AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

[Panos]

But on stock settings I get 4150ish boost on one core and temps keep spiking to 60’s as soon as there is some light activity going due to vcore jumping to 1.44/1.46v. I can easily do all core 4200MHz and not have the temps go above 50C for the same workloads.

Going all core OC you get more performance and runs cooler. So am I missing something, why would I run it at stock?



Agreed.

I think he is applying is logic to having a 3900X which has high clocks out of the box relative to the lower the stack CPU’s like my 3600. What he is saying applies to him, but not me unless I am missing something?

Just looked at your wig, you have same cpu as him. Lol. What voltage you on for your all core 4.4GHz?

If you OC the Ryzen you should do per core OC not all core

 

[DragonQ]

You won't get UEFI bios on the X470 Taichi and pay a visit to Asrock reddit if you do not believe me that 1004 only on 16mb bios motherboards.

Thanks for that, hadn't seen that post. Even if X470 Taichi gets 1.0.0.4 in November, that means Ultimate version will be December most likely.

 

[TNA]

If you OC the Ryzen you should do per core OC not all core

But why though? So far no one explained that. No good saying things without backing up the reasoning behind it.

Is there a reason why you should not oc all cores?

 

[Panos]

But why though? So far no one explained that. No good saying things without backing up the reasoning behind it.

Is there a reason why you should not oc all cores?

Not all cores on all CCX/CCDs are equal. Especially on 3900X with 2 CCDs.
You have cores doing 4625, others 4600, other 4550 while on 3900X CCD1 will do 4300-4400.

Setting per core speed, utilizes the maximum what each core can do, without the fastest cores getting slowed down by the slowest core.

And here is the irony. Many here doing all core 4200Mhz because 4174 is "too low" and promoting that tiny change of 26mhz.
Yet ignore the fact that they can get an extra 400Mhz on half the CPU if they go manually and set the values for each core.

 

[TNA]

Not all cores on all CCX/CCDs are equal. Especially on 3900X with 2 CCDs.
You have cores doing 4625, others 4600, other 4550 while on 3900X CCD1 will do 4300-4400.

Setting per core speed, utilizes the maximum what each core can do, without the fastest cores getting slowed down by the slowest core.

And here is the irony. Many here doing all core 4200Mhz because 4174 is "too low" and promoting that tiny change of 26mhz.
Yet ignore the fact that they can get an extra 400Mhz on half the CPU if they go manually and set the values for each core.

But you are ignoring the fact that if I do not manually OC all cores to the same I get a lot less performance and on stock the voltages spike to high values causing temps to go much higher than they need to. Also for me setting all my cores to say 4.4GHz nets a big performance boost vs stock.

Is there a reason I should not be doing this?

 

[Panos]

But you are ignoring the fact that if I do not manually OC all cores to the same I get a lot less performance and on stock the voltages spike to high values causing temps to go much higher than they need to. Also for me setting all my cores to say 4.4GHz nets a big performance boost vs stock.

Is there a reason I should not be doing this?

I see your misunderstanding. Per core overclock doesn't mean stock speeds. There is option in Ryzen Master to overclock each core individually

 

[TNA]

I see your misunderstanding. Per core overclock doesn't mean stock speeds. There is option in Ryzen Master to overclock each core individually

My understanding from what you said was don’t bother overclocking at all and leave everything on stock. Basically don’t even use ryzen master.

For me that would mean putting up with higher spikes in temps and lower performance. What I am wondering is, is there something actually wrong in overclocking every core if your CPU is actually capable?

Please keep in mind I only have 6 cores and a lot less heat to deal with also.

 

[Panos]

My understanding from what you said was don’t bother overclocking at all and leave everything on stock. Basically don’t even use ryzen master.

For me that would mean putting up with higher spikes in temps and lower performance. What I am wondering is, is there something actually wrong in overclocking every core if your CPU is actually capable?

Please keep in mind I only have 6 cores and a lot less heat to deal with also.

Depends what each core does. On a 3600 like yours, you won't see much different as the numbers are low. 100mhz here or there is small

But on a 3900X where CCD0 cores are doing 4550-4650 and CCD1 cores 4300-4400 is madness not to do per core overclock but all core overclock from 4200 to 4300 for example.

 

[TNA]

Depends what each core does. On a 3600 like yours, you won't see much different as the numbers are low. 100mhz here or there is small

But on a 3900X where CCD0 cores are doing 4550-4650 and CCD1 cores 4300-4400 is madness not to do per core overclock but all core overclock from 4200 to 4300 for example.

But my 3600 does a lot more than 100mhz as I explained. Example on stock it struggles to hit 4ghz on all cores when all are used, but I can get them all running at 4.4ghz no problem at a much lower voltage. 400mhz oc is not exactly small no?

 

[Illuminist]

Yes I have a 3900x and yes it's 4.4ghz all core usually. If I drop the second ccx to less, I get less performance..... Unless I'm lucky and all my cores can do 4.4.
I don't get it either, if I leave it on auto I lose about 10%...
I've been fiddling for a very long time and this is definitely the fastest, quietest, coolest way to set it up
I'm on 1.625v, max of 80.c is the highest I've seen, usually in games it's 60 ish maybe 70 of the game actually uses my cores

 

[TNA]

Yes I have a 3900x and yes it's 4.4ghz all core usually. If I drop the second ccx to less, I get less performance..... Unless I'm lucky and all my cores can do 4.4.
I don't get it either, if I leave it on auto I lose about 10%...
I've been fiddling for a very long time and this is definitely the fastest, quietest, coolest way to set it up
I'm on 3.625v, max of 80.c is the highest I've seen, usually in games it's 60 ish maybe 70 of the game actually uses my cores

How many volts?

Yeah, it does seem that it comes down to silicon lottery. Like you I am much better of using Ryan master and undervolting and overclocking.

 

[ljt]

How is stability on that overclock @TNA ? How does it cope with AIDA64/OCCT Linpak/AVX etc?

 

[TNA]

How is stability on that overclock @TNA ? How does it cope with AIDA64/OCCT Linpak/AVX etc?

Only tried AIDA64 and it was ok, but I did not run it for hours or anything. Way I see it is as long as it is stable for what I use it for, I am happy. If no games crash, it passes 3DMark, cinebench and blender benchmarks then it’s all good

Long gone are the days where I would run Prime95 etc for hours on end to test. That will only serve to degrade the chip imo. If I get a crash now, I have so much headroom in volts I can just up it a little at a time.

I am only using Ryzen Master for the moment. Not touching bios as I don’t see the point as they will keep bringing out new bios’s over the next year or two which may mess things up anyways.

Lets see what the next AGESA brings along. If it was not for it wanting to use 1.46v to boost for small loads which causes temps to soar into the 60’s then I would stay on stock to be honest. Nothing that I do really even needs the extra power most of the time and when I do need it Ryzen Master is a few clicks away



Oh and @Panos, it was a typo on post 25680, I meant to say light loads on my all core 4.2GHz OC does not break 40C, I said 50C my mistake. So that is around 25C difference there from stock.

 

[ljt]

Only tried AIDA64 and it was ok, but I did not run it for hours or anything. Way I see it is as long as it is stable for what I use it for, I am happy. If no games crash, it passes 3DMark, cinebench and blender benchmarks then it’s all good

Long gone are the days where I would run Prime95 etc for hours on end to test. That will only serve to degrade the chip imo. If I get a crash now, I have so much headroom in volts I can just up it a little at a time.

I am only using Ryzen Master for the moment. Not touching bios as I don’t see the point as they will keep bringing out new bios’s over the next year or two which may mess things up anyways.

Lets see what the next AGESA brings along. If it was not for it wanting to use 1.46v to boost for small loads which causes temps to soar into the 60’s then I would stay on stock to be honest. Nothing that I do really even needs the extra power most of the time and when I do need it Ryzen Master is a few clicks away



Oh and @Panos, it was a typo, I meant to say light loads on my all core 4.2GHz OC does not break 40C, I said 50C my mistake. So that is around 25C difference there from stock.

Fair enough, your voltage is low enough not to cause any damage anyway, in fact 1.25v is lower than the stock all core voltage lol.

The 1.46v is normal though, low core, light workload, low current is perfectly fine. If your fans start to ramp up, just do a custom curve so those small spikes dont initiate a fan spin up. It's how I've set mine, I watched what temps it spikes to under max boost single/dual core, then set a constant 30% speed until about 5 degrees above that temperature before the fans change speed.

 

[Illuminist]

How many volts?

Yeah, it does seem that it comes down to silicon lottery. Like you I am much better of using Ryan master and undervolting and overclocking.

Corrected
Ryan is the master but he also can interfere with bios settings even when you set them to default in Ryan

 

[TNA]

Corrected
Ryan is the master but he also can interfere with bios settings even when you set them to default in Ryan

My reaction remains the same

Haha

Fair enough, your voltage is low enough not to cause any damage anyway, in fact 1.25v is lower than the stock all core voltage lol.

The 1.46v is normal though, low core, light workload, low current is perfectly fine. If your fans start to ramp up, just do a custom curve so those small spikes dont initiate a fan spin up. It's how I've set mine, I watched what temps it spikes to under max boost single/dual core, then set a constant 30% speed until about 5 degrees above that temperature before the fans change speed.

I think on my custom curve it is set to 3C above, will change it to 5C and see how it deals with the stock spiking issue

What would you say is the max safe all core 4.4GHz voltage for 100% workloads like video encoding? Not that I do that.

 

[ljt]

I think on my custom curve it is set to 3C above, will change it to 5C and see how it deals with the stock spiking issue

What would you say is the max safe all core 4.4GHz voltage for 100% workloads like video encoding? Not that I do that.

AMD hasn't really hinted at a "safe" constant voltage, so it's all speculation.

Going off some testing I did with my stock 3700x it settles around 1.35v on none AVX all core load and drops to around 1.22v with heavy AVX load (Intel Burn Test on maximum)

If you say you don't do any intense AVX workloads then with your current settings you might have some room to up the voltage ever so slightly as games on the whole don't really use AVX instructions

 

[TNA]

AMD hasn't really hinted at a "safe" constant voltage, so it's all speculation.

Going off some testing I did with my stock 3700x it settles around 1.35v on none AVX all core load and drops to around 1.22v with heavy AVX load (Intel Burn Test on maximum)

If you say you don't do any intense AVX workloads then with your current settings you might have some room to up the voltage ever so slightly as games on the whole don't really use AVX instructions

Yeah, I hardly used AVX in the 6+ years I had my 4770K.

What I do is look at what the constant voltage is on stock while on like cinebench or something. From what I recall that was around like 1.3/1.35v. So I am happy to go up to 1.3v if I need to. But would rather keep it below as then you get a lot less heat.

 

[4K8KW10]

Exclusive: AMD’s Ryzen 9 3950X Cinebench, No-OC Score – Achieves A 32% Performance Increase Over 1950X
https://wccftech.com/amds-ryzen-9-3...chieves-a-32-performance-increase-over-1950x/

 

[alec]

single thread score 549 looks tasty