AMD Raven Ridge 'Zen APU' Thread

[T1beriu]

Although, how can one tell whether a motherboard is friendly to integrated GPU overclocking?

As for no power limit on the motherboard, isn't that ultimately decided by the CPU?

Raven Ridge will use separate VRMs on the motherboard to power the integrated GPU. Low priced cards have only 2 VRMs for the SoC part (that's how it's labeled on the mobos for GPU + Memory + IO bits) while the expensive ones mostly have 4 VRMs.

 

[Valantar]

For the last 4 years the Pro lineup had the same frequency as the non-Pro CPUs. I expect the same to continue with Ryzen Pro.

I know I said Pro would not have XFR, I made the mistake to link the lack of X CPUs with XFR, even though I knew that wasn't the case. I can't find no reasons why it should not be present on a non-X CPU.

If they should lack XFR, I would guess the reasoning to be added stability (a stretch, but ...) and the fact that probably 99.999% of Pro SKUs being installed in OEM SMB/enterprise PCs with stock coolers and relatively limited airflow.

While there might be minor clock speed differences, they're not big enough to matter in the long run. I expect the whole point of buing/using Pro CPUs is for the added validation, long term support, and generally better treatment from AMD.

Raven Ridge will use separate VRMs from the motherboard to power the GPUs. Low priced cards have only 2 VRMs for the SoC part (that's how it's labeled on the mobos for GPU + Memory + IO bits) while the expensive ones mostly have 4 VRMs.

Yeah, I've seen that on all AM4 motherboards. Still, even 2 VRMs of somewhat decent quality should be able to power a reasonably powerful iGPU (if not, AMD would make a better mandatory minimum spec for this, I suspect). But in the end, does this touch on BIOS/CPU specific power limits? After all, on previous architectures, placing a 65W CPU in a motherboard with VRMs for 95W CPUs (as in "all of them") didn't affect the CPU power limit. VRMs are a "hard" limit common to anything installed on the board, but my impression is that most chips lay down their own restrictions below this.

 

[Topweasel]

For the last 4 years the Pro lineup had the same frequency as the non-Pro CPUs. I expect the same to continue with Ryzen Pro.

I know I said Pro would not have XFR, I made the mistake to link the lack of X CPUs with XFR, even though I knew that wasn't the case. I can't find no reasons why it should not be present on a non-X CPU.

Well you were one of the ones who brought it up when I was thinking about it. But I mostly stated the performance angle on the 1500 vs 7500 slide. There was also a mentioning of not throttling which means a more aggressive base clock with reduced turbos. Running XFR on top of that seems even less likely.

But if it really is clocked exactly the same again even more pointless of branding. I guess they will get a little more increased margin from companies that buy them. But like I said earlier even for the business model of most companies they tend not to require anything other than a normal desktop CPU.

 

[T1beriu]

There was also a mentioning of not throttling which means a more aggressive base clock with reduced turbos.

I haven't seen anything AMD mentioned about "not throttling" in their materials or in their talks. Source?

But if it really is clocked exactly the same again even more pointless of branding.

OMG.
When it comes with added features (min 25:05) of course it's normal to change the brand a bit.
It's not called a new completely different name. It just has a "Pro" at the end. That's what's called market segmentation branding.
Valantar explained it very well.

But like I said earlier even for the business model of most companies they tend not to require anything other than a normal desktop CPU.

Well, AMD's results with the Pro lineup completely disagrees with your statement. More than 350 large enterprises/public sector customers use products from the Pro lineup. Source (min 24:10)

 

[Topweasel]

Frankly I don't know why it's this important to you guys to agree with me. The throttling was a suggestion to why the 1500 did so well in the their three tests against the 7500.

I am looking for something, just one thing actually done on the Ryzen CPU, that would require the name change. Higher base, lower turbo was a grasp as straws. The supposed security features already exist on the CPU or require implementation on the motherboard. Which is something that could be done with Ryzen ala a normal i7 in a Latitude or Precision.

What this really comes down to is that I think this is more of a platform branding and should have stayed that way instead of using it as a CPU branding solution.

The use of the CPU by businesses is second to the OEM that supplied them. Sure 300 unique businesses might be using it. But pretty much the only one or two families of HP were ever offered. Every comparison is made to VPro. Vpro isn't a CPU line from Intel. It's a feature support option available on on OEM systems.

But this is going nowhere and is detracting from the regular topic. You don't mind it. I don't like it. Done. I don't know why I can't be allowed to not like a product.

 

[lobz]

I dont think this topic could go any more off :/

Sent from my VTR-L09 using Tapatalk

 

[Valantar]

I dont think this topic could go any more off :/

Sent from my VTR-L09 using Tapatalk

We did try to bring it back on topic a bit up this page, but apparently the awful, awful practice of having a separate business/enterprise line of products took precedence here. Oh well. He said he's done, after all. Perhaps we can all move on?

 

[T1beriu]

First good quality of Raven Ridge die that we can use to count the pixels. Do you know what's that coin? Do you know what resistors are on the package so we can use them as reference?

Source.

 

[tamz_msc]

First good quality of Raven Ridge die that we can use to count the pixels. Do you know what's that coin? Do you know what resistors are on the package so we can use them as reference?

Given that it's Taiwan, maybe it's their currency?

EDIT: New Taiwan Dollar 1 coin.

 

[T1beriu]

Indeed. It's a $1NT with a 20mm r̶a̶d̶i̶u̶s̶ (LE: brainfart, diameter).

LE: I did some photoshop work using the coin as reference. I think the die is around 20 mm x 11 mm, ~220 mm².

As a comparison the Zeppelin die is 22.01 mm x 8.87 mm, 195.228 mm². (Source)

LE2: The package is around 38 mm x 26 mm, ~988 mm².

AM4 is 1600mm². (Source)

L3: Can anyone help with the die areas for the Kaby Lake 4C and Kaby Lake 2C for mobile?

 

[krumme]

While someone calculates mm2; I wondered how they can roll out so many new products so fast. Its like a bulk comming now? But perhaps its just so timely spaced there is enough for the different job specializations?

 

[tamz_msc]

Indeed. It's a $1NT with a 20mm radius.

Diameter I suppose. Otherwise that guy has a huge hand.

 

[Hans de Vries]

Indeed. It's a $1NT with a 20mm radius.

LE: I did some photoshop work using the coin as reference. I think the die is around 20 mm x 11 mm, ~220 mm².

As a comparison the Zeppelin die, in Ryzen, is 22.01 mm x 8.87 mm, 195.228 mm². (Source)

Yes agree, I had roughly 19.5 x 11.5 ~ 224 mm2

 

[krumme]

Yes agree, I had roughly 19.5 x 11.5 ~ 224 mm2

Ok what does that hint to?

 

[Hans de Vries]

Ok what does that hint to?

I think that's what one should expect given AMD's official specs

 

[Valantar]

L3: Can anyone help with the die areas for the Kaby Lake 4C and Kaby Lake 2C for mobile?

I was googling around for that for a post a few weeks ago, it doesn't seem like any numbers (either official or not) really exist, but Skylake GT2 is roughly 122.4 mm². Kaby Lake should be roughly the same, no? According to this TechReport forum post, Skylake GT4 is ~234mm² (they link to an AT news article, but that doesn't contain any numbers - I guess they counted pixels too).

If Raven Ridge with 11CUs is 224mm², I would call that a huge win. I can't imagine that being any less than twice as fast as the Iris Pro 580 (if not quite a bit more, considering the Sorry state of Intel GPU drivers) and with less total area? Holy cow! Now, we know the Zen core is smaller than a KBL/SKL core, but that still means that AMD has some serious density going on here. At the same time, this makes me hope they have a smaller die too, for lower cost SKUs.

 

[krumme]

I think that's what one should expect given AMD's official specs

I meant eg how many CU, performance guess and so on ..
But i see Valantar have a guess.
I didnt see cu in official specs?? Lol i missed it.

 

[imported_jjj]

Given that it's Taiwan, maybe it's their currency?

EDIT: New Taiwan Dollar 1 coin.

Doesn't seem to be the same coin, the writing is different.
The one in the pic is like this

While the one on wiki is

 

[Bouowmx]

Die areas (mm^2) of Intel Kaby Lake small integrated graphics (192 shaders) from Ian Cutress: http://www.anandtech.com/show/11083/the-intel-core-i3-7350k-60w-review
2 cores: 99
4 cores: 126

Possible that Raven Ridge's 11 compute units are from a die that features 12?

 

[tamz_msc]

Doesn't seem to be the same coin, the writing is different.
The one in the pic is like this

While the one on wiki is

Are the diameters same? Because that's what ultimately matters.

 

[imported_jjj]

Are the diameters same? Because that's what ultimately matters.

No idea looking for info but not finding.
Found this pic, the one in the upper right corner has same letters and seems to be 26.5 or 28mm

 

[tamz_msc]

No idea looking for info but not finding.
Found this pic, the one in the upper right corner has same letters and seems to be 26.5 or 28mm

In all probability, it varies by the year it was minted. That's why you are having a hard time.

 

[imported_jjj]

In all probability, it varies by the year it was minted. That's why you are having a hard time.

Could be the other way, the large one could be 20mm and the other one smaller.
If the coin is larger,so is the die and that's not possible.
I was expecting this die at 150-156mm2 or so, 200+ is already plenty large.

 

[T1beriu]

Doesn't seem to be the same coin, the writing is different.
The one in the pic is like this
Snip.
While the one on wiki is

I think it's still the same coin, even if it has different writings. It has the same value and I'm very sure it has the same size & weight because it would be silly to change it. Every machine that uses coins would have to be modified.

2 cores: 99 (2+2)
4 cores: 126 (4+2)

Good find, thanks! I forgot I read that article.

~220 vs 126 mm². Not a good number for the margins.

For mobile, I expect the CPU performance to be around 10% lower that Kaby Lake 7700HQ, but the GPU performance to be 3x better (Kaby Lake has 0.4 TFLOPS on GT2, I expect RR to be clocked around 850 Mhz - 11*64*2*0.850=1.2 TFLOPS - similar to RX 550).

This doesn't sound very good for the mobile version. Not that many mobile CPUs are used for gaming. The margins will be a lot lower when going against Intel in this sector.

For the desktop version (even though I don't remember AMD publicly saying anything about a desktop APU for Ryzen) I expect the CPU to be 5% slower than Kaby Lake 7400 (LE: and 20% slower than 7700K), but 4x better at 1.6 TFLOPS (~1.1 Ghz).

Raven Ridge seems a great buy for e-sports laptops.

 

[imported_jjj]

I think it's still the same coin, even if it has different writings. It has the same value and I'm very sure it has the same size & weight because it would be silly to change it. Every machine that uses coins would have to be modified.



Good find, thanks! I forgot I read that article.

~220 vs 126 mm².

This doesn't sound very good for the mobile version. I expect the CPU performance to be around 10% lower that Kaby Lake, but the GPU performance to be 3x better (Kaby Lake has 0.4 TFLOPS on GT2, I expect RR to be clocked around 850 Mhz - 11*64*2*0.850=1.2 TFLOPS - similar to RX 550).

For the desktop version (even though I don't remember AMD saying anything about a desktop APU for Ryzen) I expect the CPU 5% slower than with Kaby Lake and 4x better at 1.6 TFLOPS (~1.1 Ghz).

As shown in the pic with many coins, there are diff size.

CPU perf is TDP limited not clock limited here (and TDP is about MT not ST) and Ryzen is quite efficient plus we can't quite know if they improve latency and how the uncore power consumption compares to Intel. AMD could end up having the lead.
Vs Kaby Lake, it will wipe the floor with it , that's certain because Kaby Lake is focused on dual core while RR is quad all the way or almost all the way. Next year vs Coffelake, remains to be seen if Intel drops prices for quads and if they manage to deliver 10nm in Q1 so just a quarter after RR.

As for Vega, hard to imagine that it's clocked bellow 1GHz if it was designed for high clocks and ofc clocks will depend on TDP - 12W to 35 or 45W. EDIT: the 15W SKUs likely have 512 cores enabled with 704 at 25W and above.