[Techpowerup] AMD "Zen" CPU Prototypes Tested, "Meet all Expectations"

[Dresdenboy]

Recall that WCCFTech also said that Skylake would have MorphCore...

Would be more credible if they gave specifics that we could then compare to the actual dies when they are released.

Checked that for you. That was a different author and even then he used a very cautious wording, like "might", "suggest", "educated guess".

I mentioned possible flaws found in the MophCore paper here in this very forum.

 

[Phynaz]

They never said it would have MorphCore for certain, it was just speculation from their side IIRC.

Anyway, you could use your argument for everything, which would mean that once a site has published a single piece of information that later proves to be incorrect, no info on that site can ever be trusted anymore. If you go by that standard, I don't think there are many sites on the Internet that can be trusted, if any.

Yeah, because Pakistan is the center of the tech world.

 

[Abwx]

Perhaps huge in terms of that it has the performance and design properties of a typical high performance "big core" CPU, but the die itself should be quite small according to WccfTech:

"The Zen core itself is surprisingly compact and power efficient. The CPU die itself isn’t particularly large and the Zen FX CPUs will target mainstream price points in which they will compete very aggressively with their upcoming Intel counterparts."

I was talking transistor count wise of course.

 

[Fjodor2001]

Yeah, because Pakistan is the center of the tech world.

Mr BS is back. Didn't you get a vacation due to your latest forum violation only yesterday?

 

[Fjodor2001]

I was talking transistor count wise of course.

Ok, I see. But normally that would result in a large die. Unless AMD is using some kind of high density libraries.

 

[Shehriazad]

Sheesh...can we stop to always blow up AMD threads like this...not to mention stuff always drifts off topic for no reason.


Either you got something to say about AMD Zen that is relevant (or you got a question about it)...or you simply avoid the AMD thread.

Not sure what is so hard here.

I for one would love seeing either new facts/informations or people delivering some well thought out guesstimations based on what we know....that kind of stuff.

 

[cbn]

I'm not sure why a 1024 shader 4C/8T APU shouldn't be possible from AMD within a reasonable die size. Kaveri is 245mm^2 with two Steamroller modules and 512 GCN1.1 shaders on a 28nm process. The move to GloFo's 14nm process probably won't give the expected 4x density, but a doubling of cores and shaders in a 200mm^2 isn't unreasonable.

I believe GloFo's 14nm uses a 20nm BEOL.

20nm BEOL means 2x density increase (assuming perfect scaling).

 

[Phynaz]

Mr BS is back. Didn't you get a vacation due to your latest forum violation only yesterday?

Had to attack me I see, because you couldn't argue with the content of my post.

 

[DrMrLordX]

Sheesh...can we stop to always blow up AMD threads like this...not to mention stuff always drifts off topic for no reason.


Either you got something to say about AMD Zen that is relevant (or you got a question about it)...or you simply avoid the AMD thread.

Not sure what is so hard here.

I for one would love seeing either new facts/informations or people delivering some well thought out guesstimations based on what we know....that kind of stuff.

It is difficult to provide too much in the way of useful speculation about Zen at this point. I will add that HBM/HBM2 is highly probable on any Raven Ridge product, except maybe not on the lowest-priced models. Delta color compression will make higher-speed DDR4 more-acceptable as a memory solution, though I agree that even that will not be enough to keep 1024+ shaders happy.

I fully anticipate that the 200W+ TDP APU that AMD has hinted at all along will have more shaders than that!

 

[Fjodor2001]

Had to attack me I see, because you couldn't argue with the content of my post.

Your post did not have any content, just BS as usual.

 

[Abwx]

Ok, I see. But normally that would result in a large die. Unless AMD is using some kind of high density libraries.

As pointed by some members with twice the density they can keep a module like area, at 14nm this would be 8mm2/core, so about 64mm2 for 8 cores excluding caches and other uncore parts.

A compressed 16T Vishera at same density than Richland would be about 350mm2 so we can expect a 8C/16T to be around 175-200mm2.

 

[jpiniero]

I fully anticipate that the 200W+ TDP APU that AMD has hinted at all along will have more shaders than that!

For sure, but you should assume that's a Phi-like HPC product though and not a product aimed at consumers.

 

[Fjodor2001]

As pointed by some members with twice the density they can keep a module like area, at 14nm this would be 8mm2/core, so about 64mm2 for 8 cores excluding caches and other uncore parts.

A compressed 16T Vishera at same density than Richland would be about 350mm2 so we can expect a 8C/16T to be around 175-200mm2.

Well, then you have to ask the question: What do they mean by that the CPU core is small?

Is it smaller because it's on 14 nm vs 28 nm for previous AMD CPU cores? Or is it small even when considering it's on 14 nm (e.g. if comparing it to a theoretical Excavator core on 14 nm)? The article is not entirely clear on that.

 

[Abwx]

Well, then you have to ask the question: What do they mean by that the CPU core is small?

Is it smaller because it's on 14 nm vs 28 nm for previous AMD CPU cores? Or is it small even when considering it's on 14 nm (e.g. if comparing it to a theoretical Excavator core on 14 nm)? The article is not entirely clear on that.

I think that they say that it s small because it s something like half the size of an EXV module or eventualy a little less.

Samsung version of the A9 has 20.8 miilions transistors per mm2, to compare with Intel s Core M s 15.85 millions/mm2.

That s 31% higher density than what is branded by some people, at odd with all evidences, as the only genuine 14nm process, numbers be damned..

Actualy intel is surely on a 22nm BEOL that was shrinked from a 32nm BEOL used for their 22nm process, the discretanpcy in the density numbers cant be explained otherwise.

 

[ShintaiDK]

I think that they say that it s small because it s something like half the size of an EXV module or eventualy a little less.

Samsung version of the A9 has 20.8 miilions transistors per mm2, to compare with Intel s Core M s 15.85 millions/mm2.

That s 31% higher density than what is branded by some people, at odd with all evidences, as the only genuine 14nm process, numbers be damned..

Actualy intel is surely on a 22nm BEOL that was shrinked from a 32nm BEOL used for their 22nm process, the discretanpcy in the density numbers cant be explained otherwise.

You sure make up things on the go. But again with SA and wccftech as your main information sites I guess its needed.

You forgot what density library is used. Also you forgot that the A9 got 3MB L2 and 8MB L3. While the Core M got 256KB L2 and 4MB L3.

https://www.semiwiki.com/forum/content/3884-who-will-lead-10nm.html

 

[cbn]

The bigger question is how well a dual channel DDR4 interface could keep that kind of APU fed. Even the 512 shader HD7750 showed a massive performance difference between the GDDR5 version and the DDR3-1600 version, and Kaveri keeps on increasing gaming performance with higher frequency RAM up past 2866. Assuming DDR4-3200 is widely available and cheap by then that will help, but four fast Zen cores and a 270X class GPU could be very bandwidth constrained even if thermals and die size look good.

I suspect AMD would target mobile first.

So they probably size the APU for 15W (and secondarily for 35W)

What that comes out to be is anyone's guess. (My prediction was 768sp, but I do think 1024sp would be possible.....but the clocks will be very low on the iGPU @ 15W cTDP)

 

[TechGod123]

Yeah, because Pakistan is the center of the tech world.

Do you have a life?

 

[fourdegrees11]

Speculating on the Zen lineup based on the current FX lineup of 4c/6c/8c. With AMD supposedly moving toward an aggressive pricing strategy vs Intel and their usually no nonsense approach to their line-up I can see Zen being

i3 competitor - 4c/8t with unlocked top models
i5 competitor - 6c/12t with unlocked top models
i7 competitor - 8c/16t with unlocked top models

Priced right at Intel levels for relative clock speeds

 

[jpiniero]

With AMD supposedly moving toward an aggressive pricing strategy vs Intel

Given what AMD is saying right now you have to assume it will be the exact opposite of aggressive. They want to charge as much as they can. You should assume the 8 core will be $999 at launch regardless of performance... they can always cut the price later if (when) it doesn't sell. Intel can pretty easily ruin their plans but I don't think that'll happen since it would only cost Intel money.

 

[Fjodor2001]

Given what AMD is saying right now you have to assume it will be the exact opposite of aggressive. They want to charge as much as they can.

What is it AMD is saying that makes you arrive at that conclusion?

You should assume the 8 core will be $999 at launch regardless of performance...

Really!? Are you assuming it'll be faster than the 5960X (or equivalent available at the time Zen is released) which costs $999, or why else would they price it above that CPU?

Oh, and BTW, the info from the WccfTech article states just the opposite of your personal speculations:

"The CPU die itself isn’t particularly large and the Zen FX CPUs will target mainstream price points in which they will compete very aggressively with their upcoming Intel counterparts. This is a space that’s been traditionally occupied by Intel’s quadcore I7, i5 and dual core i3 CPUs."

 

[ShintaiDK]

The article by wccftech is made up fairy tales. Just like their Prescott to Conroe for Broadwell to Skylake. And all the other rubbish they write. Its a 100% clickbait site for ignorant people.

 

[dark zero]

Speculating on the Zen lineup based on the current FX lineup of 4c/6c/8c. With AMD supposedly moving toward an aggressive pricing strategy vs Intel and their usually no nonsense approach to their line-up I can see Zen being

i3 competitor - 4c/8t with unlocked top models
i5 competitor - 6c/12t with unlocked top models
i7 competitor - 8c/16t with unlocked top models

Priced right at Intel levels for relative clock speeds

I feel that we might see it soon with Zen and Canonlake
All FX are unlocked.

Intel i7 (Hexa) - Zen FX (Octa) -- Both With HT
Intel i5 (Hexa) Without HT -- Zen Phenom Black (Hexa) With HT
Intel i3 (Quad) With HT -- Phenom (Quad) With HT
Intel Pentium (Quad) Without HT - Athlon (Quad) Without HT
Intel Celeron (Dual) With HT - Sempron (Dual) With HT
No competitor from Intel --- Duron (Dual) Without HT

Yeah, resurrecting Duron won't be a bad idea after all.

And calling the tiers with names like Intel could be interesting to see.

---

Err.... Prescott to Conroe was indeed REAL, like the traisition to K8 from AMD side.

 

[Fjodor2001]

The article by wccftech is made up fairy tales

The info comes from their own sources, as mentioned in the article. Got any evidence indicating the opposite?

Also, it's funny you only tend to believe articles from WccfTech and other sites when they suit your own agenda. You're having double standards as usual.

 

[Fjodor2001]

I feel that we might see it soon with Zen and Canonlake
All FX are unlocked.

Intel i7 (Hexa) - Zen FX (Octa) -- Both With HT
Intel i5 (Hexa) Without HT -- Zen Phenom Black (Hexa) With HT
Intel i3 (Quad) With HT -- Phenom (Quad) With HT
Intel Pentium (Quad) Without HT - Athlon (Quad) Without HT
Intel Celeron (Dual) With HT - Sempron (Dual) With HT
No competitor from Intel --- Duron (Dual) Without HT

Just remember that Cannonlake is 2018Q1, while Zen is 2016Q4.

 

[mikk]

The info comes from their own sources, as mentioned in the article. Got any evidence indicating the opposite?

Past clickbait articles are evidence enough. In general, without slides and when they say "own sources" it's most likely rubbish from them.