600W PSU, even SLI RTX 3090 and threadripper won't take 1KW ] I've just asked a simple question whether poor voltage regulation will affect component lifespan, sadly didn't get sufficent answer. I hoped that someone made a an actual empirical study on that, or at least tests running PSUs for months and checking the degradation of components under microscope and with measurement equipment[mutlimeter checking for capacitance etc]. I would also like to see same study for PSU ripple regulation if possible. Truth be told, the biggest offenders are capacitors and then varistors, they take the blunt of all this problems, in theory replacing caps and varistors on an old psu should prolong its lifetime drastically.

Even a best security system is defeated with enough stupidity of its users. You can make much simpler and safe system for users that are knowledgeable, but if you have bunch of illiterate normies who will use even the best system, they will still spill the beans somewhere. 1. Keycards, encrypted pendrives -if user lose it because he/she is forgetful 2. Two factor authentication -same for forgetful users 3. Phishing, gaining personal information to reset the passwords, get clues -if users are naive, uninformed, stupid 4. Server storage methods that depends on it specialist input and mainteneace -negligence and laziness 5. Truly automated systems that can be circumvented by only inside knowledge -blackmail 6. Multiple layers, automated security systems -hardware spy chip hoisted that will nullify any encryption taking place Therefore a good system has multiple layers of security. The fact that they don't used such system means they are either stupid, or truth be told -the meeting didn't have REALLY important information after all

Get a cheap multimeter, put a serious load on your psu, and short the ground and 12v wires on spare pciex. Make sure to set it on direct current voltage setting[or your multimeter will blow :D]

some psu have thermal venting implemented, and it will keep spinning fans even after you turn off power to cool down PSU.

Check ram and cpu, the error say that checksum doesn't match, a compute error in either CPU or ram can do that. Im 99% sure its ram.

Im kinda scared to push into "limit" with furmark/prime95 but truth be told it only takes 355w of power during occt, and it already have some trouble regulating. Would never run it to a literal max of 550w or even 500w. Its a cheap psu after all Nephew wont do any of that, if it goes in smoke i just hope it wont kill his stuff.

I wouldnt put it over the rating, but if you will ran it to a max of 400w it should be fine. Some gpu like to trip protections with sudden power spike. Make sure you use ups to filter voltage spikes, the caps are old in this.

Download ryzen dram calc and trim some of these down. tRC and tFRC still have noticeable impact and you have them quite high.