Question Susquehanna's second chat with Charlie Demerjian

[uzzi38]

https://www.reddit.com/r/AMD_Stock/comments/eczmzx

This is not a good picture Charlie's painting of whatever's left of Intel's roadmap.

For the record, this is also significantly worse than any expectations I ever had as well. Far, far worse.

 

[Glo.]

I won't and don't tell others what they can or can't listen to, but I choose to pay attention to people that seem most capable of calling the truth the truth no matter what. Anand Shimpi before moving on to other things was that guy. Now it's people like Stephen Burke.

Im sorry to say this to you, but Stephen Burke has nowhere near the same type of sources, that Charlie has.

And its not only about Intel. Its about a lot of this tech industry.

 

[lobz]

There was a lot more than Charlie to tell us 10nm was not going well, it was widely analyzed and reported.

Biased sources only tell you the truth when it aligns with their biases (and even then may be the part of the truth which best serves their agenda). It's like Rush Limbaugh or Rachel Maddow, the agenda just oozes like an irrepressible slime.

If I have a broken clock, once in a while it will look perfectly accurate. Still doesn't mean I should rely on it.

I won't and don't tell others what they can or can't listen to, but I choose to pay attention to people that seem most capable of calling the truth the truth no matter what. Anand Shimpi before moving on to other things was that guy. Now it's people like Stephen Burke.

YMMV. Best holiday wishes

Stephen Burke is one of the most awesome reviewers out there. Little too all-in on horribly optimized adobe products, but that's not his fault. Other than that, he doesn't have much idea of what's going on behind the scenes (or just opts never to talk about that) and I believe it is by his choice, so he can keep himself out of this all. Nevertheless, they're not in the same category, so even though what you say about him is absolutely true, it has nothing to do with the quality of Charlie's sources or what S/A does.

 

[Arkaign]

Im sorry to say this to you, but Stephen Burke has nowhere near the same type of sources, that Charlie has.

And its not only about Intel. Its about a lot of this tech industry.

And that may be a philosophical difference. Which is fine. It's like politics, "super secret sources say XYZ! Subscribe to find out more!!", it just doesn't interest me much if I know the person is extremely motivated to spin things a certain way.

Listen to what you prefer. What honest difference should it make, for me, or for you? I'm not offended, you do you.

 

[mikk]

6-core samples - still goes to show that yields cannot be that terrific .

No there was a 14C sample as well and we have no idea what samples Intel is sampling. The point is Intel is already sampling ICL-SP to some OEMs obviously which DrMrLordX denied.

 

[Glo.]

And that may be a philosophical difference. Which is fine. It's like politics, "super secret sources say XYZ! Subscribe to find out more!!", it just doesn't interest me much if I know the person is extremely motivated to spin things a certain way.

Listen to what you prefer. What honest difference should it make, for me, or for you? I'm not offended, you do you.

But you realize that if Charlie will blatantly lie, or be dishonest he will lose subscribers?

Very high-level people even from Intel are subscribed to his site. Why do you think that is?

 

[ksec]

Reading this... How can the man live with himself.... The sheer amount of negativity and joy over their failures

The man creates nothing

From the content point of view I dont Get the icelake status
Icelake notebooks are everywhere it doesnt look like capacity problem

I dont understand why people hate negativity. Do you want everything to be sugar coated? Especially for a Company's business practice that was made famous because of its Only Paranoid Survives mentality.

There are things he said that is wrong, like 10nm is dead, ( as in Intel will scrap it and go to 7nm ), or Nokia / Ericsson will go bankrupt. But there is no denying in Intel 10nm is having its problem. The Icelake volume we are getting now for Xmas has been in the work for the whole of 2019.

And if quantity or yield were good enough, you can trust one of the premium vendor that practice Just in Time Stock management to perfection would have updated their Laptop.

 

[Arkaign]

But you realize that if Charlie will blatantly lie, or be dishonest he will lose subscribers?

Very high-level people even from Intel are subscribed to his site. Why do you think that is?

First off, it's no secret that Charlie has far from a spotless track record, and regularly infuses things with hyperbole and frankly silly hostility and drama. That's neither here nor there though.

As for 'high level subscribers', well of course. If you are some kind of bigwig at Intel/etc in the industry, you obviously have plenty of money, in that respect his fees are no more than a regular person would think of buying a bag of cookies from a vending machine. It would obviously have some interest if what the person was saying could fuel rumor/news and potentially stock valuation and so forth.

TMZ is very often correct as well. Doesn't mean I care much about front running rumor mongering or the style of journalism it entails. Different philosophical perspective.

 

[Glo.]

As for 'high level subscribers', well of course. If you are some kind of bigwig at Intel/etc in the industry, you obviously have plenty of money, in that respect his fees are no more than a regular person would think of buying a bag of cookies from a vending machine. It would obviously have some interest if what the person was saying could fuel rumor/news and potentially stock valuation and so forth.

It has zero to do with having more money than brain.

It has everything to do with information Charlie provides.

 

[Arkaign]

It has zero to do with having more money than brain.

It has everything to do with information Charlie provides.

Didn't say anything about it being more money than brain.

#1 - As a 'high level' Intel/Industry person, the money charged for 'Semi' accurate is beyond negligible. Non factor.

#2 - If S/A does indeed have some influence on news/rumor/Industry views, which I believe it can, given how sourcing can spread like a virus (remember AdoredTV and the 'leaked' Zen2 lineup and prices? Repeated like a game of telephone and ended up being fanfiction), then what S/A puts out there can have an impact on consumer, partner, and investor attitudes. Therefore, stock prices, OEM contracts, etc. Keeping up with their grist mill becomes almost compulsory. You can bet the Democratic candidates watch Fox News, if only to keep up with the attitude and spin of the opposing perspective.

I just don't have much personal interest or value in content put forth from what I feel is a fundamentally dishonest motivation. Just as I dislike partisan politics, or similar tribal attitudes that crop up all over the place. From cars to sports teams to hamburgers to video cards to game consoles and beyond, you have no shortage of people viewing it and spinning it in a needlessly combative, tribal manner.

 

[lobz]

Didn't say anything about it being more money than brain.

#1 - As a 'high level' Intel/Industry person, the money charged for 'Semi' accurate is beyond negligible. Non factor.

#2 - If S/A does indeed have some influence on news/rumor/Industry views, which I believe it can, given how sourcing can spread like a virus (remember AdoredTV and the 'leaked' Zen2 lineup and prices? Repeated like a game of telephone and ended up being fanfiction), then what S/A puts out there can have an impact on consumer, partner, and investor attitudes. Therefore, stock prices, OEM contracts, etc. Keeping up with their grist mill becomes almost compulsory. You can bet the Democratic candidates watch Fox News, if only to keep up with the attitude and spin of the opposing perspective.

I just don't have much personal interest or value in content put forth from what I feel is a fundamentally dishonest motivation. Just as I dislike partisan politics, or similar tribal attitudes that crop up all over the place. From cars to sports teams to hamburgers to video cards to game consoles and beyond, you have no shortage of people viewing it and spinning it in a needlessly combative, tribal manner.

You didn't understand Glo, I think. Money is indeed a non-factor and if Charlie doesn't have actual insider sources, nobody subscribes, high-level or not, or the sub being $100,000,000 or free. No reason to bring up AdoredTV here, because he got more things right from the Zen 2 launch than anybody else. I mean ANYBODY. People were laughing at him on all forums when he first heard and told us that chiplets are coming to the desktop, and articles have been written on tech sites, explaining why is that not possible and financially not viable.
As for the final Ryzen 3000 specs and prices, it's not like he told anyone to bet money on the specs, they were clearly stated by him as leaks, which are always subject to change, depending on the success of the launch stepping. The best thing is, when things changed and his sources told him otherwise (Zen 2 clocks are lower than expected), he made a video about that too. Before launch. As for the prices? Numbers in an excel sheet. I'm not sure you've heard about Jensen deciding Pascal pricing not too long before he entered the stage...
Getting a bit 'fashion to follow'-like to hate on Charlie and Jim here, it's like if people could feel themselves smarter because of that... seems very strange to me.

 

[LikeLinus]

Sounds like some people are misinterpreting bias. It's not about how accurate or inaccurate his claims are. They go both ways. The point is that he does not post with an open mind. There's an agenda to it. He really only post negative comments about Intel. He doesn't point out the positives, but focuses solely on the negatives.

The first 3 questions:

He goes from "10nm is still technically non-existent."
Following that up with "10+ nm Ice Lake are technically on the shelves"

Well that's a whole lot of "technical" comments that are completely contradicting, lol.

 

[lobz]

Sounds like some people are misinterpreting bias. It's not about how accurate or inaccurate his claims are. They go both ways. The point is that he does not post with an open mind. There's an agenda to it. He really only post negative comments about Intel. He doesn't point out the positives, but focuses solely on the negatives.

The first 3 questions:

He goes from "10nm is still technically non-existent."
Following that up with "10+ nm Ice Lake are technically on the shelves"

Well that's a whole lot of "technical" comments that are completely contradicting, lol.

Again......................................... the conversation was about confirmation bias, and how is it confirmation bias when his sources are legit.

For the other parts of your comment: if you say so, lol.

 

[soresu]

From cars to sports teams to hamburgers to video cards to game consoles and beyond, you have no shortage of people viewing it and spinning it in a needlessly combative, tribal manner.

I've got to find out what hamburgers can inspire tribal thinking!

 

[LikeLinus]

Again......................................... the conversation was about confirmation bias, and how is it confirmation bias when his sources are legit.

For the other parts of your comment: if you say so, lol.

Again.

Confirmation bias is the tendency to search for, interpret, favor, and recall information in a way that affirms one's prior beliefs or hypotheses.

He most definitely does this, as mentioned in my post. He actively post negative Intel articles and never anything towards the positive.

 

[moinmoin]

He actively post negative Intel articles and never anything towards the positive.

"Intel's engineers are very good."

 

[lobz]

Again.

Confirmation bias is the tendency to search for, interpret, favor, and recall information in a way that affirms one's prior beliefs or hypotheses.

He most definitely does this, as mentioned in my post. He actively post negative Intel articles and never anything towards the positive.

The word spinner has spoken, I see. Have fun talking to yourself.

 

[Failnaught]

OK, so maybe Charlie is biased. Maybe not. Whatever...

The crucial question is: how long will it take Intel to design a chiplet-based CPU? However good Intel 10+++ or 7 nm nodes are, yields would have to be *really* good before you can scale up to (say) 64 cores and match AMD's pricing on one die. How long do you think this will take? I'm willing to bet that Intel is already working on it. In fact, some of their work in interconnects seems to be pointing in this direction already. What do you think?

 

[jpiniero]

OK, so maybe Charlie is biased. Maybe not. Whatever...

The crucial question is: how long will it take Intel to design a chiplet-based CPU?

Pretty soon actually but still massive dies. Cooper Lake is rumored to be two 14 nm 28 core dies and a 10 nm IO die all connected via EMIB.

 

[Markfw]

Pretty soon actually but still massive dies. Cooper Lake is rumored to be two 14 nm 28 core dies and a 10 nm IO die all connected via EMIB.

Now see here is the problem IMO....

Intel is RUMORED to have 56 cores on a chip that is no where close to real production, and AMD has had 64 cores for 6 months, and by the time Intel has that one in real production, AMD will be live with their next generation.

 

[DrMrLordX]

No there was a 14C sample as well and we have no idea what samples Intel is sampling. The point is Intel is already sampling ICL-SP to some OEMs obviously which DrMrLordX denied.

AMD stated publicly that they were sampling Milan back in October 2019. Intel has never claimed publicly to be sampling IceLake-SP. Only IceLake (there is a difference). We have some leaks of silicon, but those may be ES samples rather than QS. And again, Intel isn't talking about those. Unless you think they were sampling IceLake-SP all the way back in May? Even optimists estimate IceLake-SP won't launch until Q3 2020. How do you go from QS CPUs in May 2019 to a launch in Q3 2020? And we still haven't seen any 38c dice yet, which Intel teased earlier this year.

edit: and while Intel "doesn't have to" announce sampling status to anyone, it's really in their best interest to be honest with the public about what they are planning to sell in the near future. When your product works and you are proud of it, you showcase it for everyone to see. You can sandbag a little - AMD is notorious for such tactics. But you still talk up the product. Intel's guerilla marketing back in 2005/2006 was peerless, drumming up hype for Conroe which wound up being a smash hit. Meanwhile, Intel is assuring everyone that IceLake-SP 38c will come out in 2020 (now delayed to Q3 I guess) but doesn't want to talk about silicon samples? That doesn't add up.

Intel is RUMORED to have 56 cores on a chip that is no where close to real production

That's actually Cooper Lake. It's 14nm so it will come to market more-or-less "on time", barring wafer shortages. Cooper Lake is more likely to cause the shortages than be the actual victim.

 

[Arkaign]

Now see here is the problem IMO....

Intel is RUMORED to have 56 cores on a chip that is no where close to real production, and AMD has had 64 cores for 6 months, and by the time Intel has that one in real production, AMD will be live with their next generation.

That's the crux of it. For heavy MT purposes, AMD has opened an absolutely overwhelming advantage. Production is about their only limitation at this point, and my guess as to why it's taking so long to get 7nm laptop stuff out the door.

 

[Markfw]

That's the crux of it. For heavy MT purposes, AMD has opened an absolutely overwhelming advantage. Production is about their only limitation at this point, and my guess as to why it's taking so long to get 7nm laptop stuff out the door.

Then why are 3600s, 3700x's, 3800x's and now 3900x's all available and selling like hotcakes ?

 

[JasonLD]

Then why are 3600s, 3700x's, 3800x's and now 3900x's all available and selling like hotcakes ?

retail CPU market is very small market compared to overall PC market in terms of volume. It doesn't take a huge amount of volume to satisfy that market demand vs laptop market that needs millions of processors in volume.

 

[Arkaign]

Then why are 3600s, 3700x's, 3800x's and now 3900x's all available and selling like hotcakes ?

Huh? I didn't say they weren't It was basically a compliment : in other words they can and do sell basically everything they can get out the door. I think if they had even greater capacity, they'd have even greater sales, and that includes 7nm mobile, where I think they'd really clean up.

I think you may have misunderstood my post.

 

[uzzi38]

Pretty soon actually but still massive dies. Cooper Lake is rumored to be two 14 nm 28 core dies and a 10 nm IO die all connected via EMIB.

No I/O die afaik. Just the two 28 core dies.

At one point it was rumoured to have an I/O die, but only go up to 48 cores. Whatever that design was, it got scrapped before they announced the 56 core Cooper Lake.

At that point, afaik the I/O die was scrapped for extra power budget for the actual cores. Which they'll need, because Cooper Lake-SP is now a socketed 300W Cascade Lake-AP.