AMD teases third-generation Ryzen CPU during CES keynote

[Shawn Knight]

In brief: During its CES keynote AMD confirmed plans to launch third-generation Ryzen desktop processors by mid-2019. The upcoming 7nm chip will be backward compatible with current AM4 boards, so you won’t need to run out and grab a new board to upgrade.

AMD previewed its third-gen Ryzen platform, based on its Zen 2 core architecture and built on a 7nm process, in a live demo running Forza Horizon 4 at 1080p with max graphics settings. During the brief showcase, the system never dipped below 100 FPS.

In another demonstration (watch below), AMD pitted its 8-core / 16-thread Ryzen chip against Intel’s Core i9-9900K in Cinebench R15. The AMD chip claimed the victory in addition to demonstrating more efficient power consumption. Notably, frequencies haven’t been finalized and we don’t know what the chip was operating at during this test.

CEO Lisa Su also shared a first look at what the third-gen Ryzen CPU looks like under the hood. Notably, it utilizes a chiplet design with two distinctive dies – the smaller one for the 7nm processing cores and the other for I/O hardware.

AnandTech further points out that there could be some additional room for another CPU or GPU chiplet on the package, giving AMD some room to grow down the road.

Permalink to story.

https://www.techspot.com/news/78199-amd-teases-third-generation-ryzen-cpu-during-ces.html

 

[torchyear]

"The upcoming 7nm chip will be backward compatible with current AM4 boards"

Yay

 

[rrwards]

"The upcoming 7nm chip will be backward compatible with current AM4 boards"

Yay

One of the best reasons to go AMD when building a budget desktop. In a year or so when I have the cash, just plop down a new CPU. With Intel I'd need to get a new motherboard too.

 

[Evernessince]

So the package has room for another chip, meaning 16 core chips could be possible on the consumer platform.

 

[wiyosaya]

The chip also supports pci-e 4.0, too.

 

[toooooot]

"Compatible with AM4"
I bless the rains that fall on AMD's sanctuaries.

 

[IAMTHESTIG]

One of the best reasons to go AMD when building a budget desktop. In a year or so when I have the cash, just plop down a new CPU. With Intel I'd need to get a new motherboard too.

And half the time new RAM, too.

 

[Vulcanproject]

The great mystery remains. The clocks.

AMD demonstrated that 8 Zen 2 cores are at least 10 percent faster than a 2700x in that same test. That would indeed be enough to match a 9900k in almost all areas.

How much of that gain is IPC, and how much comes from improved clock speeds. I hope that the hint these were not finalised clocks was intended as a positive teaser, that this chip is possibly still operating below it's maximum frequency potential.

 

[Danny101]

"AnandTech further points out that there could be some additional room for another CPU or GPU chiplet on the package, giving AMD some room to grow down the road."

Wouldn't it be cool if all you had to do to upgrade the GPU is to drop it into a chiplet slot?

 

[EEatGDL]

Wouldn't it be cool if all you had to do to upgrade the GPU is to drop it into a chiplet slot?

It sounds very far-fetched right now, but we can only dream.

 

[toooooot]

Is she..Is she wearing a sort of leather jacket with short sleeves? Not cool stealing Huang's outfit ideas.

 

[texasrattler]

AMD making claims that their new 3rd gen Ryzen beat a 9900K but offer no proof, just their word LOL. We've all hurt that before from companies. If these claims are to be taken for real, they need to show off the specs the system is running at. Otherwise it's all meaningless and even more so if the actual product doesn't perform the same.
Im not saying AMD can't beat Intel, Im just saying they have yet to prove it. So far it's all talk which all companies do.

 

[Evernessince]

The great mystery remains. The clocks.

AMD demonstrated that 8 Zen 2 cores are at least 10 percent faster than a 2700x in that same test. That would indeed be enough to match a 9900k in almost all areas.

How much of that gain is IPC, and how much comes from improved clock speeds. I hope that the hint these were not finalised clocks was intended as a positive teaser, that this chip is possibly still operating below it's maximum frequency potential.

Yes, the clocks are not finalized and she stated it was an early sample. It definitely bodes well for AMD that an early sample can beat Intel's fastest processor.

Is she..Is she wearing a sort of leather jacket with short sleeves? Not cool stealing Huang's outfit ideas.

Technically that's not a jacket and she's related to Huang so yeah, family trait I guess.

AMD making claims that their new 3rd gen Ryzen beat a 9900K but offer no proof, just their word LOL. We've all hurt that before from companies. If these claims are to be taken for real, they need to show off the specs the system is running at. Otherwise it's all meaningless and even more so if the actual product doesn't perform the same.
Im not saying AMD can't beat Intel, Im just saying they have yet to prove it. So far it's all talk which all companies do.

Please watch the presentation. They showed the Ryzen 2 beating the 9900K in Cinebench. Don't expect specs yet, AMD specifically stated that it was an early sample of Ryzen 2 and that specs are not yet finalized.

AMD is not like Nvidia, who are currently being investigated for lying to their investors.

 

[Sausagemeat]

It won’t be cheap if it beats a 9900K.

 

[texasrattler]

Yes, the clocks are not finalized and she stated it was an early sample. It definitely bodes well for AMD that an early sample can beat Intel's fastest processor.



Technically that's not a jacket and she's related to Huang so yeah, family trait I guess.



Please watch the presentation. They showed the Ryzen 2 beating the 9900K in Cinebench. Don't expect specs yet, AMD specifically stated that it was an early sample of Ryzen 2 and that specs are not yet finalized.

AMD is not like Nvidia, who are currently being investigated for lying to their investors.

The only thing that matters is real world results not synthetic benchmarks that mean absolutely nothing.
For all you know they did stuff to the proc that wont be in the final product. Any company can say they beat another which both AMD and Intel even Nvidia have all made false claims before.

Also if you think these investigations mean anything, they dont. All of these companies have been investigated for one thing or another. Its very common in their business. It wont matter in the end. Nvidia will still do business n make a ton of money, afterall everyone wants what they offer.

 

[Evernessince]

The only thing that matters is real world results not synthetic benchmarks that mean absolutely nothing.
For all you know they did stuff to the proc that wont be in the final product. Any company can say they beat another which both AMD and Intel even Nvidia have all made false claims before.

Also if you think these investigations mean anything, they dont. All of these companies have been investigated for one thing or another. Its very common in their business. It wont matter in the end. Nvidia will still do business n make a ton of money, afterall everyone wants what they offer.

No, it's not very common in the business just like it's not common for your company to be sued by investors for making misleading statements. Don't try to play it off as commonplace. It isn't.

What could AMD possibly do to the processor to make an engineering sample faster then the final product? It's not even aggressively clocked judging by the power consumption. You are suggesting that AMD cooked a Cinebench run with zero proof. AMD isn't making statements like "10x the ray tracing performance" or made up numbers like "20 gigarays". They showed everyone one of the most common benchmarks out there, nothing misleading to it and I fail to see any correlation to anything Intel or Nvidia have done in 2018. You are going to have to explain how they would even do such a thing first.

 

[meric]

If that "previewed" sample is a R5, this will be a great year for AMD

 

[Maxiking]

The only thing that matters is real world results not synthetic benchmarks that mean absolutely nothing.
For all you know they did stuff to the proc that wont be in the final product. Any company can say they beat another which both AMD and Intel even Nvidia have all made false claims before.

Also if you think these investigations mean anything, they dont. All of these companies have been investigated for one thing or another. Its very common in their business. It wont matter in the end. Nvidia will still do business n make a ton of money, afterall everyone wants what they offer.

No, it's not very common in the business just like it's not common for your company to be sued by investors for making misleading statements. Don't try to play it off as commonplace. It isn't.

What could AMD possibly do to the processor to make an engineering sample faster then the final product? It's not even aggressively clocked judging by the power consumption. You are suggesting that AMD cooked a Cinebench run with zero proof. AMD isn't making statements like "10x the ray tracing performance" or made up numbers like "20 gigarays". They showed everyone one of the most common benchmarks out there, nothing misleading to it and I fail to see any correlation to anything Intel or Nvidia have done in 2018. You are going to have to explain how they would even do such a thing first.

Companies make misleading statements all the time, Intel and their 10nm being on track, AMD and their Vega.

2600 and 2700 are faster in the benchmark by 5% when being clock to clock with 8700k and 9900k. So yeah, it is the same trap as Nvidias 10x the ray tracing performance. Well, actually, Nvidia numbers are actually correct and will be the same in real life usage. While AMD presentation, it is just one poorly coded benchmark which does not represent RL performance of their cpus at all.

 

[Humza]

Is she..Is she wearing a sort of leather jacket with short sleeves? Not cool stealing Huang's outfit ideas.

Huang's jacket looks slightly different because... ray tracing

 

[Puiu]

The only thing that matters is real world results not synthetic benchmarks that mean absolutely nothing.
For all you know they did stuff to the proc that wont be in the final product. Any company can say they beat another which both AMD and Intel even Nvidia have all made false claims before.

Also if you think these investigations mean anything, they dont. All of these companies have been investigated for one thing or another. Its very common in their business. It wont matter in the end. Nvidia will still do business n make a ton of money, afterall everyone wants what they offer.

No, it's not very common in the business just like it's not common for your company to be sued by investors for making misleading statements. Don't try to play it off as commonplace. It isn't.

What could AMD possibly do to the processor to make an engineering sample faster then the final product? It's not even aggressively clocked judging by the power consumption. You are suggesting that AMD cooked a Cinebench run with zero proof. AMD isn't making statements like "10x the ray tracing performance" or made up numbers like "20 gigarays". They showed everyone one of the most common benchmarks out there, nothing misleading to it and I fail to see any correlation to anything Intel or Nvidia have done in 2018. You are going to have to explain how they would even do such a thing first.

Companies make misleading statements all the time, Intel and their 10nm being on track, AMD and their Vega.

2600 and 2700 are faster in the benchmark by 5% when being clock to clock with 8700k and 9900k. So yeah, it is the same trap as Nvidias 10x the ray tracing performance. Well, actually, Nvidia numbers are actually correct and will be the same in real life usage. While AMD presentation, it is just one poorly coded benchmark which does not represent RL performance of their cpus at all.

If Cinebench is so poorly coded then why is it used by everybody that actually does proper benchmarks like Techspot? It is used because it is well coded. It scales with both clocks and core counts in a realistic fashion. If a CPU has better IPC/Single threaded performance then the test will show that. If a CPU has better multithreading then the results will reflect exactly that. It also generally represents what you will see in many CPU rendering workloads.

An example of a poorly coded professional software are the ones from Adobe where their scaling is all over the place.

You entire statement can easily disproved by this: https://www.techspot.com/article/1616-4ghz-ryzen-2nd-gen-vs-core-8th-gen/

It clearly shows that the 8700k has better IPC vs the 2600/2700 at the same clocks in the CB benchmark. You can also see that exact same scaling in the Corona benchmark, which is also a rendering task where the 8700k is slightly faster than the 2600 with both at 4GHz. Blender and V-Ray are the same.

TL;DR the CB benchmark is perfect for those that are looking to use the CPU in Corona, Blender, V-Ray or other rendering software.

 

[senketsu]

I just want to non-stop phone Intel and just say ….. 7nm, that's right AMD's at 7nm

 

[wiyosaya]

I just want to non-stop phone Intel and just say ….. 7nm, that's right AMD's at 7nm

Hmmm, Vicious, are you?

 

[Evernessince]

Companies make misleading statements all the time, Intel and their 10nm being on track, AMD and their Vega.

2600 and 2700 are faster in the benchmark by 5% when being clock to clock with 8700k and 9900k. So yeah, it is the same trap as Nvidias 10x the ray tracing performance. Well, actually, Nvidia numbers are actually correct and will be the same in real life usage. While AMD presentation, it is just one poorly coded benchmark which does not represent RL performance of their cpus at all.

Intel predicted the 10nm timeline. It's a prediction, not a hard performance number. Big difference. What exactly are you referring to with Vega? The "poor volta" sign on the wall in the single ad? Bad marketing yes but once again nowhere near hard performance numbers and it was up to the viewer to assume what it meant. I see a big difference between a small sign in a single ad vs Nvidia making a whole media presentation so lets no forget that not only is the content of the supposed infraction different but the scale is most certainly different. The whole 20xx series was sold heavily based off ray tracing at the presentation. Nvidia never even released consumer volta.

And, no Nvidia's numbers on Ray-Tracing are a completely made up brand new metric as they themselves admitted. After all, there was no measurement of real time ray tracing performance before. The problem wasn't "were they correct" because obviously anything you yourself make up is going to be correct. The problem was "10x the performance" is clearly a misleading statement given that the card launched with zero games able to use ray tracing and that 10x number is actually a metric of ray tracing performance against a card that can't real time ray trace. It'd be like saying AMD cards can use AMD settings 100% more then Nvidia cards. It's completely misleading.

Cinebench is poorly coded? Really? Yeah, must be why many major tech outlets use it for benchmarking. Sorry, but if it's your word vs that of professional review outlets like TechSpot I'm going to say TechSpot know their stuff more then you. It's another ludicrous claim to just say one of the standard benchmarking apps is "poorly coded".

 

[Berticus]

The worst release for a new product could be not having enough stock on launch day.

Mama Su probably had to chose between launching sooner with only a single chiplet CPU, or launch at a later date in the year offering the full range of single and multi-chiplet CPUs. No brainer then that she only showed off the single chiplet CPU at CES. Perhaps disappointing to some hyped up fans but at least there would be enough ZEN2 CPUs available on store shelves.
I would expect those 1-chiplet CPU SKUs to be revealed in May with a tease of what the 2-chiplet CPUs will be capable of at the same time - likely on store shelves before Christmas.

 

[hahahanoobs]

"The upcoming 7nm chip will be backward compatible with current AM4 boards"

Yay

One of the best reasons to go AMD when building a budget desktop. In a year or so when I have the cash, just plop down a new CPU. With Intel I'd need to get a new motherboard too.

lol Why would anyone want to?
Ryzen 2000 series = 3% IPC over previous gen Ryzen and no 2800(X)...