AMD Curve Optimizer any guides / experience

[outpt]

Software to find out how well my Ryzen cores are for PBO2 CO (Negative)?

Is there a software that can find out how well all my Ryzen cores are for PBO2 CO (Negative)? Because Ryzen master only shows the best 2 cores not how good the other ones are. Should I use Clocktuner for Ryzen (4.18 in) : to see how good all my Ryzen cores are? (5800x) Thanks

www.techpowerup.com

2nd video.

 

[jesdals]

i just gave up on curve optimizer and started running all core at 4.7ghz using mb voltage settings. which are around 1.3v using gooseberry benchmark to check temps and voltages. at idle about 1.1v or lower gets about 80-81C fill tilt. Or run it pbo enabled and it run 4.85ghz. have not had blue screen error in a week.
there is one thing i have noticed is that i can set it to -30 all cores and it runs fine until i put it to sleep. next morning coming out of sleep it's blue screens to the max. somebody explain that.

I am haveing some issues with cold boots after disconnection the power to my system - do switch of the power as well or only sleep mode?

 

[Cowboystrekk]

Try minus -15 on all cores and -30 on your best core then -29 on the other one then check your long term stability.

Problem is that the second best core is the - 29 core, the ranking doesnt add up.

 

[tabascosauz]

Problem is that the second best core is the - 29 core, the ranking doesnt add up.

Obviously the core quality ranking doesn't mean anything once you've already changed the V-F curves by applying CO offsets.

Makes no sense comparing between cores, because every core comes binned very differently and per-core CO changes each core only relative to its own stock V-F. On mine Core 1 only does -5, it takes Core 2 @ -30 just to reach similar 4.925-4.95GHz clocks to Core 1 @ -5. But if you were comparing the two you'd think that Core 2 was significantly "better" for having a much bigger undervolt right?

If anything, it should be the other way around. I'd assume the top rated cores are already performing close to their full potential (since they are always preferred by CPPC) - so minimal room for undervolting. Whereas the other cores (esp. on CCD2) can be relaxed and won't matter whether AMD pushes them hard or not out of the box, since they'll never see more than all-core speeds. I've been 3 out of 3 on this so far, 5600G, 5700G and 5900X all have lower achievable offsets on the 2 best cores.

Which is why while a broad, say, -15 undervolt across the board is understandable if you don't have time to test, it also logically makes no sense because -15 on Core 0 is not going to mean the same thing as -15 on Core 5 for example. It's a good place to start before you fine-tune, though, because it gives you a good sense of where every core stands currently.

 

[Cowboystrekk]

Obviously the core quality ranking doesn't mean anything once you've already changed the V-F curves by applying CO offsets.

Makes no sense comparing between cores, because every core comes binned very differently and per-core CO changes each core only relative to its own stock V-F. On mine Core 1 only does -5, it takes Core 2 @ -30 just to reach similar 4.925-4.95GHz clocks to Core 1 @ -5. But if you were comparing the two you'd think that Core 2 was significantly "better" for having a much bigger undervolt right?

If anything, it should be the other way around. I'd assume the top rated cores are already performing close to their full potential (since they are always preferred by CPPC) - so minimal room for undervolting. Whereas the other cores (esp. on CCD2) can be relaxed and won't matter whether AMD pushes them hard or not out of the box, since they'll never see more than all-core speeds. I've been 3 out of 3 on this so far, 5600G, 5700G and 5900X all have lower achievable offsets on the 2 best cores.

Which is why while a broad, say, -15 undervolt across the board is understandable if you don't have time to test, it also logically makes no sense because -15 on Core 0 is not going to mean the same thing as -15 on Core 5 for example. It's a good place to start before you fine-tune, though, because it gives you a good sense of where every core stands currently.

I have the time to test and - 15 all core would reduce performance vs - 30 and +50 which I know us stable i get about 200MHz more allcore vs stock using - 30 on all. Stock AC lies at 4.4, - 30 gives me 4.6 avg. I can try reducing a bit on my best core and see if this resolves the issue If anyone has a tip for a low liad pr core test which invokes maximum voltage/core clock I bet that could find out which core is the slightly unstable one.

 

[HD64G]

Different type of workloads allows more voltage and frequency for Ryzen CPUs. Needs some experimenting (time and patience) to find the best balance. Don't try for the ultimate. Go for the stable first and work from there a bit more. Settle after a few tests and have fun with your PC. Don't waste many hours for practically nothing. Friendly advice.

 

[Cowboystrekk]

Different type of workloads allows more voltage and frequency for Ryzen CPUs. Needs some experimenting (time and patience) to find the best balance. Don't try for the ultimate. Go for the stable first and work from there a bit more. Settle after a few tests and have fun with your PC. Don't waste many hours for practically nothing. Friendly advice.

I tweak for fun, and have a lot of patience, only things misding is the tools to identify which core is the problem I have a safe profile to fall back on

Got a tip on another forum, eventviewer from reboots. That worked! Found out both core 0 and 1 had caused a couple of reboots, but I only adjusted core 0. Also adjusted core 1 now, hopefully that fixes it.

 

[outpt]

I am haveing some issues with cold boots after disconnection the power to my system - do switch of the power as well or only sleep mode?

coming out of sleep mode or complete shut down. it's runs for hrs. or days till i do any kind of restart/shut down. at this point i have thrown in the towel. besides i only have a xfx 5600xt so its not really being pushed that hard. at this point pbo or ac overclock works just fine. to be honest i haven't seen it do anything at all. O'Well

 

[purecain]

What a thread, amazing work overclocking guys. I want the higher single core performance now , and that isnt possible on air.imo

Lucky for me I have decent silicon. Since AMD released a (beta)bios that literally turned the pump off, on two of my water-cooling kits before I realised the culprit ive been on air.

But after reading only half of this thread I've decided I need that single core performance.

I'll check out the rest of the thread and see which cooling has been the most successful. Arctic cooler 2 AIO stuck in my head, and I'll probably give that company a try.

 

[Ibizadr]

Hey guys I have a r7 5800x and play with some bios settings to try to achieve the best of this cpu. In my CO I'm with
Max boost: +200mhz
PPT:200
TDC:145
EDC:145
CO: - 15 all core except 2 best one with - 10
Try these settings today on cinebench and got this any way to improve since and don't get full pbo in all cores? I have an artic aio 240mm and got in multi thread almost 80ºc.
Its possible to achieve 5.0ghz in a single thread with 5800x?
Some advice to improve?

 

[Mussels]

Hey guys I have a r7 5800x and play with some bios settings to try to achieve the best of this cpu. In my CO I'm with
Max boost: +200mhz
PPT:200
TDC:145
EDC:145
CO: - 15 all core except 2 best one with - 10
Try these settings today on cinebench and got this any way to improve since and don't get full pbo in all cores? I have an artic aio 240mm and got in multi thread almost 80ºc.
Its possible to achieve 5.0ghz in a single thread with 5800x?
Some advice to improve?
View attachment 218442

I can get 5050Mhz on mine, remember that the curve undervolt going too far can actually lower the clocks, as the chip doesnt get enough voltage to ramp up

your numbers may simply be too high, for your cooling and VRMs. lower them.

 

[tabascosauz]

Hey guys I have a r7 5800x and play with some bios settings to try to achieve the best of this cpu. In my CO I'm with
Max boost: +200mhz
PPT:200
TDC:145
EDC:145
CO: - 15 all core except 2 best one with - 10
Try these settings today on cinebench and got this any way to improve since and don't get full pbo in all cores? I have an artic aio 240mm and got in multi thread almost 80ºc.
Its possible to achieve 5.0ghz in a single thread with 5800x?
Some advice to improve?

Your EDC is like, stock, but what's up with 200W? ~180W is a metric ton of power to be putting through the cores on even a 5950X, let alone a single chiplet, don't matter if you have an AIO.....

I'm betting a 5800X can achieve peak performance with a smart combination of PBO settings at just ~120W. Stock 142W is honestly already too much per-core power for a 5800X, that's why you can usually lop off a whole bunch of PPT and be no slower (or even faster).

Having a high PPT does nothing for your single threaded performance. Higher PPT enables higher multi thread perf if you have the cooling for it, but scalar, curve optimizer on the relevant 1 or 2 cores, and EDC makes a bigger difference to ST perf. PPT you only need to make sure is "high enough" not to hold back your ST (which is like 75W for single chiplet, easy peasy)

PBO +200 should cap out at 5050 for 5800X, so if your chip can handle it then there's no reason why it can't be done.

get @mtcn77 in here, not sure what the rule is, is it like EDC = PPT x 0.9?

 

[Mussels]

the post in my sig has some info from GerKNG, he suggested using:

PPT: 105W
TDC:70A
EDC: 95A

stock performance, but doesnt let the chiplet blow out into the thermal throttling ranges which can actually let it boost higher
clearly you can go higher, those are very efficient settings not the max performance ones (add 20 to each, for that)

 

[RandomJerk]

Anyone here has any experience with overclocking a 5950x and testing stability on Linux? I tried PBO+CO settings and tested stability on a Windows install, and everything seemed fine, till I went to Linux and started seeing crashes immediately. So, all my CO values ended up being thrown out of the window on Linux. FYI, I have a Gigabyte X570 AORUS PRO WIFI.

 

[Cowboystrekk]

Hey guys I have a r7 5800x and play with some bios settings to try to achieve the best of this cpu. In my CO I'm with
Max boost: +200mhz
PPT:200
TDC:145
EDC:145
CO: - 15 all core except 2 best one with - 10
Try these settings today on cinebench and got this any way to improve since and don't get full pbo in all cores? I have an artic aio 240mm and got in multi thread almost 80ºc.
Its possible to achieve 5.0ghz in a single thread with 5800x?
Some advice to improve?
View attachment 218442

Try the limits on auto (142W should be more than enough) and rather see if you can improve the UV. Try core cycler and find lowest values, if you get intermittent crashes during low loads even with 3 passes in corecycler, check the event viewer for whea 18 and find out which lower, lower the UV by 1.

 

[purecain]

This is my starting point. I dont how good these scores are as i havnt compared them yet but this is on air as

I have it set up now.

I'm looking forwards to seeing how close I can get to higher single core performance on a Noctua NH-U12a.

Water-cooling should give me 200-400mhz hopefully at across all cores. My single core run had the cpu running at 4.69 to 4.72ghz through out on air.

 

[outpt]

Hey guys I have a r7 5800x and play with some bios settings to try to achieve the best of this cpu. In my CO I'm with
Max boost: +200mhz
PPT:200
TDC:145
EDC:145
CO: - 15 all core except 2 best one with - 10
Try these settings today on cinebench and got this any way to improve since and don't get full pbo in all cores? I have an artic aio 240mm and got in multi thread almost 80ºc.
Its possible to achieve 5.0ghz in a single thread with 5800x?
Some advice to improve?
View attachment 218442

pushing those kind of numbers means(probably) nothing. you need to scroll down further until you find ppt,edc,tdc and see what is actually being used. bring those numbers down.
PPT-125W TDC-90A EDC-115A are good starting points. EDC should be about 90% of PPT.

 

[Ibizadr]

So after some advice of you guys i change some settings. Now I'm at:
Max boost : 200mhz
PPT: 180
TDC: 95
EDC: 140

Only have time to do a test with cinebench in single core. I'm happy with the temp it reaches in single core I think 60 was not too hot and in game RTSS never show temps high.
For now I want to improve Single core since this pc its most for gaming.
I post a ss of hwinfo that I run with cinebench.

Do you guys have some advices for my single core UV+Pbo?

I see now some more advices and later I will go to turn down my settings to see if I got improvments

 

[jesdals]

Hey guys I have a r7 5800x and play with some bios settings to try to achieve the best of this cpu. In my CO I'm with
Max boost: +200mhz
PPT:200
TDC:145
EDC:145
CO: - 15 all core except 2 best one with - 10
Try these settings today on cinebench and got this any way to improve since and don't get full pbo in all cores? I have an artic aio 240mm and got in multi thread almost 80ºc.
Its possible to achieve 5.0ghz in a single thread with 5800x?
Some advice to improve?

Remember Alt+Printscreen thats easier than the phone. And please fill in your system specs - perhaps some one have the same motherboard etc.

I used to ad to one core at the time ( a pain on the 5950x) and wrote down how the changes affected in intervals of 5 to begin with and then 2.

 

[Ibizadr]

Remember Alt+Printscreen thats easier than the phone. And please fill in your system specs - perhaps some one have the same motherboard etc.

I used to ad to one core at the time ( a pain on the 5950x) and wrote down how the changes affected in intervals of 5 to begin with and then 2.

If you see my last post before your reply you see a screenshot. After that I change my settings after a couple of people telling me 180w was too much for my 5800x. I change the settings for the screenshot in this post but not tested yet with these settings

 

[outpt]

have a msi mag b550 tomahawk and my bios is very similar.

 

[jesdals]

@Ibizadr
I would suggest that you try to look at your Infinity settings - if your memory allow for 3600MHz settings it would be a large boost - with the right memory 1900/3800MHz Infinity settings should be possible on that board.

 

[Ibizadr]

@Ibizadr
I would suggest that you try to look at your Infinity settings - if your memory allow for 3600MHz settings it would be a large boost - with the right memory 1900/3800MHz Infinity settings should be possible on that board.

Yeah I think it's possible since it's a Samsung b die. Ryzen dram calculator only have zen 2, the settings for zen 3 works? I will try to overclock my ram since my system it's stable at practically stock settings.
Best way to overclock ram its disabling xmp or use xmp and only change my Infinity fabric dram speed with dram calculator?

 

[outpt]

I think I may have stumbled on to may bsod problems. I think my gpu is on the way out. Today I started getting the GSOD it is my understanding that this may be gpu and or gpu memory related. I have turned down the power limits to 80% and memory timings down also. I have 3 cores running between -20 to -28. All other cores are at 0. I haven’t had any death screens in 5 hrs knock on wood. This at the moment means nothing but I have this gpu go belly up once before and took it apart and cleaned it up the thermal paste was as hard as a brick and it worked. After trying everything with the cpu and no whea errors no memory errors and so forth. If this gpu fails I am going to walk out into 5pm Dallas traffic

sumbitch my gpu is failing. all i have is a gtx660 backup

 

[Mussels]

Windows key + shift + s uses the new snipping tool in W10 and W11, you can then paste that immediately into TPU with no effort (it's awesome)

answering all the mess of posts above: a 5800x just cant use some of those numbers that high, so letting it max out for no gain isn't helping (0.01% faster for 10C, when the power and heat savings could give a bigger gain) - and black/grey screen crashes.... yeah core undervolting can be like that.