Tuning RAM question

[3x0]

1usmus until you weed out the errors since there's documented reasons for each error for that config, afterward anta777 extreme and Universal profiles to verify stability

50 cycles of 1usmus is the "gold standard"

 

[lance36]

Fan suggestions, I would prefer something that's not outrageous in style but I prefer function so if it's ridiculous but it performs ridiculously then I'll be ridiculous. @1.47v it'll boot but WHEA errors. I then messed around with the settings that have an impact on phase and power quality. No cigar, however Tfrc did impact temps. Loosening timings only caused it to reach 50C slower.... So it's fan time.

 

[3x0]

WHEA 19 errors are usually related to VSOC voltage, VDDP/G or CPU 1.8v (either too high voltages or too low). I wouldn't get a fan unless you're getting specific errors in TM5. Or rather, ghetto a normal fan over the RAM and see if it makes a noticeable difference in what timing you can achieve before spending money on a RAM fan

 

[lance36]

I made a custom view for WHEA errors, I'll check it when I get home.

 

[tabascosauz]

Fan suggestions, I would prefer something that's not outrageous in style but I prefer function so if it's ridiculous but it performs ridiculously then I'll be ridiculous. @1.47v it'll boot but WHEA errors. I then messed around with the settings that have an impact on phase and power quality. No cigar, however Tfrc did impact temps. Loosening timings only caused it to reach 50C slower.... So it's fan time.

Don't get dedicated RAM coolers, they don't work well at all. I had the Corsair single fan one and I can get similar results out of just having an exhaust fan in the top of the case.

Just use any old 92mm, 120mm fan and stand it up on the GPU. Your temps look fine in 1usmusv3.

1usmus is a good fast test. For longer test I do 15 cycles of anta777 (edit config file to change cycles), but you can do the same for the 1usmus test

Temp destabilization should be quite obvious (as in, errors suddenly start appearing during test once past temp threshold). But it is always good practice to have a separate fan on AIO/water.

Bus/Interconnect WHEAs have nothing to do with RAM stability, it's Fabric stability. Mostly need to up your VSOC and/or VDDG_IOD, some say higher PLL helps

 

[lance36]

Don't get dedicated RAM coolers, they don't work well at all. I had the Corsair single fan one and I can get similar results out of just having an exhaust fan in the top of the case.

Just use any old 92mm, 120mm fan and stand it up on the GPU. Your temps look fine in 1usmusv3.

1usmus is a good fast test. For longer test I do 15 cycles of anta777 (edit config file to change cycles), but you can do the same for the 1usmus test

Temp destabilization should be quite obvious (as in, errors suddenly start appearing during test once past temp threshold). But it is always good practice to have a separate fan on AIO/water.

Bus/Interconnect WHEAs have nothing to do with RAM stability, it's Fabric stability. Mostly need to up your VSOC and/or VDDG_IOD, some say higher PLL helps

What setting is ppl? I haven't seen that. I always stop the tests at 50c. So I haven't seen it become unstable. Should I let it go and see how far the temp's climb. I stopped because I didn't wanna damage the ram.

 

[tabascosauz]

What setting is ppl? I haven't seen that. I always stop the tests at 50c. So I haven't seen it become on stable. Should I let it go and see how far the temp's climb. I stopped because I didn't wanna damage the ram.

High temps don't damage B-die, it just becomes unstable. Keep it going and find out where your temp threshold for your current settings is.

I think it's called CPU_PLL, but different companies call it different things. On my Unify-X there are also 2 different PLLs......try to solve WHEAs through VSOC and VDDG_IOD first. I've never seen VDDP fix WHEAs, it should be fine just left between 0.9-1.0V.

 

[3x0]

What setting is ppl?

CPU 1P8. Not named the same across MoBo vendors, but its default value is 1.8v

 

[lance36]

When I get home I'll start my testing.

 

[tabascosauz]

CPU 1P8

What setting is ppl? I haven't seen that. I always stop the tests at 50c. So I haven't seen it become unstable. Should I let it go and see how far the temp's climb. I stopped because I didn't wanna damage the ram.

He doesn't have an MSI board

Asus in AGESA 1206:

Unfortunately, I know Asus intentionally hides a lot of adjustments for PRIME/TUF/STRIX. Been a while since I had my TUF mATX, not sure if it has PLL

 

[lance36]

He doesn't have an MSI board

Asus in AGESA 1206:

View attachment 268038

Unfortunately, I know Asus intentionally hides a lot of adjustments for PRIME/TUF/STRIX. Been a while since I had my TUF mATX, not sure if it has PLL

I have the latest bios but if it doesn't have it lame. I've been slowly upgrading as prices have dropped so maybe a aboard is next, we'll after gpu

He doesn't have an MSI board

Asus in AGESA 1206:

View attachment 268038

Unfortunately, I know Asus intentionally hides a lot of adjustments for PRIME/TUF/STRIX. Been a while since I had my TUF mATX, not sure if it has PLL

What do you think?

 

[3x0]

That's the one

 

[lance36]

Okay, what's the max? And it's for stabilizing the infinity fabric correct? Oh and at what temps should I stop testing to not damage the ram?

 

[tabascosauz]

Okay, what's the max? And it's If to stabilize the infinity fabric correct? Oh and at what temps should I stop testing to not damage the ram?

Try running 1.9V. Some gone up to 2.0V but 1.9V should be safe

What's your other voltages look like right now? Grab em from Zentimings

Need to make sure your WHEAs are actually Bus/Interconnect, otherwise not even Fabric related

 

[lance36]

Try running 1.9V.

What's your other voltages look like right now? Grab em from Zentimings

Let me finish this test, I changed vdimm to 1.47, soc to 1.125 CCD and IOD to 1.0 and CLDO to I think .960 or .980. 7 minutes in and an error temps are stable at 46.8c

Alright next step is pll, I ran it for 10 minutes with one error. I would like to clean that up. Temps did get to 48c. BUT no crash at 1.47v. I'll post a pick of all my voltages from bios.

Try running 1.9V. Some gone up to 2.0V but 1.9V should be safe

What's your other voltages look like right now? Grab em from Zentimings

Need to make sure your WHEAs are actually Bus/Interconnect, otherwise not even Fabric related

 

[tabascosauz]

Let me finish this test, I changed vdimm to 1.47, soc to 1.125 CCD and IOD to 1.0 and CLDO to I think .960 or .980. 7 minutes in and an error temps are stable at 46.8c

Alright next step is pll, I ran it for 10 minutes with one error. I would like to clean that up. Temps did get to 48c. BUT no crash at 1.47v. I'll post a pick of all my voltages from bios.

What errors? WHEAs or errors from TM5? The two are not the same.

VSOC, VDDG_IOD, PLL are to address Fabric stability - WHEAs. They don't help you with TM5 errors (generally).

RAM stability you fix by improving temps, more VDIMM, or working on your timings.

I think you just need to continue this when you're able to get a hold of a fan.

 

[lance36]

No more WHEA errors TM5. I bumped the 1.8 to 1.85 and trying again

 

[tabascosauz]

No more WHEA errors TM5. I bumped the 1.8 to 1.85 and trying again

read my last reply

I thought you said you loosened tRFC?? Tighter tRFC makes temp sensitivity worse

90% convinced that what you're seeing is temp sensitivity, not a lack of VDIMM. We have the same bin of kit, and 1.47V is high for 3600CL14 even considering your tRFC and 14-13-13.

If you run 1.5V and still get errors in TM5, it's 100% temp sensitivity, and you should just stop until you can point a fan at the RAM.

50C is just an estimate. Some people can run up to 55C without destabilizing, some can barely do 40C.

 

[lance36]

What errors? WHEAs or errors from TM5? The two are not the same.

VSOC, VDDG_IOD, PLL are to address Fabric stability - WHEAs. They don't help you with TM5 errors (generally).

RAM stability you fix by improving temps, more VDIMM, or working on your timings.

I think you just need to continue this when you're able to get a hold of a fan.

I was getting zero TM5 errors at 1.48v just 50c.I wonder and I'm no expert but if my low trfc is causing IF issues. If I increase the IF at my current timings or older ones, from the start of this post it caused memtests to error. However you have much more experience than I do, so your most likely correct. I want to try a little more testing before calling it quits

 

[tabascosauz]

I think that I was getting zero TM5 errors at 1.48v just 50c.I wonder and I'm no expert but if my low trfc is causing IF issues. If I increase the IF at my current timings or older ones, from the start of this post it caused memtests to error.

tRFC has nothing to do with IF. I just told you. Just run looser tRFC for now if it's causing errors.

1usmusv3 can provide some indication as to what error is what:

1usmus Error Diagnosis - Google Sheets

 

[lance36]

tRFC has nothing to do with IF. I just told you. Just run looser tRFC for now if it's causing errors.

1usmusv3 can provide some indication as to what error is what:

1usmus Error Diagnosis - Google Sheets

idk the only thing I changed was the PLL from the attempt that got a TM5 error. I want to push the IF higher I'm gonna bump it to the next freq up and try again. The only reason I mentioned tRFC is because I had to loosen them to get 3733 to boot. This was after I changed them from 288 to what they are now.

 

[tabascosauz]

idk the only thing I changed was the PLL from the attempt that got a TM5 error. I want to push the IF higher I'm gonna bump it to the next freq up and try again. The only reason I mentioned tRFC is because I had to loosen them to get 3733 to boot. This was after I changed them from 288 to what they are now.

Again.........RAM =! Fabric, PLL has nothing to do with RAM stability

Measure tRFC in terms of ns. 272 tRFC @ 3600 (~152ns?) is not the same as 272 @ 3733 (~143ns?). So obviously you would have to bump VDIMM with that tRFC ns jump, in addition to having to bump VDIMM anyway because of the speed increase

If you're just running default 3 cycle 1usmusv3, getting an error in one run and none in the next doesn't mean a lot, not when temps are this high.

 

[lance36]

Again.........RAM =! Fabric, PLL has nothing to do with RAM stability

Measure tRFC in terms of ns. 272 tRFC @ 3600 (~152ns?) is not the same as 272 @ 3733 (~143ns?). So obviously you would have to bump VDIMM with that tRFC ns jump, in addition to having to bump VDIMM anyway because of the speed increase

If you're just running default 3 cycle 1usmusv3, getting an error in one run and none in the next doesn't mean a lot, not when temps are this high.

Got u,I think, thanks for being patient. I feel like this is still progression

Again.........RAM =! Fabric, PLL has nothing to do with RAM stability

Measure tRFC in terms of ns. 272 tRFC @ 3600 (~152ns?) is not the same as 272 @ 3733 (~143ns?). So obviously you would have to bump VDIMM with that tRFC ns jump, in addition to having to bump VDIMM anyway because of the speed increase

If you're just running default 3 cycle 1usmusv3, getting an error in one run and none in the next doesn't mean a lot, not when temps are this high.

Alright, I have a goal. I'm not sure if I can do it but with you I have the best chance. I'll dig up a fan tomorrow. I'll shortly test and post the errors that I get from what's in the pics.

That didn't take long

 

[tabascosauz]

Got u,I think, thanks for being patient. I feel like this is still progression


Alright, I have a goal. I'm not sure if I can do it but with you I have the best chance. I'll dig up a fan tomorrow. I'll shortly test and post the errors that I get from what's in the pics.

Eventually running long anta777 when you get a fan will be good, but if you can at least pass 1usmusv3 then you should honestly be fine in terms of daily stability. Biggest worry about instability is data/Windows corruption so if sfc/scannow, DISM health come up clean and you don't have weird buggy behaviour then don't be too concerned. Like you said, gaming temps are fine

@lance36 test #10 huh

1 minute in, probably not temp sensitivity then. Not really much reason to believe it's tWR either.

Could run tWR 16 and tRTP 8 though, just to see.

Maybe you just need to run 14-14-14. Otherwise increase VDIMM to support 14-13-13 (honestly 13s not easy to do) but temps are high already so I'd just go to flat 14s. It's still respectable.

 

[lance36]

Eventually running long anta777 when you get a fan will be good, but if you can at least pass 1usmusv3 then you should honestly be fine in terms of daily stability. Biggest worry about instability is data/Windows corruption so if sfc/scannow, DISM health come up clean and you don't have weird buggy behaviour then don't be too concerned. Like you said, gaming temps are fine

Oh I've bone windows before, it's not fun doing passwords and reinstalling but I'll live. I'll get a fan for sure.

I'll try 14s too