Ryzen Owners Zen Garden

[Space Lynx]

Humor me. Run OCCT for an hour both on cpu and memory test and see if you get any WHEA errors. I tested over 4000/2000 just like you did here running for hours successfully but failed OCCT with around 100 WHEA errors after 45 minutes on both tests.

running world of warcraft for even just 15 minutes will detect errors that memtest, 3dmark tests, and cinebench didn't detect.

 

[tabascosauz]

running world of warcraft for even just 15 minutes will detect errors that memtest, 3dmark tests, and cinebench didn't detect.

Well yeah lol......there is no "test" out there for IF WHEAs......that's much less a reflection on WoW as a stress test and more on the frustration of troubleshooting something that's a pain in the ASS that can't be tested

Though in all fairness, gaming does seem to be the closest you can get to a "test" for it. My 3700X's Infinity Fabric HATED Insurgency Sandstorm. It isn't even that demanding of a game on the CPU.

any recommendations for me tightening my b-die ram? 2x16gb 3200 cas 14-14-14 OC'd to 4000 2000 fclk 16-16-16

what values should i try to mess with a bit to tighten it a little?

try tRP+tRAS for an easy to achieve tRC

tRFC needs to be 160ns or lower at the very least if you don't want to torpedo your performance, which is tRFC 320 or lower at 4000 (I think 4000 is the only speed where tRFC is literally tRFC ns x 2)

hard to tell anything if you don't provide a zentimings picture

 

[freeagent]

That's the thing though, I had no crashing, no instability, no windows error logs, nothing but great scores and lickety split performance. I used Linpack Xtreme, Memtest, Super Pi etc, everything was faneffing tastic.. till occt trolled my hood. Now I have to dial things back a bit.. spoil my good mood lol.. Ahh that's ok. It works, and its still faster than my 3770K.

 

[Space Lynx]

Well yeah lol......there is no "test" out there for IF WHEAs......that's much less a reflection on WoW as a stress test and more on the frustration of troubleshooting something that's a pain in the ASS that can't be tested



try tRP+tRAS for an easy to achieve tRC

tRFC needs to be 160ns or lower at the very least if you don't want to torpedo your performance, which is tRFC 320 or lower at 4000 (I think 4000 is the only speed where tRFC is literally tRFC ns x 2)

hard to tell anything if you don't provide a zentimings picture

I didn't like the 4000 OC, it was stable but eh I don't want to push it that hard. if you can help me tighten my 3600 cas 14-14-14 up though would be helpful. nothing hardcore. i want a safe tightening. just tell me what numbers to change and I will give it a shot

 

[tabascosauz]

I didn't like the 4000 OC, it was stable but eh I don't want to push it that hard. if you can help me tighten my 3600 cas 14-14-14 up though would be helpful. nothing hardcore. i want a safe tightening. just tell me what numbers to change and I will give it a shot

How much VDIMM are you pushing? For [relatively] "safe", try these at 3600:



Do any tRC between (inclusive) 50 and 60.

Then try a tRFC of 300 to start.

Can't really know for sure if you'll need more VDIMM, because if your current VDIMM is "barely" stable for those super loose timings, anything respectably tight might need more VDIMM. But I don't know what VDIMM you're running.

 

[freeagent]

If it helps I can use these settings up to 1900/1900

I can feels the dogs waiting to attack..

 

[Gmr_Chick]

Here's mine. I'm assuming it's running @ stock

 

[freeagent]

View attachment 191777

Here's mine. I'm assuming it's running @ stock

Fit as a fiddle

 

[dont whant to set it"']

I did give OCCT a try and found errors it did witch I ruled out to be some RAM timmings, for all other real world testing my go to is to start two games and alt-tab between the two after a mission (or missions) in each : SC2 and GTA V with some cine R23 here and there trown in for a mix.

 

[tabascosauz]

New HWInfo now v7.0! Everybody spam that Update button

One thing I noticed right off the bat is that Renoir iGPUs are showing a new voltage reading called VDDCR_GFX. I'm guessing it's derived from VSOC. Looks like it maxes out at about 0.981V, which makes a lot more sense than the previous "voltage" that topped out at 1.419V - clearly neither VSOC nor the actual voltage fed to the iGPU.

Oh and tooltips galore for every field, much info to look at

 

[Space Lynx]

New HWInfo now v7.0! Everybody spam that Update button

One thing I noticed right off the bat is that Renoir iGPUs are showing a new voltage reading called VDDCR_GFX. I'm guessing it's derived from VSOC. Looks like it maxes out at about 0.981V, which makes a lot more sense than the previous "voltage" that topped out at 1.419V - clearly neither VSOC nor the actual voltage fed to the iGPU.

Oh and tooltips galore for every field, much info to look at

View attachment 191804

just downloaded the portable variant of it, is there any benefit to 64 bit over 32 bit? i just use 32 bit because i figure it uses less resources but achieves the same goal... no idea though thats just me guessing all these years LOL

 

[Mussels]

just downloaded the portable variant of it, is there any benefit to 64 bit over 32 bit? i just use 32 bit because i figure it uses less resources but achieves the same goal... no idea though thats just me guessing all these years LOL

probably next to no difference, 32 bit needs to be emulated on 64 bit OS and likely uses more resources

 

[Octopuss]

I love Hwinfo. If I wasn't broke I would buy it right away even though there's no need to. I've been using it, holy shit!, since the DOS days, and I just love it.

 

[INSTG8R]

I love Hwinfo. If I wasn't broke I would buy it right away even though there's no need to. I've been using it, holy shit!, since the DOS days, and I just love it.

Like me and AIDA64 I’ve been paying for it since it was Everest

 

[PaulieG]

Fyi, newly released OCCT 8.0 now officially has stability testing built in for curve optimizer users. I guess that's both good and bad for some of us.

 

[freeagent]

Fyi, newly released OCCT 8.0 now officially has stability testing built in for curve optimizer users. I guess that's both good and bad for some of us.

Got it yesterday

I nailed 1933 down but hurt my back this morning so not going to be hanging out at the desk much today.

 

[PaulieG]

Got it yesterday

I nailed 1933 down but hurt my back this morning so not going to be hanging out at the desk much today.

Sorry to hear about your back. Feel better man.

 

[Octopuss]

Fyi, newly released OCCT 8.0 now officially has stability testing built in for curve optimizer users. I guess that's both good and bad for some of us.

Not very useful with 1 hour limit

 

[PaulieG]

Not very useful with 1 hour limit

Simply not true. OCCT will throw errors or fail if your OC is not stable, and it doesn't need an hour. I'm pretty confident about this... but prove me wrong.

 

[Octopuss]

I presume the concept is similar to Prime95. If a test fails right away or within minutes, your settings are very, very wrong and barely past the point of booting successfully. You need at least a few hours to make sure it's stable.
It's like with the Linpack. Just because you do a few passes that take 10 minutes doesn't mean the PC is stable.

 

[freeagent]

No not similar to prime or anything else really. You can use it like that sure, you can generate some impressive heat. But its not like other stress tests that I have used before. I haven't used it for 15 years or so because it wasn't like prime or similar. I kinda like it, I wouldn't pay for a subscription though. I would consider the option to purchase if it was like 5 bucks or something. I bet an AGESA update could fix the high speed errors. Its so close.

 

[Mussels]

What he's saying is that for whatever reason, OCCT throws errors faster (especially WHEA) so 1 hour of it might equal 10 others of another program

Use OCCT and its 1 hour test to start, then throw your other tests at it later

 

[PaulieG]

STABLE

 

[freeagent]

STABLE

Thanks for posting your zen timings shot! I was looking to see what was different other than timings and noticed the voltage.. I copied those settings and has been running for 10 minutes, usually spits errors within seconds.. yay

 

[Octopuss]

I would test cores individually, because all core boost is always lower than single, thus possibly "hiding" instability with at higher frequencies.