VideoCardz: AMD Ryzen 9 3950X to become world’s first 16-core gaming CPU

Status
Not open for further replies.
They probably just mean the first 16 core on a mainstream socket, thus more aimed at the gaming community than Threadripper, and of course, this is just marketing. Convincing gamers to buy its $600+ CPUs is a huge win for AMD.

If it's $600+ then very few gamers and/or mainstream users are going to buy that processor and motherboard. It will sell as well as Intel HEDT systems...not very well. With that said I plan on getting one, but this isn't where AMD is going to make much money. Their 8C and below processors are where the money is.
 
My general experience with Ryzen 1000 and 2000 series chips is that you can can get all cores to pretty close to the boost clock. It won't be at the rated TDP, but I have not ever really heard of this not being a thing. I do not expect this to change for the 3000 series chips.

That being said, if the new PBO for the 3000 series chips is as good as PBO for the 2000 series chips, then this conversation is largely moot. The mainboard and processor will simply automatically overclock all cores as needed as high as they can go based on cooling, available power, and headroom from the mainboard. And if you need a lot of juice for lightly threaded workloads, those will actually clock a little higher than the official boost while the other cores downclock to idle. No muss, no fuss.

Seriously, PBO is God's gift to my 2700x systems, and it works on my Crosshair VI Hero & Prime X370-A x370 boards just fine.
 
If it's $600+ then very few gamers and/or mainstream users are going to buy that processor and motherboard. It will sell as well as Intel HEDT systems...not very well. With that said I plan on getting one, but this isn't where AMD is going to make much money. Their 8C and below processors are where the money is.

If they are willing to spend that kind of money on an 8c Intel system, they would do it with a 16c AMD part.....the 9900k is not cheap, and the 9900ks i think its called will be even more.

If AMD can bring out a 16c part for $600, the 9900k is in trouble with the $529-$599 price
 
No gaming in title = 0 clickies.

If they left it at mainstream it would have been fine. Are there any games that utilize 16 / 32 ?
in my opinion this is designed for workstations , that is also able to run games very well.
man I´m getting old :oldman:, just bugs me that everything is labeled gaming. end rant.

Things have come along fast , seems not that long ago on a 100w tdp you would be lucky to get half a dozen cores


wow, ESO, a lot of the games I play run 60+ threads. Not all the threads do heavy lifting so I am thinking that having more cores and a few that boost works really well for the MMO/RPG type games. I just loaded up Cities and the Windows Resource Monitor shows it using 135 threads.
 
This is great. Under water the CPU should be able to do 4.5+.

I'm going to start out with a 3800x (or lower 8 core if OCing turns out to be the same) and then maybe in a year bump it up to the 3950x if the price comes down enough.

It's a very good thing when you can buy a great CPU and have room for way more without a full system overhaul..
 
  • Like
Reactions: N4CR
like this

Why would it have to be the 16C part anyway? Looks like the announcement of the 3800x alone knocked the 9900k right off of it's throne. If a 5ghz 9900k isn't enough to get a clear lead over a 3800x, the chip is doomed to drop under $399 in short order..

Capture.PNG
[newegg]
 
Why would it have to be the 16C part anyway? Looks like the announcement of the 3800x alone knocked the 9900k right off of it's throne. If a 5ghz 9900k isn't enough to get a clear lead over a 3800x, the chip is doomed to drop under $399 in short order..

View attachment 166764
[newegg]

Price drops have begun!
 
Two reasons:

1) Building fewer cores doesn't change the clock speed a few of them will reach. There is a limit that specific design can reach on a specific process. So core count has no real impact on the boost clock.

Depends on whether it is thermally limited or not. A overclock limited by thermals would certainly be able to clock higher with fewer cores, but as dies shrink, thermal limits are less and less common.

2) Boost clock is a choice, to be higher on higher core count parts, so when someone shells out the big bucks, they get a processor that is better at everything. A higher clock lower count part would undermine that.

That's sort of like saying that if you want the fastest possible sports car, you have to buy the dump truck.

They are different products for different purposes.
 
What I don't understand is, if they are doing this, why didn't they just announce it at Computex along with all the other models?

It seems awfully weird to have a major launch announcement, wait a few weeks, and then announce an even higher spec:ed model in the same family.
 
Aye aye mate


I agree with the idea (from that video) that Intel knows it will not compete on the desktop (especially high end mainstream) but would happily dominate on mobile and server. AMD has a great product in the server however it takes time to transition. In the mobile the APUs are not good enough to compete with Intel except for the low end where ARM competes. AMD needs to get 7nm APUs out ASAP.
 
An 8C/8T part clocked as high as it can would probably be best for gaming. Gaming gets weird / unpredictable benefits from threading (unless devs are really good - most aren't), and engines generally take very harsh diminishing returns from threads beyond even 6.

But that aside - that's a beast of a processor for highly parallel workloads. I will be eyeballing this for my build stations for sure.
 
You really thought this would do 4.7 ghz all core with a 105 TDP? REALLY?!

not the 105TDP, but at least all-core 4.7GHz, yah

do you all think the existence of the "AMD Ryzen 9 3950X" is legitimate or is this all potentially fake rumors?
 
If you use intel rules where the TDP is essentially meaningless...yes.

That isn't a fair assessment, from my experience. I make embedded systems where things must work all the time, and often with extremely confined specs. Some even blow up, by design. <hint>

I have seen the chips do what they are stated to do. If you want to run beyond that, well, that's a different thing, and requires different power and cooling.
 
If it's $600+ then very few gamers and/or mainstream users are going to buy that processor and motherboard. It will sell as well as Intel HEDT systems...not very well. With that said I plan on getting one, but this isn't where AMD is going to make much money. Their 8C and below processors are where the money is.

$600 is the extremely optimistic low end. It could easily be much more than that.


The 6c 3600x is $250

Two of those in one chip with higher clock is double the price:
The 12c 3900X is $500.


The 8c 3800x is $400

Going by the 6c -> 12c precedent they could be charging double that for the 16C., So
The 16c 3950X is $800.


Price will be somewhere from $600 to $800. As I said this is NOT a mainstream part.
 
Depends on whether it is thermally limited or not. A overclock limited by thermals would certainly be able to clock higher with fewer cores, but as dies shrink, thermal limits are less and less common.

Maybe you just didn't follow the thread back to read the question.

It was about getting 2 out 4 cores to higher clocks.

There is absolutely no advantage to achieving that, vs getting 2 out of 8 cores to higher clocks. We aren't talking about 4 cores all maxing vs 8 all maxing. This is 2/4 vs 2/8.

The 8 core has more area to dissipate heat, it also has more cores to swap between to pick the best two cores, and more cores to rest cores more often, to further spread heat around.


That's sort of like saying that if you want the fastest possible sports car, you have to buy the dump truck.

I like cars and car analogies, so I will play.

Instead of selling you a Sports Car, or Dump truck, AMD is selling a Transformer, that is capable of being a both a Sports car (High clocks on few cores) and a Dump Truck (many cores).

It's a way to make sure top spenders get the best of both worlds. Both AMD and Intel do this, it seems like the most reasonable way to partition, especially when you consider for top parts, they are likely binning.

If you aren't happy with this, you can buy a lower core part and overclock it. Think of this as putting a Turbo in your Honda Civic to make it a fast sports car...
 
Glad to see this was really something hidden up AMD's sleeve and they weren't going to stop at 12. I am curious to see how it will OC compared to the 12-core version, and if similar or better all-core turbo OCs will be possible. Really hope they don't wait long to launch this after the first Ryzen 3000s officially announced this far - a month or two at most.

However, the other question is regarding Threadripper - will there be a Zen2 based TR kit anytime soon or within this cycle? Or will they wait to bring it back in 2020 with Zen2+? Some think that a 16 core proc with high performance will mean that they may dispatch with Threadripper for now, because most of its features (ie more cores, more PCI-E lanes, etc) are being moved into Ryzen or (in the case of things like quad channel RAM or more, ECC official support, tons-o-cores etc..) Epyc? Though AMD has said something to the effect of them "not being done" with Threadripper, I wonder if it is a long time plan or not. I wouldn't like to spend a considerable amount building a top-level Zen2 16 core Ryzen machine thinking TR is gone for at least the next year or two, only to see it arrive.
 
Glad to see this was really something hidden up AMD's sleeve and they weren't going to stop at 12. I am curious to see how it will OC compared to the 12-core version, and if similar or better all-core turbo OCs will be possible. Really hope they don't wait long to launch this after the first Ryzen 3000s officially announced this far - a month or two at most.

However, the other question is regarding Threadripper - will there be a Zen2 based TR kit anytime soon or within this cycle? Or will they wait to bring it back in 2020 with Zen2+? Some think that a 16 core proc with high performance will mean that they may dispatch with Threadripper for now, because most of its features (ie more cores, more PCI-E lanes, etc) are being moved into Ryzen or (in the case of things like quad channel RAM or more, ECC official support, tons-o-cores etc..) Epyc? Though AMD has said something to the effect of them "not being done" with Threadripper, I wonder if it is a long time plan or not. I wouldn't like to spend a considerable amount building a top-level Zen2 16 core Ryzen machine thinking TR is gone for at least the next year or two, only to see it arrive.

Lisa Su said during a press brief after the computeX keynote that there will be a zen 2 threadripper but didn't say exactly when.. my guess is it'll either be Q4 2019 or early 2020.. big thing to look at is when the full launch roll out happens for Rome and they announce the entire processor lineup for that, threadripper will probably be announced/launched soon after that given they are pretty much the same exact processors just with the 8 channel DDR4 limited to 4 channels and running full clock speeds.

also if they use the epyc chipset it'll be at ~15w vs the 11w from the x570 chipset so might see some crazy board designs to keep that thing cool since they're far more likely to see pcie 4.0 raid setups on a threadripper system so cooling that chipset will definitely be a must.
 
Once you exceed stock, all TDP is meaningless. Each architecture scales differently- hell, each SKU scales differently, and each individual part has its quirks.

Have you seen the power draw of an overclocked Threadripper?

:D
Like I say at the end of this poking fun, but Threadripper with PB2 stays within the rated TDP while PBO blows right past it with nary a care.
https://m.hardocp.com/article/2018/11/13/intel_core_i99980xe_vs_amd_ryzen_threadripper

That isn't a fair assessment, from my experience. I make embedded systems where things must work all the time, and often with extremely confined specs. Some even blow up, by design. <hint>

I have seen the chips do what they are stated to do. If you want to run beyond that, well, that's a different thing, and requires different power and cooling.

I was just poking fun at it all. I agree with [H] assessment that most enthusiasts are blast past TDP regardless.

It’s also all in how each company has a different idea of what TDP is. To Intel it’s an average (which....doesn’t compute when you break down the acronym but I digress) and they have boost power limits past the stated TDP that have limit times to be sustained.
 
Like I say at the end of this poking fun, but Threadripper with PB2 stays within the rated TDP while PBO blows right past it with nary a care.
https://m.hardocp.com/article/2018/11/13/intel_core_i99980xe_vs_amd_ryzen_threadripper

Which is fair, if one tries to apply the same functionality to both. However, the Intel parts have more headroom, and if the platform can supply the power and the heat can be dealt with, so be it. Perhaps we'll see equivalents of each from both manufacturers, one from Intel that respects TDP, and one from AMD that wrings out more clockspeed, should it be available.
 
  • Like
Reactions: N4CR
like this
I don't understand what all the fuss is about. 1-2 cores at 4.5+ ghz is well into the area where you are seeing little in the way of better returns on most games and the other cores are still very capable for doing actual pc work for those of us that need it. This seems to occupy a weird space between gaming and productivity and I am seriously debating this over another threadripper.
 
Lol @ all of the people complaining about too many cores..... Just buy an 8 core and OC it!!!!!!!!
It's not like any of the Ryzens will be locked.
I don't see the problem...
 
So, AMD did their E3 thing, and there's no mention at all about those 3950X, so i'm thinking this rumor is totally bogus ?

Re: https://www.techpowerup.com/256393/intel-challenges-amd-to-beat-it-in-real-world-gaming

"AMD's latest stab at Intel is the "Zen 2" microarchitecture at the heart of the new Ryzen 9 3900X 12-core/24-thread processor, leading a gaming-focused processor lineup."

AMD's event is tomorrow.
https://techboostdaily.com/2019/06/...2019-livestream-how-to-watch-the-gpu-keynote/

"When: AMD’s Next Horizon Gaming livestream starts Monday, June 10 at 3pm PT, 6PM ET and 11PM BST"
 
What I don't get is why Intel and AMD won't offer CPU models with lower core counts and high clocks. As a gamer, I would very much prefer a cheaper 4-core CPU that reaches 5.5 GHz on at least two cores over a 16-core CPU that only reaches 4.7 GHz. Even a dual-core i3 @ 5.5 GHz would be better for playing single-threaded simulation flight simulators and games like Cities Skylines, Oxygen Not Included, or Factorio.

Because AMD is using binning to select the best dies and achieve 4.6--4.7GHz on the more expensive R9 models. So there is no 4-core 5.5GHz model.
 
If you use intel rules where the TDP is essentially meaningless...yes.

You got it backwards. Intel TDP figures are real and represent real dissipation [1]. AMD is the one using meaningless marketing values. E.g. that 105W 2700X has a real 140W TDP.

[1] Don't link to some review measuring auto-overclocked chip.
 
:'(

i wasn't trolling, man that blows ... 4.7GHz on 1 or 2 cores ... that blows


i'm hoping to upgrade from my Haswell-EP 18-core which boosts to 3.8GHz on 1 to 2 cores

Typically when really pushing ([H]ard), you can get to boost on all cores with existing Zen and Zen+. But it's not the best way to go for thermals, stress, efficiency, longevity etc. Drop it down a little maybe 4.5 or 4.6 and you might be looking at something realistic all-core on a very good water cooled or crazy huge custom air cooled setup. Power use could get a bit crazy though, wouldn't be suprised to see 200-300W range.

So yes that will be an upgrade in every way over your existing Haswell-EP, even at stock.
 
Alright, I learned somethibg new today:


I knew they had daylight savings, but I didn't know it had its own name and time code

News to me as well.

I find it amusing how formal and official Greenwich Mean Time sounds compared to how colloquial British Summer Time sounds.

We have switched from boring time, to party time. :)
 
What I don't understand is, if they are doing this, why didn't they just announce it at Computex along with all the other models?

It seems awfully weird to have a major launch announcement, wait a few weeks, and then announce an even higher spec:ed model in the same family.
Just like when you give a girl an amazing night and she comes back a few times more, AMD is looking to keep that girl by giving her multiple amazing nights so she is hooked for ever.
Replace girl with you and night with marketing and you get the idea. AMD needs mindshare above all and what they are doing now, creating discussion about their new parts is how they get mindshare and shitloads of articles, rumours and youtube videos.


That isn't a fair assessment, from my experience. I make embedded systems where things must work all the time, and often with extremely confined specs. Some even blow up, by design. <hint>

I have seen the chips do what they are stated to do. If you want to run beyond that, well, that's a different thing, and requires different power and cooling.

You should check out the power draw of 9900 in turbo which goes well beyond TDP. Run them TDP to TDP and the 2700x is same speed... Same with Xeons, they almost double their TDP in some use cases.
Embedded stuff with strict TDP isn't relevant to high end server/desktop use cases and turbo beyond TDP rating. But if you are making space/mil stuff it's a little different I guess.. (hint taken).
TR also goes over TDP as mentioned earlier in PBO (O = overdrive..) by IdiotInCharge

So far only reason I'd see to get TR beyond crazy network/drive/semi-pro specialised WS setups is for the binned sillicon and extra 100-200MHz it will get. It's quite tempting though.
 
Lol @ all of the people complaining about too many cores..... Just buy an 8 core and OC it!!!!!!!!
It's not like any of the Ryzens will be locked.
I don't see the problem...
The "problem" is that the 8C CPU chiplets won't be the same high level binned chiplets found in the 12C and 16C CPUs, so getting the 8C CPU to overclock to 4.7 GHz on 1-2 cores cores like the 16C does out of the box won't be easy or might be impossible due to silicon lottery.
 
You got it backwards. Intel TDP figures are real and represent real dissipation [1]. AMD is the one using meaningless marketing values. E.g. that 105W 2700X has a real 140W TDP.

[1] Don't link to some review measuring auto-overclocked chip.

"Intel TDP figures are real and represent real dissipation as long as you don't exceed the base clock and you have a motherboard that respects the stock PL1 as defined by Intel, and that's a totally real scenario for the end user! Trust me!"

Meanwhile, AMD uses a published formula for determining the TDP of their processors based on set load and idle temperatures and the heat capacity of their most effective stock cooler. The relevance of this number is debatable, but at least it's not up the end user to investigate and determine whether or not they need to hobble their processor (they do) to hit that number.

Also, where are you getting that 140W number? I can't find evidence of a 2700X consuming more than right around 100W in normal stress testing loads. The worst I can find is 117W at Anandtech (a test where the Intel processors exceed their stated TDPs by an average of almost 20% vs the AMD parts which come in at 11% LOWER than their stated TDP on average.)

*EDIT* Even Kyle only hit 167W total package power on an overclocked 2700 running ~1.4V and all cores at 4.2GHz...

1527715968fh9328n4uq_5_1.png


Anyhow, to keep this post somewhat on-topic, I don't expect the 105W number to be completely realistic for the 3950X (especially if PB2 is still a thing), but I also don't try to pretend that I think TDP means anything for the enthusiast anymore.
 
Last edited:
Status
Not open for further replies.
Back
Top