Updated AMD Ryzen 3000 chipset drivers and power profile

[TheLostSwede]

The base clock is 300MHz higher, VS the 3700X.

The thing is, with the design of Ryzen 3000, it seems no chip really runs at that base clock, as all the Ryzen 7 and Ryzen 9 parts at least, seem to hover around 4.0-4.2GHz, regardless of what you're running. It doesn't matter if you load one core, or all cores, as you showed above.
So even a higher base clock, seems to have no impact on the average clock speeds of the chips.

The thing is that this is really no different than the first gen Ryzen chips. The 1700 was a 65W TDP and the 1800X was a 95W TDP. Even though the 1800x was faster as base OC the 1700 to 3.9 would put it near to the 1800x which had a maximum 4.1 GHZ OC based on what I have seen.

Well, the 1700 had worse silicon than the 1700X and 1800X, as it simply wouldn't clock as high. Again, it was 100-200MHz difference, but it was a very hard ceiling at 3.9GHz for the 1700, whereas the X parts could hit 4.0-4.1GHz.

The situation is seemingly reversed now, with the 3700X hitting the same speeds as the 3800X, which is just plain odd imho, or the first batch of 3700X chips are actually 3800X parts or something along that line...

 

[HTC]

The thing is, with the design of Ryzen 3000, it seems no chip really runs at that base clock, as all the Ryzen 7 and Ryzen 9 parts at least, seem to hover around 4.0-4.2GHz, regardless of what you're running. It doesn't matter if you load one core, or all cores, as you showed above.
So even a higher base clock, seems to have no impact on the average clock speeds of the chips.

It will if both PBO and XFR are disabled.

Think of it like the 9900K being forced to "obey" the 95W TDP: pretty much the same thing.

 

[TheLostSwede]

That 65W TDP means jack squat.

Out of the box, no tweaks or whatever, the CPU consumes up to ~95-100W or so on 100% load, measured by the motherboard (complete SoC, not just cores) ... so basically the amount of heat that the cooler needs to dissipate.

You can restrict it to 65W intentionally, but that would also make it A LOT slower.
I tested that, with 65W it only boosts all-core to about 3.7-3.8, and single core below 4.2. Nowhere near the 4.4 written on the box.

Add 10W and you get 3800X - that is, if it's not a dud like the one from TheLostSwede
The difference between them is extremely minor, it's not that one is better quality than the other... it's not.


But that is not what happens. The CPU boosts between 4.0 to 4.1 during CB R20. My sample anyway. Yes, with PBO disabled. And it consumes more than 65 for sure while doing that...
Either my ASUS board is "dumb" for not intentionally choking the CPU to that 65W, or the CPU is simply designed to go higher than 65W from the factory.

(And that seems to be inline with reviewers, including TPU - The difference is that they have AC rooms, not my hot 30+ degrees attic.
I'm quite sure that during the winter my CPU will boost to 4.2 as well during CB run)

This is exactly the 9900K story with it's 95W TDP which in reality is about 150+ when actually using the CPU...
(which btw it would lose badly to 3700X if it would actually run at that 95W instead of getting hotter than the surface of the sun to clock to that magical 5.0 Ghz)

Basically, both vendors are selling now "factory overclocked" CPUs.
End of the story.

Keep in mind that AMD's TDP and what they say the actual max power draw is, differs quite a bit, as pointed out above.
You shouldn't break 88W though, unless you enable PBO.

My problem is that my chip, regardless of PBO on or off, just won't do what it's supposed to do, which is annoying and not at all what I expected. If i enabled PBO, it draws more power, but doesn't clock higher...

It will if both PBO and XFR are disabled.

Think of it like the 9900K being forced to "obey" the 95W TDP: pretty much the same thing.

Never seen my CPU run at 3900MHz.
Idle looks like this

A light load looks like this

 

[jesdals]

It seems like the speed also depends on the workload tested with HWiNFO64 during CPUid Bench and it never went above 4200 on all cores, then did an our of gaming and got these timings

During the CPUid Benchmarks the picture where as follows

 

[TheLostSwede]

It will if both PBO and XFR are disabled.

Think of it like the 9900K being forced to "obey" the 95W TDP: pretty much the same thing.

For some reason PBO was still enabled, apparently auto in the UEFI means on for Gigabyte...
No difference at idle though.

1 core load

All core load

 

[kapone32]

For some reason PBO was still enabled, apparently auto in the UEFI means on for Gigabyte...
No difference at idle though.

View attachment 128169

1 core load

View attachment 128170

All core load

View attachment 128171

I seriously thing your problem comes down to temp. 74 degrees is pretty high for an AM4 chip. 1.45 volts on a single core is really high too.

 

[Zareek]

It seems like the speed also depends on the workload tested with HWiNFO64 during CPUid Bench and it never went above 4200 on all cores, then did an our of gaming and got these timings
View attachment 128166
During the CPUid Benchmarks the picture where as follows
View attachment 128168

Yeah, I see the same kind of thing with my 3800X. I think @TheLostSwede either got a dud or one of the programs on his machine is stealing too much cpu time to allow single core boosting. I regularly hit 4.5Ghz and even 4.55Ghz but it doesn't last long. I did a little experiment last week, I watched the clocks on my second screen while playing F1 2018. I had an average of over 4.3Ghz all core for my over 4 hour gaming session.

 

[EarthDog]

I seriously thing your problem comes down to temp. 74 degrees is pretty high for an AM4 chip.

No, it isn't.

Ryzen from the beginning could easily handle more than that. Now, previous AMD CPUs had a much lower limit, but Ryzen was all 100C Tjmax IIRC....

 

[TheLostSwede]

It seems like the speed also depends on the workload tested with HWiNFO64 during CPUid Bench and it never went above 4200 on all cores, then did an our of gaming and got these timings

During the CPUid Benchmarks the picture where as follows

Gaming makes no difference, still stuck at 4,425MHz as the highest boost speed...

I seriously thing your problem comes down to temp. 74 degrees is pretty high for an AM4 chip. 1.45 volts on a single core is really high too.

That's 100% load on all cores...

As you can see, it's not that hot during gaming.

Yeah, I see the same kind of thing with my 3800X. I think @TheLostSwede either got a dud or one of the programs on his machine is stealing too much cpu time to allow single core boosting. I regularly hit 4.5Ghz and even 4.55Ghz but it doesn't last long. I did a little experiment last week, I watched the clocks on my second screen while playing F1 2018. I had an average of over 4.3Ghz all core for my over 4 hour gaming session.

I have killed off all the background tasks that I can and the remaining uses almost zero CPU.

Then there's this... Although most of it is known already.

Our Tests Show Not All Ryzen 3000 Cores Are Created Equal

Not all cores in AMD's Ryzen 3000-series processors can reach the boost frequency.

www.tomshardware.com

 

[Wavetrex]

Article about this.

Our Tests Show Not All Ryzen 3000 Cores Are Created Equal

Not all cores in AMD's Ryzen 3000-series processors can reach the boost frequency.

www.tomshardware.com

Edit... was that an edit or it was just posted as I was posting this ?
Anyway...

----

The thing is, it appears to boost to that advertised frequency only on the "fastest core".
If your single thread software is running on any other core than the best one, it will not go to max boost.

So this is not a general "max", but a very specialized max, that is relevant to only one core out of all.

Very very shady marketing this time...

@TheLostSwede maybe you can try to Set Affinity to each one of the cores individually and run tests ( 8 times ? ), to see if there's any difference between cores ?

(p.s. - I see that in your case that is Core 8, the golden star)

 

[TheLostSwede]

Our Tests Show Not All Ryzen 3000 Cores Are Created Equal

Not all cores in AMD's Ryzen 3000-series processors can reach the boost frequency.

www.tomshardware.com

Article about this.
(not sure if it was posted by anyone else before)

The thing is, it appears to boost to that advertised frequency only on the "fastest core".
If your single thread software is running on any other core than the best one, it will not go to max boost.

So this is not a general "max", but a very specialized max, that is relevant to only one core out of all.

Very very shady marketing this time...

@TheLostSwede maybe you can try to Set Affinity to each one of the cores individually and run tests ( 8 times ? ), to see if there's any difference between cores ?

(p.s. - I see that in your case that is Core 8, the golden star)

I've tried that, makes no difference...

 

[Wavetrex]

I've tried that, makes no difference...

Oh man, that reminds me of my previous 6800K, which only went to 4.1 before crashing and had a terrible voltage wall, while other people were OC'ing those to 4.5.
Wanted to pull my eyes out when seeing those results...

Feels so bad to have a dud...

But in the end, it's just $70 and 100Mhz, so no big deal really... or you can try to return it on the reason that it doesn't work as advertised.. and see what happens.

 

[TheLostSwede]

Oh man, that reminds me of my previous 6800K, which only went to 4.1 before crashing and had a terrible voltage wall, while other people were OC'ing those to 4.5.
Wanted to pull my eyes out when seeing those results...

Feels so bad to have a dud...

But in the end, it's just $70 and 100Mhz, so no big deal really... or you can try to return it on the reason that it doesn't work as advertised.. and see what happens.

It's a $100 here...
So $1 per MHz...

Returning is not going to happen, as I'm past that deadline, unless AMD accepts it as an RMA.

And yes, my 8th core is the fastest one, but I have yet to see it boost any higher than the supposedly not so fast.
In fact, core 2 and 7 (if you count 1-8 rather than 0-7) are the fastest when it comes to boost clocks.

Also, the so called "thread pinning" is clearly not working in my case, as the fastest cores are not the ones being utilised the most.

 

[Zareek]

I have killed off all the background tasks that I can and the remaining uses almost zero CPU.

Okay, so you got a dud apparently. I'm very sorry and I bet you aren't the only one, hopefully AMD will resolve your problem. The only reason I said something about it again was in your screenshot two other cores were awake. You won't see the the full boost unless they are asleep.

 

[Wavetrex]

So $1 per MHz...

Still cheaper than Inkjet ink...

Well if returning is not an option, we can then make fun of your dud for the next 10 pages

The pains and sorrows of early adopters ...

 

[TheLostSwede]

Okay, so you got a dud apparently. I'm very sorry and I bet you aren't the only one, hopefully AMD will resolve your problem. The only reason I said something about it again was in your screenshot two other cores were awake. You won't see the the full boost unless they are asleep.

View attachment 128173

Eh? Other people have clearly had more than one core awake to be able to boost that high, just look at post #29. He was playing games while his CPU was boosting to 4.5GHz.

Also, at idle on the desktop...

And here's repeating your test with the affinity set to what's supposed to be my fastest core...
In other words, a bunch of bollocks.

 

[Zareek]

Eh? Other people have clearly had more than one core awake to be able to boost that high, just look at post #29. He was playing games while his CPU was boosting to 4.5GHz.

That is a good point but we don't really know what else was happening when it hit those boosts. I just noticed your screenshot had the two awake and mine had 7 asleep. I'm not even sure how I even ran Ryzen Master, took a screenshot and ran SuperPi with 7 cores asleep. Like I said, I think you got a dud. Hopefully, AMD gives you a replacement chip. This new boosting system they have is pretty cool but it really makes these higher clocked SKUs worthless. I wouldn't have paid an extra $70 for a 1% improvement in performance. That being said, I still think if the 3700X didn't exist, I would think the 3800X was a great value. I think the real impact is on the people who upgrade just for the sake of their next overclocking fix. This boosting system basically makes overclocking worthless unless you want to do LN2 but that is a whole different sort of beast.

 

[jesdals]

Btw gaming is done in post 29 with Division 2 in 5700x1200 with all settings on max except for FOG, thats on the second highst setting.

Early review etc said that Ryzen Master could be a bug, so I run without and I dont use the Gigabyte Apps software either, everything else is setup according to the optimization from this video

Though I dont overclock but use the PBO as he recommend in the video

 

[TheLostSwede]

That is a good point but we don't really know what else was happening when it hit those boosts. I just noticed your screenshot had the two awake and mine had 7 asleep. I'm not even sure how I even ran Ryzen Master, took a screenshot and ran SuperPi with 7 cores asleep. Like I said, I think you got a dud. Hopefully, AMD gives you a replacement chip. This new boosting system they have is pretty cool but it really makes these higher clocked SKUs worthless. I wouldn't have paid an extra $70 for a 1% improvement in performance. That being said, I still think if the 3700X didn't exist, I would think the 3800X was a great value. I think the real impact is on the people who upgrade just for the sake of their next overclocking fix. This boosting system basically makes overclocking worthless unless you want to do LN2 but that is a whole different sort of beast.

See updates above, with the second screenshot being Super Pi with nothing else loading the CPU.

I don't have an issue with how the CPUs work, or the overall performance of the CPU I bought, I have an issue with the fact that it doesn't deliver what AMD says it should deliver.
Had I known there was no discernible difference, I wouldn't have wasted the extra $100 either and that's what's pissing me off.

 

[kapone32]

No, it isn't.

Ryzen from the beginning could easily handle more than that. Now, previous AMD CPUs had a much lower limit, but Ryzen was all 100C Tjmax IIRC....

Understood I personally have never had a CPU go that high but 100C is very comforting.

 

[TheLostSwede]

Btw gaming is done in post 29 with Division 2 in 5700x1200 with all settings on max except for FOG, thats on the second highst setting.

Early review etc said that Ryzen Master could be a bug, so I run without and I dont use the Gigabyte Apps software either, everything else is setup according to the optimization from this video

Though I dont overclock but use the PBO as he recommend in the video

First post in this thread has a link to the latest version, as well as the latest drivers, which is what I'm using.
Anything else, doesn't report the proper clocks either, so I don't know what to trust any more.
Hardware Info can't see the low clocks or cores that are asleep...

I guess I should ask Steponz, he might have an idea or two what I might be able to try...

 

[kapone32]

First post in this thread has a link to the latest version, as well as the latest drivers, which is what I'm using.
Anything else, doesn't report the proper clocks either, so I don't know what to trust any more.
Hardware Info can't see the low clocks or cores that are asleep...

I guess I should ask Steponz, he might have an idea or two what I might be able to try...

You have a Corsair H115!. You could download the Corsair Link software (First version) and it does a pretty good job of providing the voltage, clock and temps of the CPU. In terms of raw CPU performance I use Windows task manager performance tab and set the CPU view for all cores. It provides clock speed and usage of each individual core.

 

[EarthDog]

You have a Corsair H115!. You could download the Corsair Link software (First version) and it does a pretty good job of providing the voltage, clock and temps of the CPU. In terms of raw CPU performance I use Windows task manager performance tab and set the CPU view for all cores. It provides clock speed and usage of each individual core.

I've found Hwinfo64 to be the most accurate, next to Ryzen master...... which doesn't seem to want to work properly on the Biostar X570GT8.

ick on 3rd party software like that being accurate... those tend to be hit or miss.

Where do you see each clock speed listed for each core in task manager/performance? It shows a general speed, but I don't think it lists individual core speeds.....

 

[Zareek]

See updates above, with the second screenshot being Super Pi with nothing else loading the CPU.

I don't have an issue with how the CPUs work, or the overall performance of the CPU I bought, I have an issue with the fact that it doesn't deliver what AMD says it should deliver.
Had I known there was no discernible difference, I wouldn't have wasted the extra $100 either and that's what's pissing me off.

Well I obviously don't have an answer for you then. It seems to me you didn't get what you paid for and you have a right to be pissed. I would ask AMD to make it right.

 

[kapone32]

I've found Hwinfo64 to be the most accurate, next to Ryzen master...... which doesn't seem to want to work properly on the Biostar X570GT8.

ick on 3rd party software like that being accurate... those tend to be hit or miss.

Where do you see each clock speed listed for each core in task manager/performance? It shows a general speed, but I don't think it lists individual core speeds.....

It won't give you the individuals clocks but you will see usage for each core. You have to change the view to all logical processors. I will check again but I think it might give you core speeds too. I will confirm when i get home today.