5800X Boost set to 5050Mhz all cores with AMD Curve Optimizer

[gerardfraser]

I do not know man, I have like 3 NVME and M.2 SSD in the Tomahawk motherboard. I use a PCI-E Adaptor for extra NVME in case that internet sleuth is watching what I type.

 

[schoenzy]

I also tried -15 on all cores except strong cores (0 & 6) which I set -5, LLC on mode 4, +200Mhz offset

This held out in R20, but crashed on the menu in a game.

I would be interested swapping chips with Gerard / anyone who seems to have a 'good one' if only to see if the different motherboard has any affect. Of course, this is not possible (well it is but isnt going to happen)

I dont know how AMD / Intel can get away selling chips that are really good / really crap for the same price. Surely just test each one to make sure it can hit X without failing. Basic quality control

I've seen a lot people not able to get stable with -15 all core or 2 best cores or any cores, and that's including other forums. -5 seems to be the universal given on all cores. As suggested in this thread earlier, I got it stable at -5 best 2 cores, -15 the rest, but that was only after setting PBO scalar to 10x. I also have my DRAM/SOC/IOD/CCD/VDDP voltages manually tuned to support my mem and FCLK overclock. LLC is auto. Cranking out pretty solid numbers

Spoiler

Ryzen DRAM calc FAST preset with Rec voltages etc.

 

[ViRuS2k]

fixed my issue with a bios update lol

1.225v @ 4.575ghz fast enough for me with a 29k multi r23 score @79c
and single core performance all cores reaching @5.05 and lots exceeding 5.1ghz and 5.224ghz with a ingame temp of around 50c

now i have a new issue HAHAHA i can no longer get my Infinity fabric @1900mhz for some reason with this bios so im going to have to tune a new fabric stable speed + memory damnt there is always something you fix one thing something else breaks...

 

[Guest2]

I've seen a lot people not able to get stable with -15 all core or 2 best cores or any cores, and that's including other forums. -5 seems to be the universal given on all cores. As suggested in this thread earlier, I got it stable at -5 best 2 cores, -15 the rest, but that was only after setting PBO scalar to 10x. I also have my DRAM/SOC/IOD/CCD/VDDP voltages manually tuned to support my mem and FCLK overclock. LLC is auto. Cranking out pretty solid numbers

Spoiler

Ryzen DRAM calc FAST preset with Rec voltages etc.

Thanks. I tried similar with my voltages from DRAM calc. Stable for a CPUZ validation but crashing instantly running R20.
I'm still thinking this is a BIOS thing. PBO limits to motherboard set really high VDDG voltages, a lot more than DRAM calc max.

fixed my issue with a bios update lol

1.225v @ 4.575ghz fast enough for me with a 29k multi r23 score @79c
and single core performance all cores reaching @5.05 and lots exceeding 5.1ghz and 5.224ghz with a ingame temp of around 50c

now i have a new issue HAHAHA i can no longer get my Infinity fabric @1900mhz for some reason with this bios so im going to have to tune a new fabric stable speed + memory damnt there is always something you fix one thing something else breaks...

What motherboard do you have out of interest?

 

[Robzere31]

fixed my issue with a bios update lol

1.225v @ 4.575ghz fast enough for me with a 29k multi r23 score @79c
and single core performance all cores reaching @5.05 and lots exceeding 5.1ghz and 5.224ghz with a ingame temp of around 50c

now i have a new issue HAHAHA i can no longer get my Infinity fabric @1900mhz for some reason with this bios so im going to have to tune a new fabric stable speed + memory damnt there is always something you fix one thing something else breaks...

So is your single core boosting to 5ghz+ automatically when your set at 4.7ghz all core?
Once I change any setting on mine the single core speed gets locked to the all core speed.

 

[ubasti]

I've seen a lot people not able to get stable with -15 all core or 2 best cores or any cores, and that's including other forums. -5 seems to be the universal given on all cores. As suggested in this thread earlier, I got it stable at -5 best 2 cores, -15 the rest, but that was only after setting PBO scalar to 10x. I also have my DRAM/SOC/IOD/CCD/VDDP voltages manually tuned to support my mem and FCLK overclock. LLC is auto. Cranking out pretty solid numbers

Spoiler

Ryzen DRAM calc FAST preset with Rec voltages etc.

Interesting. A larger negative offset on the weaker cores seems counter intuitive to what AMD is suggesting in their presentation but it's obviously working for you so I may give it a try to see if the multi-core performance improves.

My memory is next on the list. I have it running at 1900 on an untuned XMP profile, was dithering over whether to tune further incase the next AGESA would better support 2000 FCLK (crashes my machine right now), but might do it anyway.

 

[ubasti]

I also tried -15 on all cores except strong cores (0 & 6) which I set -5, LLC on mode 4, +200Mhz offset

This held out in R20, but crashed on the menu in a game.

I would be interested swapping chips with Gerard / anyone who seems to have a 'good one' if only to see if the different motherboard has any affect. Of course, this is not possible (well it is but isnt going to happen)

I dont know how AMD / Intel can get away selling chips that are really good / really crap for the same price. Surely just test each one to make sure it can hit X without failing. Basic quality control

I guess the devil's advocate / manufacturer's perspective is if it runs to the advertised performance (e.g. 4.8 on a 5900x or 4.9 on a 5950x) you are getting what you paid for and anything else is a bonus. It is still frustrating to hit a brick wall though.

 

[Guest2]

So is your single core boosting to 5ghz+ automatically when your set at 4.7ghz all core?
Once I change any setting on mine the single core speed gets locked to the all core speed.

Exactly the same for me. Manual OC locks single core boost to 4.7Ghz if 4700Mhz is set

Interesting. A larger negative offset on the weaker cores seems counter intuitive to what AMD is suggesting in their presentation but it's obviously working for you so I may give it a try to see if the multi-core performance improves.

My memory is next on the list. I have it running at 1900 on an untuned XMP profile, was dithering over whether to tune further incase the next AGESA would better support 2000 FCLK (crashes my machine right now), but might do it anyway.

I thought that too. AMD presentation was to set a higher negative curve on stronger cores, not weaker. I'm going to try -15 on my 2 strong cores and -5 on the rest

I guess the devil's advocate / manufacturer's perspective is if it runs to the advertised performance (e.g. 4.8 on a 5900x or 4.9 on a 5950x) you are getting what you paid for and anything else is a bonus. It is still frustrating to hit a brick wall though.

If i bought a shirt that was £50, I would expect the same shirt that someone else bought for £50 to be identical, not have better stitching than mine. CPUs should be no different.
Same goes for other tech, if I bought speakers for £500 i would expect the components to be the same in the same models. Not different speakers but the same model having more capable components that can be driven to produce better quality sound

 

[ubasti]

If i bought a shirt that was £50, I would expect the same shirt that someone else bought for £50 to be identical, not have better stitching than mine. CPUs should be no different.
Same goes for other tech, if I bought speakers for £500 i would expect the components to be the same in the same models. Not different speakers but the same model having more capable components that can be driven to produce better quality sound

I don't disagree but it does pose some interesting dilemas. Assuming that manufacturers can't completely eradicate variation in the silicon without tanking their yields and their entire business model, I guess is the answer for AMD and Intel to bin for golden samples and price them accordingly? How different fundamentally is to what is currently done with say, 3600 / 3600x / 3600xt? Is there a risk you end up paying the same for an 'average' CPU and a premium for the golden sample?

 

[ViRuS2k]

Thanks. I tried similar with my voltages from DRAM calc. Stable for a CPUZ validation but crashing instantly running R20.
I'm still thinking this is a BIOS thing. PBO limits to motherboard set really high VDDG voltages, a lot more than DRAM calc max.




What motherboard do you have out of interest?

MSI x570 godlike

 

[ViRuS2k]

So is your single core boosting to 5ghz+ automatically when your set at 4.7ghz all core?
Once I change any setting on mine the single core speed gets locked to the all core speed.

Yeah thats the same on mine i had to use a combination of scaler and pbo limits motherboard + curve optimizer and 200mhz max to get those clocks @1.225v
before the bios update fix i couldnt use use vcore voltage @ override mode as soon as i used override in any capacity my clocks would instantly stick at base clocks *3.44ghz* with no boost or all core speeds at all... but i updated the bios out of chance and that fixed the issue some how.

 

[Guest2]

I don't disagree but it does pose some interesting dilemas. Assuming that manufacturers can't completely eradicate variation in the silicon without tanking their yields and their entire business model, I guess is the answer for AMD and Intel to bin for golden samples and price them accordingly? How different fundamentally is to what is currently done with say, 3600 / 3600x / 3600xt? Is there a risk you end up paying the same for an 'average' CPU and a premium for the golden sample?

Yes, I guess you could say I have paid for x and be happy you are getting that. I wouldbe happier paying for something I knew i would get more / less from

MSI x570 godlike

Thanks

Well, I tried with -5 on all and -15 on two best cores. R20 instant crash.
Tried the same with positive 0 and -15 on two best cores. R20 instant crash.

Also tried with PBO motherboard limits, auto limits and manual voltages.

I'll wait to see what a new BIOS brings when one is released for the MEG Unify. PBO on Auto is working fine (which is wasnt on the previous BIOS)

 

[alec]

If i bought a shirt that was £50, I would expect the same shirt that someone else bought for £50 to be identical, not have better stitching than mine. CPUs should be no different

Thats why there is spec and there is silicon lottery. This is overclockers, nothing is guaranteed.

Besides, I am yet to see anyone with a true lemon CPU. All reports of bad cases can be chalked up to motherboard, bios, bios settings, cooling. All Zen 3 are wildly overperforming their spec.
I would even go on a limb and say gerards seemingly stellar results are not so much because of silicon lottery, but because of a very good motherboard.

Btw setting more negative offset on good cores is not always a good idea, whatever the slides say. For me, two good cores I keep at 0 and experiment with worse ones, to bring their actual clocks higher.

 

[ViRuS2k]

Managed to get 4.650 all core and 5.2ghz single core @1.225 with all core temps @73c and single at 46c

i must have a really good chip. i can go as far back as neg 20-30 with 200mhz and motherboard limits enabled and still have no instability.

i need to fix my Infinity fabric clock cant lock it at 1900 anymore on this bios though lol 30+k on all core r23
40k on winrar benchmark lol

 

[alec]

i can go as far back as neg 20-30 with 200mhz and motherboard limits enabled and still have no instability.

Don't want to be the one to burst your bubble. But have you checked your performance matches the apparent clock speeds?
Whats your SC MC Cinebench scores at those settings?

I suspect you have a case of clock stretching. CPU sees that it is not getting enough volts and drops performance, although clocks seem to be high. It was covered in depth last year with Ryzen 3000 undervolting hype.

 

[ViRuS2k]

Don't want to be the one to burst your bubble. But have you checked your performance matches the apparent clock speeds?
Whats your SC MC Cinebench scores at those settings?

I suspect you have a case of clock stretching. CPU sees that it is not getting enough volts and drops performance, although clocks seem to be high. It was covered in depth last year with Ryzen 3000 undervolting hype.

well im gettinng over 30+k on r23 seems right for 4.650 all core.
and boost it boosting way past its 4.9 to 5.1 and 5.2 almost 5.3 on some cores lol and most on 5.1-5.2

single core r20 668

boost seems to be working also im getting a lot of ranges. though your probably right i might be getting a little clock stretching but im still happy with the results and temps.

 

[alec]

over 30+k on r23 seems right for 4.650 all core.

Think that is stock 5950X performance territory, at 4.3-4.4GHz. So it does look like clock stretching.

But your single core looks better than fine... And that is stable? No random reboots?

 

[ViRuS2k]

Think that is stock 5950X performance territory, at 4.3-4.4GHz. So it does look like clock stretching.

But your single core looks better than fine... And that is stable? No random reboots?

no random reboots no,
so how do i fix this clock stretching inn all core then i have never heard of this clock stretching stuff lol am i just not supplying enough vcore ?? as i have headroom

 

[Guest2]

I tried negative 5 on 6 cores and positive 0 on my two known good cores. Motherboard PBO limits, 200Mhz offset
Everything seemed fine, R23 passed, CPUZ validation, R20 passed, Prime95 max temps 86c after ~15 mins.
5050Mhz boost, 4.7Ghz all core
Seemed great, scores were all about right then failed a third run of R20.

So, not stable and something is crapping it out.

 

[LtMatt]

I tried negative 5 on 6 cores and positive 0 on my two known good cores. Motherboard PBO limits, 200Mhz offset
Everything seemed fine, R23 passed, CPUZ validation, R20 passed, Prime95 max temps 86c after ~15 mins.
5050Mhz boost, 4.7Ghz all core
Seemed great, scores were all about right then failed a third run of R20.

So, not stable and something is crapping it out.

Might take a while, but why don't you try testing 1 core at a time if you are struggling to get stability by making mass changes.