TR Forums

So heise.de got this picture from AMD's Computex press event:



It purports to show a Raven Ridge APU chip (below the coin) next to our friend The Threadripper's package.

The coin appears to be a $1 Taiwanese coin with a known diameter of 20mm ( https://en.wikipedia.org/wiki/New_Taiwan_dollar )

Thanks AMD! That's all I need to do some pixel counting!

And the numbers I'm coming up with, which are of course not 100% precise but are probably a good ballpark estimate, are that the Ryzen APU die is about 20 mm wide (e.g. very close to the diameter of that coin) and about 11 mm in height. That of course gives us an area of about 220 mm^2 for a quad-core APU.

To put that into perspective, a quad-core Kaby Lake is about 121 mm^2, and the best estimates I've seen of the RyZen (8-core) die are putting it in the 200 to 210 mm^2 range. So this APU is not tiny by any stretch of the imagination. One would presume that most of the area is the onboard graphics.

It's bound to thump the preceding generation handily, so long as the clock speeds are reasonable. I do wonder what the performance delta between GCN 1.0 and the integrated Vega graphics will amount to. According to some things I've seen around the DDR4 max (supported) clock will be 2400 MHz. Dual-channel, that gives 38.4 GB/second, so *maybe* GTX 750 levels? Ish? That wouldn't be half bad, and a competent quad core part with that IGP would be a very nice backbone for a home theater PC.

chuckula wrote:

To put that into perspective, a quad-core Kaby Lake is about 121 mm^2

Is that the Kaby Lake-S die size? That may not be a fair comparison. If this is AMD's largest Raven Ridge then it would be competing with Intel's GT4 dies.

EndlessWaves wrote:

chuckula wrote:

To put that into perspective, a quad-core Kaby Lake is about 121 mm^2

Is that the Kaby Lake-S die size? That may not be a fair comparison. If this is AMD's largest Raven Ridge then it would be competing with Intel's GT4 dies.

The standard Kaby Lake quad-core die size (e.g. any chip that you've seen reviewed on TR like a 7700K) is about 121mm^2. I'm not sure what "fair" means here. I have absolutely no doubt that the Raven Ridge parts will have substantially stronger IGP performance compared to a 7700K, for whatever that's worth (the 7700K was never a "high end" part even by Intel IGP standards).

How and if Intel responds (and you could say Raven Ridge is a response to Intel GT4 parts in the first place, they sure launched first) on the graphics front remains to be seen.

The tech press has been calling all of these launches "responses" when in fact they've been in various stages of life for way too long for them to be a response to anything. How does Intel respond to Ryzen? With stuff they've been working on for 24 months or more. How does AMD respond to Intel? With stuff that's been in the pipe just as long. I mean geez.

derFunkenstein wrote:

The tech press has been calling all of these launches "responses" when in fact they've been in various stages of life for way too long for them to be a response to anything. How does Intel respond to Ryzen? With stuff they've been working on for 24 months or more. How does AMD respond to Intel? With stuff that's been in the pipe just as long. I mean geez.

It's the state of news today: when there are two parties involved, everything turns into football and short attention span theater.

Concupiscence wrote:

derFunkenstein wrote:

The tech press has been calling all of these launches "responses" when in fact they've been in various stages of life for way too long for them to be a response to anything. How does Intel respond to Ryzen? With stuff they've been working on for 24 months or more. How does AMD respond to Intel? With stuff that's been in the pipe just as long. I mean geez.

It's the state of news today: when there are two parties involved, everything turns into football and short attention span theater.

It's not just the press. Go read the comments on virtually any AMD or Intel article here.

DancinJack wrote:

Concupiscence wrote:

derFunkenstein wrote:

The tech press has been calling all of these launches "responses" when in fact they've been in various stages of life for way too long for them to be a response to anything. How does Intel respond to Ryzen? With stuff they've been working on for 24 months or more. How does AMD respond to Intel? With stuff that's been in the pipe just as long. I mean geez.

It's the state of news today: when there are two parties involved, everything turns into football and short attention span theater.

It's not just the press. Go read the comments on virtually any AMD or Intel article here.

These companies are great! Nobody ever does anything unless its to respond to something somebody else did.
It makes you wonder how the industry ever got started in first place.... who were they responding to!?!?!

It was the Big Bang of technology. Some caveman got angry with another caveman. He picked up a rock. Bang!

The rest is history.

DancinJack wrote:

It's not just the press. Go read the comments on virtually any AMD or Intel article here.

In fairness to the readers, they kinda get sucked into the narratives in the press. In fairness to you and your point, readers are generally kinda stupid.

derFunkenstein wrote:

DancinJack wrote:

It's not just the press. Go read the comments on virtually any AMD or Intel article here.

In fairness to the readers, they kinda get sucked into the narratives in the press. In fairness to you and your point, readers are generally kinda stupid.

Agreed on all counts. I don't just mean TR either, but if they really understood how this industry works they would have a different viewpoint (I hope). That's not to say there aren't actual reactions from some big companies. Pricing is a good one that fluctuates depending on competition, albeit not always as much as some would hope. But to think Intel is going to respond!!!! to Raven Ridge with CannonLake parts this fall is just foolish.

Absolutely. Pricing is how companies "respond". AMD "responded" to reality rather than its hype when setting Ryzen's prices. The reality is that it's still not quite up to Intel in per-thread performance so their CPUs reflect it, regardless of what the hype machine's Blender results say. In graphics, it's "responding" to the reality of Nvidia's superior efficiency, so roughly-equivalent performance is a bit cheaper because it uses more power. In other words, it prices its stuff to sell based on what is competing. That's not a great place to be with either market; you'd much rather be the standard. AMD just isn't.

edit: also, without pricing info, performance, and power consumption, we're all just kind of spinning our wheels here. Yes it's bigger than Kaby Lake quads, but it's also going to have some quantity of graphics performance. Presumably better graphics performance.

chuckula wrote:

EndlessWaves wrote:

chuckula wrote:

To put that into perspective, a quad-core Kaby Lake is about 121 mm^2

Is that the Kaby Lake-S die size? That may not be a fair comparison. If this is AMD's largest Raven Ridge then it would be competing with Intel's GT4 dies.

The standard Kaby Lake quad-core die size (e.g. any chip that you've seen reviewed on TR like a 7700K) is about 121mm^2. I'm not sure what "fair" means here. I have absolutely no doubt that the Raven Ridge parts will have substantially stronger IGP performance compared to a 7700K, for whatever that's worth (the 7700K was never a "high end" part even by Intel IGP standards).

The chip pictured is apparently either a mobile or embedded part as AM4 packages are square. As such it's most likely aimed to compete with Intel's HQ processors which I believe are larger dies than the 7700k et al.

It just seemed a bit misleading to compare it to desktop parts.

EndlessWaves wrote:

chuckula wrote:

EndlessWaves wrote:

Is that the Kaby Lake-S die size? That may not be a fair comparison. If this is AMD's largest Raven Ridge then it would be competing with Intel's GT4 dies.

The standard Kaby Lake quad-core die size (e.g. any chip that you've seen reviewed on TR like a 7700K) is about 121mm^2. I'm not sure what "fair" means here. I have absolutely no doubt that the Raven Ridge parts will have substantially stronger IGP performance compared to a 7700K, for whatever that's worth (the 7700K was never a "high end" part even by Intel IGP standards).

The chip pictured is apparently either a mobile or embedded part as AM4 packages are square. As such it's most likely aimed to compete with Intel's HQ processors which I believe are larger dies than the 7700k et al.

It just seemed a bit misleading to compare it to desktop parts.

The package is most certainly mobile, although the chip itself could of course be put on an AM4 package for a desktop system.

As for the "HQ" processors are you actually referring to the larger GPU parts that have the crystalwell cache? Those exist but as you point out they are not desktop processors with the exception of the Broadwell i5 and i7 parts. I don't have the link handy but I remember doing an analysis that came up with a die size on the order of 180 mm^2 or so for an i7 5775C, which is of course bigger than Kaby Lake.

Most "HQ" processors don't actually have the large GPU and the silicon is pretty much identical to a 7700K. The package, however, is larger because Intel puts the southbridge chip on the package.

Here's the "7700HQ": https://ark.intel.com/products/97185/In ... o-3_80-GHz

Here's the plain 7700K: https://ark.intel.com/products/97129/In ... o-4_50-GHz

derFunkenstein wrote:

The tech press has been calling all of these launches "responses" when in fact they've been in various stages of life for way too long for them to be a response to anything. How does Intel respond to Ryzen? With stuff they've been working on for 24 months or more. How does AMD respond to Intel? With stuff that's been in the pipe just as long. I mean geez.

I would disagree to a certain extent. The 18 core i9 is a response to Thread Ripper. Intel most certainly had that die in the wings for the server market under the Xeon name. However, the consumer version is very much a reaction to Thread Ripper. I strongly suspect that back in March Intel had zero plans for releasing the Core i9 up to 18 cores.

the wrote:

I would disagree to a certain extent. The 18 core i9 is a response to Thread Ripper. Intel most certainly had that die in the wings for the server market under the Xeon name. However, the consumer version is very much a reaction to Thread Ripper. I strongly suspect that back in March Intel had zero plans for releasing the Core i9 up to 18 cores.

Agree with all but that last sentence. Pretty sure Intel always had a contingency plan for a "Core i9", a brand that could be reserved for if AMD ever got its **** together again and produced something sufficiently good that Intel's strategy of increasing margins (by sandbagging on core count for as long as possible) would cease to cut it and they'd need to compete again.

I've always thought that the i9 nomenclature was going to happen eventually due to Intel marketing insanity. It does have some meaning now that the actual core design is slightly different than the main stream designs (see cache topology and AVX-512). I would have thought that even if Intel were to remain unchallenged in the high end, we'd still see the i9 branding as their marketing department would have demanded the change to make something appear better than the previous i7 generation. Rebrand ahoy!