AMD Ryzen 4000 7nm+ Zen 3 CPUs Spotted

I thought many consoles were sold as a loss and the profits were all in the games and (licensed) accessories?

Most consoles have been sold at a small margin even at launch... with the expectation that refined processes will drive those margins up over the consoles life span. They make the majority of the cash from royalties and accessories no doubt but they don't loose money on consoles... no matter what any one guesses. No one is jamming a thousand dollars worth of tech into a 500 dollar package. Worst case for Sony and MS they are getting everything for $450-490 and selling it for 500. lol :)
 
You would have to think they have something in the pipeline as next gen consoles will be Navi based. I can't imagine a separate graphics chip alongside a CPU is a very cost effective way to build a $500 console.

If you think about APU (or SoC) design as something like Legos, it would make sense for AMD to release a desktop and mobile counterpart to the part they're shipping MS and Sony.

But we've been expecting that from AMD for a decade, and generally speaking, they're extremely conservative as to how they release their products.

My guess is that while AMD looks good in the news, they're still on the razors edge when it comes to operating costs. A failed product today would hurt them significantly, so they're not really taking risks yet.

Cause seriously, I'd be all over an ultrabook APU that could hook up to a 120Hz+ Freesync panel and crank out decent performance when plugged in. Or a desktop CPU with an eight core CCD and an HBM-based GPU package in place of the other CCD?
 
The cool thing is, AMD's chips are a serious temptation to upgrade from the first gen Zen to either Zen+ or Zen 2, just for gaming alone. :) The only reason I have not done so is because I have 3 desktop gaming computers and would not benefit on it overall, even if I would most definitely see increased performance across the board. ;)
 
I think with 20/20 hindsight the bulldozer generation wasn't really that bad. They fell victim to Intel cheating. (I'm being very serious) That time frame is exactly when Intel started building their cheating cache systems that didn't bother doing basic security checks. So sure they gained a ton of performance... we just didn't know how unsecure Intel systems where at that time. Although no doubt the Gov knew.

Consumers referenced those terrible launch reviews for years and wrote AMD off. If you take any of those supposed junk FX chips and bench them today vs their Intel rivels from the time with all mitigation for both platforms. All of a sudden Bulldozer vs Core doesn't look all that bad.
I'd like to see some of those modern comparisons. Do you have links to any?
 
If you take any of those supposed junk FX chips and bench them today vs their Intel rivels from the time with all mitigation for both platforms. All of a sudden Bulldozer vs Core doesn't look all that bad.

FX-8350 (which is Pile Driver architecture) is a very fine gaming CPU. Loads games noticeably slower than say an i7-7700 (but still acceptably fast enough) but when it comes to gaming I can run ALL games on high settings smoothly (except perhaps RDR2, haven't tried it yet) at 1080p @ 60 fps or better using a GTX 970
 
  • Like
Reactions: ChadD
like this
I'd like to see some of those modern comparisons. Do you have links to any?

https://www.phoronix.com/scan.php?page=article&item=sandy-fx-zombieload&num=1

Intel still wins but its not the pure blow out that was reported at the FX launch. Its funny seeing some of the server tests where the FX chips gained performance with mitigation. Intel still wins overall... but there are some examples such as context switching (which would be very much like some of the old heavy multi tasking tests you used to see people include in reviews) where Intel went from being 3-4x faster then AMD to basically the exact opposite.

Michaels conclusion;
"Lastly is a look at the geometric mean for all of the tests carried out successfully on all of the processors/combinations. The Intel Sandybridge/Ivybridge CPUs all basically remained faster than the AMD FX CPUs benchmarked even when mitigated and disabling SMT with the exception of the Core i3 2120. The Core i3 Sandybridge processor's geometric mean went from being ahead of all the FX processors to around the speed of the FX-8320E with the default mitigations and behind if disabling Hyper Threading. The Intel Core i5/i7 CPUs remained ahead but the Core i5 2500K saw a 14% hit to the performance in the benchmarks ran with the default mitigations. The Core i7 2700K and i7 3770K each saw around a 14% hit to the default mitigated performance as well but that extended to 18~19% if disabling Hyper Threading in the name of security. The AMD FX CPUs saw minimal change to performance with their relevant Spectre mitigations."

So sure Intel would still have been in the lead... but FX wouldn't have been the train wreck it was portrayed as either. Instead of it looking really bad it would have been more like Zen one... where sure it was a bit slower then Intel but at the price... and bonus core count who knows perhaps it would have been seen differently.
 
Last edited:
FX-8350 (which is Pile Driver architecture) is a very fine gaming CPU. Loads games noticeably slower than say an i7-7700 (but still acceptably fast enough) but when it comes to gaming I can run ALL games on high settings smoothly (except perhaps RDR2, haven't tried it yet) at 1080p @ 60 fps or better using a GTX 970

They aged surprising well. :)
 
Last edited:
FX-8350 (which is Pile Driver architecture) is a very fine gaming CPU. Loads games noticeably slower than say an i7-7700 (but still acceptably fast enough) but when it comes to gaming I can run ALL games on high settings smoothly (except perhaps RDR2, haven't tried it yet) at 1080p @ 60 fps or better using a GTX 970
Definitely can't run PUBG smoothly with those specs... But that game runs like shit on all hardware... Stuttering mess... Awful frame times.... My 3700x runs it pretty well but my 1700x and i5 2500k both had major stutter problems.
 
I am highly interested in Zen 2 based (cores) in the next APU. Should make APUS absolutely hauil ass even over this generation .
 
I wonder if this means we'll see Zen 3 within 6 months time?

Will probably be a couple months before I can get a 3950x for my next upgrade, and I'm debating if I might as well just wait for Zen 3
 
I wonder if this means we'll see Zen 3 within 6 months time?

Will probably be a couple months before I can get a 3950x for my next upgrade, and I'm debating if I might as well just wait for Zen 3

If intel doesn't do anything competitive it is expected AMD will slow down release of new tech so fast.
 
I wonder if this means we'll see Zen 3 within 6 months time?

Will probably be a couple months before I can get a 3950x for my next upgrade, and I'm debating if I might as well just wait for Zen 3
1st half 2020, rumor is June, 2020
 
If intel doesn't do anything competitive it is expected AMD will slow down release of new tech so fast.

Don't think they will stop.

AMD needs to recover market share as fast as possible, and nothing will entice more users, consumer or datacenter, to upgrade more than a giant power/speed increase.

That said, I want Intel to release 10nm already.
 
Don't think they will stop.

AMD needs to recover market share as fast as possible, and nothing will entice more users, consumer or datacenter, to upgrade more than a giant power/speed increase.

That said, I want Intel to release 10nm already.

another thing to remember is that this roadmap was laid out years in advance with the expectation that intel was going to be releasing newer chips in the process. rome was created to compete with intels 10nm server chips they were promising, that didn't happen zen 3 was meant to compete with the 10nm refresh which hasn't happened and zen 4 was likely to compete with intels 7nm which won't be until 2021 maybe 2022(my prediction). so to then decide to slow that down because intels not releasing anything puts you in a position where you have all this potential profit margin sitting there for no reason. they've seen what that mentality does by looking at the last 6-7 years with intel, they would have to be fools to make that same mistake.
 
so to then decide to slow that down because intels not releasing anything puts you in a position where you have all this potential profit margin sitting there for no reason

The cost of switching die production will eat profit margin too, especially to newer process nodes. Remember that AMD still has to compete with the likes of Apple for wafers, and Apple is as much or more interested in increasing performance and reducing power consumption.

The only thing AMD can gain is marketshare; if they keep an exhausting release cadence, they'll pay for it at the bottom line, and well, marketshare isn't worth much considering their history of gaining crumbs only to be wiped out again when they inevitably stumble.
 
Don't think they will stop.

AMD needs to recover market share as fast as possible, and nothing will entice more users, consumer or datacenter, to upgrade more than a giant power/speed increase.

That said, I want Intel to release 10nm already.

Intel has abandoned 10nm.

Murthy Renduchintala, Intel's chief engineering officer said a few months back; "We have humble pie to eat right now, and we're eating it," "My view on [Intel's] 10nm is that brilliant engineers took a risk, and now they're retracing their steps and getting it right."
He also told investors 7nm was progressing well and basically they where back to using tech with 7nm that they understood better. He basically talked to investors for 20 min and danced around 10nm pretty much admitting they will never bother making real desktop 10nm parts. Their focus is clearly on getting 7nm up as fast as possible now. Although he also said it took about 4 years in general to get a new node up... so they are shooting for 2021, but by their own admission that would put them about a year ahead of scheduled so well see.
It sounds like Intel brought Renduchintala in to fix 10nm.... and after managing to get a few low performance laptop parts out he has basically admitted 10nm is a lost cause. In a more recent interview he said something like they went down a rabbit hole with 10nm pursuing advances that just weren't working for far to long... by the time they retrace everything back to where it was and got it working the 7nm team will basically be ready.

Worse case for Intel 7nm isn't ready to go till early 2021 and 2020 is basically the year AMD really beats them with a stick... and they loose a lot of OEM contracts. I fully expect 2020 Intel is going to dip into their dirty trick book to try and keep OEMs tied up till their 7nm is actually ready. The enthusiast market is pretty much already lost and zen 3 is going to make that even more the case. For Intel now it will be all about the sales dept and how well they do to hold on to their big OEM and server clients.
 
The cost of switching die production will eat profit margin too, especially to newer process nodes. Remember that AMD still has to compete with the likes of Apple for wafers, and Apple is as much or more interested in increasing performance and reducing power consumption.

The only thing AMD can gain is marketshare; if they keep an exhausting release cadence, they'll pay for it at the bottom line, and well, marketshare isn't worth much considering their history of gaining crumbs only to be wiped out again when they inevitably stumble.

LOL! You know that most are thinking people who see through stuff like this.
 
They've more recently said that 10nm for the desktop is on track.

You mean Intel telling investors all is well. (they didn't in fact they never even mentioned 10nm desktop parts to investors... just started talking about what comes after)

There have been reports back and forth of them abandoning and having it on track. I tend to believe the on track reports are BS, posted by Intel fanboys or worse leaked by Intel people that know different.

This was an official Intel response a few weeks ago to one of the sites reporting 10nm abandonment.
“We continue to make great progress on 10nm and our roadmap of 10nm products will include desktop. However, we have not yet announced specific timing for desktop products”

Yep sure sounds like they have it all in hand. lol :rolleyes: They only released laptop parts cause their reduced size gates don't work properly. So all the talk of 10nm being = to 7nm is out the window. Their head engineer has said they had to go back over 10nm and increase the gate sizes to make it work... which means a complete redesign. Which I would imagine is why after admitting that they talk about nothing but 7/5 nm and 3D stacking tech for 3-4 years out.
 
  • Like
Reactions: N4CR
like this
So your opinion on their cautious statement is that they're doomed.

lol

In general when your getting your ass handed to you if you can say just wait for Q1 / Q2 or what have you. That is what you do. Intel has even been willing to cannibalize their own sales to hype future products. They don't talk about 10nm desktop anymore outside of rumors or 1 line answers to articles that gain too much traction.

Its clear 10nm is dead. Their head engineer as much as said so months back already... saying they basically wasted 2 years trying to make reduced gate size work when they should have looked at it with a sane eye and said this isn't working. Even if they do manage to get a desktop 10nm part out in the later half of 2020... there 7nm team will be months away from tapping out. So the 10nm parts which will no doubt require a new chipset will be dead end parts, I can't imagine Intel even bothering to release something that OEMs aren't likely to support outside basic/OEM MOBO skus. The better move is to simply sit it out another 6-9 months for their 7nm to be ready.
 
Last edited:
  • Like
Reactions: Auer
like this
No offense but I hope they dont release a Threadripper 4 until at least 2021. Gawd I'm so tired of being obsolete 3 months after buying a new platform.
 
  • Like
Reactions: Auer
like this
In general when your getting your ass handed to you if you can say just wait for Q1 / Q2 or what have you. That is what you do. Intel has even been willing to cannibalize their own sales to hype future products. They don't talk about 10nm desktop anymore outside of rumors or 1 line answers to articles that gain too much traction.

Its clear 10nm is dead. Their head engineer as much as said so months back already... saying they basically wasted 2 years trying to make reduced gate size work when they should have looked at it with a sane eye and said this isn't working. Even if they do manage to get a desktop 10nm part out in the later half of 2020... there 7nm team will be months away from tapping out. So the 10nm parts which will no doubt require a new chipset will be dead end parts, I can't imagine Intel even bothering to release something that OEMs aren't likely to support outside basic/OEM MOBO skus. The better move is to simply sit it out another 6-9 months for their 7nm to be ready.

Based on where they are now, I expect them to produce SKUs on both nodes side by side.
 
  • Like
Reactions: ChadD
like this
So the 10nm parts which will no doubt require a new chipset will be dead end parts, I can't imagine Intel even bothering to release something that OEMs aren't likely to support outside basic/OEM MOBO skus. The better move is to simply sit it out another 6-9 months for their 7nm to be ready.

Sounds reasonable.
 
No offense but I hope they dont release a Threadripper 4 until at least 2021. Gawd I'm so tired of being obsolete 3 months after buying a new platform.

If AMD keeps on track, we will see fourth generation Threadripper CPUs about September or so next year. However, they are supposed to work with existing sTRX4 motherboards.
 
Based on where they are now, I expect them to produce SKUs on both nodes side by side.

Your expectations does not mean Intel will do so, however. They are in trouble at this time and their size and monetary advantage does not guarantee success for them, at all.
 
Your expectations does not mean Intel will do so, however. They are in trouble at this time and their size and monetary advantage does not guarantee success for them, at all.

Trouble is a relative term. It is unlikely that a company that size with that kind of monetary advantage doesn't figure out a way to make things work (likely at 7nm not 10nm) sometime in the next 2 years for HEDT/Server. In the meantime, it's not like their notebook chips aren't selling well.
 
  • Like
Reactions: Auer
like this
Nope, he simply said Intel has more or less confirmed that 10nm desktop is dead.

...while quoting them as saying the opposite.

Your expectations does not mean Intel will do so, however. They are in trouble at this time and their size and monetary advantage does not guarantee success for them, at all.

My expectations are based on their history, not your religion.
 
Based on where they are now, I expect them to produce SKUs on both nodes side by side.

Its possible but I am not sure trying to produce enthusiest level parts on 10nm will make much sense. Your probably right though they will try and recoup their investment I am sure... who knows perhaps they will build a ton of SOC 10nm arm parts. ;) Not being a joker on that one... the Intel ARM SOC market they have actually been growing, and we know they have got 10nm to work seemingly all right there.
 
Nope, he simply said Intel has more or less confirmed that 10nm desktop is dead.

I don't think even the hardest AMD fanboy would say Intel is done. They'll be back... just seems like for the first time ever AMD might have an entire year to rub Intels incompetence in. As IC pointed out earlier Intel is still not in a bad position from a business standpoint. Their 2020 product (and perhaps even 2021) is going to be inferior. They do however have a long history with OEMs, and products in the pipe. Intel wrangling OEMS up and having them promote/spin Intel for 12-18 months isn't out of the realm of possibility.

Those thinking AMD is going to sit back now and try and make margin are being silly. AMD still ships far less to the major OEMS that account for the vast majority of CPUs sold. They will deliver on performance that seems clear... but they are going to have to continue bringing the pricing hammer, OEMs don't care if AMD is a few % points better they care about their own % points.
 
Trouble is a relative term. It is unlikely that a company that size with that kind of monetary advantage doesn't figure out a way to make things work (likely at 7nm not 10nm) sometime in the next 2 years for HEDT/Server. In the meantime, it's not like their notebook chips aren't selling well.

Intel isn't ordained to be the #1 CPU company forever.

We heard the same thing 30 years ago about IBM. There huge, they have armies of lawyers and engineers... and a sale force that takes no prisoners. Sometimes just throwing more and more good money after bad doesn't solve the problem. 10nm was a massive investment that appears to be a fail. Which means Intel has next to nothing to move their product forward for close to 2 years. They brought in some new head Engineers to try and fix things but really they waited to long to backtrack. Which means any redesigned 10nm now won't hit market in time to make any impact on the market. Its a minor miracle they got 14nm+++ to a point where they are still even in the game at all on the high end.

Anyway if it happened to IBM it can happen to anyone.
 
Back
Top