Ryzen Owners Zen Garden

[MxPhenom 216]

TM5 with the anta777 extreme1 .cfg is basically the standard yes. Some like using 1usmus v3 .cfg instead, I stick to anta. It's one of the quickest tests.

But you have to keep running it manually every 1h20m or so which makes it a bit of a pain. So I only use it for testing during the day. For overnight I use HCI Memtest Pro, paid $5 for it I think. A lot more convenient than the free HCI, though that still has all the functionality that matters.

Overall anta777 seems more aggressive than HCI. Therefore usually more effective.

But relying on TM5 alone is stupid, like any other test alone, regardless of how good it is. If you want daily stability, always run a bunch of different tests.

e.g. a few runs of TM5 anta + 1000-2000% HCI coverage + LinpackXtreme or Prime95 Large FFT

Even TM5 certainly isn't foolproof. It's better than HCI at detecting most timings, but is fucking terrible at evaluating tRFC on APUs. Vega iGPU are very sensitive to tRFC, so while it might pass any # of TM5 runs at 133ns it'll instantly bluescreen a game.

Fuck, i just had my buddy with a 3700x and first gen ddr4 shit box memory run the membench and got something like 90ns latency and his time was 145s. But can't even get through first 2 tests of TM5. But has no issues in games or anything.

His settings are 3200 at CAS18. Idk how that system is even functional lol. He upgraded from a 6700k and kept the RAM.

 

[tabascosauz]

Fuck, i just had my buddy with a 3700x and first gen ddr4 shit box memory run the membench and got something like 90ns latency and his time was 145s. But can't even get through first 2 tests of TM5. But has no issues in games or anything.

His settings are 3200 at CAS18. Idk how that system is even functional lol. He upgraded from a 6700k and kept the RAM.

You'd be surprised what 2133 jedec looks like in membench then i had one of my friends run his memory through membench when he upgraded 1600(AE) to 3700X. His time in Easy was slower than my time in Normal (supposed to be about double, test is twice the coverage), and that was when I was on CJR

Yet there's not really anything gamebreaking about his gaming performance. Whereas there definitely something off about it when he was using R5 1600......Just goes to show that as shitty as the cheapo chiplet design is, a healthy dollop of L3 goes a long way

I'm guessing if it's that unstable in TM5 it probably reboots or crashes from instability once every now and then and he just thinks it's a game bug. Happened to me with Arma, always thought it was the game. OS corruption will come for him eventually.

 

[freeagent]

Oh yeah I still like Linpack Xtreme too, its pretty good.. just gotta make sure your cooling is good. SuperPi 32m is also pretty decent.. I find myself running OCCT these days, seems kind of realistic.. but I dont take its word for gospel.. just another program to test with. I use some old benchmarks because they are still relevant, and they help me determine if my settings are beneficial or just for show. I wouldn't try to daily that 4466 shot.. because C17 is like eww. And one more bump up to 4533 wouldn't post even with C18 so not sure if that's the ram limit or the board. Not bad for 3200C14 though. I could probably daily 4400.. but I didn't test too much. I have it in my head that Intel is better with memory, but I haven't actually run a new one to know for myself.

 

[tabascosauz]

I have it in my head that Intel is better with memory, but I haven't actually run a new one to know for myself.

Certainly not the garbage that is rocket lake lol

Comet Lake is the dream for B-die though. 4800 CL17 daily with no Gear2 bullshit yes please. They say 5700G can also do that 1:1 with an dGPU, but how stable idk and it barely matches even rocket lake on latency

Gear2 is such a meme, I wasn't aware that it actually affects command rate. On AMD being desync will never affect GDM/2T, but on Intel being Gear 2 turns 1T into effective 2T and 2T into effective 4T

I'm still surprised at how cool Comet Lake is. People running 5GHz all-core 10900K on my C14S in cases smaller than mine, still holding their own on temp under stress tests. Meanwhile I'm out here barely managing all-core below 80C at stock

 

[freeagent]

Comet Lake is the dream for B-die though. 4800 CL17 daily with no Gear2 bullshit yes please. They say 5700G can also do that 1:1 with an dGPU, but how stable idk and it barely matches even rocket lake on latency

Yes that's the Intel I mean lol, forgot about the meme though.. the newest Intel I have run is 3rd gen.. not very indicative to what's going on it the real world lol. And I seem to forget about those APU's.. not too many people run them that's probably why.

I didn't think Gear 2 affected command rate either.. I figured they brought dividers back, and you could mess around with it like X58. 2T into 4T is a bit of a letdown.. I didn't really look into it much as I was exploring what this thing can and cant do. I was also never able to get 5GHz stable with Intel, believe me I tried.. since like 2008.. on a bunch of different CPU's lol. Main reason why I got the 5900X.. I wanted that 5GHz lol..

So stupid I know. Whatever

 

[MxPhenom 216]

Just got an update for HWiNFO to v7.10 and the sensors are much more informative now, rather than having generic names like it was in the previous version i had:

EC_TEMP1 is my water coolant temp, before it was just labeled as Temperature 2:

 

[tabascosauz]

There's some pretty cool features in v7.1 that I'll be keeping an eye on.

I'd be surprised if that IO die ever goes above 45C though. Haven't seen it exceed 23W or so ever. It's not like it's got an iGPU to support. So far I've gotten it up to about 40C in HCI.

That active core count indicator will be intriguing to compare between games.

 

[Taraquin]

So after a week of stability with curve optimizer and +200 pbo and - 30 on all cores except one with - 28 I got a random reboot while moneromining (I use 45W ppt and voltage stays at 850mv, but get a few rare spikes to 1200-1300mv during mining, temp is at 52-55C. I used core cycler to check stability, and it quickly found the one bad core, hence the - 28 on one, but apparently it's a tad unstable still. What other programs/settings would you use to check CO stability? I kniw ram, IF etc is stable since I had no crash/reboot for over 2 months with - 30 CO while using +50 pbo. Only changed variable is the extra 150 pbo.

 

[dont whant to set it"']

Is it paying for the power drawn?

 

[droopyRO]

A fatal hardware error has occurred.

Reported by component: Processor Core
Error Source: Machine Check Exception
Error Type: Cache Hierarchy Error
Processor APIC ID: 8

I have been getting this reboots from idle or standby. The errors look like this, just the APIC ID varies, it could be 0, 3, 5, 8 etc. I have been using Google to find an answer after i tried a lot of things, including different RAM sticks, BIOS update and downgrade, updating AMD drivers etc.. No matter what game or app or stress test i throw at it, it is stable. The moment i leave the PC at idle and go do something else for 10 minutes i am greeted with the system restarted and the error above in Event Viewer.

Some people suggested it was the CPU(mine is a Ryzen 5600X), and that they RMA'ed it, and that fixed it for them, all were 5xxx series. Have any of you encounterd this issue ? i do have the old 2700X i could put in place. But at the same time i want to exhaust all possible fixes. Thanks.

 

[Taraquin]

Is it paying for the power drawn?

Yeah, I earn about 0.4usd a day, idle consumption is 20watts anyways so since I'm gpu mining adding the cpu using 45W costs me about 0.6kW extra energy a day, which costs me 0.1usd in electricity.

 

[tabascosauz]

I have been getting this reboots from idle or standby. The errors look like this, just the APIC ID varies, it could be 0, 3, 5, 8 etc. I have been using Google to find an answer after i tried a lot of things, including different RAM sticks, BIOS update and downgrade, updating AMD drivers etc.. No matter what game or app or stress test i throw at it, it is stable. The moment i leave the PC at idle and go do something else for 10 minutes i am greeted with the system restarted and the error above in Event Viewer.

Some people suggested it was the CPU(mine is a Ryzen 5600X), and that they RMA'ed it, and that fixed it for them, all were 5xxx series. Have any of you encounterd this issue ? i do have the old 2700X i could put in place. But at the same time i want to exhaust all possible fixes. Thanks.

Stock CPU? Assuming stock,

Cache hierarchy is core-related, APIC identifies the core#. I get these occasionally when using unstable Curve Optimizer settings. But if it's idle-related:

• Disable global C-states, or
• Try using Curve Optimizer in the other direction by adding positive offset
• Try different BIOSes

First two are easy but Band-Aids that don't solve the real problem.

If you haven't touched any kind of OC at all and you don't mind RMA, just send it for warranty. Early Ryzen 5000 and some Ryzen 3000 were plagued with cache hierarchy, that went away around early 2021. Sometimes the board vendor will fix their shitty BIOS and solve the problem, but severe cases across a lot of cores should just be RMA

If you're not running stock settings, obviously go back to stock

 

[droopyRO]

Yeah stock settings. Just XMP enabled. I tried curve optimizer -5 all core and to disable it, also tried with PBO off. It started happening about a month ago, it was fine up until then. As said above i tried to up and downgrade my BIOS, F34, F35 and the latest F36c. My CPU was bought at launch i got it in November 2020. I think i will put the "old" 2700X back and see if it dose the same and RMA the 5600X :/

PS: it did it again a few minutes ago. I left the PC for about 10 minutes.

 

[tabascosauz]

Surprised to see 5GHz in daily usage, not a bug, and not benchmarks at unstable curve optimizer settings:

For a poop quality CPU it is regularly clocking higher every day after fixing my C14S mount. Obviously runs cooler too without changing anything else. So perhaps it really would be a 5GHz chip all the time with an AIO?

 

[PaulieG]

Surprised to see 5GHz in daily usage, not a bug, and not benchmarks at unstable curve optimizer settings:

View attachment 217939

For a poop quality CPU it is regularly clocking higher every day after fixing my C14S mount. Obviously runs cooler too without changing anything else. So perhaps it really would be a 5GHz chip all the time with an AIO?

Over that C14S? Only with one of the top AIO's. One of the EK's or Arctic Cooling.

 

[freeagent]

I don’t think my chip is stellar either brotha! But I don’t think yours is poop. I bet it would run just like mine

 

[LifeOnMars]

I need help with squeezing the last bit of juice out of my system now that I have adequate cooling. I have a basic 4.2Ghz overclock on my 3600(old chip not the higher clocking newer ones) and my RAM is just running standard XMP timings cl16@3200. So I'm just looking to push the ram a bit and the IF as I have a little room to grow fps wise with the 3060 Ti@1080p. Not sure if I'll get much out of my board as it's only a Matx Asrock B450 but looking for some guidance on where to start guys please.

 

[QuietBob]

I have a basic 4.2Ghz overclock on my 3600

What voltage do you need for 4.2 all core? What temps do you get with the Vetroo V5 after half an hour in Cinebench/Blender?

 

[freeagent]

My 3600XT did 4400 all core with 1.2875v and 4500 with 1.3375v. For all loads I probably wouldn't go further than 1.2875..

 

[LifeOnMars]

What voltage do you need for 4.2 all core? What temps do you get with the Vetroo V5 after half an hour in Cinebench/Blender?

Don't know about cinebench/blender but it tops out at 1.309v/Average of 1.15v after a 9hr Prime 95 Blend test. Temps maxed out at 80c with all my fans on quiet. In intensive games using all cores to a very high percentage per core it hovers around 64c. Happy with the stability of CPU just looking for RAM and IF clocking.

My 3600XT did 4400 all core with 1.2875v and 4500 with 1.3375v. For all loads I probably wouldn't go further than 1.2875..

Yeh the newer chips will do those clocks with ease at good voltages but I have an OG 3600, they all generally topped out at 4.2 needing slightly higher volts.

 

[Kovoet]

I went from a hassle free i7 7700k with Asus Maximus Formula to a AMD 5900x and the Hero VIII. I like it in game and everything except the start. Intially I throught it was the ram which I had 3200mhz corsair sticks to now having Viper 3733mhz and still got the same problem. But as I mentioned in game briilant and Saturday installing my new 3080Ti

 

[QuietBob]

Don't know about cinebench/blender but it tops out at 1.309v/Average of 1.15v after a 9hr Prime 95 Blend test. Temps maxed out at 80c with all my fans on quiet. In intensive games using all cores to a very high percentage per core it hovers around 64c. Happy with the stability of CPU just looking for RAM and IF clocking.

80c in Prime is OK, you shouldn't be thermally limited then. I assume you set a fixed voltage value for the CPU? Prime is good for testing cooling capacity, but your CPU may be able to attain higher clocks with more realistic workloads.

To the point, you'll need to decouple RAM and IF and test each separately. I'd suggest finding your max stable IF clock first. What are your current SOC, VDDP and VDDG voltages?

My 3600XT did 4400 all core with 1.2875v and 4500 with 1.3375v. For all loads I probably wouldn't go further than 1.2875..

Yeah, the XTs are top bins. My 3300X does 4.5 @ 1.275, but not in Prime. Yet it's stable otherwise.

 

[HD64G]

I went from a hassle free i7 7700k with Asus Maximus Formula to a AMD 5900x and the Hero VIII. I like it in game and everything except the start. Intially I throught it was the ram which I had 3200mhz corsair sticks to now having Viper 3733mhz and still got the same problem. But as I mentioned in game briilant and Saturday installing my new 3080Ti

Increasing a bit the SOC voltage should solve the problem. We have the same RAM.

 

[Kovoet]

Thanks will give that a bash

 

[tabascosauz]

Over that C14S? Only with one of the top AIO's. One of the EK's or Arctic Cooling.

I don’t think my chip is stellar either brotha! But I don’t think yours is poop. I bet it would run just like mine

Kinda tempted to temporarily buy a D15 from Amazon and see exactly how far off I am in performance, exact same brackets I think, I can fit it as long as I don't put the side panel on but the rational side of my brain is screaming after the failed outcome of the last CPU cooler escapade

When the C14S has two fans, the two coolers weigh in about the same at 1.3kg

@freeagent was beating me by like 10c with his FC140 but he blasts air through the case at 3000rpm so he's a sneaky cheater