Ryzen Owners Zen Garden

[Mussels]

Damn, using OCCT to test out a curve and seeing 80C with just two cores loaded

these chips realllllllly like to just go balls to the wall for whatever load you have

 

[freeagent]

I've been tuning my ancient 3000 series a little. It took about a month of tinkering and experimenting. Running 2 sticks and 4 sticks and playing with all of the buttons while keeping vdimm @ 1.5v or under though I was tempted..

I settled at 1.46v on four stix for this overclock

Man those 5000s look hot, and not just the temps but holy balls man they pack a punch.

 

[Mussels]

I feel like we need coolers that specifically cool the CCX part of hte chip, not the center

that and/or i need some better thermal paste (but shit, kryonaut is one of the best isnt it?)

Ahaha, core 3 is my weak one - at -30 on the curve OCCT is all HERRRRROOOO

 

[biffzinker]

I feel like we need coolers that specifically cool the CCX part of hte chip, not the center

Considering the die is Indium soldered to the heatspreader there shouldn't be a heat transfer issue other than the surface area of the die is small.

Thermal conductivity	81.8 W/(m·K)

 

[dgianstefani]

I'm going to change my cryonaut extreme to conductonaut on my CPU.

 

[tabascosauz]

Considering the die is Indium soldered to the heatspreader there shouldn't be a heat transfer issue other than the surface area of the die is small.

Thermal conductivity	81.8 W/(m·K)

Maybe Intel was onto something with the die thinning and thick IHS on Comet Lake. It sure improved temps across the board for that entire family of CPUs. Wouldn't be surprised if AMD seized the new AM5 socket opportunity to shake things up a bit, maybe different die footprints and and some of those IHS tricks.

Can't remember if it was bz or luumi who said it but at a certain point high clock frequency itself becomes as much a contributor to heat as voltage.

The extra die area on Renoir seems to help soak up some heat, so I'm keen to see when the Cezanne APUs come to desktop and what they can do. Since the iGPU part is going to be a complete borefest

 

[Space Lynx]

I'm so glad I got a stable all core OC of 4.625 at 1.350v - reading this thread gives me a headache i don't care if i lose out on single thread boosts, im quite happy with this.

 

[dgianstefani]

1.35v is high for static voltage.

 

[8BitZ80]

New owner of a Ryzen 5800x!!!! IT'S A HOT CPU! Wow lol. Performance is amazing, but even with a Noctua NH-U14S, and a Noctua ippc 2000 pwm, prime is insane (90C), cinebench gets low 80C, fan at 100%, speed of 4.475/4.490ghz all core, and I set a negative offset of 0.06v.

I am thing maybe if an arctic freezer II 240mm would help a bit !!

It's a shame the 5800X is so hot. Some owners can tolerate it but it would heatsoak my room far too quickly and make playing games uncomfortable. That's why I didn't buy one. Even my 5600X gets a bit warm after an hour or so and I have to turn a fan on.

 

[Space Lynx]

1.35v is high for static voltage.

i don't even break 60 celsius in games. and even w1zzard tpu review uses a 1.35v oc. its not that bad really. its my property so ill do as i please though, thanks

 

[dgianstefani]

Your loss.

 

[Space Lynx]

Your loss.

Do you have any evidence that 1.35v is too high for a zen 3 chip? every youtube review i saw of it used a similar voltage. people who say they are getting same overclocks at 1.28v or w.e are full of bs and not doing proper stress tests for stability.

 

[Durvelle27]

I feel like we need coolers that specifically cool the CCX part of hte chip, not the center

that and/or i need some better thermal paste (but shit, kryonaut is one of the best isnt it?)

Ahaha, core 3 is my weak one - at -30 on the curve OCCT is all HERRRRROOOO

I’ll try OCCT on mine as this far temps are very good sitting around 77*C loaded

 

[ChristTheGreat]

It's a shame the 5800X is so hot. Some owners can tolerate it but it would heatsoak my room far too quickly and make playing games uncomfortable. That's why I didn't buy one. Even my 5600X gets a bit warm after an hour or so and I have to turn a fan on.

I see that the 5900x is colder than the 5800x, du to CCX. The CPU gets hot, but, it looks like if heat stays in the CPU and not beeing dissipate.

I know, all my ryzen snce my 2600x (3600x, 3700x, 5800x), they were all host with NH-U12S and NH-U14S. Repasted ALOT of time, used MX4, NT-H1, NT-H2.

Sometimes, I'm thinking if it'S not my ROG Strix x470-F that report over?

 

[8BitZ80]

Finally got an MSI Afterburner beta version that shows individual CPU temperatures, clocks, usage and power draw for each core on Zen 3 processors.

Not sure how long this 4.6.3 Beta 4 version has been available, but somehow I didn't find it until now.

 

[Alyjen]

Oh! I need to check this, I was able to work around it (there were some guides how to add this support) but the results were lacking.

 

[tabascosauz]

I see that the 5900x is colder than the 5800x, du to CCX. The CPU gets hot, but, it looks like if heat stays in the CPU and not beeing dissipate.

I know, all my ryzen snce my 2600x (3600x, 3700x, 5800x), they were all host with NH-U12S and NH-U14S. Repasted ALOT of time, used MX4, NT-H1, NT-H2.

Sometimes, I'm thinking if it'S not my ROG Strix x470-F that report over?

What does HWInfo's Power Reporting Deviation metric say when you're in the middle of an all-core benchmark?

I think more of the 300- and 400-series motherboards are prone to underreporting power draw from being older, creating higher boost clocks but more current and power draw and heat; a lot of the X570/B550 boards came out after the community discovered the intentional telemetry underreporting.

I can understand the 5800X and 3600X, but a 3700X that properly reports current draw and respects power limits is a pretty cool running chip in the context of the other chips.

 

[ChristTheGreat]

What does HWInfo's Power Reporting Deviation metric say when you're in the middle of an all-core benchmark?

I think more of the 300- and 400-series motherboards are prone to underreporting power draw from being older, creating higher boost clocks but more current and power draw and heat; a lot of the X570/B550 boards came out after the community discovered the intentional telemetry underreporting.

I can understand the 5800X and 3600X, but a 3700X that properly reports current draw and respects power limits is a pretty cool running chip in the context of the other chips.

I'll check!

I did a few test, installed ryzen master, and started an manual overclock. 3.8ghz start, at 1.15v, no issue, low 60C temp Cinebench (suprised?? LOL)

4.5ghz @ 1.25v isn't stable, ran prime and after a few sec, a couple of core failed.

if I leave stock (I am running 0.04v negative offset), everything is stable but high temp. vcore is like between 1.35 and 1.4v to reach 4.475ghz. I can try to reduce voltage for the all core clock. Right now, -0.06v can crash because of PBO.


Maybe I have again, a bad clocker, like 99% of all my CPU I bought in my life (the only one I had great, was a pre binned opteron 165 hahaha)

 

[tabascosauz]

Anyone else get these low L3 read results? Could've sworn I've seen something along the lines of this somewhere in one of the billion Zen 3 threads on the forum.

I get this most of the time at 4000 and 4200. I think I got normal (550-650GB/s) L3 read like, twice, out of 10 runs. Upping SOC doesn't change anything, the memory itself is stable and so is the IF.

Next step I gotta see if it still does this at 3800.

I don't think I've ever seen this on my 3700X, but then again I can only run 3600 on there.

 

[biffzinker]

Anyone else get these low L3 read results? Could've sworn I've seen something along the lines of this somewhere in one of the billion Zen 3 threads on the forum.

My 3800X at 42.75/43 gets this for the L3 Cache.

 

[Mussels]

i have no idea whats good or bad, and i didnt quit all background tasks

edit: quit all my RGB stuffs. these scores seem to be all over the place? lots of variance in this i guess.

 

[thesmokingman]

Finally got an MSI Afterburner beta version that shows individual CPU temperatures, clocks, usage and power draw for each core on Zen 3 processors.

Not sure how long this 4.6.3 Beta 4 version has been available, but somehow I didn't find it until now.

View attachment 185489

You shouldn't use it for monitoring anyways. Use hwinfo and set it to send whatever sensors to OSD via RTSS. That said you should still use AB for oc'ing your gpu.

 

[dgianstefani]

i have no idea whats good or bad, and i didnt quit all background tasks

That's... Really bad. I'm sorry to say. Your latency should be under 60 ideally, mine is 54. Your read and copy speeds should be above 60gb/s. L3 cache is also slower than it should be.

 

[Mussels]

That's... Really bad. I'm sorry to say. Your latency should be under 60 ideally, mine is 54. Your read and copy speeds should be above 60gb/s. L3 cache is also slower than it should be.

Feel free to recommend me some settings, i have weird RAM - Hynix something

clean boot not much better

This is 2x32GB 3600 C18 sticks, you may be expecting too much from them

 

[thesmokingman]

Anyone else get these low L3 read results? Could've sworn I've seen something along the lines of this somewhere in one of the billion Zen 3 threads on the forum.

I get this most of the time at 4000 and 4200. I think I got normal (550-650GB/s) L3 read like, twice, out of 10 runs. Upping SOC doesn't change anything, the memory itself is stable and so is the IF.

Next step I gotta see if it still does this at 3800.

I don't think I've ever seen this on my 3700X, but then again I can only run 3600 on there.

The latency is good but the load on the imc is probably too much hence the really poor L3 perf.

i have no idea whats good or bad, and i didnt quit all background tasks

edit: quit all my RGB stuffs. these scores seem to be all over the place? lots of variance in this i guess.

Base Ryzen 5k latency 62ns. You're well above that and yea Hynix or other. I believe base memory speed is 3200mhz and cas14. As for timings... no idea for that ram. You'll have to play with dram calc, and use thaiphoon to export your xmp to dram calc so you can get medieval on your timings.

Here's my slightly tightened 4000mhz C17 2x16g b-dies running 3600mhz. The latency is ok-ish at best but the cache perf is really good.