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

[Potatowithearsontheside]

ACT = All Core Turbo, yes.
Yes, I'm fully aware of silicon lottery, XFR2 functionality, and cooling effects on the latter (including how the correct definition of TDP relates directly to cooling).
What I'm getting at is whether there is anything inherent the higher clocked and core count CPUs that is resulting in AMD feeling the need to slap that 135w TDP on the 3850X. As I said previously, it could be just that they've rated it conservatively. However, if they haven't then there could be a clockspeed issue (like the wall that the 2nd gen hit) with those higher core count CPUs. That's where I'm getting the 4.6GHz figure from. Maybe it is conservative, and maybe there is no wall (which the single core turbo figures might imply doesn't exist), and sure, better cooling without a wall could provide better clocks.
I'm trying to work out if there is anything stopping them from being manually overclocked beyond a certain point. You know, tempering my own expectations.
I don't deny that it is conceivable that these Zen 2 CPUs could clock immensely well. I just won't blind myself into believing it is true until I see it for real. In the meantime, I'll try to pick apart the leaks to see how credible they seem.
At the moment I'm stuck on an ACT of 4.6GHz at 135w.
To be clear, I'm talking about sustainable clocks within the rated TDP.
The reason that I prefer this type of metric is because it is a more honest assessment of how a CPU both clocks and consumes power under a 24/7 load.

 

[humbug]

ACT = All Core Turbo, yes.
Yes, I'm fully aware of silicon lottery, XFR2 functionality, and cooling effects on the latter (including how the correct definition of TDP relates directly to cooling).
What I'm getting at is whether there is anything inherent the higher clocked and core count CPUs that is resulting in AMD feeling the need to slap that 135w TDP on the 3850X. As I said previously, it could be just that they've rated it conservatively. However, if they haven't then there could be a clockspeed issue (like the wall that the 2nd gen hit) with those higher core count CPUs. That's where I'm getting the 4.6GHz figure from. Maybe it is conservative, and maybe there is no wall (which the single core turbo figures might imply doesn't exist), and sure, better cooling without a wall could provide better clocks.
I'm trying to work out if there is anything stopping them from being manually overclocked beyond a certain point. You know, tempering my own expectations.
I don't deny that it is conceivable that these Zen 2 CPUs could clock immensely well. I just won't blind myself into believing it is true until I see it for real. In the meantime, I'll try to pick apart the leaks to see how credible they seem.
At the moment I'm stuck on an ACT of 4.6GHz at 135w.
To be clear, I'm talking about sustainable clocks within the rated TDP.
The reason that I prefer this type of metric is because it is a more honest assessment of how a CPU both clocks and consumes power under a 24/7 load.

They could very well be taking a leaf out of Intel's marketing book, at the end of the day if your competition markets their products with power consumption figures that have hidden caveats and you don't you're putting yourself at a disadvantage.

Having said that why not then just have a lower base clock that matches Intel's quoted TDP but actually run much higher? as Intel do.

Thus far Ryzen TDP has been roughly at its quoted TDP under high loads with no shenanigans, give or take 10%.

Personally i don't think 135 or 125 Watts for the 7nm 16 core parts are outside of the realm of possibilities with AMD's current true power consumption TDP quotes.

 

[Murphy]

At the moment I'm stuck on an ACT of 4.6GHz at 135w.
To be clear, I'm talking about sustainable clocks within the rated TDP.
The reason that I prefer this type of metric is because it is a more honest assessment of how a CPU both clocks and consumes power under a 24/7 load.

I think you maybe reading to much into these leaks and expecting Zen 2 to be similar to Zen+ when for all we know these leaks could be way off the mark.

Trying to draw comparison between Zen2 and Zen+ is pointless IMO, the two are simply too different, one is fabricated on 12nm while the other is 7nm, one incorporates all the I/O within the die the other possibly moves the I/O into a separate die (some people say it may not but I'd be surprised if they drastically altered the design from EPYC 2) and added to all that is we don't know how much dark silicon is going to be used or what the distances are between the active, and in turn heat generating, silicon.

Put it this way, if we took 16 cores and packed them tightly together we wouldn't be able to clock them as high as 16 cores spread over a larger area.

 

[humbug]

They could very well be taking a leaf out of Intel's marketing book, at the end of the day if your competition markets their products with power consumption figures that have hidden caveats and you don't you're putting yourself at a disadvantage.

Having said that why not then just have a lower base clock that matches Intel's quoted TDP but actually run much higher? as Intel do.

Thus far Ryzen TDP has been roughly at its quoted TDP under high loads with no shenanigans, give or take 10%.

Personally i don't think 135 or 125 Watts for the 7nm 16 core parts are outside of the realm of possibilities with AMD's current true power consumption TDP quotes.

I'll put this simply.

The Quoted 105 Watt 2700X are high load runs at about 105 to 115 Watts depending on which reviewer you look at

The 9900K is Quoted 95 Watts, that looks better than 105 Watts, it makes people think Intel CPU's are more power effisient than AMD's, a feather in Intel's cap, the reality is while AMD's 16 thread Mainstream CPU actually runs at around 110 Watt's the Intel 16 threader is 160 Watts and as high as 200 Watts when not throttled by a cheap board or cooling.

Intel are not lying there just not telling you the whole truth.

 

[LePhuronn]

I guarantee the second AMD officially announce Ryzen 3000, there will be people like der8auer who will break open their engineering samples and show the world how it's made.

 

[LePhuronn]

Intel are not lying there just not telling you the whole truth.

Lie by omission, although you could argue that rating the 9900K as a 95W part is a full-fat lie given it never, ever runs that low.

 

[Potatowithearsontheside]

We have:
8c base 4.0GHz 95w
12c base 4.2GHz 105w
16c base 4.3GHz 135w
Each jump has +4c and a minor clock boost.
The additional +4c can only be adding a trivial amount of power (read the IF consuming loads of power and barely increasing with core count).
So why then a big jump of 30w for +4c and 0.1GHz? If we're confident that the additional +4c adds very little power, why is that 0.1GHz so power hungry?
Either the 135w TDP is very conservative, or its the clocks that are kicking up the power.
If the clocks kick up the power by ~20w of that for 0.1GHz, and we know that power efficiency drops as clocks increase, then how much more power is needed to get beyond 4.6GHz?
Even if we kept linear scaling from 4.3-4.6GHz, that'd still be ~20w per 0.1GHz, so we're hitting 195w for 4.6GHz? I don't think so.
It has to be a very conservative TDP IMO.
I appreciate that these are worst case figures in my examples, since they'd be expected to operate below TDP at base frequency, but still, the trend seems a concern.

 

[ivu]

No idea if it's real or not but...from techpowerup forum:

 

[Potatowithearsontheside]

Lie by omission, although you could argue that rating the 9900K as a 95W part is a full-fat lie given it never, ever runs that low.

It can. You simply need the Maximus Hero XI, with XMP On, and the other enhancement Off.
Who would gimp their CPU in such a way though, especially since we know that is able to comfortably operate at higher power with just about any cooling solution you'd likely pair up with such an expensive CPU?
So yes, more honest marketing of its TDP wouldn't go amiss.

 

[LePhuronn]

Why do you assume TDP scales linearly? Like I said before, it's entirely possible that having all 16 cores in an AM4 package saturates the IHS to the point thermals increase higher than the other SKUs. The 8 and 12 core models have some headroom because they're not fully functioning so a linear 10W is fine, but heating up 16 cores gives zero headroom and things bump up a bit more.

 

[Murphy]

We have:
8c base 4.0GHz 95w
12c base 4.2GHz 105w
16c base 4.3GHz 135w
Each jump has +4c and a minor clock boost.
The additional +4c can only be adding a trivial amount of power (read the IF consuming loads of power and barely increasing with core count).
So why then a big jump of 30w for +4c and 0.1GHz? If we're confident that the additional +4c adds very little power, why is that 0.1GHz so power hungry?
Either the 135w TDP is very conservative, or its the clocks that are kicking up the power.
If the clocks kick up the power by ~20w of that for 0.1GHz, and we know that power efficiency drops as clocks increase, then how much more power is needed to get beyond 4.6GHz?
Even if we kept linear scaling from 4.3-4.6GHz, that'd still be ~20w per 0.1GHz, so we're hitting 195w for 4.6GHz? I don't think so.
It has to be a very conservative TDP IMO.
I appreciate that these are worst case figures in my examples, since they'd be expected to operate below TDP at base frequency, but still, the trend seems a concern.

If I've got this right are you asking why there isn't a liner increase in TDP?

 

[humbug]

No idea if it's real or not but...from techpowerup forum:

How long ago?

Its not like that now...

https://www.gigabyte.com/Motherboard/X470-AORUS-GAMING-7-WIFI-rev-10#support-cpu

Bucket of salt with this one.

 

[d_brennen]

Binning. AMD are putting the best performing die in Epyc 2 (low power leakage, best frequency scaling) and progressively "worse" die in consumer parts. I don't really see anything unbelievable in the quoted figures above...

 

[theone8181]

We have:
8c base 4.0GHz 95w
12c base 4.2GHz 105w
16c base 4.3GHz 135w
Each jump has +4c and a minor clock boost.
The additional +4c can only be adding a trivial amount of power (read the IF consuming loads of power and barely increasing with core count).
So why then a big jump of 30w for +4c and 0.1GHz? If we're confident that the additional +4c adds very little power, why is that 0.1GHz so power hungry?
Either the 135w TDP is very conservative, or its the clocks that are kicking up the power.
If the clocks kick up the power by ~20w of that for 0.1GHz, and we know that power efficiency drops as clocks increase, then how much more power is needed to get beyond 4.6GHz?
Even if we kept linear scaling from 4.3-4.6GHz, that'd still be ~20w per 0.1GHz, so we're hitting 195w for 4.6GHz? I don't think so.
It has to be a very conservative TDP IMO.
I appreciate that these are worst case figures in my examples, since they'd be expected to operate below TDP at base frequency, but still, the trend seems a concern.

I was thinking if you compare 8c to 16c theres not double the power draw so seems ok to me?

 

[ivu]

Yes, the post from the colleague from techpowerup who found it was some hours ago and he said at that time that it was already removed from their website.

As i said, just wanted to post it here, but haven't seen it with my own eyes, so...

 

[humbug]

Yes, the post from the colleague from techpowerup who found it was some hours ago and he said at that time that it was already removed from their website.

As i said, just wanted to post it here, but haven't seen it with my own eyes, so...

Ok, thanks for the post.

 

[LePhuronn]

No idea if it's real or not but...from techpowerup forum...

I'm going to be cynical web professional and say fake because there are UI elements present on the Gigabyte website that do not appear in that screen grab.

I've just flicked through all of the English language sites and they all have an additional breadcrumb navigation under the product title (top left) and a compare check box under the top-right menu. This grab has neither. So either this has come from a internal test page which does not have an identical template to the live site, or it's a half-assed Photoshop job - even adding table rows through the browser inspector would retain the navigation elements missing here.

I can speculate and fantasise about what little dubious information we have on Ryzen 3000, but I can dissect the hell out of a web page

Edit: if you scroll the page enough, the menu hides taking those elements with it.

 

[Jokester]

I'm a bit dubious, the F6 BIOS was only released last week.

 

[humbug]

I'm a bit dubious, the F6 BIOS was only released last week.

Yeah i'm putting it down to trolling, mind you its too good of a photoshop tho, the font is identical and exactly the same colour/shade/resolution, no cut and paste lines either, i've had a close look at it, so who knows...

 

[AMDPower]

What is the probability the higher end AM4 board like ASUSVII croshair and Gigabyte X470 7 can take the 16 core cpu? I am assuming its not going to be a VRM issue with these board but could be possibly other issues?