Ryzen 3000 Chiplet Socket

[ch3w2oy]

7 minutes ago, BobbyPdue said:

I didn't do any typing I just quoted from your post. I try to avoid quoting an entire post of quotes to avoid these types of mistakes.  

 

I don't know why you would think I'm biased toward one CPU manufacturer since the entire point of my post was "don't get excited until there is an official release and third party benchmarks" my posts are against blindly believing rumors and speculation. 
 

I would be excited if the CPU could run solitaire. Seriously. What we already have to go on, is good enough to be happy about. AMD is worth being happy about. When was the last time Intel made you feel all fuzzy inside?

 

You're coming off as biased because you're making it a big deal that people are getting excited that AMD may deliver a worthy product. Who the hell cares if it's not faster than the 9900k. IT IS worth getting EXCITED over. Period.

 

It's really that simple. You're the only person overthinking this.

[pas008]

31 minutes ago, ChewToy! said:

I would be excited if the CPU could run solitaire. Seriously. What we already have to go on, is good enough to be happy about. AMD is worth being happy about. When was the last time Intel made you feel all fuzzy inside?

 

You're coming off as biased because you're making it a big deal that people are getting excited that AMD may deliver a worthy product. Who the hell cares if it's not faster than the 9900k. IT IS worth getting EXCITED over. Period.

 

It's really that simple. You're the only person overthinking this.

not even a big deal unless you think about intel only needed 6 cores to match amd's 8

 

this shows its even playing ground considering no avx talk but thats another discussion on another playground

 

8core cb15 mt beating  8 core intel that is big deal

should be to us all

might been a demo but that chip is late 2nd qtr

intel only has kf products in their lineup at that time

which renames a king after 12 ish yrs

that is a big deal

 

 

edit

think phenom took over single core slightly for abit and we got sandy bridge after nephaelm

 

but amd wasnt smart binning either, which now constitutes huge difference

why my 550 dual core became quads then in bios unlocker, amd now is playing the binning game

claps claps claps

they will  sooon do the same on gpus and funny thing is it will be multi chip too but  no crossfire problems

 

[Stefan Payne]

8 hours ago, BobbyPdue said:

Don't be so snarky.  Did she say "This is our mid range CPU" If she did please quote it.

RLY?!
Look at the Video

Look at Anandtech.

Look at the Leaks.

 

What she said is that there will be more than 8 Cores...

 

Why do you still try to claim that we're talking about a high end AMD Chip when its pretty clear we are not...

 

8 hours ago, BobbyPdue said:

Samsung and Apple both have 7nm consumer products. The Crimson Canyon NUCs models NUC8i3CYSN and NUC8i3CYSM were introduced to the market back at the beginning of December they have 10nm processors.

Oh I forgot.

The holy Intel everyone has to worship did some low volume, low performance, low volume Chip that really nobody cares about...

 

Yeah, totally comparable with a high performance Desktop Chip that's available to the masses and the flag ship product...

Samsung and Apple both have lower power Designs that are also not comparable to the Consumer Desktop stuff...

 

But here's the question:
How can AMD do something that Intel failed at?!

Meaning sub 10nm Products for the Desktop/Consumer Market...

 

8 hours ago, BobbyPdue said:

What does that do for consumers?

Higher bandwith for graphics card and SSD, need for less Lanes for the same performance.

With PCIe 4.0 a 1 Lane USB 3.1 Gen2 Card with a couple of ports is possible without problems or limitations.

Right now you need a couple of that as the Speed of USB 3.1 gen 2 is higher than a single PCIe 3.0 Lane...

 

8 hours ago, BobbyPdue said:

A poster on The Rzyen Reddit page scored 1964 on a Ryzen 7 2700X with air cooling. 

With 130W System Power Consumption?
The i9-9900K consumed around 180W. 

AMD did show the Power Consumption on screen.

 

Aren't you Intel guys the ones claiming that Power Consumption is important??

Now the Power Consumption shown is pretty impressive, don't you agree??

And also it allows AMD to add another 8 Core DIE, don't you agree??

8 hours ago, BobbyPdue said:

What performs amazing?  What is it?  Is 90 more points in Cinebench amazing?

More Points in Cinebench with less Power Consumption, while also crippling the Memory.

You Know, we're talking abotu DDR-4 2666 here...

No Cinebench here but look at that:

https://www.legitreviews.com/ddr4-memory-scaling-performance-with-ryzen-7-2700x-on-the-amd-x470-platform_205154

 

7% Gain from DDR4-2666 to DDR4-3200 in Far Cry 5...

 

So the Conclusin is that AMD didn't show much and only showed the bare minimum!

With Optimization the CPU could have done better! 

But that's not what they wanted to show...

 

And again, there will be 12 and 16 Core Versions available...

That's secured.

AMD Just didn't talk (much) about it on CES.
But they did mention it!

 

Lisa said:

Quote

„There is some extra room on that package and I think you might expect we will have more than eight cores“

Yeah, that is absolutely the TOP Chip in that lineup - N O T.

Obviously...

 

8 hours ago, BobbyPdue said:

Most if not all desktop CPU's consume a less power than a 9900k it's not an optimized CPU it was a quick answer to AMD. 

Why you defend Intel here? Ain't it about AMD?
And that's a rather shitty claim as well because its up to Intel to optimize that.

Ryzen was introduced in February 2017 the i9-9900K was introduced in October 2018.

That gave Intel more than 1,5 years to make this CPU.

 

So you're saying that Intel is not able to optimize a CPU with 1,5 Years development time as an Answer to AMD??
RLY?!
 

That would mean that, according to you, Intel is incompetent and can't compete with AMD with the Designs they have right now...

Even with 1,5 Years time to develop a competiting Product to the Ryzen Architecture...

 

Now AMD comes with a new Ryzen core that is completely redesigned - when did Intel do THAT?!

Its not "just" an incremental improvement like Sandy -> Ivy -> Haswell -> Broadwell -> Skylake like Intel, it is a complete redesign of the CPU Core that AMD brings. 

 

Or to cite Anandtech again: 

We know it has an improved branch predictor unit, and improved prefetcher, better micro-op cache management, a larger micro-op cache, increased dispatch bandwidth, increased retire bandwidth, native support for 256-bit floating point math, double size FMA units, and double size load-store units. 

 

 

Thus the frequency is not known. And AMD knows why!

They might have run the CPU at 5GHz.

They might have run the CPU at 4,3-4,5GHz. Wich means that it is possible that the IPC is better than Intel. And they didn't want to tell anybody right now...

 

8 hours ago, BobbyPdue said:

An unamed, unspecificed CPU runs Cinebench 90 points better than a 4.3Ghz 2700X, there's no pricing, there's no specs, only speculation.  What uses PCIe 4.0? As far as I've read no GPU is maxing out PCIe 3.0 at this time.

Oh now you're downplaying the advantages because "your guy" doesn't have it. Figures...

Its not that it is 90 points better than a 4,3GHz 2700X it is that it is better than a 4,7GHz Intel Chip.

 

And to make it worse: The Frequency is not known. And its still a couple of Months to release. And it works awesomely well.

That means that they can still improve things. They can still do another revision or two (wich takes ~3 Months) to the Silicone. Wich means additional Bug Fixing and maybe even Performance. Or Power Consumption as well...

 

Its the 90 Points at (almost) half the power consumption!

THAT is the amazing part.

And it might even be possible that the CPU ran at the same 4,3GHz!!

The 2700x scores 1754 in Anandtec's Reviews.

The new Ryzen 3000 scores 2023. That is a 15% Improvement.

So it is entirely possible that that is entirely due to Architectural improvements.

 

 

So why didn't AMD tell us the Clockrates? You ever thought about that?
Maybe there is a reason. And that is that the Improvements are performing better than expected. And they don't want others to knew (yet)...

8 hours ago, BobbyPdue said:

What does that do for the consumer? 

Smaller form factor boards possible without problems. Though they haven't been seen.

And if the Chipset dies the Rest might actually still work. 

 

8 hours ago, BobbyPdue said:

What am I supposed to get out of the article you posted? 

Information...

But you don't seem to want to see positive Information about AMD anyway...

It shows how much an edge AMD has right now...

Because AMD has the Chipset already integrated onto the CPU!!

The whole thing...

8 hours ago, BobbyPdue said:

Here is what I got from it:

 

"This press release contains forward-looking statements concerning Advanced Micro Devices, Inc. (AMD) including the features, functionality, availability, timing, deployment, benefits and expectations of AMD future products and technologies and growth in the industry, which are made pursuant to the Safe Harbor provisions of the Private Securities Litigation Reform Act of 1995. Forward-looking statements are commonly identified by words such as "would," "intends," "believes," "expects," "may," "will," "should," "seeks," "intends," "plans," "pro forma," "estimates," "anticipates," or the negative of these words and phrases, other variations of these words and phrases or comparable terminology. Investors are cautioned that the forward-looking statements in this document are based on current beliefs, assumptions and expectations, speak only as of the date of this document and involve risks and uncertainties that could cause actual results to differ materially from current expectations. Such statements are subject to certain known and unknown risks and uncertainties, many of which are difficult to predict and generally beyond AMD's control, that could cause actual results and other future events to differ materially from those expressed in, or implied or projected by, the forward-looking information and statements."

Yeah, because can't be happening that AMD has an edge and the more advanced product, now can it?

Didn't happen in the past...

 

8 hours ago, BobbyPdue said:

Harley Davidson released a new electric motorcycle at CES it supposedly does 0 - 60 in under 3.5 seconds is that worth being excited about? 

If you want to believe in Intel, probably not.

If you don't care about Manufacturers and want the best Product, hell yeah...

[Stefan Payne]

8 hours ago, ChewToy! said:

What we already have to go on, is good enough to be happy about. AMD is worth being happy about.

Exactly!

It was a short Presentation.

But we can interpolate some other Information with the Rome presentations. You know the "one Rome beats TWO of the best Xeons!"

8 hours ago, ChewToy! said:

When was the last time Intel made you feel all fuzzy inside?

Hm, never? OK, maybe Nehalem. That was over 10 years ago.
With the "another 2-5% incremental Improvement" stuff, naa....

 

 

Oh, Itanic was awesome as well...

[nhill638]

So, how long are we waiting for these to come out again? And if it's Q3, when's that?

[Stefan Payne]

1 hour ago, nhill638 said:

So, how long are we waiting for these to come out again? And if it's Q3, when's that?

"middle of 2019". 

So anywhere between may (optimistic) and August (pessimistic)....

[PianoPlayer88Key]

So I saw the AMD Matisse die shot, and got to thinking...

 

 

Ya know, if they moved that I/O die up a bit, maybe 3 chiplets could fit on there...  

Also I'd love it if they leapfrog as far ahead of Intel on IPC (single-core performance at a fixed clock) by at least as much as how far Bulldozer/Excavator was behind Skylake/Kaby Lake before Ryzen 1st-gen launched.

 

Yeah, I know, I'm dreaming.  

 

I'm guessing that what they showed off (in Cinebench) was actually a Ryzen 5 part, probably running at about 4.5 or 4.7 GHz or something like that.

[Stefan Payne]

Lets restart the Hype-Train...

According to unofficial Reports, the AMD Chip is said to be have worked at around 4,5GHz or so.

The Intel Chip is speculated to be have run at 4,7GHz...

 

So meaning AMD beat Intel with 200MHz _LESS_.

Ähm, Huston, we have a Problem...

At 11:07 its shown:

 

[WallacEngineering]

In case nobody else knows of the information we are referring to, we are of course talking in regards to AMD's official announcement of Ryzen 3000 at CES:

 

https://youtu.be/Jc7lqFaTvPo

 

Now I fully believe this benchmark could in fact be a midrange Ryzen 3000 that is going up against one of Intels strongest desktop chips. But how is this possible? 7nm vs 14nm, its that simple.

 

AMD CEO Lisa Su clearly states in the video that the new Ryzen 3000 chips are 7nm-based. This means twice the transistor capacity per square mm than the Intel counterpart, and we all know that more transistors (of course amongst other things) leads to more performance.

 

I also fully believe the released market chips will be just as fast, if not faster. Remember the original Ryzen 1000 launch? The demo clock speeds were actually slightly lower than the final product, simply because AMD had the time to finalize the design before the official launch.

 

Whats even more interesting, however, is that this launch marks the beginning of the end. Computer Scientists are still struggling to find a physically possible way to create 5nm-based CPUs that actually functions and can dissapate heat ina realistic fashion, even right now. It has been previously stated by said scientists that after 7nm, we will be forced to fundamentally rethink the CPU and technologies behind computers all together before CPUs can get any faster. Some reports say that other than classified military technologies, we will be stuck in stagnation where we are for the next 5-10 years while a solution is found. Some people even say 20+ years or until software catches up and most everyday programs and games can actually take full advantage of a 16+ thread CPUs performance, because as most of you are probably aware, most people dont even use half of a modern high-performance CPU's potential, ever in their lives.

 

You might find that some applications load all threads but not to 100% while others will load threads to 100% but not all of them at once. This is of course excluding synthetic benchmarks.

 

The most logical next step is quantum computing but if you have seen Linus's video where he visits a quantum computing facility, Star Trek is more likely to become a reality than getting this kind of technology small and affordable enough to get into our homes and everyday lives.

 

This 7nm release is actually quite alot earlier than I first predicted in my thread that discussed the "7nm barrier". I always imagined it would not happen until late 2020 - early 2021. Thread here:

 

[Stefan Payne]

1 minute ago, PianoPlayer88Key said:

 by at least as much as how far Bulldozer/Excavator was behind Skylake/Kaby Lake before Ryzen 1st-gen launched.

Seen what AMD got out of the Bulldozer ARchitecture??

Well, except for the fact that they fucked up the I/O (=L3) Part of the CPU...

 

ATTENTION; GERMAN INCOMING:

https://www.planet3dnow.de/cms/18564-amd-piledriver-vs-steamroller-vs-excavator-leistungsvergleich-der-architekturen/

 

I haven't seen anything like that anywhere else...

[Vorg]

Spoiler

 

What about the G versions regarding extra spot? The gpu would need some space, so maybe that 3rd chip spot is for the gpu on G versions. 

[WallacEngineering]

6 minutes ago, Vorg said:

  Reveal hidden contents

 

What about the G versions regarding extra spot? The gpu would need some space, so maybe that 3rd chip spot is for the gpu on G versions. 

Would make total sense, safe assumption

 

Regardless if you ever wanted to build a craptastically overpowered AMD rig that will last for many years to come, it looks like the time is soon to be upon us

 

I may still wait until 2020 and AMD releases a matured 7nm Chip where all the kinks and bugs are worked out and efficency is top-notch.

 

In other words: 2020 should be when Zen 2+ refined and perfected 7nm tech comes out.

[Stefan Payne]

3 minutes ago, Vorg said:

  Reveal hidden contents

 

What about the G versions regarding extra spot? The gpu would need some space, so maybe that 3rd chip spot is for the gpu on G versions. 

You can design the GPU Chip for that spot so that the pinout is compatible with the spot...

What does it need?
a) Power

b) an interface...

Since there is no space for memory anyway, you can use the same pinout...

 

Ergo: It might be used for both...

The 2nd CPU is more likely though as there is a need for higher core counts.

[WallacEngineering]

8 minutes ago, Stefan Payne said:

You can design the GPU Chip for that spot so that the pinout is compatible with the spot...

What does it need?
a) Power

b) an interface...

Since there is no space for memory anyway, you can use the same pinout...

 

Ergo: It might be used for both...

The 2nd CPU is more likely though as there is a need for higher core counts.

Im hoping for a solid 10-Core/20-Thread option from Ryzen 7 this time around. Im thinking that with the 7nm transistors it shouldnt be too difficult to offer one.

[Stefan Payne]

9 minutes ago, WallacEngineering said:

Im hoping for a solid 10-Core/20-Thread option from Ryzen 7 this time around. Im thinking that with the 7nm transistors it shouldnt be too difficult to offer one.

Impossible.

You can't divide 10 by 4 and its not possible with one Die either.

 

So you have to choose between either an 8 core or a 12 core. Due to the way its made, nothing in between is possible...

 

So how would you do a 10 core anyway??

2+2+3+3??

 

As we know both CCX of a Ryzen processor have to have an equal amount of active Cores, if the other is active (ie 3+0 is possible but 3+2 is not)...

 

So its safe to assume that all CCX have to have the same amount of cores active. 
With 2 Dies we are talking about 4 CCX, so it has to be a product of 4...

[WallacEngineering]

56 minutes ago, Stefan Payne said:

Impossible.

You can't divide 10 by 4 and its not possible with one Die either.

 

So you have to choose between either an 8 core or a 12 core. Due to the way its made, nothing in between is possible...

 

So how would you do a 10 core anyway??

2+2+3+3??

 

As we know both CCX of a Ryzen processor have to have an equal amount of active Cores, if the other is active (ie 3+0 is possible but 3+2 is not)...

 

So its safe to assume that all CCX have to have the same amount of cores active. 
With 2 Dies we are talking about 4 CCX, so it has to be a product of 4...

Well thats a pain. I was unaware if this limitation. 10 cores would be perfect!

[Stefan Payne]

7 hours ago, WallacEngineering said:

Well thats a pain. I was unaware if this limitation. 10 cores would be perfect!

12 Cores are better though

[AndrewCC]

Couldn't you take 2 really bad 8 core chiplets with 3 defective cores each, stick them togheter with some Infinity Fabric, and call the result a 10 core? 

I don't know how good AMD's 7nm yield is. It might be that chiplets with that many defective cores are very rare. They might just chuck them outright because it's not economically viable.

[Stefan Payne]

4 hours ago, AndrewCC said:

Couldn't you take 2 really bad 8 core chiplets with 3 defective cores each, stick them togheter with some Infinity Fabric, and call the result a 10 core? 

I don't know how good AMD's 7nm yield is. It might be that chiplets with that many defective cores are very rare. They might just chuck them outright because it's not economically viable.

No, as said, the Size of the CCX has to be the same.

And it looks like each CCX on a Chip also has to be the same.

[WallacEngineering]

12 hours ago, Stefan Payne said:

12 Cores are better though

True that but I am fairly certain AMD would not consider a 12-Core CPU for the Ryzen 7 batch. Pretty sure 12+ Core will remain as Threadripper SKUs for some time to come.

[WallacEngineering]

11 hours ago, AndrewCC said:

Couldn't you take 2 really bad 8 core chiplets with 3 defective cores each, stick them togheter with some Infinity Fabric, and call the result a 10 core? 

I don't know how good AMD's 7nm yield is. It might be that chiplets with that many defective cores are very rare. They might just chuck them outright because it's not economically viable.

Agreed, technically either AMD or Intel can make as many cores as they want, but it would pretty much make them not very viable on the market. Maybe Ill just build a 12-Core threadripper machine in 2020 and just bite the money bullet.

[Stefan Payne]

1 hour ago, WallacEngineering said:

True that but I am fairly certain AMD would not consider a 12-Core CPU for the Ryzen 7 batch. Pretty sure 12+ Core will remain as Threadripper SKUs for some time to come.

You're wrong.

Remember that we had 10 years standstill and 4 Cores in Desktop.

Or rather 11 -> Q6600 was released back in 06, Ryzen came in 2017 - the first "true" 8 Core Desktop CPU. Hell even mobile already had 8 Core CPUs...

And 4 Cores with SMT was introduced back in 2008 as well (Nehalem, LGA1366)...

[WallacEngineering]

15 hours ago, Stefan Payne said:

You're wrong.

Remember that we had 10 years standstill and 4 Cores in Desktop.

Or rather 11 -> Q6600 was released back in 06, Ryzen came in 2017 - the first "true" 8 Core Desktop CPU. Hell even mobile already had 8 Core CPUs...

And 4 Cores with SMT was introduced back in 2008 as well (Nehalem, LGA1366)...

I wasnt referring to them being able to make an unlimited about of cores but rather that they could come up with any sort of weird number of cores such as even a 9-core or an 11-core. Its just that it wouldnt make any sense to do so as you would need to purposely combine a bunch of defective cores with an eneven amount of good cores on a die joined by infinity fabric, which would just cost way more than the profit would yield compared to products that are already on the market.

 

To be honest though, massive core count isnt too far out of reach these days. Threadripper already has 32 cores, it wouldnt be too hard for them to build a 64-Core chip on 7nm. Is there a reason to even build such an expensive chip? Probably not, there isnt really going to be a need for such a CPU, but Im sure its possible. 128 Cores on a single CPU seems to be out of reach for awhile due to the 7nm barrier, but perhaps 10+ years in the future it will become a reality.

[WallacEngineering]

Im now wondering if second gen sTR4 Socket Motherboards will finally arrive. I always found it wierd that second gen Threadripper was never followed by new motherboards.

[Stefan Payne]

14 hours ago, WallacEngineering said:

Im now wondering if second gen sTR4 Socket Motherboards will finally arrive. I always found it wierd that second gen Threadripper was never followed by new motherboards.

Because the Market is too small as too many people still buy the totally overpriced Intel crap with Toothpaste between the Lid and the Die.

But there are a couple "2nd Gen Threadripper" Boards, they aren't just as mentioned as the other side...

 

And there's Ryzen 3000 series wich requires new Boards anyway because PCIe 4.0...