Zen 2 - a lesson in hypetrains

[Metroid]

Both of the things that @Captain_Tom dude said came true on the desktop. -- I mean AMD stock alone in the last 3 months is up 44%. Intel is down 14% for the same time period. They added the cores... now they have a 16 core desktop piece in the works.

Considering the relative size of both companies that's about as hard of a curb-stomp as can happen.

He was off on IPC gains but pretty spot-on otherwise.

Intel now has to hold the server market or the pain will get worse until they release in 2020.

Overclocking wise we dont know yet but it seems a lot better than zen+. IPC is more than I expected, I expected 10%, amd stated 15%. The reason I will buy the ryzen 3700x or 3900x is power consumption. I use the cpu 24/7 and any savings on electricity is important for me. 12 cores 24 threads and a 200 watts max is very good.

A good article to read on tdp.

Why Intel TDP Measurements Don't Reflect Real-World Power Draw

Intel publishes TDP figures for each of its CPUs, but the chip's actual behavior often doesn't conform to what this metric implies.

www.extremetech.com

 

[RichF]

12 cores 24 threads and a 200 watts max is very good.

I wonder of a three stage setup would be better, along the lines of big.LITTLE + GPU, but not designed for battery life as much as for maximizing serial performance not just parallel. Basically, have one CPU that has a low core count with robust core(s), a second that's like the tiny core spew model we're getting as the current trend, and the discreet GPU — three tiers of parallelism rather than two. Putting all three into the same socket would reduce latency, of course, but it would also concentrate heat and reduce cost flexibility.

 

[Metroid]

I wonder of a three stage setup would be better, along the lines of big.LITTLE + GPU, but not designed for battery life as much as for maximizing serial performance not just parallel. Basically, have one CPU that has a low core count with robust core(s), a second that's like the tiny core spew model we're getting as the current trend, and the discreet GPU — three tiers of parallelism rather than two. Putting all three into the same socket would reduce latency, of course, but it would also concentrate heat and reduce cost flexibility.

Well, if performance is justified then I'm all for it. I mean choosing a 3700x x 3900x for me still debatable, reason why I will wait reviews. Chipset wise, nothing wrong with the increase in tdp as long as it justifies performance.

 

[phanbuey]

Well, if performance is justified then I'm all for it. I mean choosing a 3700x x 3900x for me still debatable, reason why I will wait reviews. Chipset wise, nothing wrong with the increase in tdp as long as it justifies performance.

Doubtful. This chipset is more to D*slap intel than anything else. The first technological leap from AMD in over a decade. The PCIE 3.0 bus is not saturated and with amd promising their chips will perform the same on x470 vs 5 series, you would have to have a collection of nvme drives/ want to run an impressive raid, as a normal consumer, to *NEED* a 5 series board.

I think it's more to dumpster Intel's HEDT lineup, as they offer the bandwidth, connectivity, and performance at a lower price on a consumer platform.

 

[R-T-B]

I meant it as far as social engineers go:

Ok, where I come from "socially engineered" influencing is called "marketing." AMD has a whole department for that.

 

[mtcn77]

Wow. You don't see the megairony in your post?

The part where you don't take me on my word?

 

[Xzibit]

Ok, where I come from "socially engineered" influencing is called "marketing." AMD has a whole department for that.

Nvidias department is under "Community Advisor Program" CAP

For the past year, Mod Op has shepherded NVIDIA’s influencer program to generate awareness and online buzz around SHIELD and their gaming service, GeForce Now for Mac.
We enlisted 20+ influencers who produced 70+ pieces of SHIELD and GeForce NOW for Mac
content to over 5.3MM+ followers online. With over 1.4MM in engagement across all social
channels, our average cost per influencer was also 70% lower than our proposed cost. Additionally, we surpassed our yearly content goal by 110%. Influencer coverage had such a major impact that it
spread around the online tech community as well as media publications like Business Insider. Service profiles and 'how-to' features showcasing our influencers started sprouting up across the web.

There was a recent opening

Nvidia: Manager - Influencer Marketing, GeForce Global Consumer Marketing

 

[Deleted member 157276]

Nvidias department is unders "Community Advisor Program" CAP

Correct. They're part of the same corporate system, which was what R-T-B was making a point of; it wasn't a jab at AMD specifically.

Why are people in here constantly getting offended, and feel the need to do the "but [insert company] aren't the only ones bad, so and so do it as well". THIS IS NOT ABOUT AMD, THEY'RE JUST BEING USED AS A CASE EXAMPLE. Rooting for actors whose interests are contradictory to your interest as a person, and who have no relation to you, is wholly irrelevant and completely contrived.

If we are struggling to move past this immature puerile mindset, where everything is turned into a fanboy spit fest, and no criticism or praise can be had without it being taken as biased, maybe 95Viper should just lock up this thread again. I've seen more fanatical zealous support (to corporations, in this case) in here than an average political discussion on Reddit. Consumerism at its absolute worst.

 

[bogmali]

Thread cleaned and reply bans issued. If you cannot conduct yourself appropriately while discussing with others, don't post and stay out of it.

 

[Deleted member 157276]

A very good and relevant summary from Hardware Unboxed about the Zen 2 saga the past year:

 

[Xzibit]

A very good and relevant summary from Hardware Unboxed about the Zen 2 saga the past year:

I pointed this out already but at the 20:10. They had been contacted about the information but went ahead to include it in a few videos for their benefit. A few others have said they were contacted about the misleading rumor before hand but still ended up making videos and helping it along.

 

[Deleted member 157276]

Which one? AdoredTV's?

 

[Darmok N Jalad]

Well, if performance is justified then I'm all for it. I mean choosing a 3700x x 3900x for me still debatable, reason why I will wait reviews. Chipset wise, nothing wrong with the increase in tdp as long as it justifies performance.

If I recall right from an Anandtech article, the CCX's are actually pretty darn efficient. It's the Infinity Fabric that consumes a lot, even at idle. I'm curious if Zen 2 and the Chiplets (sounds like a good band name) will resolve some of that inefficiency. I think it's why AMD can slap 12C on one CPU, clock it high, and yet still keep thermals somewhat manageable. It's not the CCX's burning up the budget, so the more the merrier, so long as the IO die can keep everything well fed.

I'm still going somewhat out on a limb to say that there is more clockspeed headroom than what the SKUs suggest, for two reasons. One, AMD is setting a lineup to out-compete what they believe Intel can manufacture. Two, I think that the best of the best of these chiplets aren't seeing lowly desktop duty. I bet EPYC has absorbed all of the early best yields, and the chiplet concept makes it very easy for AMD to bin and route the best chiplets to the most profitable application. Reviews will tell the tale, especially overclocking results.

 

[moproblems99]

The OP is talking about the entire field of marketing.

Marketing is about confusing people with emotion and lies to get them to part with more of their money than they should. That is, of course, what we call profit. It's profit for some but not necessarily for others. More broadly, the topic becomes our entire existence. From the point of view of the universe, nothing we do makes the slightest bit of difference. Our lives are utterly insignificant so we have to live based on illusions. So, really, hype is what keeps us alive.

You missed the point of this thread. The point of this thread is not about marketing. It is about people (consumers) speculating (hyping, or making up far-fetched shit) that blows products out of proportion and leads to general disappointment within the community when if people just looked at the information in front of them then realistic expectations would have stayed and the products would have had a better reception.

And my point, is that average Janes and Joes on a forum have no impact on expectations and reception and should speculate away as long as everyone understands the context of the discussion. Most people don't pay any attention to what a nobody says on the internet. So what is the point of even bothering with this thread?

People like AdoredTV fall into a completely different category. They are clearly influencers (for better or for worse) and need to be categorized as media or journalists (extremely loosely using the term here). Yes, people like AdoredTV blurting out dumb shit isn't the best thing but anyone with half a brain knows better. The rest, well it really doesn't matter what is said for them does it?

And who knows, maybe that information is more accurate than we know and AMD didn't need to release 5GHZ chips because they can beat Intel with 4.4. I don't know and I don't care but speculating about it isn't a problem.

 

[Metroid]

If I recall right from an Anandtech article, the CCX's are actually pretty darn efficient. It's the Infinity Fabric that consumes a lot, even at idle. I'm curious if Zen 2 and the Chiplets (sounds like a good band name) will resolve some of that inefficiency. I think it's why AMD can slap 12C on one CPU, clock it high, and yet still keep thermals somewhat manageable. It's not the CCX's burning up the budget, so the more the merrier, so long as the IO die can keep everything well fed.

I'm still going somewhat out on a limb to say that there is more clockspeed headroom than what the SKUs suggest, for two reasons. One, AMD is setting a lineup to out-compete what they believe Intel can manufacture. Two, I think that the best of the best of these chiplets aren't seeing lowly desktop duty. I bet EPYC has absorbed all of the early best yields, and the chiplet concept makes it very easy for AMD to bin and route the best chiplets to the most profitable application. Reviews will tell the tale, especially overclocking results.

If we take a look at multithread power consumption, 2700x x 9900k. They are pretty much equal, 12nm x14nm++.

https://tpucdn.com/reviews/Intel/Core_i9_9900K/images/power-multithread.png

2700x x 9700k, 10% less for the 9700k.

https://tpucdn.com/reviews/Intel/Core_i7_9700K/images/power-multithread.png

In January at ces, the likely 3700x scored equally x 9900k on cinebench while using 30% less power, 130w x 180w

AMD Ryzen 3rd Gen 'Matisse' Coming Mid 2019: Eight Core Zen 2 with PCIe 4.0 on Desktop

www.anandtech.com

The question remains, a 3700x will be better for overclocking as it will give more room for cooling or a 3900x will be better, more cores, more productivity. 3700x = max 120w, 3900x = max 200w.

 

[Deleted member 157276]

That's system power. The 3700X used substantially less power by itself, at 75W, vs. the 125W of 9900K. That's 65% less power. The 3700X is a dream CPU for SFF builders. Perfect sweet spot of performance and efficiency.

 

[Darmok N Jalad]

If we take a look at multithread power consumption, 2700x x 9900k. They are pretty much equal, 12nm x14nm++.

https://tpucdn.com/reviews/Intel/Core_i9_9900K/images/power-multithread.png

2700x x 9700k, 10% less for the 9700k.

https://tpucdn.com/reviews/Intel/Core_i7_9700K/images/power-multithread.png

In January at ces, the likely 3700x scored equally x 9900k on cinebench while using 30% less power, 130w x 180w

AMD Ryzen 3rd Gen 'Matisse' Coming Mid 2019: Eight Core Zen 2 with PCIe 4.0 on Desktop

www.anandtech.com

The question remains, a 3700x will be better for overclocking as it will give more room for cooling or a 3900x will be better, more cores, more productivity. 3700x = max 120w, 3900x = max 200w.

I’ve seen conflicting results. The 2700X consumes considerably less than the 9900K at load. It leads to a discussion on just how misleading the TDP rating can be.

The Intel 9th Gen Review: Core i9-9900K, Core i7-9700K and Core i5-9600K Tested

www.anandtech.com

 

[Metroid]

I’ve seen conflicting results. The 2700X consumes considerably less than the 9900K at load. It leads to a discussion on just how misleading the TDP rating can be.

The Intel 9th Gen Review: Core i9-9900K, Core i7-9700K and Core i5-9600K Tested

www.anandtech.com

And the interesting thing is that I used that chart many times. I used the techpowerup chart because since we are here but yeah I get your point. We clearly see a 50 watts difference. Since we have conflicting results, there is nothing much we can do about it other than to test it ourselves and take our own conclusions but even then my or your conclusions can be either close to anandtech or techpowerup. It's all a mess. 2700x super binned and a terrible 9900k? It can be anything.

 

[Darmok N Jalad]

And the interesting thing is that I used that chart many times. I used the techpowerup chart because since we are here but yeah I get your point. We clearly see a 50 watts difference. Since we have conflicting results, there is nothing much we can do about it other than to test it ourselves and take our own conclusions but even then my or your conclusions can be either close to anandtech or techpowerup. It's all a mess. 2700x super binned and a terrible 9900k? It can be anything.

They use different programs to load the system (POV-Ray vs Cinebench), so that could be a factor. Nothing against TPU, but Anandtech goes pretty deep into testing on these things. They also have a really good dive on Ryzen and the power budget cost of Infinity Fabric.

 

[Deleted member 157276]

People like AdoredTV fall into a completely different category. They are clearly influencers (for better or for worse) and need to be categorized as media or journalists (extremely loosely using the term here). Yes, people like AdoredTV blurting out dumb shit isn't the best thing but anyone with half a brain knows better. The rest, well it really doesn't matter what is said for them does it?

Well AdoredTV makes it highly relevant, as most of the hype could be traced back to his stuff. The point being that when these people with influence, who have a responsibility to be critical and properly informative, are not, then issues like what we saw the last 6 months arise. And it fires up a community who we know are very easy to rile up, in terms of fantastic expectations. AdoredTV's video(s) was so impactful that when Techspot released an article on December 8th, 2018, a week after AdoredTV's leak, titled "Our take on AMD Zen 2 CPU and Navi GPU rumors" (video version here) in which they played down the hype and made arguments for why they found Jim's claims unreliable, they recieved heavy negative feedback. Interestingly, Jim took up arms against criticisms against his "source", of which that Techspot article was in the centre, and said that the Zen 2 release would "redeem" him and prove everyone else wrong. Needless to say, that never happened.

The whole saga from then on is really interesting. As the HU guys explained the first video I linked:

For us, we were just trying to inform our viewers that 'look, I know you guys are excited. We're excited as well, but maybe put excitement in check', because it's going to be quite a few months before you really hear anything concrete, and it's going to be a Q3 thing for when you you buy them. The reason we covered it in News Corners is because we were getting hundreds of hundreds of comments saying "why aren't you covering this [AdoredTV leak]. $100 6 core Ryzen 3 CPU is definitely going to be my next buy"

This was the kind of hype the Reddit leak and Jim from AdoredTV had helped spread. When Techspot releasd this video back in December 7th, contradicting Jim and saying that Zen or Navi would not be announced at CES in January, 2019, the negative backlash in the comment section was enough for them to finally conclude:

In the future we'll probably just keep what we hear from industry sources to ourselves, it's too much hassle to deal with the rubbish from the community.

Of course, the negative feeedback was far worse in their "Our take on AMD Zen 2 CPU and Navi GPU rumors" article and video, where they downplayed many of his claims as "insane", "unlikely" and "not possible". There were over 2000 comments there, of which a large portion of them were defending Jim's leaks.

They sum it very well up in the latest "Too Good to be True" video, and Tim echoed the same sentiments posed in this very thread, near the ned of the video:

I don't agree with [the talk of Zen 2 being disappointing] at all (...) This is kind of why we don't cover leaks or rumour, or like these sort of leaks in general. Imagine if this leak hadn't happened, and there was no information of this degree about the Ryzen 3000 line, and then it launched (...) people would actually be amazed that AMD has a 9900K-capable part that is not only 65W (TDP), but is also costing you $200 less than a 9900K (....) But when you come from the perspective suggesting there'll be a cheap part [it'll disappapoint].

And my point, is that average Janes and Joes on a forum have no impact on expectations and reception and should speculate away as long as everyone understands the context of the discussion. Most people don't pay any attention to what a nobody says on the internet. So what is the point of even bothering with this thread?

I disagree. People coming together and having discussions, like on this forum, is precisely about making for more informed human beings -- that's the entire purpose of having public forums like these. We might not carry the same weight of influence as AdoredTV or Hardware Unboxed, but we nevertheless have some influence and some impact; namely on those reading and participating on TPU. If you think that's insignificant, then you are also disqualifying the entire function of the forum of TPU, and other similarly-sized forums, as a whole. Why have any discussion at all to begin with, if it won't matter much on a larger scale?

Although I completely understand your point, and it certainly touches upon an issue of constraint and marginalization, the pessimism won't get you far.

 

[Metroid]

They use different programs to load the system (POV-Ray vs Cinebench), so that could be a factor. Nothing against TPU, but Anandtech goes pretty deep into testing on these things. They also have a really good dive on Ryzen and the power budget cost of Infinity Fabric.

I do think that too. The difference is huge, if i had any of them i could go deep testing power consumption too, my old prehistoric i7 920 2.6 base clock 130 watts, overclocked to 4.2, used 310 watts 1.30v. I usually tend to double power consumption when I consider overclocking but that was before, 50% or more overclock, things seem different right now to some extent. The overclock people are happy nowadays is 300 mhs or so hehe and they tend to increase the vcore a lot, my idea for the new ryzen cpu is to undervolt while leave base clock intact and see how much further I can reduce the voltage then I will see how much room for overcloking it will leave me to.

Well AdoredTV is highly relevant.

There are some rumours AMD will announce a 16 cores at E3 for $699. They will be smart if they do, even if they launch for $999, there will be plenty people that will buy it and maybe that is where they might be going. Nothing wrong about what adoredtv has done, any rumours/leak have always to be taken with a huge barrel of salt. I must admit adoredtv fantasied very low prices for something that not even intel has in the mainstream market.

 

[TheLostSwede]

Except it looks like the 16 core part is just around the corner...

Source: https://videocardz.com/newz/amd-ryzen-9-3950x-to-become-worlds-first-16-core-gaming-cpu

A good place to start with an upcoming product is, as I mentioned in my first thread, Navi. Most of the hype existing around it today have branched out from Jim at AdoredTV and his "source". Jim presented the information he got regarding both Zen 2 and Navi in a video in December, 2018. Now that we can resolutely say he was reasonably wrong in the information he received regarding Zen 2 (but also Navi in terms of release dates), as well as its inconsistency with later information regarding Navi's supposed efficiency, his Navi predictions should be completely forgotten. The reliability of what he provided has verified itself worthless. Furthermore, the price points he assumed Navi to be at are also quite ambitious; Navi 10 is/was, according to Jim, matching the RTX 2070 in performance at 150W TDP, for only $250.

You might think so, but he went with data that was correct at that point in time. The fact that things change over time, is clearly not considered a valid thing on the internet.
People go on and on about how wrong he was at the time, yet people have had access to information about Ryzen 3000 for quite some time.
AMD pushed things back, as the A revision had issues and they're now on the B revision of chips. Since AMD had never publicly disclosed a launch date, they had every opportunity to delay things without upsetting anyone. On top of that, the original plan was to launch without a new chipset, as the X570 was much later than planned, not only from AMDs side, but also from a motherboard availability perspective, as AMD relies a lot more on the board makers than Intel does when it comes to developing the actual motherboards. As such, it takes longer time for the final product to be ready. It doesn't mean that AMD isn't actively working with the board makers to help them create good products, but it does mean it takes much more time to get to the point of a finished product. But yeah, it's easier to call him names and say that he made it all up, than to try and understand what's gone on behind the scenes. I suggest getting some actual sources in the industry, rather than making up some story that fits your narrative.

As I shared elsewhere here, I've sat on details since October 2018 about Ryzen 3000, including this.

October_2018_1.jpg

drive.google.com

As for Navi, I don't have any comments, as I have no insider knowledge.

 

[Tomgang]

Except it looks like the 16 core part is just around the corner...

Source: https://videocardz.com/newz/amd-ryzen-9-3950x-to-become-worlds-first-16-core-gaming-cpu

Ah you where faster than me to post it. Darned.

Well here is a ES model of Ryzen that maybe cut just be the 3950X in an early stage and it also have a Cinibench R15 score.

AMD Ryzen 9 16-cores CB15 Benchmark (4346 points)?

I'm pretty sure you've followed the Livestream or read our articles on yesterday's Ryzen 3000 announcements. You will have noticed that the Ryzen 9 3900X is the proc with the most cores, 12 of the...

www.guru3d.com

I begin to think that second half of 2019 might gonna be a pain in the ass for intel as if rumors are true that they only will have 10 cores as max core count on the little socket or gaming platform. Off cause we will have to wait and see until official release and review of ryzen 3000 series. But what i have seen so far about ryzen 3000, it really looks promising for AMD this round.

 

[Deleted member 157276]

You might think so, but he went with data that was correct at that point in time.

You, me or even AdoredTV, have no way to say or know that it was correct at that time. His source claimed that information was correct. That's all. And it was information that, according to Techspot, contradicted the information they had from their "sources" from AMD, as well as those of other parts of the tech press that they knew of.

The fact that things change over time, is clearly not considered a valid thing on the internet.

It is actually considered quite valid. It's just not considered as valid by Jim at AdoredTV, who didn't seem to take it into account when making his firm and pretty detailed prediction about Zen 2 and Navi (like which specific CPUs would be announced on CES, which in May, and so on and so forth). He actually does mention that "things might change" in his video, but only says so in relation to "minor changes" in dates, frequencies and prices. Of course the release showed pretty significant changes to what he claimed.

Also, this argument has an lesser leg of a stand on when you consider the dates of Jim's videos. He published the Zen 2 and Navi leak predictions on December 4th, a mere month before CES, and repeated his stance on December 12th. That's far too close up to the CES launch for AMD to make any significant changes, like deciding not to announce Zen 2 processors and a Navi GPU.

Let me also remind you again that all this was when Jim's source was providing information that was in contradiction with the sources that Techspot and others had. Also, whenever AMD releases a new line of CPUs or GPUs they host a Tech Day, where all the enthusiast media gather to learn about these new products. It’s at these events we usually hear about the products in detail and then there are several NDA dates afterwards for announcements, reviews and so forth. This is the way AMD has been operating for the last few years. Jim still made his claims well aware of this.

People go on and on about how wrong he was at the time, yet people have had access to information about Ryzen 3000 for quite some time.
AMD pushed things back, as the A revision had issues and they're now on the B revision of chips. Since AMD had never publicly disclosed a launch date, they had every opportunity to delay things without upsetting anyone.

All this is baseless speculation.

it's easier to call him names and say that he made it all up

What names did I call Jim? And where did I say he made anything up?

I said his leaks turned out to be untrue/wrong, and that they were responsible for unecessary hype. The first one is factually undeniable, the other is an opinion that's hard to disagree with.

than to try and understand what's gone on behind the scenes. I suggest getting some actual sources in the industry, rather than making up some story that fits your narrative.

I agree, empirical data is what I'm all about. Let's start with you providing actual sources for your baseless claims for how AMD had actually intended to launch on CES, but delayed it (and that they at least made this decision as late as 1 month before). That includes evidence of the actual prices and frequencies of AMD's products, which ended up different than Jim's predictions in the end.

 

[TheLostSwede]

You, me or even AdoredTV, have no way to say or know that it was correct at that time. His source claimed that information was correct. That's all. And it was information that, according to Techspot, contradicted the information they had from their "sources" from AMD, as well as those of other parts of the tech press that they knew of.

It is actually considered quite valid. It's just not considered valid by Jim at AdoredTV, who didn't seem to take it into account when making this firm and pretty detailed claims of Zen 2 and Navi (like which specific CPUs will be announced on CES, which in May, and so on and so forth). He actually does mention that "things might change" in his videos, but only says so in relation to "minor changes" in dates, frequencies and prices. Of course the release showed pretty significant changes to what he claimed.

Also, this argument has an lesser leg of a stand on when you consider the dates of Jim's videos. He published the Zen 2 and Navi leak predictions on December 4th, a mere month before CES, and repeated his stance on December 12th. That's far too close up to the CES launch for AMD to make any significant changes, like deciding not to announce Zen 2 processors and a Navi GPU.

Let me also remind you again that all this was when Jim's source was providing information that was in contradiction with the sources that Techspot and others had. Also, as Techspot said, whenever AMD releases a new line of CPUs or GPUs they host a Tech Day, where all the enthusiast media gather to learn about these new products. It’s at these events we usually hear about the products in detail and then there are several NDA dates afterwards for announcements, reviews and so forth. This is the way AMD has been operating for the last few years:

"So far there is no planned Tech Day for AMD at CES [in January]. In fact, at last year’s CES AMD did host a Tech Day, it’s where they announced the Ryzen APU and a few new Ryzen Mobile SKUs. But we heard about that event as early as September, so nearly four months ahead of time, which is fairly typical for planning these sorts of trips, so if there was a Tech Day, we would have heard about it by now."

I don't know you, you don't know me, yet you're making assumptions that I had no knowledge of what was going on. Seriously, did you even look at the attachment on my previous post?
I have sat on details about Ryzen 3000 since Q3 2018, but apparently I don't know shit...
Dude, get a life.