Ryzen Owners Zen Garden

[tabascosauz]

Did I miss where his rep took a hit somewhere? A bit out of the loop here.

It's not so much to do with the software (which are a very mixed bag), as it is how he brags about his infinite intellect but flames and refuses to engage with anyone who thinks his software is bugged/flawed. Which leads to the software being forever sad like it is (DRAM calc/CTR/Hydra).

But that's beside the point, because it's AMD's perennially hilarious software engineers at work here. I'll test it out on my 5900X soon.

 

[dont whant to set it"']

Well, aparently my x3d variant of the 5800 is capable of 2000MHz IF clock , while so far , so stable in game and no, I am not playing linpack , p95 and the likes.

Current setup is in dualcore config because a cooler remount must be done and I've used the last dash of MX5 the tube was holding.

Fixed multiplier is doable , tried 30 odd some + the other and 46 was a no go, so far.

Started with D.O.C.P. but now , after a brief stable half hour game, I am trying with tweaked primaryes and a bump up of 0.05V to 1.4V dimm.

le: game running minimized might explain the low memory benchmarks.

 

[QuietBob]

Well, aparently my x3d variant of the 5800 is capable of 2000MHz IF clock , while so far , so stable in game and no, I am not playing linpack , p95 and the likes.

Current setup is in dualcore config because a cooler remount must be done and I've used the last dash of MX5 the tube was holding.

Fixed multiplier is doable , tried 30 odd some + the other and 46 was a no go, so far.

Started with D.O.C.P. but now , after a brief stable half hour game, I am trying with tweaked primaryes and a bump up of 0.05V to 1.4V dimm.

le: game running minimized might explain the low memory benchmarks.

Thanks for posting your results! I'm also hoping to see 2000 IF on mine, got some nice B-dies to go 1:1. Still waiting for delivery.
BTW, your dual core 5800X3D gives me a crazy idea...
"Gaming on the fastest single core CPU"

 

[lexluthermiester]

Well, aparently my x3d variant of the 5800 is capable of 2000MHz IF clock , while so far , so stable in game and no, I am not playing linpack , p95 and the likes.

You're going to have to run something that stress tests the CPU. Otherwise you have no way of knowing if you have a stable OC. A 10 minute run of Prime95 should be all you need.

 

[RJARRRPCGP]

Made it through 1/8 of an AIDA run before hard rebooting. AIDA.

Sounds just like a WHEA Windows crash out for "Cache Hierarchy Error". So far, every time Windows crashed out, it was that reason given on my Ryzens.

1usmus doing coding for AMD now?

In 1usmus' defense, the only issue with CTR, for me, was that it was a bit too positive about my chip, LOL.
A +6mV or similar Vcore increase, wasn't enough Vcore (step-wise) for my Ryzen 5 5600X. When I set it manually and had it run Prime95, then it can run without an error, but for PX, right when it looked like there wasn't any whoopsie, I then suddenly started getting random new access violation crashes with Chromium tabs, which was clearly a whoops with the auto-tuned PX profile.

I remembered getting an access violation error when just scrolling down a YouTube page, IIRC.

I honestly think 1usmus did a very good job with CTR.

 

[freeagent]

I have been playing with the latest beta bios for my board a little bit today.. It is coded with the latest AGESA V2 PI 1.2.0.7. As with 1.2.0.6 it limits voltage. But performance is still very strong regardless of the lower boosting action with all of the cores. Even still, when presented with a heavy all core workload, it somehow pushes more wattage, more current with less voltage. For instance, I have never seen 157 TDC, nor have I seen 237w CPU Package power, or over 230w PPT.. yet manages to keep itself under 85c.

Works pretty good..

 

[tabascosauz]

I have been playing with the latest beta bios for my board a little bit today.. It is coded with the latest AGESA V2 PI 1.2.0.7. As with 1.2.0.6 it limits voltage. But performance is still very strong regardless of the lower boosting action with all of the cores. Even still, when presented with a heavy all core workload, it somehow pushes more wattage, more current with less voltage. For instance, I have never seen 157 TDC, nor have I seen 237w CPU Package power, or over 230w PPT.. yet manages to keep itself under 85c.

Works pretty good..

I'll take your endorsement then, installing 4201 now

I noticed 4006 (1206) isn't all that bad at stock and CO. CPU performance actually strong compared to 1200-1203. It's just that for some reason the power limits have been borked since BIOS 3801 - 142W PPT becomes 145W for some reason.

 

[freeagent]

I will play with it some more after I get the kids to bed

 

[Mussels]

4403 is out for my X570-F, Aggie-see-saw of 1.2.0.7

I'll give it a shot and let the system run on "stock" settings and see how temps are, then try with PBO on - i want to see what the boosts are like, vs the max temps with my heat monster

Edit: oh and i love that i can BIOS flash straight from NTFS drives and not even need USB drives - modern BIOS tools are AWESOME.

Older BIOS 4.6GHz locked temps: (also handy for me since it shows my RAM timings, voltages etc)

Scuse the wobbles, motion photo from phone converted to gif

This is what drives 5800x owners nuts

 

[tabascosauz]

I will play with it some more after I get the kids to bed

Think you're right. 1207 does seem to do the same with less clock, scores the same as 1206 (ST is low because Windows 11). New AGESA takes a few days to settle in, so I'll check back.

 

[Ibizadr]

@freeagent with 1.2.0.7 we can push ppt after 140? Im still on 1.2.0.3c because of that.

 

[freeagent]

@freeagent with 1.2.0.7 we can push ppt after 140? Im still on 1.2.0.3c because of that.

Yessir

 

[Ibizadr]

Yessir

OH boy let's save some ss of settings in bios and ram to do it all again. Here we go.

 

[Mussels]

backing up my wobbly GIF above: somethings different with idles.
Yes, i get those sudden spikes of 20C changes (honestly i think it changes which core its measuring, hence the sudden changes) - but i never used to get minimum idles that low previously.

Todays update with the new AGESA: stock settings, other than TRFC and ram to 3900/1800

2 hours of Dying light (1440p 165Hz pew pew + 3 hours of desktop use/idling)

I'm seeing 50Mhz less boost (Keep in mind those boosts could be for single digit milliseconds) with 4.8Ghz peak vs 4.85
--------------------------------------------------(Current - min - max - average)

Average idle temps are lower than before, by a good margin. 72C peak is perfectly fine combined with a 3090 (Notice that Tdie hit 72C, but the cores themselves are 58C-65C)

At present on 1.2.0.7, i'm seeing what i wish i had all along: core temps under 65C with more or less the maximum low thread boost clocks


And an R23 run with current values only - not a clean OS, background tasks running (Avoiding the clocks from the gaming history)
4.55GHz (measuring error it seems) vs the 4.3 i'd see on the older AGESA, 68C at 145W (ouch, needs that curve undervolt)

Temps are definitely better, performance is good. They changed something.

 

[Ibizadr]

I need to try this bios version, I'm on 1.2.0.3c and don't have any problems with it, but what I think something not good in your bios @Mussels cause normally voltage goes to 1.5 v on my 1.2.0.3c bios version and in yours only got 1.33v. you change your Edc limit past 140?

Edit: msi don't launch yet 1.2.0.7 to my b450 tomahawk max II, only 1.2.0.6c from 20/04. Waiting more days to see if they launch 1.2.0.7

 

[tabascosauz]

I need to try this bios version, I'm on 1.2.0.3c and don't have any problems with it, but what I think something not good in your bios @Mussels cause normally voltage goes to 1.5 v on my 1.2.0.3c bios version and in yours only got 1.33v. you change your Edc limit past 140?

Edit: msi don't launch yet 1.2.0.7 to my b450 tomahawk max II, only 1.2.0.6c from 20/04. Waiting more days to see if they launch 1.2.0.7

VID =! Vcore

1.487V SVI2 vcore which is normal

afaik VID is only useful for seeing if curve optimizer is working, or which cores are being used the most for ST. If you see 1.5V VID then you may have some big board problems on your hands. I've never seen >1.35V VID.

@Mussels are you at stock power limits when reaching 145W? It's not just my Impact that thinks 145=142?

 

[Ibizadr]

VID =! Vcore

1.487V SVI2 vcore which is normal

afaik VID is only useful for seeing if curve optimizer is working, or which cores are being used the most for ST. If you see 1.5V VID then you may have some big board problems on your hands. I've never seen >1.35V VID.

@Mussels are you at stock power limits when reaching 145W? It's not just my Impact that thinks 145=142?

My core vid goes to 1. 5v I will do a ss and post it later. I think I don't have any problems with board or I'm not seeing any problems, no reboots no weird behavior. I'm not on stock limits

 

[tabascosauz]

My core vid goes to 1. 5v I will do a ss and post it later. I think I don't have any problems with board or I'm not seeing any problems, no reboots no weird behavior. I'm not on stock limits

Ah, no. I see what you mean. Disable Snapshot Polling, and VID goes to 1.5V. It also makes the VID pretty much the same across all cores, doesn't seem to be any per-core VID behaviour anymore, seems to make VID useless.

Personally I think I prefer Snapshot Polling on, because VID illustrates load distribution on cores. But on ROG boards we also get a die-sense voltage circuit, which is much more accurate all the time than Vcore SVI2 TFN or VID (close to Vcore that Ryzen Master reports).

 

[LifeOnMars]

Meanwhile on my old shitty Asrock B450 board they only have a bios up to 1.2.0.2 AGESA

Anyone know how I can contact Asrock and get a new bios from them? Otherwise, I might just get a nice new board. I'm not bothered about upgrades or anything/EOL for at least three years. I've only recently got the 5700x and 3080 12Gb and my games backlog is huge which this setup will easily handle.

 

[Mussels]

I need to try this bios version, I'm on 1.2.0.3c and don't have any problems with it, but what I think something not good in your bios @Mussels cause normally voltage goes to 1.5 v on my 1.2.0.3c bios version and in yours only got 1.33v. you change your Edc limit past 140?

Edit: msi don't launch yet 1.2.0.7 to my b450 tomahawk max II, only 1.2.0.6c from 20/04. Waiting more days to see if they launch 1.2.0.7

Mine was pure stock settings, and IMO they're working great

Today saw a peak of 1.487v on the vcore, so that 1.33 was just low for that gaming session i guess

Heres a longer 5 hour time period, have some numbers

 

[Chomiq]

Meanwhile on my old shitty Asrock B450 board they only have a bios up to 1.2.0.2 AGESA

Anyone know how I can contact Asrock and get a new bios from them? Otherwise, I might just get a nice new board. I'm not bothered about upgrades or anything/EOL for at least three years. I've only recently got the 5700x and 3080 12Gb and my games backlog is huge which this setup will easily handle.

Looks like they just dumped this board, you might want to contact them using this form:

ASRock > Support Request Form

event.asrock.com

Otherwise you might wait for AM5 to drop and pick up some B550 board on sale or a used X570 from someone upgrading.

 

[LifeOnMars]

Looks like they just dumped this board, you might want to contact them using this form:

ASRock > Support Request Form

event.asrock.com

Otherwise you might wait for AM5 to drop and pick up some B550 board on sale or a used X570 from someone upgrading.

Yeh tbh that makes more sense, prices should crash for a little bit on those boards before they then become 'expensive' again due to rarity. Sounds like a plan, performance is good for me anyway so I'm not really sacrificing much but would be nice to maximize potential.

 

[freeagent]

Man, I am torn. I had good results with the newest AGESA, but rolling back to an older one really woke up these cores, and I picked up a few GBP/s. Don't let those high clocks fool you though, she wont pass core cycler like this.. but I get no crashes and it passes everything except CC lol

TM5 used to throw thread handling exception errors until I tweaked it enough to run it at +200, Super Pi 32M is good too for the single core boosts. Getting it to run like that in 11 took some figgering, but I got it. All so I didn't have to revert to a static OC just to run TM5

Also, that spike on CCD1 was a freak, 97% of the time it was sitting at 80-81.

I am dual booting 10 and 11, Aida doesn't show such nice L3 numbers with this AGESA.

10 passes at 10GB just to warm it up a little.. the uneven ram temps are from playing a game.. the system is quiet ish.

 

[tabascosauz]

Man, I am torn. I had good results with the newest AGESA, but rolling back to an older one really woke up these cores, and I picked up a few GBP/s. Don't let those high clocks fool you though, she wont pass core cycler like this.. but I get no crashes and it passes everything except CC lol

TM5 used to throw thread handling exception errors until I tweaked it enough to run it at +200, Super Pi 32M is good too for the single core boosts. Getting it to run like that in 11 took some figgering, but I got it. All so I didn't have to revert to a static OC just to run TM5

Also, that spike on CCD1 was a freak, 97% of the time it was sitting at 80-81.

I am dual booting 10 and 11, Aida doesn't show such nice L3 numbers with this AGESA.

10 passes at 10GB just to warm it up a little.. the uneven ram temps are from playing a game.. the system is quiet ish.

You really like living on the edge huh I've never seen TM5 error out from CPU instability before

Interestingly enough L3 numbers over here on 11 are the highest they've ever been on 1206 and 1207. About a TB/sec or more on all 3, used to be like 800/sec on Win10 1203 and like 140 when bugged

 

[freeagent]

You really like living on the edge huh I've never seen TM5 error out from CPU instability before

Interestingly enough L3 numbers over here on 11 are the highest they've ever been on 1206 and 1207. About a TB/sec or more on all 3, used to be like 800/sec on Win10 1203 and like 140 when bugged

Yup I was seeing very high L3 numbers in 11 using 1207. 10 was about the same. About the same as running an all core clock of somewhere around 4600-4700MHz all core. Never saw TM5 throw that error? Oh man it was my bane for weeks after I got the cpu lol. To the point I just resorted to a static oc lol. I could get 25 passes in no problem now though, takes a bit with 32GB though..