AMD's Ryzen CPUs (Ryzen/TR/Epyc) & Vega/Polaris/Navi GPUs

Th blog wasa from an exec, it promised zen compliance to 2020, not AM4, Epyc or TR4.

 

So, you want a little research. As of Feb. 2017, AMD was saying that Ryzen was supported until 2020. By July 2017, AMD was saying that AM4 was supported until 2020. Why the change? Because Threadripper is TECHNICALLY RYZEN! The name is RYZEN THREADRIPPER. So, trying to alleviate any consumer confusion that may occur, a mere couple months before, they changed the framing of the promise for support. Now, technically, when using Ryzen in those earlier instances, they discussed it more as there being that many generations of Zen guaranteed, with no promise of Zen design beyond 2020, as I read those articles. But lining the statement on Ryzen support to 2020 and AM4 support until 2020, it can be said consumers could draw an inference that is reasonable that the meaning of support, even though the statements were different, suggested they were referencing generally the same promise.

Now think to the argument AMD just lost about what counts as a "core" in a recent court case. This is about consumer expectations and beliefs on the basis of statements and marketing. Do you need a line drawn for what I just laid out and the legal implications?

You focusing so hard on whether the specific socket was uttered in the promises ignores completely the potential arguments otherwise, and liability can go BEYOND just simply whether a proactive statement with the specific words were uttered.

Now I don't want to go looking up statutes and court cases, providing specific authority and laying out a timeline of statements, when they changed, and the potential understandings of early purchasers of those statements and whether or not there was the potential for consumers to have been mislead or have a different understanding due to the comments and whether there was sufficient time for more definite statements to have changed the understanding of purchasers. But, to suggest that your argument is anything but legal exculpatory cover for whether consumers were right or wrong to understand support as they did is naive at best and disingenuous at worst.

You do not have to overtly state something is about a legal nature for it to have legal implications. You do understand that, right? This is why I find the overt nature of you pushing SOLELY for a statement of support on the specific socket to be, well, wanting.

Do you want me to do more research on this point for you?

Stop trying to make this about price shock! THIS IS NOT ABOUT PRICE AT ALL! THIS IS ABOUT CONSUMER EXPECTATIONS! When a person is already prepared to spend $1K+ for a CPU, and has done so in the past, price is NOT the problem with these consumers. And telling people to buy old hardware, which they could have purchased ANY TIME IN THE PAST YEAR and which has significant problems in some use cases is insulting.

There are NOT additional memory channels on the TRX40 boards. X399 and TRX40 BOTH have quad channel memory support. Now, which pads on the silicon substrate are used for memory channels would change as instead of the two channels being at diagonal ends, they are now on the central I/O die. This does raise questions on whether they could have used the two channels of memory closest to the dies and put the dies two deep in each caddy-corner, which would cause significant increase in heat causing lower speeds, but have the connectivity be fine on latency, etc. These are design decisions. Or even spreading them to each corner, by using a group of two channels instead of one channel on each of the two pairs on the I/O die, which they may have done anyways, would have needed routing to a pin under the dies, making it less efficient and increase mem latency a bit, etc. But I do not buy for a second that this was the only way it could have been accomplished. And centralizing the memory to mask the node situation IS what makes Zen 2 so good. Let's be clear on that. So I do not fully buy that argument.

Now what I'm surprised no one has brought up, which I have been waiting for, is that the difference in TR and Epyc is Epyc has an integrated chipset, TR doesn't. This allows for some nuanced distinctions in how things are laid out. Or that even getting Rome to work on Naples isn't straight forward (a point made by Wendell on that GN video). But pin changes were not needed.

I may not fully agree on the chipset analysis of others here (although I will point out that the X570 chipset IS the I/O die AND costs more than the alternative options, but no one else has a PCIe 4.0 chipset (looking at Asmedia)). But, I will say that I also do not buy the need for more lanes to it to be the issue. They were considering the 8 lanes to be used on the X570 boards, but decided against it not because of compatibility, rather due to heat considerations and power draw of the chipset. So to hear this now pushed for X399 compatibility I find a hard pill to swallow.

So, I'm going to call bull on your explanation of NEEDING a new board. Sure, if you need superfast drives based on PCIe 4.0, then you have that. But even with the newer Navi based cards, or the 2080 Ti, PCIe 3.0x16 is NOT being fully saturated, so you won't get that improvement on I/O, with Navi support not showing a significant increase in projects in 4.0 mode yet AND no other cards running PCIe 4.0. Sure, there is inefficiency in not utilizing it, but until add-in cards catch up, like LSI HBAs and RAID cards, etc., the use of PCIe 4.0 is future proofing, not exactly speeding up the CPU.

Knowing the above, you can see the skepticism in the official story of what is happening. Also, I watched that interview when it was live streamed. I asked questions in the youtube chat. I was NOT satisfied with his answers, generally, nor the selection of which questions to ask him.

And once again, the reason Zen 2 is improved IS THE I/O DIE! That is what made it better and sidestep ALL the issues the Zen+ TRs faced and IS the reason Intel cut the costs in half for their HEDT lineup. AMD should have been up front and direct as soon as they knew the problem: THE VRM.

We are looking at increased power draw over the prior gen. Period. It was known that for the Zenith Extreme, you had to have active VRM cooling to support the 2970WX and 2990WX. The Asrock Taichi used the same VRM, roughly, so would also need active cooling. Then there was the MEG Creation board. That could handle it, but had a price premium.

Now, with ONLY 24 and 32 core variants, while having rumors of above that level, the VRM is the issue, hence my initial analysis of saying it was about power delivery and their initial sin was not properly explaining the situation to MB MFRS at the release of gen 1 (they knew gen 2 Zen+ would go above the 16-core count at that point and said nothing, and if they didn't, then how did they not realize that was the natural progression/are we to believe when chip designs take 9-15 months for tape out to release, with 18 months at the longest end, that they did not know at the time of giving power needs to MB MFRS that these chips, arriving in Aug. a year after an Aug. release, with the boards shown off two months before that at computex and the specs given a couple months before that, that there was going to be a change coming to try to beat Intel's 12-core (although the 18-core was not announced until Intel found out about AMD's 16-core HEDT chip)?). I just find the whole thing to be bad planning on their part and a lack of foresight and vision, if they are truly saying they had no idea. Especially since specs are usually done in the planning stage, which is about 2 years out, like how they are currently planning and designing Zen 4 and 5. TR was rushed and done last minute and as a side project, but that doesn't mean they shouldn't have looked at the most powerful Epyc, asked how much power a 32-core clocked that high would draw, and used an insane over-estimate for what would be needed, much in the way MB MFRS went VRM phase count crazy on the X570. This also would have forced an actual design of the boards rather than a copy paste situation of X399 boards from their designs for the X299 platform.

So I clearly find the statements and explanations unsatisfactory.

 

The thing is for 48 and 64 core yes, trx40 may be needed for VRM and scalability. This means pinouts could have been mapped out too allow x399 on the lower core counts. This though would not have pushed the sale of Zen+ CPU's. As it is the 2970wx and 2990wx are Linux specific use CPU's so are a hard sell, as far as 2950x it is just too little a gain to justify.

So because of marketing needs we suffer.

 

Market needs, not marketing needs. In the end AMD has to offer what enough people will buy to pay down the cost of development and production and make a profit to stay in business.

The market doesn't need 2 new 16c CPU's - one AM4 and one TR4, just like AMD had a hard time selling a TR4 8c CPU at the same time they offered 8c AM4 CPU's, Robert Hallock discusses this in that video.

Robert said AMD found that there were essentially 2 kinds of buyers. Either the buyer wanted the top core count available or they wanted the best deal available.

The top core count buyers didn't want to buy the 8c TR4 CPU they wanted to buy the top core count TR4 CPU, so they didn't buy the 8c ThreadRipper 1900x.

And, the best deal on an 8c Zen CPU was on the AM4 lineup, so those buyers didn't buy TR4 motherboards and buy the 8c ThreadRipper 1900x.

We discussed it here at the time, wondering why AMD would offer 8c in both AM4 and TR4. It turns out that AMD was trying it out to see if covering the lower end core count on TR4 was what buyers would want. They didn't.

I get that at least one Zen 2 CPU for TR4 x299 / x399 owners would have been nice, even though AMD is already actively offering and selling 3 16c CPU's: 1950x 2950x 3950x.

Again, that would have meant doubling up the 3950x design, one for TR4 and one for TRX40, doubling the development cost for the Gen 3 16c CPU and splitting that cost over about the same number of buyers.

AMD needed to choose AM4 or TR4 for their new 16c CPU and it makes sense AMD would choose AM4 as that's their largest pool of buyers. A new 16c CPU in the AM4 world is a big deal - the highest core count CPU ever offered on AM4. While a 16c CPU has already been done 2x on TR4.

If a 16c TR4 CPU were released as the only TR4 Zen 2 CPU, then the complaints would have been why weren't the 24c and 32c CPU's also offered on TR4 now, and complaints again if / when higher core count CPU's arrive for TRX40.

While a 16c Zen 2 CPU on TR4 would have made the 1900x / 1950x / 2950x(?) TR4 owners happy, what about the other higher core count TR4 owners? AMD would also need to double up on the 24c / 32c and higher core count CPU's for both TR4 and TRX40 to offer to essentially the same group of buyers.

Once the decision for a change to a new socket is made there's no case for duplicating the effort to develop essentially the same new CPU's for the previous generation socket at the same time as the new generation socket.

That would have been a huge costly mess.

The decision was made long ago to move to a new socket. It's too late to change it. All we can do is continue on from here with the understanding that we can't count on assumptions. Unless AMD explicitly states something we can't assume it.

Hopefully we'll get 2+ generations on TRX40. I wouldn't plan for more than 2 generations given the TR4 experience.

" Stop trying to make this about price shock! THIS IS NOT ABOUT PRICE AT ALL! THIS IS ABOUT CONSUMER EXPECTATIONS!"

Of course it's about price shock, it's about nothing but price shock. You had expectations of spending a fixed amount you had set aside for a CPU upgrade and now you need more $ for the CPU *and* more $ for a new motherboard.

The ThreadRipper 3 upgrade is shockingly expensive.

Socket changes usually happen unexpectedly. Justifications are moot, the decision is already done and it's too late to change it.

We'll find out if the new TRX40 CPU's work in the old TR4 socket, I'm sure a few will risk finding out by trying it, and then hack away to see if they can make it work.

Maybe someone will discover a pin re-mapping that works and we'll get one of those CPU socket adapters like eon's ago. Those worked pretty well at the time:

https://www.anandtech.com/show/1650/2
https://www.google.com/search?q=cpu+adapter+socket+478

I recall a couple of earlier generation socket adapters too - less elaborately implemented due to fewer differences between sockets. A simple pin re-mapping that requires a simple socket adapter would be quick and easy to implement.

The Powerleap PL ProMMX - Powerleap PL ProMMX K6-2 333
https://images.app.goo.gl/DBHwd7vUaM5Gs6jJ8
http://www.powerleap.com/


There were also Slot-1 adapters like the Slocket:

https://www.anandtech.com/show/456

There were lots of socket converters and adapters long ago:
http://www.cpushack.com/UpgradeProcessors.html
MORE

IDK if it's possible to make something like this to adapt the new ThreadRipper 3 CPU's to the x299 / x399 motherboards, but similar products have been offered many times before.

 

Could be possible, I'm running my 4980hq in my t440p. The question would be would there be Adapter's or would you have to buy the cpu w/ adapter as a package deal?

Would be interesting though.

 

Intel used to sell "Overdrive" packages with the CPU and adapter. Cyrix did the same. To make the upgrade simple for users often the adapter makers would have packages that included the CPU too.

There were plenty of "adapter only" solutions - I used several over the years for myself and upgrades for friends that wanted to save $ over a complete new computer.

It was a conceptual issue, a physical component issue, and a component availability issue. To cut through the mess of options - or lack of options - for a number of years a plug-in adapter to allow an old computer use a new CPU seemed like the most expedient solution.

Today we can simply swap in a standard fitting motherboard upgrade and keep all of our other components.

Laptops motherboard's are still mostly proprietary in shape and size so an adapter would make installing a new CPU generation easier than swapping in a new motherboard. But, then there are the height and socket area clearance issues.

If you look at the old adapters / converters they were pretty thick and wide, depending on the conversion circuitry needed. There likely isn't enough room in and around the laptop CPU socket to fit an adapter.

Those adapters often came with a new heatsinks and cooling solutions for the new CPU. Laptops have proprietary cooling plates over the CPU that would get in the way of fitting an adapter.

Even the slimmest form fitting adapter would still add height to the CPU / socket and displace the heatplate at the socket and elsewhere.

 

this is the position AMD takes, is that these cpus are still very capable so their prices should be kept up. we see them from enthusiast pov and those dont worth all that much, but not from a performance per dollar perspective.

given that you can buy a bunch of cheap ivy/haswell xeons from ebay vs a 2990wx for $1.5k, the cheap xeons probably the better choice.

 

You are mistaking price shock and consumer expectations. Hell, if they made clear they were not using the same socket, a decision made at least 15 months before, if not longer, then they would have sold a LOT FEWER second gen and first gen CPUs during the past year. Those clients expected socket compatibility.

Believe it or not, but with Intel, socket compatibility does NOT change suddenly. They have 1 gen on mainstream. Everyone knows and complains about the bilking. Some modify firmware to get around it. etc. With server and HEDT, they get two gens. This is because they use the same socket (although this is changing and so may be subject to change moving forward) for HEDT and servers, and the server guys banded together to force platform support for 2 generations of CPUs. But with Intel, you darn well know what you get.

With AMD doing this, if they dare have TRX40 for Zen 4, I will NOT give it a consideration. Period. I will buy Intel or AMD server (and if their server products don't OC and no one has found a work around, Intel by default). I'm darn serious about that because THAT is when a new socket is needed. Right now, it is VRM bullshirt that caused the decision for the change, NOT a true need to remap the pads.

And no, if they had told us even during CES new MBs are needed, zero backwards compatibility, I would be waiting two weeks to click buy now.

But it is people like you trying to tell me what my problem is that makes me want to say certain things to you and tell AMD to ....

And as for socket adapters, it shouldn't be needed.

Also, as for the explanation how this is different than the 8-core fiasco, go read my darn posts from a week or two ago. I went through excruciating detail there on similarities and differences.

Edit: and to be clear, sTR4 is a variant of their server socket. TRX40 is another variant on the server SP3 socket. Zen 4 server is looking at socket SP5. If they try to cram new features like DDR5 and PCIe 5 onto that, either or both, when THAT is when a new socket is likely needed, and they stick with the SP3 revision they just come out with, THAT is unacceptable!

 

You know AMD and Intel can't directly tell us months in advance they are going to change the socket / motherboard requirements. Their sales of the current generation CPU / motherboards would plummet.

Have you heard of the "Osborne Effect"?

"The Osborne effect is a social phenomenon of customers canceling or deferring orders for the current soon-to-be-obsolete product as an unexpected drawback of a company's announcing a future product prematurely."
https://en.wikipedia.org/wiki/Osborne_effect

I experienced that at the time, and it was a disaster for everyone involved. There were other contributing factors to Osborne's failure, like competition from Kaypro, but that complete stoppage of Osborne 1 sales due to the premature release of detailed information and demonstrations of their upcoming new Osborne models was the last straw that tipped them into bankruptcy.

It's the way of the technology world. We know as customers the next generation might just obsolete what we are buying today, but we buy it anyway because we need it or want it, and we will do the same when the next generation comes out.

If you want the new ThreadRipper 3 performance you were expecting to get, the entry price is a new motherboard along with the new CPU. Why not go for it and enjoy it? If you need to, save up for a couple of months. By then the bugs will be worked out (hopefully) and you can upgrade to a cool new rig without problems. That's how I like to do it, wait for the bugs to be fixed *then* get it.

I didn't say you had a problem. I am trying to help you get through a tough time you are having trouble dealing with. These abrupt changes happen all of the time, and have been happening in the computer industry for decades.

 

Guess what, Intel doesn't have to because it is KNOWN 1 gen on mainstream, 2 gens on server and HEDT. It is a happy extra if you get an extra gen on a socket.

So I call BULL. AMD was expected to honor the same compatibility they did for am4 and SP3. They did not. They are now doing stupid crap and in damage control. We'll see who is right, the customer telling the company how they messed up or the company trying to tell the customer what to believe.

And this isn't a rough time, this is me outright saying how AMD is wrong. It is you and rlk trying to spin what I am saying that has resulted in the more emotional and visceral reactions, as you guys are trying to manipulate public perception of the event and trying to downplay actual consumers trying to explain why it is a mistake.

And as soon as you have something WORTH adding, we can discuss it. Instead of doing PR, try community outreach. But my point is you are trying to pump smoke up people's bums here and call it sunshine and rainbows.

And to repost my edit in case you didn't see it: and to be clear, sTR4 is a variant of their server socket. TRX40 is another variant on the server SP3 socket. Zen 4 server is looking at socket SP5. If they try to cram new features like DDR5 and PCIe 5 onto [SP3 Rev. x (I forgot which socket revision was listed in one of their slide shows)], either or both, when THAT is when a new socket is likely needed, and they stick with the SP3 revision they just come out with, THAT is unacceptable!

 

It's not so much "known" as "established" by past behavior. Unfortunately it's not promised at all and can change again depending on the circumstances. Right now Intel is stuck on 14nm for at least another couple of years, so Intel might stay on the sockets they have for much longer than "normal" - your expectations not withstanding.

BULL on what? AMD never promised explicitly they would have any kind of fixed calander date or even year through which TR4 would have new generations of CPU's. Nor was there a specific number / count of generations of CPU's to be released on TR4.

AMD isn't in damage control, they aren't even talking about it.

You said you were watching that Live Interview with Robert Hallock - did you hear anyone ask about ThreadRipper 3 on TR4 / x399 / x299 as I sure didn't.

AMD released a new CPU product, it requires a new motherboard, buy it or don't buy it. If the results customers hear from application benchmarks are sufficient to benefit them including the cost of a new motherboard they will buy ThreadRipper 3 CPU's and the TRX40 motherboard to run them.

I can't speak for @rlk, but I can tell you I am not twisting what you are saying, if anything I am trying to get from you what you mean, but you keep changing the angle you are attacking from so I have to keep responding to new things you say.

I am trying to help you focus on the reality of the here and now instead of the what if's from countless possible pasts that never existed.

We have new ThreadRipper 3 CPU's and they run on new motherboards. That's it. That's all we have to work with. Buy them or don't buy them.

No amount of discussion about how this came about and how it could be different is going to change anything.

I started out giving you guys alternatives to spending that huge sum to buy a new motherboard and CPU, but you rejected it. Instead you wanted the resale market for TR1/Tr2 to die.

That makes no sense to me as that's exactly what you own - why would you want your own TR1 (CPU) TR4 system to be worthless at a time you would benefit from selling it to buy the new TRX40 motherboard and ThreadRipper 3 CPU?

All the points you are arguing are speculation, speculation that might have been useful at AMD over a year ago, but you and I weren't there so we couldn't influence their decisions. We don't have all the technical details or the list of options available to AMD at that time available to us now to help us understand exactly why AMD decided to change the socket.

Again, the argument or discussion is moot - already been decided, we are in the here now and the only options available are to either follow through with an upgrade to the TRX40 motherboard and ThreadRipper 3 CPU, or stick with what you have and upgrade to the best CPU available on TR4 now or over time - then upgrade to something else down the road.

I don't think ThreadRipper is going to get cheaper moving forward. AM4 and it's successor will continue to be the best cost performance option for us as an individual.

Unless you have a corporate budget to support the high cost of HEDT it will continue to be cost prohibitive to buy as an individual.

 

FYI - Microcenter will let you build your own bundle deal, if the motherboard + CPU you want isn't explicitly listed as an existing bundle you can pick both yourself and ask for an additional $30 off for buying a motherboard + CPU at purchase:

If anyone happens to live near a Micro Center and is looking for a deal, found this while looking around. $80 below Best Buy and NewEgg!
Submitted 1 day ago by jacobatemyowl
https://www.reddit.com/r/Amd/comments/dv6mok/if_anyone_happens_to_live_near_a_micro_center_and/


With the additional $30 discount for a bundle deal Microcenter has the cheapest Ryzen and ThreadRipper 1 and 2 prices of anyone that I've found right now (so far).

 

The point being is AMD could have, but opted not too, make x399 zen 2 compliant. Proof of this was self evident in many a leak. It does seem the choice to segment out x399 was made long ago and my bet is about the same time as TR was removed from the road map. Why people are upset is AMD knowing this for apparently a long time kept it hidden. Everyone makes future plans based on best known info.

Nothing to do with price shock, assuming the performance is there I would rather give AMD $2,000 for the 24 core than a person trying to price gouge. I doubt anyone has trouble with spending just trouble with being lied too by hiding information or outright deception.

Edit; as an example, I have the 1950x. If instead of removing the TR from the roadmap it was explained for marketing integrity of up to Zen+ x399 would not be compatible I would have just held off till the dust settled on Zen2 and then maybe planned a new system build. Again working with best known info. As it is now everyone feels led on by AMD and we have been giving recommendations to others based on this as well.

Toms Hardware so stated that now knowing about the lack of x399 they can not yet recommend TRX4 as there is no guaranty of how long it now will be supported. AMD screwed themselves here with the people they need to support and endorse them.

 

Me and my friends dont feel led on at all, Whens the last time we had long term support for an HEDT system?

 

Agreed. Hell, if AMD said moving forward HEDT will be a new socket every two years. That is fine. With Intel, we knew it was that. It's about managing consumer expectations (corporation's POV), and consumers getting what they expect. I don't think anyone would worry about knowing you have two years per socket. And as mentioned, I'd switch to purchasing on the second, rather than first, year of the two year cycle (buying EOL platform with no upgrade, but getting a more stable platform firmware with whatever benefits the second year brought over the first gen would be OK).

It really is like going to buy a car, then they start slapping on hidden fees not disclosed until you start the paperwork. It is a sense that you no longer are where you were before on expectations. Just trying to make an analogy.

It is basically a question of reliance.

Also, notice how suddenly he is pushing 2+ years for TRX40, after also saying "did the company say that?"

 

The problem is all HEDT prior was Intel and assumed no long term support, ZEN was supposed to be different. Also you are not an AMD consumer so of course you were not led on. Stop speaking of what you have no experience or at least some skin in.

 

Um you might want to check yourself sir.

I've been building systems for my friends local and worldwide for the past 5 years, If they had any complaints they would've been more than happy to curb check me as that would've been a violation of trust. I dont personally have a need for anything greater and there is only 2 laptops with ryzen in them, but both fail to replace 1:1 what I have or peak my interest enough to forgo my justification check list.

I was genuinely interested in the citation you claimed to have but its seemingly more enjoyable to continue to rant and speak for "everyone" instead of simply putting up the evidence. If you had I would see your justification quite clearly.

Didnt think to see argument fallacy employed to remove someones opinion and inquiry because you dont like where it could lead.

 

If you build for others, as I have as well, they usually are not having the systems built with doing upgrades in mind. You failed though to state how many of those were AMD based HEDT's as well.

As far as justification, it seems QMD has been busy, just like with the Whitehaven leaks, plugging them all up. That is fine. I know what was there, eventually I will find a screen shot somewhere. I did not remove squat, I would not do that! This discussion is besides me being a mod but a user.

 

They chose those systems because of the upgrade path they could pursue, after giving them contextual information about both vendors and different tiers of product offering and expected budgets for each subset. Also, this thread isnt limited to HEDT's, maybe you need an HEDT thread just for TR owners (who are willing to prove it with receipts of course, shouldnt be able to participate without "Skin in the game")

I didnt imply direct action on your part, but your use of argument fallacy to remove a point of view you obviously dont agree with. If you had, I wouldve simply reported and see what the other mods would've thought of it.

 

AMD justifies the shift to sTRX4 socket for 3rd gen Threadrippers and promises long-term support, but current Threadripper owners appear to be unconvinced notebookcheck.com | Nov 14, 2019

"Even with socket AM4, AMD had to route the smaller circuits on the 7nm Ryzen 3000 processors to the existing AM4 pins to ensure compatibility" From this... Everything was/is possible if they really wanted "ensure compatibility" with their older TR boards.

 

In that article it states that it never promised the same support as they have for am4?

 

Of course. But from the beginning it was about Ryzen and Zen. TR was just a Luck draw from some AMD engineers. Many ways to hide themself from the truth.

 

Then I guess we just have to wait on the citation then...

 

Yeah, only two choices. Eat the rotten apple or move on (Either use what you already have or migrate over on the blue side). I know who will win if you stay at AMD. They will win either you buy their old inventory of chips or their new package (chips+board).

 

Some might see it that way surely, I dont currently have that issue nor do my friends (currently). Time will tell I guess.

 

Again, the discussion at hand is of AMD HEDT, where is your reported skin here? We are not complaining of AM4 or Epyc, just TR. So what AMD HEDT systems do you have under you belt, how many do you support etc.? And none of this is POV but you again fail to show where you have a valid POV other than someone on the outside looking in, and I hope this not to be true as then it means you are trolling the thread and that is AGAINST TOS! Remember you opened the door by claiming all the systems and expertise.

 

Agreed, they could have done it but did not. Does not bother me to be right. But as stated by others, probably too late now.

 

Just because you dont like my opinion doesnt make it trolling, YOU are still utilizing argument fallacy, just because you are doubling down doesnt make it any less of a fallacy dude. That isnt trolling, thats proper debating. I have no need to entertain your appeal to authority.

If you want to claim sequence of events were still waiting on this claim that AMD would support Zen and not AM4. You made the claim, now back it up and we would all be good as there would be no more argument to be had.

EDIT: I am going to walk away from this thread, I fear being banned at this point.

 

what is long term in this case. zen 2 and zen 3 gonna go with DDR4 so thats 2 years, same with zen & zen+. zen 4 gonna be on ddr5 with pcie5 i don't see how that can be made easily compatible with the current boards.

so basically treated like same as older TR systems, 2yrs only.

 

I'm sure AMD knew this. Hence they pushed the change now and not right before major hardware changes will come

 

yep. pretty sure their "long term" is just 2 yrs. they should just say 2 years lmao. zen 2 isn't all that much of a performance uplift vs intel's CFL chips though so i can still wait where as zen 3 might be great, but end of platform life sucks.

2021 is where its at, possibly zen4 vs tigerlake/golden cove i just can't wait. both sides bring big IPC/core count jump over my current 8700k with DDR5/PCIe5.

 

Yeah, AMD for the times floating on more cores. And they are already on new arch.
AMD Ryzen 7 3800X @ HWBOT
Intel Core i9 9900K @ HWBOT

---------------------------------------------------

Ryzen DRAM Calculator Tool Author Not Receiving Ryzen 9 3950X or sTRX Samples wccftech.com | Nov 14, 2019



I wonder how many will get hold of the chips within end of this year.

 

Looks like Gigabyte's TRX40 prices leaked:

Gigabyte TRX40 Motherboards
(Image credit: Bottom Line Telecommunications)

So it's possible to get one of the TRX40 motherboards for as low as $408, hopefully the other vendors will drop into the low $300's. Maybe Microcenter will have some nice holiday savings deals too.

AMD TRX40 Threadripper Motherboards Will Probably Break Your Piggy Bank
By Zhiye Liu 5 days ago Motherboards
If you think X399 motherboards were expensive, then think again.
https://www.tomshardware.com/news/amd-trx40-threadripper-motherboard-price-availability-retail

"In a sense, it shouldn't have come as a surprise that TRX40 motherboards would be equally or more expensive than the prior X399 motherboards. It's just the price you have to pay to play in the HEDT league."

For the x570 build I'm liking the minimal approach to the MB with PCIE plug-in options for the Wifib/BT / Sound / Storage, option it out with what I want instead of settling for the on board options in each higher tier motherboard.
https://www.asus.com/Motherboards/PRIME-X570-P/

Perhaps that won't be available for the TRX40, I'll have to wait to see all the available motherboards, the spec sheets, and then the reviews.

 

lmao $870. talk about milk, even the mobo guys want piece of the pie, a huge piece at that.

 

https://wccftech.com/amd-ryzen-threadripper-roadmap-confirms-4th-next-gen-hedt-cpus/

Although not the graphic I seek here is proof by omission. The roadmap shows TR4 LGA compliance, well if they had noted sTRX4 instead (denoting a socket change) we would have had better info back then. I am sure there is a lot of other things too.

This is why I can no longer trust even their printed info! I mean look at no promised 15% from Zen+, not even 10%. Moral of the story, research and build a system today but do not go by AMD promises for the future.

 

The most expensive Gigabyte TRX40, must have a lot of cool options, can't wait to see the specs on all of them.

I don't think I need the top model, just enough for full features for IO / Compute - power regulation / not sure I need much else on the motherboard, the rest I can add later.

I wonder if there are PCIE 4.0 expansion chassis out there yet?

Hey check this out, Gigabyte Aorus Extreme z390 is as expensive as that TRX40 Aorus Extreme, it looks like they are both Waterforce motherboards, that explains the price:

GIGABYTE Z390 AORUS Xtreme WATERFORCE (Intel LGA1151/Z390/E-ATX/All-In-One Monoblock/3xM.2 Thermal Guard/Onboard Intel CNVI 802.11ac 2x2 Wave 2 Wi-Fi/ESS Sabre DAC/Motherboard) - $899.99
https://www.amazon.com/Z390-AORUS-WATERFORCE-Monoblock-Motherboard/dp/B07NDNMD8B

GIGABYTE Z390 AORUS XTREME WATERFORCE LGA 1151 (300 Series) Intel Z390 HDMI SATA 6Gb/s USB 3.1 Extended ATX Intel Motherboard - $899.99
https://www.newegg.com/p/N82E16813145131

 

That's interesting, when I read what they said:

" New platform features to take TR4 to the next level"

I take that as saying AMD are leaving TR4 behind to go to a new platform on the next level - as in a new socket platform will be taking over from TR4!!

And, yes I am serious, I am not joking. To me that reads as clear as anything can be as a heads up that in 2019 ThreadRipper will have a new platform / socket.

Of course AMD isn't going to give the name or any details of the new socket / platform as it was likely still being worked out way back then - about a year and a half ago. Maybe AMD had decided those details then or shortly thereafter, but this is quite typical in that it tells us without naming it explicitly - AMD isn't going to give the "street name" of the new socket / platform until it's announced.



I think you found the smoking gun, this is AMD telling us 1.5 years ahead of time that TR4 was going to be left behind in 2019 and "Castle Peak" was taking us to the next level.

 

Nope, to me it reads platform features, not specifically platform (you are taking the statement out of context). This also does not in the same vein exclude x399 just like pcie-4 did not kill x470 or AM4. so again this is not them saying x399 is dead. You really are trying your hardest to excuse AMD, almost too hard. IMHO, they really blew it and can't just PR this one away.

Edit; now they will still get the business sales as those consumers are all about entire new machines. Where they have lost out is the enthusiast fans of their HEDT, you know the ones they said they would be making happy. Try and get them now to ever trust AMD again!

 

See, and if they say that HEDT will be their new feature playground, which needs a two year socket cadence to constantly get new features, that would be fine. It is about trust and honesty.

As I've said, people would prepare to get on platform, proper advice would be given, etc.

Also, did you see the new Zombieload exploits (V2). Intel's fixes didn't fix the issue. But that is more something to consider for HEDT. Sucks (but still is a toss up on what a person can tolerate).

I agree with TANWare. You are reaching and trying too hard to justify your position and excuse AMD. Similar language was used regarding X570 without a similar incompatibility.

 

yep i did. i am reconsidering if i should spent so much into a 9900ks from SL. cost 1250 just for 2 more cores isn't all that worth it though i only do plan on just having meltdown protection which is in the silicon.

at this point might as well just keep waiting till 2021. zen 4 TR or golden cove and save up big for the next system that might come with 3dxpoint dimm support.

 

Seriously, nobody is talking about the 3950x reviews in here? I think AMD lifted the embargo early.

 

Nah, I've read many of those "slides" over many years, and I'm 100% sure that's exactly what AMD intended from that wording.

x570's are the only PCIE 4.0 capable AM4 motherboards, if not a new socket it is a new platform that leaves the previous AM4 motherboards behind as AM4+ + PCIE 4.0 takes over from AM4.

AMD already promised AM4 would support new CPU's through 2020, no such promise was made for TR4, so we get a whole new socket platform in 2019 for ThreadRipper as TRX40 with PCIE 4.0 and 2x PCIE lanes between the CPU and Chipset.

AMD wanted to tell us exactly what was going on - those are exactly the kind of marketing speak "code words" with a "wink and a nudge" that AMD would use to telegraph a socket change, when they can't explicitly say it.

If I had read that back then I would have gotten the same translation out of it.

Anyway, what's done is done. And, I think I am ready to be dazzled by the performance of the 3960x and 3970x - maybe if there is a nice entry level TRX40 motherboard I'll be able to join in for 2-3 generations of ThreadRipper goodness.

You do realize that TRX40 is exactly the time to buy into a CPU product line, right at the first implementation of a new socket inception so that you have the maximum run time with the platform? Nows the time to buy in and enjoy the ride.

'Tis the season to be Jolly. Here's AMD's Thanksgiving Cheer from 2018, I wonder what AMD will do for this Thanksgiving?

https://www.facebook.com/AMD/posts/10156922209911473:0

Here's where the AMD Holiday Deals were last year:
https://www.amd.com/en/where-to-buy/promotions-comp

 

3950x run cooler than 3900x thats all i cared about. finally seeing some true prowess of 7nm here, binned 7nm. this gives hope to 7nm+ or 6nm w/e its called thats coming up from TSMC and if AMD can lessen those density on consumer/TR chips for higher frequency zen3 might turn out to be a real monster that has no equal.

give us minimum 4.6 or 4.7ghz all core OC with 16-24 cores and im game.

 

More 3950x reviews, here's PCWorld's review and live panel discussing a number of topics starting with the 3950x Review - "Stupidly Fast!":

Ryzen 9 3950X review: AMD's 16-core CPUs is an epic end-zone dance over Intel
AMD's 16-core CPU reigns supreme as the most powerful consumer CPU we've ever seen
https://www.pcworld.com/article/3453320/amd-ryzen-9-3950x-review.html

Ryzen 9 3950X review, Black Friday PC deals, MacBook Pro/Mac Pro, Q&A | The Full Nerd ep. 114
PCWorld
Streamed live 2 hours ago
Join The Full Nerd gang as they talk about the latest PC hardware topics. Today's show covers Gordon's review of AMD's Ryzen 9 3950X, Alaina dives into the PC deals she's seeing so far for Black Friday, and the announcement of Apple's 16-inch MacBook Pro and Mac Pro. As always we will be answering your live questions so speak up in the chat.
Start - 4:03
3950X review - 5:44
PC deals - 46:56
MacBook Pro - 1:05:22






OMG, the 3950x is 10% slower than the 9900k/ks in a game! And, shockingly the 9900k is 1 FPS faster than the 9900ks!! -


Nice performance bump from the 3900x to the 3950x:


BOOM!, as the 3950x enters Ludicrous Speed at 8 threads, and quickly goes to Plaid.

 

Exactly, and what will even better silicon bring with the ThreadRipper 3960x and 3970x?

Robert Hallock seemed pretty smug over the 3960x and 3970x performance results AMD are seeing, and they are anxious for us to see it.

 

It was to expensive this time put with a proper air cooler, Oh’well can’t expect it would lasted. And the 105w TDP isn’t target anymore for Ryzen chips Welcome after. Not much OC headroom left. Manuel overclocking means goodbye to the higher single core boost.
https://www.tomshardware.com/reviews/amd-ryzen-9-3950x-review

 

What Intel HEDT CPU comes with a water cooler in the box?

Robert Hallock said AMD recommends at least a 280mm water cooler, but he plans on using a Noctua NH-D15 to cool his 3950x build at home (Noctua air cooling is also what I've used for years), and he goes into the reason AMD didn't include a water cooler in the box in that interview.

You did watch the Robert Hallock Interview video I posted, right? You wouldn't be criticizing AMD's choices without educating yourself on the subject?

AMD did included water coolers in the past when there were quality and reliability problems with the whole water cooling world. AMD wanted to QC their coolers to the point of reliable operation for their customers.

Now there are lots of reliable choices for water cooling, and AMD doesn't need to limit the customer with their choice - when that choice may not fit the design of the customers case or build.

 

I’m sure a proper air cooler could work. But too expensive. I remember when someone said the extra feature added from AMD increased the value. Everything forgotten

On top the most expensive one the last years. Can’t remember seen mainstream chips around $750 before. Only those binned from Caseking, Silicon lottery etc.

Isn’t 3950X mainstream chips as last year?

Defending all changes is quite amusing

 

You didn't even bother to check prices on recommended air vs watering cooling, did you? Why waste our time with smack when you can't back it up with facts?

Here are some of my choices for cooling the 3950x, and air comes out much cheaper - especially with the holiday sales - now is a great time to buy and build a 3950x rig:

Noctua NH-D15, Premium CPU Cooler with 2x NF-A15 PWM 140mm Fans - $89.95
https://www.amazon.com/Noctua-NH-D15-heatpipe-NF-A15-140mm/dp/B00L7UZMAK

be quiet! Dark Rock Pro 4, BK022, 250W TDP, CPU Cooler - $89.90
https://www.amazon.com/quiet-Dark-Rock-BK022-Cooler/dp/B07BY6F8D9

Ooo, this ones pretty for the looks - hopefully the same performance - I'd pay the $10 premium!:

NOCTUA NH-D15 chromax.Black, 140mm Dual-Tower CPU Cooler (Black) - $99.95
https://www.amazon.com/NOCTUA-NH-D15-chromax-Black-Dual-Tower-Cooler/dp/B07Y3CTQNT

Height (with fans): 165 mm

 

I talk about all the changes from AMD Not what you can buy afterwards or when you order your chips.

For the records. I have posted a lot about that AMD should lower prices instead for the following cooler. But instead we now get higher prices and no cooler