Https://wccftech.com/exclusive-amds-plans-for-7nm-ryzen-apus/amp/

>Ryzen 2 APUs in November
Get hype

Attached: AMD-Radeon-RX-Navi-Mockup-e1552291307270.jpg (1200x675, 143K)

Other urls found in this thread:

wccftech.com/exclusive-amds-plans-for-7nm-ryzen-apus/amp/
amd.com/en/press-releases/semi-custom-graphics-chip-2017nov6
en.wikipedia.org/wiki/Double_data_rate
tomshardware.com/news/sk-hynix-first-ddr5-ram,38079.html
anandtech.com/show/13605/sk-hynix-develops-first-16-gb-ddr5-chip-demos-ddr5-rdimm
anandtech.com/show/13771/amd-ces-2019-ryzen-mobile-3000-series-launched
tomshardware.com/news/ddr4-memory-overclocking-world-record-ram-micron,39358.html
twitter.com/SFWRedditImages

Unless you're building your own nas or an office pc there's no point in going for an apu. Especially waiting almost half a year for that shouldn't be that exciting.

a320's will only support 3000 apus

This. OP is a viral marketer shilling for AMD. He's not even trying to hide it.

I'm waiting for Zen 2 APUs because I'm hoping with the chiplet design and all they'll be able to put more than four cores on a chip, and maybe more than 10CUs, because between those two things and lower power consumption OEMs should actually start using Ryzen parts for their business machines, including mobile parts for their laptops.

All I really fucking want is a Latitude 7490 (only the 7000 series get the magnesium alloy chassis) with a non-touch FHD display and a Ryzen APU. Is that too much to ask?

>wccftech.com/exclusive-amds-plans-for-7nm-ryzen-apus/amp/
The 3000 mobile series just launched, OEMs have just started building up inventory for their new systems. The 4000 series 7nm APUs aren't going to be anywhere until mid 2020.
As usual WCCFtech is low IQ Paki clickbait.

The APUs are not going to be massive package chipet designs. The socket FP5 isn't getting replaced. Its a monolithic die. IGPs aren't going to dramatically increase in size because there is no economically viable solution to their bandwidth bottleneck.

i don't care nothing about amd, they are shit

We're not talking about Ryzen 3xxx series APUs like the 3200G/3400G. This would be Zen 2/Navi based, so 7nm chiplet designs. 7nm increases density, which means more transistors in less space, i.e. 6/8 cores and 12-16CUs fitting in the same socket the current monolithic 12nm 4-core 10CU parts do.

No, I'm well aware of what we're talking about. 7nm APUs are not chiplet designs, and they still use FP5 for mobile. It is a monolithic die still using the same package that current Raven Ridge does.
AMD is not putting a fucking retarded massive MCM package inside of thin cheap laptops. Its socket FP5. Take a look at that socket since you're not familiar with it. Thats not housing 3 different dies.

Why do you think the Zen 2 APUs will be monolithic when they separated so many things already with the I/O die and chiplets? Wouldn't making a tweaked I/O die be better for their yields given that they're already doing core chiplets and GloFo's process is very mature for their I/O? Matisse's I/O die is about 122.6mm2 and the core chiplet is about 70mm, putting in a GPU component on 7nm doesn't look like a huge chiplet if we go by Raven Ridge's 209mm2 die.

Attached: Screenshot from 2019-06-14 14-21-36.png (998x751, 1.21M)

As someone who buys lots of office machines, an 8c/16t offering from AMD with an iGPU would be good competition for the i7 machines. It's the biggest sector of the market they don't currently have a product in.

These are products that aren't even close to being announced, as far as I'm aware exactly no one knows anything about 4xxx series APUs and it wouldn't make sense for them to try to split up the 8-core Zen 2 chiplet and re-integrate the I/O anyway, so unless your dad works at Nintendo where the fuck are you getting your information, shill?

>Why do you think the Zen 2 APUs will be monolithic
Because the FP5 BGA socket isn't deprecated, future APUs are still using it.
It is not a chiplet design. The IGP is still modest in size. It has no on package memory. It is essentially identical to Raven Ridge but with Zen2 cores and upgraded Navi GPU instead of Vega.

>it wouldn't make sense for them to try to split up the 8-core Zen 2 chiplet
Sherlock, APUs aren't made from cut apart desktop dies. Jesus. They are their own completely independent dies.
You're so painfully ignorant I don't even know where to begin addressing all the misconceptions you just laid out.
Socket specs are something that gets added to things like BIOS updates, coreboot, and Linux Kernel updates as soon as test chips come back to the fab, thats how we knew Kaveri was getting an updated socket with FM2+
The future 7nm parts are still using the existing sockets and package, and they are not physically big enough for an MCM.
Face it, Adoredtv is entirely full of shit.

Attached: kaveri sockets in coreboot.png (576x617, 54K)

Jesus Christ this will set me up. I almost gave up and was thinking of 2400G/3400G but now I can hold on a little.
That's not true, Vega 11 in 2400G/3400G is already fairly powerful, if not for shit RAM it'd probably beat RX 550, and as it is now it's enough for light gaming as long as you don't need to max out things.

This is this a Raven Ridge FP5 BGA package for reference, the same package 7nm parts will utilize.

Attached: raven-ridge.jpg (725x506, 45K)

>maybe more than 10CUs
Cool and all but 2400G and 3400G are 11CU

This, going with a huge die is contradictory with what APUs are about, and it's got to be huge because if they don't increase core/CU count from current gen then there's no point even making them.

Let's assume you are the all-knowing god you claim to be and there aren't going to be MCM designs for FP5 (which, btw, doesn't rule out MCM APU designs for desktop which almost certainly will happen based on the amount of space available on AM4 and the way the chips are laid out). That does not mean more than four cores and 11CU would be impossible for FP5 with 7nm density increase.

And I don't watch Adored or any other Youtuber so have fun with your assumptions faggot.

There's plenty of space on it, for example that metal rim around it doesn't even need to be there or can be narrower.

Well, that goes for every pro-AMD thread, nobody intelligent to go on Jow Forums would fall for their lies, this is a bastion of blue.

>not ready to labdobs
what's the boint

It does in fact rule out an MCM. package specs are not just about space, my uninformed eceleb following friend. Its also about the electrical properties of the package. If AMD had a new package spec people would know about it, exactly as with every prior package and socket change that was leaked a 9 months~ before product announcement.

APUs aren't performance parts, they're budget solutions. They're keeping this as cheap as possible. There is no separate IO die. There is no separate GPU die. There is no separate core logic die.
Its a monolithic design, exactly as all APUs have been, and its reusing the exact same socket and package.

Do you mean the 3400g or something better altogether?

3000 mobile series actually launched back in January. We're just now seeing it in products because integration takes a lot of time, so does building stock. So basically we can expect these in actual products around march-june 2020

Probably 4000 mobile series. This isn't even news, it's about the time scale we expected for an yearly release

>9 months
We don't even have Zen+ APUs on the desktop yet. I think we're getting Zen 2 APUs next year Q1 at the earliest.

fake news

>Probably 4000 mobile series
A huge, huge, disappointment. I thought AMD were good friends of us in the battle against Intel. Now? They fool APU buyers. Really sad.

You think they'll sell a consumer version of the APU they'll be using for ps5/Scarlett? That would kick ass.

Of course not. Semi-custom design work belongs to the client.

That was true for the previous gen, ps3, 360, but not for this one. They manage production and ship directly to the console companies, which isn't typically what happens when the client owns the IP or design. For PS3 and 360 Sony and Microsoft handled the manufacturing and all the integration. Although I'm sure they have an exclusivity agreement for those particular chips, because who would want to buy something that can get sold behind their backs after they invest a lot of money in it

he meant the 7nm Renoir APUs with Zen2 and Navi.

Nigger, the PS3 and 360 weren't semi-custom designs from AMD.
The APUs AMD designs for a client are their property as a completed product. AMD owns most of the IP contained within, but as they are semi-custom, they contain IP from the client as well. The completed design for the PS4 APU is something only Sony can sell. Same deal with the Xbone.
It is the exact same scenario with the PS5 and Nu-Boner. AMD cannot and will never sell these semi-custom designs on the consumer market.

The 360 gpu was from ati

I don't know the marketing terminology. What does that mean?

Not even remotely relevant, retard. Semi-custom is a specific business strategy of AMD which has legal ramifications and strict guidelines.
The business model would not exist if AMD could turn around and sell their client's protected IP because no one would agree to it.

Bullshit. Absolute horseshit. The 7 nm apus release 5 mo after the July ones? Not a chance.

What kind of dumb son of a bitch believes this? You aren't seeing 7 nm apus until like q1 or 2 2020. And you won't see a super jump in perf until ddr5.

Doesn't run OSX so who cares. (Only poor people)

You sure are redefining a bunch of terms to fit your narrow interpretation

What do you expect from the same group of Pakistanis who said Kaveri was going to have 6 cores and 12 threads?
They post baseless clickbait constantly because Pakistan is an even more shit poor country than India and these people need ad revenue to buy a pittance of rice each month.

I'm not redefining anything, /v/tard. AMD is not selling you console APUs.
These aren't just off the shelf parts assembled and sold to their respective buyer. They are, exactly as the name implies, "semi-custom." The PS5 APU will have IP belonging to Sony. The new Xbox APU will include IP belonging to Microsoft. AMD has no legal right to this IP, it belongs to the respective holder.

>A PU

Attached: Af3gz.png (604x531, 243K)

Semi-custom at it's base has no requirement incorporating for third party IP. That isn't what makes it semi-custom.

Also I'm not suggesting the new xb or ps5 won't have 3rd party ip in their socs. Since you'll probably try to make a post about that

the desktop APUs are some of the cheapest processors and the 7nm process is quite expensive - that's why AMD started with the Ryzen 9 and 7. They will add cheaper models later and possibly next year the new 7nm APUs.

You won't see a $100 7nm APU for quite some time.

You're just being willfully ignorant and grasping at straws.

I'm just saying that you're definition of semi-custom was wrongo

t.

Attached: 0a4.png (200x294, 68K)

You're just painfully fucking dumb.
The buy in for AMD's semi-custom business is $100 million. They don't do semi-custom designs for less than that, it is their minimum buy in.
A client seeking to have a semi-custom APU isn't going to do so just to have a generic configuration of CPU cores and GPU CU. They do it to work in their own instruction sets, their own DSPs, their own accelerators, which is exactly what happened with the PS4 and Xbone. It will not be any different for the PS5 and new Xbone.

>And you won't see a super jump in perf until ddr5.
The real jump is when they stack DRAM on top of the dies. Well fed dies leads to great performance.

Attached: amd-iedm-2017-35-1024x573.png (1024x573, 475K)

Finally, Apu’s time to shine

Attached: 52F0D5A5-AE01-4118-9489-13A6EFB99A9E.jpg (800x800, 85K)

How would this pic even look like right now with chiplets?

Thats the end goal, but its not happening any time soon. As always how economically viable the solution is determines whether or not it reaches mass market. At present even HBM is far too expensive for entry level hardware despite that very hardware needing it.
DDR5 is a decent stop gap at least.
6000mhz DIMMs in a dual channel system will let IGPs breathe easier.

So you need 12000MT/s to use an iGPU?

Hey, user, is this semi-custom

Attached: 2018-01-07-image-25.jpg (1854x964, 272K)

That's a big package

A system running 6000mhz DIMMs would have 96GB/s to work with and feed that IGP.
If you have 3200mhz DIMMs in Raven Ridge now you're only getting 51.2GB/s peak bandwidth. My 2500u laptop has a mere 38.4GB/s.

No its not. This was a cross licensed GPU, not billed through AMD's semi-custom business.
Nice shitpost, /v/tard.

AMD isn't selling you a console APU.

Whats the point of these apus? Just for consoles or what

AMD's mainstream APUs have nothing to do with the console chips.
These are for cheap laptops, OEM desktops without discrete GPU, thin client machines, point of sale machines, digital signage controllers in embedded config, etc.
Its a single chip solution which leads to lower total bill of materials for any given system an OEM puts one in.

>Whats the point of these apus?
Eventually eating up the entire mobile and desktop market for 99.99% of users.
Think of buying an APU with 64 cores, plenty of DRAM, and graphics that would make today's offerings a complete joke at mainstream prices.

I see so I should be happy apu’s are improving so they compete with low end mainstream and bring down overall prices even for higher end stuff?

APUs in the mainstream are making entry level discrete GPUs a thing of the past. They aren't going to target the high end any time soon, or ever.

>No its not. This was a cross licensed GPU, not billed through AMD's semi-custom business.
AMD refers to it as semi-custom in press releases.

amd.com/en/press-releases/semi-custom-graphics-chip-2017nov6

Yeah but whenever cheap things get cheaper, usually expensive things get less expensive. For example a 2080 ti cant be $400 cheaper because then it would be cheaper than a 2080, so if something knocks down the 2080 in price, the 2080ti has room to go down.

>Yeah but whenever cheap things get cheaper, usually expensive things get less expensive
Not really, dude. Over the past few years we've seen the price of Halo products increase.

Ok I'm lost.
I thought this generation we would only have 3200G and 3400G for APU's?
Or the article isn't talking about desktop?

Attached: 1478254254421.jpg (238x192, 9K)

6000MHz would be a world record for years to come. Are you sure you don't mean MT/s? Even 3200MHz would be a world record.

They're speculating about the next generation of APUs, not the Raven Ridge refresh.
Though in typical retard wccftech fashion they incorrectly use the code name Raven Ridge. Pakis need to just get nuked so we can be free form this shitposting. Only shit eating morons read or post their trash here.

The price of the midrange and bottom also increased

>6000MHz would be a world record for years to come.
No, no it wouldn't. Thats standard speed for DDR5.
Samsung showed off 6500mhz DIMMs, and 5500mhz low power DIMMs a year ago.

Do all of you just get spoonfed news when retards like Linus Tech Tips finally cover it or something? Read the fucking news.

I think you're the retard here.
en.wikipedia.org/wiki/Double_data_rate

You are correct, we're seeing Zen+ APUs now. We would be seeing Zen 2 APUs much later, probably first half of 2020. As always wccftech is full of shit

So you're pretty much confirming that you're an underage dipshit who can only regurgitate what he hears from youtube ecelbs. Gotcha.

tomshardware.com/news/sk-hynix-first-ddr5-ram,38079.html
anandtech.com/show/13605/sk-hynix-develops-first-16-gb-ddr5-chip-demos-ddr5-rdimm

In 2018 SK Hynix showed off 5200MT/s low power DDR5 chips.
Samsung showed off 5500MT/s, and then normal power 6000mhz DIMMs.
The max supported speed in the JEDEC standard for DDR5 is fucking 6400mhz/MT/s.

Read the fucking news, kid.

"Describing the bandwidth of a double-pumped bus can be confusing. Each clock edge is referred to as a beat, with two beats (one upbeat and one downbeat) per cycle. Technically, the hertz is a unit of cycles per second, but many people refer to the number of transfers per second. Careful usage generally talks about "500 MHz, double data rate" or "1000 MT/s", but many refer casually to a "1000 MHz bus," even though no signal cycles faster than 500 MHz."

Everyone knows how DDR works, little kid. The effective clock speed is used ubiquitously in place of the transfer rate in MT/s
Your DDR4 2400mhz kit isn't actually clocked at 2400mhz, its 1200mhz, but 2400Mt/s it its effective transfer rate.
DDR5 increases this to the range of 6000~ MT/s.

You're just grasping at straws, littl kid. Its fucking sad. Go back to your youtube comments.

Can you please show me a kit that does 5500MHz or 6500MHz?

>whoops I got proven wrong
>I'm going to keep shitposting like an immature child
Nice. Whats next? Are you going to get confused about JEDEC nomenclature for storage density vs scientific notation? lmao

>Samsung showed off 6500mhz DIMMs, and 5500mhz low power DIMMs a year ago.
Where are they?

Listed on their site.
You'd know about this if you actually read any tech news instead of watching your low IQ eceleb trash. Its not even new news. This is all for 2018.
Sad.

Can you please link me to an article that cites 5500MHz or 6500MHz DIMMs?

>tech illiterate little kid still desperately shitposting after getting proven wrong and making an ass of himself
I've already linked plenty.
I love it. I know you're just over there seething, in tears, frozen in childish impotent rage.
Samsung's DDR5, at last announcement, handles up to 7.5gbps/pin. Making these 6400/6500MT/s modules slow in comparison. Wow!

Attached: 190221_ISSCC_800_2-min.png (850x431, 90K)

We wil see 60GB/s from a single channel within the near future.

That's in megatransfers, but earlier you stated megahertz. I'm confused.

As already pointed out in colloquial usage they are interchangeable terms, you autistic brat.
The effective clock rate is the exact same as the total transfer rate, so they are used interchangeably, and have been ever since DDR came into existence.
You're just going in circles now and that fucking pathetic.

It's really confusing to use incorrect units when referencing between standards such as DDR4 and DDR5. Please use MT/s when referring to the data rate, and MHz when referring to the I/O clock.

Its not confusing at all, unless you're mentally deficient, which you likely are. The world shouldn't pander to retards. Sorry, kid.
No one else has this problem.

Well, you started using MT/s once you realized the difference, so I guess that's a small leap toward progress. 1 person down, hundreds of thousands more to go.

You mean you desperately googled around trying to find the difference as I was spoonfeeding you out of pure generosity. Count yourself lucky, Jamal. Most people wouldn't go to such lengths to educate your kind.

I'd do a shrunk monolithic APU to 7nm everything, despite the small gains. On top models I'd slap in 256MB-1G eDRAM next to the CPU to spite the faggot OEMs who sell single-channel-RAM laptops with no discrete GPU.

Attached: 61646126_p0.png (1462x1934, 1.45M)

Do you genuinely believe yourself? Why do you try to save face on an anonymous forum? Do you have somewhere else to be?

Would it be feasible to have a 4core Zen2, (10-20CU?) Navi and some on chip GDDR, even if it's just 2gb.

It'd overcome the largest flaw of APUs

The performance GPUs CPU side won't be small, they're going to be pretty big. The 2500U has a sustained base clock of just 2ghz, it can actually drop lower than this in 15w cTDP mode.
Base clocks are going to shoot way up inside of the same power envelope.

Literally everyone is aware of how DDR works, kid. No one is confused by it. No one spergs out when mhz and MT/s are used interchangeably. The only reason why you started shitposting about is that you're some sad little newfaggot retard who didn't know any of this.

Sideport memory is a thing of the past. No motherboard manufacturer wants to waste the space and routing to put one little GDDR5 chip on the board. Nobody wants to have that segmentation either.

for you

>The performance of the GPU aside*
The most dramatic improvements on the power curve for perf/watt happen at the low end.

I knew the difference, that's why I asked if you meant MT/s as 6000MHz would be a world record. I even gave you the benefit of the doubt. There's no need to lie about being wrong, it's an anonymous forum and admitting to mistakes is okay.

2400G fag here. The 11CU and 4c8t design is really good for my office PC, It runs Win10 and MS office programs with heavy web browser use very well and can even handle some light games at 60fps 1080p. I have a 4.125GHz OC and 1600MHz GPU Core/memory OC with 4GB reserved and 16GB 3200MHz Ripjaws... Delid with liquid metal running 40c at load... So its a little overkill. I definitely want to upgrade those CUs though

>I'm a newfag
>I'm so retarded
>please rape my dumb face
>thank you for educating me sempai
Thats all you needed to say, kid. You're welcome.

Doesn't this mean that they'll release mobile chips too?
If so, then I'm excited! I want laptop with 10h battery life and modern hardware. I really hope that ryzen mobile apus can deliver this!

The desktop APUs and laptop APUs are one in the same.
The laptop SKUs were detailed and launched a long time ago.

anandtech.com/show/13771/amd-ces-2019-ryzen-mobile-3000-series-launched

>this thread
tomshardware.com/news/ddr4-memory-overclocking-world-record-ram-micron,39358.html
Look at this
"5726mhz" world record
actual memory speed 2863mhz
Everyone states the effective clock as the clock speed