What is your max infinity fabric on ryzen 5000? September 2021 edition

[Mussels]

Somehow i'm stable at -20 all core as well, which never used to work
Back when i tested previously i had the joys of melting PCI-E cables throw all my testing out the window

I know this isn't an overclocking thread specifically, but why does this suddenly work so well?
Is it the RAM tuning or the SoC voltage that's made this more stable?

tried -30... so far so good???

 

[Taraquin]

Somehow i'm stable at -20 all core as well, which never used to work
Back when i tested previously i had the joys of melting PCI-E cables throw all my testing out the window

I know this isn't an overclocking thread specifically, but why does this suddenly work so well?
Is it the RAM tuning or the SoC voltage that's made this more stable?

tried -30... so far so good???

View attachment 223186

Changed bios since last time you tried? Ram tuning shouldn't matter much, but it can possibly improve multi core stability due to IO die using more power and then stealing a bit from core budget. Try corecycler, if it passes 2 6 mins runs on each core you are 99% stable.

As for curve optimizer and performance: If you can run allcore -30 stable that gives me about 250-300MHz more on my 5600X. On your 5800X I bet you get 200-250MHz, translating to 3-4% perf in CB23. If you run PBO+200 aswell -30 will probably be difficult, but that can boost fps in games by 3-5%. In SOTTR +200 PBO and - 30x4 - 29x2 (highest UV I can do) gave me 5% higher fps vs stock.

Still ram tweaking is the major boost that gave me 15-20% in some games.

CO+PBO is generally easier to stabilize at max settings, ram at max settings takes a lot of time.

 

[Mussels]

-30 crashed at idle, but even a single core running in a tester kept it going
-20 is rock solid

This is with +200, so it's genuinely impressive... could the higher SoC voltage i used for 3800/1900 (and the 4000/2000 test) be related somehow?

 

[Taraquin]

-30 crashed at idle, but even a single core running in a tester kept it going
-20 is rock solid

This is with +200, so it's genuinely impressive... could the higher SoC voltage i used for 3800/1900 (and the 4000/2000 test) be related somehow?

As I said it can improve stability at all core oc due to IO die using more with higher soc volt. Tried corecycler? Maybe all cores except 1 can run - 30.

 

[Valantar]

-30 crashed at idle, but even a single core running in a tester kept it going
-20 is rock solid

This is with +200, so it's genuinely impressive... could the higher SoC voltage i used for 3800/1900 (and the 4000/2000 test) be related somehow?

My experience with CO is that you can't trust stability until you've used it for a pretty decent amount of time. At least in my case, -30 all core seemed stable but gave random crashes at idle or in games (or waking from sleep IIRC) after a few days of normal usage. -20 was the same, but even longer between crashes. -10 has been rock solid for months.

Still, your chip seems pretty good - I haven't made much of an attempt at pushing things beyond entering a few numbers, but I've had stability issues above +100 PBO, so even preliminary stability at +200 sounds really good. Fingers crossed your current settings stay working!

On that note, I should probably look into figuring out how to tune CO per core, as the seeming stability at -30 seem to indicate that most of my cores should be able to handle it.

Going higher on voltages probably won' t fix this. You should try 3933/1966, 3866/1933 or 3800/1900, that is the only viable fix :/ Usually 5600X/5800X and ITX/mATX with 2 dimms has the highest chance of hitting 4000/2000+ without whea 19, but sometimes they don't.

3800/1900 usually work, and you can tighten timings then so performance could be close to 4000cl17. You can also lower voltages and then clockspeeds will improves since IO-die uses less power.

Neither 3933 or 3866 were error-free, sadly. If anything, I saw more errors at those speeds than 4000, but that's probably just random. 3800/1900 is error-free so far though, including gaming, where I saw consistent errors before. Any tips on tightening up timings?

 

[Taraquin]

My experience with CO is that you can't trust stability until you've used it for a pretty decent amount of time. At least in my case, -30 all core seemed stable but gave random crashes at idle or in games (or waking from sleep IIRC) after a few days of normal usage. -20 was the same, but even longer between crashes. -10 has been rock solid for months.

Still, your chip seems pretty good - I haven't made much of an attempt at pushing things beyond entering a few numbers, but I've had stability issues above +100 PBO, so even preliminary stability at +200 sounds really good. Fingers crossed your current settings stay working!

On that note, I should probably look into figuring out how to tune CO per core, as the seeming stability at -30 seem to indicate that most of my cores should be able to handle it.

Neither 3933 or 3866 were error-free, sadly. If anything, I saw more errors at those speeds than 4000, but that's probably just random. 3800/1900 is error-free so far though, including gaming, where I saw consistent errors before. Any tips on tightening up timings?

Check event viewer if you get random crashes with CO, it will identify the unstable core. Use core cycler to find 99% of the stability. Probably several cores can do - 30, on might only do - 10.

Timings: Still at 1.4V? 16 16 16 16 32 48, tRFC 288, twrl 10, 9 or 8. The rest can stay the same. Reduce SOC volt to 1.1v, iod to 1.06v, vddp 0.88 and ccd 0.92.

 

[Valantar]

Check event viewer if you get random crashes with CO, it will identify the unstable core. Use core cycler to find 99% of the stability. Probably several cores can do - 30, on might only do - 10.

Timings: Still at 1.4V? 16 16 16 16 32 48, tRFC 288, twrl 10, 9 or 8. The rest can stay the same. Reduce SOC volt to 1.1v, iod to 1.06v, vddp 0.88 and ccd 0.92.

Thanks! I'll give that a try later today. Sadly it's been too long since I was experimenting with CO, I sincerely doubt I'll be able to find those error logs now Still, good to know a method of identifying the core causing the crash

 

[Taraquin]

Thanks! I'll give that a try later today. Sadly it's been too long since I was experimenting with CO, I sincerely doubt I'll be able to find those error logs now Still, good to know a method of identifying the core causing the crash

I first set -30 and +200 pbo. I got random reboots in idle once a day. I ran corecycler and it found CPU 1 to fail after 4 mins. I set core 1 to -29 and had no problem since. Once a week I got random reboot. Check eventviewer and found whea 18 on core 0. I set it to -29 aswell and now it has been stable for a month. If you run core cycler and get instant fail you must lower more, fail after a few minutes usually means only 1-2 lower to stabilize. If you manage to stabilize all at best setting all at -30 and +200 pbo gives you up to 250-300MHz better allcore on 5600X and 200-250MHz on 5800X, this translates to 5-6% perf on 5600X and 4-5% on 5800X.

 

[Dar1usElite]

Hey @Taraquin, @Mussels directed me to you for your expertise on ram timings for my current system. attached is a screenshot from zentimings and also taiphoon burner.
System is as follows.
CPU: 5800x
Board: x570 aorus master.
ram: G skill trident z neo 2 x16gb. Hynix DJR 3600 CL16.
I have in the past successfully gotten it to 3800 but i recently lost all my profiles from my bios.
Thanks in advance

 

[Taraquin]

Hey @Taraquin, @Mussels directed me to you for your expertise on ram timings for my current system. attached is a screenshot from zentimings and also taiphoon burner.
System is as follows.
CPU: 5800x
Board: x570 aorus master.
ram: G skill trident z neo 2 x16gb. Hynix DJR 3600 CL16.
I have in the past successfully gotten it to 3800 but i recently lost all my profiles from my bios.
Thanks in advance

For 3800/1900 try 16 20 20 20, 36, 56, trrds/l 4/6, tfaw 16, twr 16, trtp 8, trfc 480, scls 4, wtrs 4, wtrl 8 if possible, but you may need 9 or 10, rdwr 10, wrrd 3, rest on auto. Set soc voltage to 1.12v, iod volt 1.04v, ccd 0.92, vddp 0.88, cad 24 20 24 24. Let me know if it does not work

 

[Valantar]

Check event viewer if you get random crashes with CO, it will identify the unstable core. Use core cycler to find 99% of the stability. Probably several cores can do - 30, on might only do - 10.

Timings: Still at 1.4V? 16 16 16 16 32 48, tRFC 288, twrl 10, 9 or 8. The rest can stay the same. Reduce SOC volt to 1.1v, iod to 1.06v, vddp 0.88 and ccd 0.92.

Haven't had much time for testing, but so far these settings seem to work just fine (I'm at 1.38V btw, not 1.4). Any tips for further tightening? Seems like these DIMMs do pretty well, though I'm not looking to ride the bleeding edge of stability here, so if it's unlikely I'll get lower I might just leave it. I'll look into some CO testing to identify my less stable cores whenever I find the time

 

[Taraquin]

Haven't had much time for testing, but so far these settings seem to work just fine (I'm at 1.38V btw, not 1.4). Any tips for further tightening? Seems like these DIMMs do pretty well, though I'm not looking to ride the bleeding edge of stability here, so if it's unlikely I'll get lower I might just leave it. I'll look into some CO testing to identify my less stable cores whenever I find the time

You must increase voltage to improve timings. You are on the limit already. If you try 1.45V which is considered safe on B-die there is lots of room for improvement

 

[Valantar]

You must increase voltage to improve timings. You are on the limit already. If you try 1.45V which is considered safe on B-die there is lots of room for improvement

Oh I know, I just meant to indicate that given that I'm still on 1.38 I thought there ought to be more headroom. Do you think I have any chance of reaching 3800c14?

 

[tabascosauz]

Oh I know, I just meant to indicate that given that I'm still on 1.38 I thought there ought to be more headroom. Do you think I have any chance of reaching 3800c14?

Average to decent B-die needs 1.5-1.6V for 3800CL14, if it can even boot CL14 at all. Maybe 3800CL15 at 1.45V if lucky. Tight 3800CL15 or CL16 is fine, CL14 is honestly overhyped for real performance.

I don't think 3200CL14 should have much trouble but more importantly, what came of that temp testing? Kinda important to know if you're looking to run 3800CL15 or 3800CL14, inadequate airflow and you can forget about it.

 

[Taraquin]

Oh I know, I just meant to indicate that given that I'm still on 1.38 I thought there ought to be more headroom. Do you think I have any chance of reaching 3800c14?

If you have a very good bin you can do 3800cl14 around 1.5V with geardownmode. My kit does 3800cl15 1T at 1.48V. 1.5V needs more cooling than you have. Try 1.45V with gdm off and 2T and tune in cl 15. Try 15 15 15 30 45, tRFC 280 or 272, wtrl 10 or 8, cwl 14, wr 14, rtp 7, rest of timings can stay the same. If your bin is mediocre try a bit higher voltage.

 

[Dar1usElite]

For 3800/1900 try 16 20 20 20, 36, 56, trrds/l 4/6, tfaw 16, twr 16, trtp 8, trfc 480, scls 4, wtrs 4, wtrl 8 if possible, but you may need 9 or 10, rdwr 10, wrrd 3, rest on auto. Set soc voltage to 1.12v, iod volt 1.04v, ccd 0.92, vddp 0.88, cad 24 20 24 24. Let me know if it does not work

Thanks for the quick reply, just need clarification on a few things since gigabyte calls them different. I just need help on what the following are called in my bios: scls, (whatever the timing for 56 was which i assume was trc?) , iod voltage, ccd voltage and which order for the cad impedances (i assume its the order of zentimings). I tried to set them best i could and it wouldn't post. i also need load line on or i get random bluescreens during short spikes of load (starting cinebench etc), i have it set to medium but nothing else. also im on bios f33, should i go to f34 or f35e or stay on f33?

 

[Taraquin]

Thanks for the quick reply, just need clarification on a few things since gigabyte calls them different. I just need help on what the following are called in my bios: scls, (whatever the timing for 56 was which i assume was trc?) , iod voltage, ccd voltage and which order for the cad impedances (i assume its the order of zentimings). I tried to set them best i could and it wouldn't post. i also need load line on or i get random bluescreens during short spikes of load (starting cinebench etc), i have it set to medium but nothing else. also im on bios f33, should i go to f34 or f35e or stay on f33?

Try bios F34, I use the same (agwsa 1.2.0.3b). Try 36 tras, 56 trc yes.

Since it does not boot try a safe approach:
16 22 22 22 38 trc 60, trrds/l 5/8, tfaw 20, twr 16, trtp 8, wtrs/l 4/12, cwl 16, trfc 512, scls 4, rdwr 12, wrrd 3. Disable spread spectrum. 1T and gear down mode on. Let me know if this works, it should

Let your cpu run stock, don't use fixed OC or pbo/co until you have stable ram oc, llc medium is usually fine then.

 

[Valantar]

Average to decent B-die needs 1.5-1.6V for 3800CL14, if it can even boot CL14 at all. Maybe 3800CL15 at 1.45V if lucky. Tight 3800CL15 or CL16 is fine, CL14 is honestly overhyped for real performance.

I don't think 3200CL14 should have much trouble but more importantly, what came of that temp testing? Kinda important to know if you're looking to run 3800CL15 or 3800CL14, inadequate airflow and you can forget about it.

I decided that my previous "as few fans as possible" approach was a bit dumb, and fished out a spare NF-A12x15. For now it's just propped up between some cabling and tubing, with the side panel preventing it from falling out, but I'm thinking I'll stick it in place with some magnets to the side panel.

Works very well though, and still essentially inaudible at ~450rpm. (Also invisible in real life, the camera greatly exaggerates things.) I wanted to set it to only turn on if the RAM got hot, but apparently Aquasuite's hardware monitoring doesn't cover VRAM temperatures For now I'm well below 40 degrees at idle though, and well below 50 even during stress testing. A bonus from this is that my VRMs are also seeing a ~10 degree drop. Not that that was necessary, but it's likely still improving things when the case is getting blown full of hot air from the rad.

Edit: apparently forgot to finish that first sentence about me being stupid and stubborn

 

[Dar1usElite]

Try bios F34, I use the same (agwsa 1.2.0.3b). Try 36 tras, 56 trc yes.

Since it does not boot try a safe approach:
16 22 22 22 38 trc 60, trrds/l 5/8, tfaw 20, twr 16, trtp 8, wtrs/l 4/12, cwl 16, trfc 512, scls 4, rdwr 12, wrrd 3. Disable spread spectrum. 1T and gear down mode on. Let me know if this works, it should

Let your cpu run stock, don't use fixed OC or pbo/co until you have stable ram oc, llc medium is usually fine then.

alright, which timing is the scls? i couldn't see anything called that. also i never set the ccd and the iod voltage because i wasnt sure what they were called in my gigabyte bios, any ideas?

 

[Valantar]

Hm, guess I should have tested more before getting so optimistic. Turns out even 3800c16 isn't actually stable - I'm getting errors (pretty quickly after starting Memtest) even up to 1.45V sadly. Goes to show that "system isn't crashing or throwing errors" isn't the same as it actually being stable, and given how quickly memtest started throwing errors it also says something about just how stable Windows is these days - it must have been juggling and correcting errors in the background like nobody's business. Guess I'll have to step back to c17.

Edit: or maybe I should double check my settings first and not run GDM off and 1T... At least now it isn't throwing errors within a few minutes.

 

[Taraquin]

alright, which timing is the scls? i couldn't see anything called that. also i never set the ccd and the iod voltage because i wasnt sure what they were called in my gigabyte bios, any ideas?

2 timings have the name scl in them. Set those to 4. Vddg Iod and ccd is under amd overclocking menu, not in timings.

 

[Deleted member 215115]

Hm, guess I should have tested more before getting so optimistic. Turns out even 3800c16 isn't actually stable - I'm getting errors (pretty quickly after starting Memtest) even up to 1.45V sadly. Goes to show that "system isn't crashing or throwing errors" isn't the same as it actually being stable, and given how quickly memtest started throwing errors it also says something about just how stable Windows is these days - it must have been juggling and correcting errors in the background like nobody's business. Guess I'll have to step back to c17.

Edit: or maybe I should double check my settings first and not run GDM off and 1T... At least now it isn't throwing errors within a few minutes.

You should post a zentimings screenshot so we can see what's wrong. I've seen 3200C14 FlareX kits do 3800 CL14 but not under 1.5V. Feel free to increase the voltage up to 1.6V. B-die can take up to 2V with a simple 120mm fan aimed at the memory.

GDM off for dual rank I think is only possible on some 2-DIMM B550 boards but I wouldn't bother too much with that. As long as you can do 1T, you're golden.

 

[Taraquin]

Hm, guess I should have tested more before getting so optimistic. Turns out even 3800c16 isn't actually stable - I'm getting errors (pretty quickly after starting Memtest) even up to 1.45V sadly. Goes to show that "system isn't crashing or throwing errors" isn't the same as it actually being stable, and given how quickly memtest started throwing errors it also says something about just how stable Windows is these days - it must have been juggling and correcting errors in the background like nobody's business. Guess I'll have to step back to c17.

Edit: or maybe I should double check my settings first and not run GDM off and 1T... At least now it isn't throwing errors within a few minutes.

1T is much harder than 2T or 1T+gdm. Most people must change cad bus etc to get 1T gdm off stable. On my setup gdm and 2T runs fine with 24 20 24 24 on cad. 1T gdm off needs 40 20 30 24, everything else gives me loads of errors.

 

[Valantar]

You should post a zentimings screenshot so we can see what's wrong. I've seen 3200C14 FlareX kits do 3800 CL14 but not under 1.5V. Feel free to increase the voltage up to 1.6V. B-die can take up to 2V with a simple 120mm fan aimed at the memory.

GDM off for dual rank I think is only possible on some 2-DIMM B550 boards but I wouldn't bother too much with that. As long as you can do 1T, you're golden.

After fixing my minor PEBKAC issue I'm currently at ~160% memtest with 0 errors after going to 2t (stepped down to 1.4V). Current settings:

Previous settings that gave errors almost immediately, which from what I can tell are identical except for the command rate:

Should I try 1T with GDM enabled instead? Is there a meaningful difference in the performance tradeoff between the two?

Also, does anyone know if ASRock B550 boards lack a software VDIMM readout? I can't find anything resembling that in HWinfo at least.

 

[Taraquin]

After fixing my minor PEBKAC issue I'm currently at ~160% memtest with 0 errors after going to 2t (stepped down to 1.4V). Current settings:
View attachment 224183
Previous settings that gave errors almost immediately, which from what I can tell are identical except for the command rate:
View attachment 224184
Should I try 1T with GDM enabled instead? Is there a meaningful difference in the performance tradeoff between the two?

Also, does anyone know if ASRock B550 boards lack a software VDIMM readout? I can't find anything resembling that in HWinfo at least.

2T is often a bit faster than gdm, but slightly less stable. Stick with 2T is my advice. You can try 15 15 15 30 45, but you will need 1.43-1.5V to do that. If you set it to 1.45V you can lower trfc to 280 or 272 as it scales with voltage. 1.5V allows for 264 or 256.