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

[Taraquin]

Requirements: Use agesa 1.2.0.0 or greater, before 1.2.0.0 pushing past 1900 IF was impossible for me, now I can do 2066.

If you have wheas, but system runs stable, please report. In many cases a sufficently high iod and soc voltage, and low ccd and vddp volt can help. 40mv stepping is suggested to help. Suggestion:
880 or 920 vddp, 920 ccd, 1040 iod past 1900 IF, 1120 soc past 1900. If above 4000 keep ccd and vddp the same, but 1080 iod and 1160 soc. At 4200 you will probably need 1200 soc or maybe 1240 which is into unsafe territory.
ProcODT should be 28-37 for 2 x single rank sticks and 28-48 for dual tank or 4 sticks, 37 or above is usually required. Resistances often work on auto.

If you have a stable 2000 IF please post zentimings for others to get tips.

This is my stable 4000-setting, no wheas, no TM5 errors. Ram us poor bin so even at 1.55V flat 15 is not errorfree. I can run 4133/2066 with no wheas, BUT I need 1200mv soc to not take a performancehit and performance is very similar to 4000 so not worth it. Newer agesas might reduce voltage reqs like it did on my 3600. It was never stable above 3733/1866, semi stable at 3800/1900 and 1150v soc and 1075v iod, but agesa 1.2.0.3b made 3800/1900 100% stable at 1100mv soc and 1050mv iod.

 

[Mussels]

I cant go beyond 1866, but i dont know if my RAM is the issue.

 

[Taraquin]

I cant go beyond 1866, but i dont know if my RAM is the issue.

Set your ram to a lower speed you know works and try 1900 IF+, they desync so latency is terrible, but you find out if your IF works/is stable. Try the volt/procodt settings and see if it boots

In general, if IF is unstable it either don't boot or you get random restarts out of nowhere. Wheas might occur too, but in some cases they don't make the system unstable.

 

[cvaldes]

I don't use ZenTimings so I don't have a screenshot.

I do have an SFF build (NZXT H210) with a Ryzen 5600X, ASUS ROG Strix X570-I motherboard and T-Force Dark Zα 32GB (2x16GB) DDR4 4000MHz CL18 RAM.

D.O.C.P. is set to Profile #1 in the BIOS. No other memory tuning was done. FCLK and UCLK are both 2,000 MHz. PBO is enabled but no other CPU overclocking adjustments have been done.

The system (with Nvidia GeForce RTX 2070 SUPER FE) plays games stably on a 1440p/165Hz monitor. I don't bother with benchmarks.

 

[freeagent]

2000 stable on my 5600X, can bench it at 2100. 1900 on my 5900X, I cant nail down 2K stable, maybe I am soo soft with voltage. Every time a new AGESA comes out I try..

 

[Mussels]

Set your ram to a lower speed you know works and try 1900 IF+, they desync so latency is terrible, but you find out if your IF works/is stable. Try the volt/procodt settings and see if it boots

In general, if IF is unstable it either don't boot or you get random restarts out of nowhere. Wheas might occur too, but in some cases they don't make the system unstable.

Fair point, but i always have that suspicion that the out of sync IF may cause issues that i mistake for instability

 

[purecain]

With 4x8gb Im stuck at 1800mhz infinity fabric. iirc. I need to check it on my next boot.

 

[freeagent]

4x8 on my 5600X is stable at 1833, anything more isn’t guaranteed. 4x8 on my 5900X is stable at 1866, 1900 is showing stable, but has a hard time cold booting occasionally.

 

[Taraquin]

Fair point, but i always have that suspicion that the out of sync IF may cause issues that i mistake for instability

I don't think it would, only issue I'm aware of is bad perf

 

[cellar door]

Even with the the newest 1.2.0.3c - my Ryzen 3600 won't go stable past 3733mhz, it will 'work' at 3800(1900IF) but will eventually error out/reboot randomly. No matter the voltage.

Good to hear, that I'm not alone being stuck on 3733mhz, I suspected this was where most chips end up. Even though my chip can do 4.5ghz allcore(most likely a chip that didn't make XT bin)

Requirements: Use agesa 1.2.0.0 or greater, before 1.2.0.0 pushing past 1900 IF was impossible for me, now I can do 2066.

If you have wheas, but system runs stable, please report. In many cases a sufficently high iod and soc voltage, and low ccd and vddp volt can help. 40mv stepping is suggested to help. Suggestion:
880 or 920 vddp, 920 ccd, 1040 iod past 1900 IF, 1120 soc past 1900. If above 4000 keep ccd and vddp the same, but 1080 iod and 1160 soc. At 4200 you will probably need 1200 soc or maybe 1240 which is into unsafe territory.
ProcODT should be 28-37 for 2 x single rank sticks and 28-48 for dual tank or 4 sticks, 37 or above is usually required. Resistances often work on auto.

If you have a stable 2000 IF please post zentimings for others to get tips.

View attachment 217227
This is my stable 4000-setting, no wheas, no TM5 errors. Ram us poor bin so even at 1.55V flat 15 is not errorfree. I can run 4133/2066 with no wheas, BUT I need 1200mv soc to not take a performancehit and performance is very similar to 4000 so not worth it. Newer agesas might reduce voltage reqs like it did on my 3600. It was never stable above 3733/1866, semi stable at 3800/1900 and 1150v soc and 1075v iod, but agesa 1.2.0.3b made 3800/1900 100% stable at 1100mv soc and 1050mv iod.

What's your latency on that setup in Aida64?

 

[Taraquin]

Even with the the newest 1.2.0.3c - my Ryzen 3600 won't go stable past 3733mhz, it will 'work' at 3800(1900IF) but will eventually error out/reboot randomly. No matter the voltage.

Good to hear, that I'm not alone being stuck on 3733mhz, I suspected this was where most chips end up. Even though my chip can do 4.5ghz allcore(most likely a chip that didn't make XT bin)


What's your latency on that setup in Aida64?

52 on the 5600X. Getting above 3800/1900 is unusual on Ryzen 3000, but quite common on Ryzen 5000.

 

[cellar door]

52 on the 5600X. Getting above 3800/1900 is unusual on Ryzen 3000, but quite common on Ryzen 5000.

How about on your R5 3600? I'm using e-die, so far have it down to 64.5ns

 

[Taraquin]

How about on your R5 3600? I'm using e-die, so far have it down to 64.5ns

Best is 66ns, sold it recently to a friend of mine. 4.1GHz all core, 3733cl15-20-8-11-21, 56 trc, 525 trfc, 1t and others tuned tight. Bad bin on cpu so can't get over 4.1GHz. Same latency on 3800 (cl16, trc 58, trfc 535) aswell since I have to loosen a few timings, but write speed is better and dram calc us 2 sec faster. Unfortunately rev E scales tRCDRD negatively with voltage so if I raise volt to 1.45V or above to lower CL, I must set tRCDRD to 21 :/ At 3733 I ran 1.44V, at 3800 I need 1.46V. With 1.4V I could do 3733cl16-19-8-11, but is was a bit slower vs 15-20-8-11.

 

[Calmmo]

Haven't pushed past 1900 as I would have to push more than 1.45v to get my ram to run at the c16 current timings, so.. i dont really know

 

[cellar door]

Best is 66ns, sold it recently to a friend of mine. 4.1GHz all core, 3733cl15-20-8-11-21, 56 trc, 525 trfc, 1t and others tuned tight. Bad bin on cpu so can't get over 4.1GHz. Same latency on 3800 (cl16, trc 58, trfc 535) aswell since I have to loosen a few timings, but write speed is better and dram calc us 2 sec faster. Unfortunately rev E scales tRCDRD negatively with voltage so if I raise volt to 1.45V or above to lower CL, I must set tRCDRD to 21 :/ At 3733 I ran 1.44V, at 3800 I need 1.46V. With 1.4V I could do 3733cl16-19-8-11, but is was a bit slower vs 15-20-8-11.

Interesting - this is all with micron e-die(rev.E)? If so, you just gave me some ideas.

 

[Taraquin]

Interesting - this is all with micron e-die(rev.E)? If so, you just gave me some ideas.

Mine might respond different than yours, mine hates low tRCDRD, I know some handles that timing better than mine, but mine likes lower trfc than most rev E.

 

[cellar door]

Mine might respond different than yours, mine hates low tRCDRD, I know some handles that timing better than mine, but mine likes lower trfc than most rev E.

Basically its the same for tRCDRD for me, 20 is stable - its a hallmark this of e-die. Ok but thanks for confirming that you were using those ICs, now I can try couple things on my end so see if I can improve the timings.

Below are my settings so right now.

 

[Taraquin]

Basically its the same for tRCDRD for me, 20 is stable - its a hallmark this of e-die. Ok but thanks for confirming that you were using those ICs, now I can try couple things on my end so see if I can improve the timings.

Below are my settings so right now.

I used to run 1.44V, cl 15, rp 11, gdm off, rrdl 4, wtrs 3, wtrl 6, trc 56, 525 trc, 10 wr, 5 rtp.
Try lower vddp, 900, iod and ccd should be 1050 or lower, soc should be 1100 or lower. I ran 1075 soc, 825 iod and ccd, and 775 vddp

At 3800 I needed 1050 iod/ccd and 1100 soc, 800 vddp.

 

[ir_cow]

I can get 2066 MHz with the SoC at 1.25 V. 2066-2166 with 1.35 V. I didn't like the SoC so high, but it is stable if I so desire to run it. I personally like 3800 / 1900. I don't have to use Gear Down Mode so the latency is actually lower and the bandwidth is higher. Also Soc is 1.15v. Much much safer for long term use.

Honestly I believe a lot of this limitation has to do with the motherboard you are using. I'm loving this MSI B550 Unify-X. I couldn't get anything to boot above 4600 mem before with a ASUS X570 Hero. Now 5066 is possible with less voltage all around. Even the IF goes to 2066 with ease.

 

[cellar door]

I used to run 1.44V, cl 15, rp 11, gdm off, rrdl 4, wtrs 3, wtrl 6, trc 56, 525 trc, 10 wr, 5 rtp.
Try lower vddp, 900, iod and ccd should be 1050 or lower, soc should be 1100 or lower. I ran 1075 soc, 825 iod and ccd, and 775 vddp

At 3800 I needed 1050 iod/ccd and 1100 soc, 800 vddp.

The soc on my chip needs 1.1 to boot 3600. vddp iod and ccd also won't go lower without eventually causing errors - this might have something to do with the fact my chip is from the newer batch(bad XT bin most likely).

But I'm going to test out some of the lower timings you were running, I only have a low bin 3000cl15 ballistix sport LT kit. Frquency wise it can go past 4000, however for ex. 525 trfc is a instant crash for me, your kit is most likely a way better bin.

 

[freeagent]

Honestly I believe a lot of this limitation has to do with the motherboard you are using.

I think it’s more CPU than anything. My 5600X is stable at 2K and can bench at 2100 1:1 but I cannot get my 5900X stable at 2K on the same board, same ram, same everything.

 

[Splinterdog]

 

[ir_cow]

I think it’s more CPU than anything. My 5600X is stable at 2K and can bench at 2100 1:1 but I cannot get my 5900X stable at 2K on the same board, same ram, same everything.

It is just the opposite for me. Both my 3800X and 5800X can achieve higher results with less voltage. Same AGESA and everything. Just different boards. Don't get my wrong, there are CPU duds out in the wild, but those who can't get 1800 MHz stable on Zen 3 either have a crap motherboard or doing something wrong 99% of the time.

Other factors like Dual Rank, vs Single, IC Density and voltages, Gear Down Mode all play a vital role in reaching 2000 or above.

 

[freeagent]

It is just the opposite for me. Both my 3800X and 5800X can achieve higher results with less voltage. Same AGESA and everything. Just different boards. Don't get my wrong, there are CPU duds out in the wild, but those who can't get 1800 MHz stable on Zen 3 either have a crap motherboard or doing something wrong 99% of the time.

Other factors like Dual Rank, vs Single, IC Density and voltages, Gear Down Mode all play a vital role in reaching 2000 or above.

I've tried all the beta bios that Asus put out over the last year.. probably 10 of them for sure. I only need 1.15625v SOC for 2000 on my 5600X. 1.1625 gets me up to 2100. As for 1800FCLK, we are talking about 2K in this thread, no? In my other post I was saying my 56 can my 59 cant, I used hella voltage too, more than I normally would have for sure. Not a big deal though.. just a little disappointing. I do prefer dual ranked smoothness over high MHs single rank for most things.. at least that's what I keep telling myself. I still enjoy playing with my 5600X, it really is a lot of fun.. the only thing wrong with it is its not a 5900X.

 

[Taraquin]

The soc on my chip needs 1.1 to boot 3600. vddp iod and ccd also won't go lower without eventually causing errors - this might have something to do with the fact my chip is from the newer batch(bad XT bin most likely).

But I'm going to test out some of the lower timings you were running, I only have a low bin 3000cl15 ballistix sport LT kit. Frquency wise it can go past 4000, however for ex. 525 trfc is a instant crash for me, your kit is most likely a way better bin.

My kit is also the 3000cl15 Usually the avg rev E can do 300ns, mine does 280. You can probably do anywhere from 530 to 580 depending on bin. Do you have agesa 1.2.0.3b? Before that I was stuck at 3733 getting random reboots, after 3800 worked fine.

I can get 2066 MHz with the SoC at 1.25 V. 2066-2166 with 1.35 V. I didn't like the SoC so high, but it is stable if I so desire to run it. I personally like 3800 / 1900. I don't have to use Gear Down Mode so the latency is actually lower and the bandwidth is higher. Also Soc is 1.15v. Much much safer for long term use.

Honestly I believe a lot of this limitation has to do with the motherboard you are using. I'm loving this MSI B550 Unify-X. I couldn't get anything to boot above 4600 mem before with a ASUS X570 Hero. Now 5066 is possible with less voltage all around. Even the IF goes to 2066 with ease.

Talking to cellar door or me? He's on ryzen 3000 which rarely can do more than 3800. I can do 4133, I might be able to do 4200+, but it would require way to much soc voltage for my comfort. I need 1.12V for 4000, 1.18V for 4066 and 1.25V for 4133. I bet 4200 would need atleast 1.32V. 1.2V is by many considered upper safe limit.

My MB is cheap, but it has only 2 dimms (helps a lot with ram oc compared to 4) and is rated for 5100MHz ram on Ryzen 3000, max 4600 on Ryzen 5000 which makes little sense...

AAs for ram oc I'm limited by bad bin ram, it has a 94% rating on dram calc, making it sub avg compared to other B-die. 1T and liw tRCDRD is very hard, trfc won't go below 141ns at 1.46V. The rev E I had was 102% rated so it beat the majority of rev E.