Jump to content
Sign in to follow this  
Dasa

AMD Zen

Recommended Posts

https://www.tomshardware.com/news/amd-fenghuang-ryzen-vega-soc,37551.html

The semi-custom SoC, codenamed Fenghuang, will power Zhongshan Subor's upcoming gaming console and PC. The Fenghuang chip harnesses the power of AMD's Zen processor and Vega graphics processing unit architectures. The SoC features a Ryzen processor with four cores and eight threads running at 3GHz and 24 Vega Compute Units operating at 1.3GHz. An Accelerated Processing Unit (APU) might be the first thing that comes to mind; however, the Fenghuang SoC is quite the opposite. APUs don't come with their own graphics memory, and the Fenghuang chip has 8GB of high-performance GDDR5 memory at its disposal. Jack Huynh also confirmed that the SoC will support the chipmaker's existing and next-generation technologies, such as AMD FreeSync, Radeon Software Adrenalin Edition software and Rapid Packed Math.

Share this post


Link to post
Share on other sites
On 8/13/2018 at 10:11 AM, Jeruselem said:

Intel have still have the OEM sales, AMD need to start putting Zen into OEM machines as you can't just rely on gamers and uber-system builders. The big money is in the servers though.

AMD are trying. They do have some product wins. Intel seems to be losing focus on OEM sales though in favour of emerging markets I think. Well... that or they are being unusually benevolent. 

Share this post


Link to post
Share on other sites
10 hours ago, SceptreCore said:

AMD are trying. They do have some product wins. Intel seems to be losing focus on OEM sales though in favour of emerging markets I think. Well... that or they are being unusually benevolent. 

I think for low end office machines and home users who don't know any better, you don't have to do much as long as it's profitable to keep sending out those machines. Changing an entire OEM line from Intel to AMD would require a bit of planning.

  • Like 1

Share this post


Link to post
Share on other sites

AMD hands TSMC its entire 7nm portfolio: Vega 20, Zen 2 and Navi right around the corner

Why it matters: This is the last piece of the puzzle in AMD's long journey to put its manufacturing strategy in order. Unfortunately this meant dropping Globalfoundries, and taking on TSMC as the go-to partner for its current portfolio. TSMC's manufacturing lead over Intel is substantial and AMD aims to capitalize on that.

Mark Papermaster, CTO of AMD, has announced that the company has finally leveled out its GPU and CPU development and unified everything under a single manufacturing node, at 7nm, handing TSMC the keys to the kingdom.

In a blog post, the exec explains how the company's long term vision has finally paid off and AMD is now in a position to put its competition between a rock and a hard place with advanced 7nm designs in both GPU and CPU segments, thanks exclusively to the help of TSMC. This announcement sets the stage for the departure of GlobalFoundries, AMD's other, oldest, foundry partner.

The announcement had already been hinted at during several corporate communications, events and calls, however, it is the first time AMD comes forward and effectively drives this narrative. Papermaster said "Our work with TSMC on their 7nm node has gone very well and we have seen excellent results from early silicon."

Papermaster goes on to report that multiple 7nm products have taped out at TSMC. This includes the workstation-class GPU, Vega 20, and the next generation CPU architecture, Zen 2, all manufactured by TSMC. Going forward, AMD will focus its entire product portfolio on TSMC's 7nm process, and that includes this year's Vega 20, followed by an early 2019 launch of the Zen 2-based server CPU, Epyc, and later that year, the Navi GPU architecture, the long-awaited 7nm follow-up to Polaris which AMD is now beginning to talk up.

For more time than most AMD fans would care to remember, the company had relied on GlobalFoundries to cater to its semiconductor manufacturing needs. The results, however, have been a mixed bag. GlobalFoundries has tripped itself up more than once while playing catch-up with Intel, having twice nixxed its developing processes. It trashed its 14nm development in favor of licensing it off Samsung, and now it has announced it has ended its 7nm node. In the meantime, TSMC, the backup fab for AMD, not only delivered where GlobalFoundries failed, but managed to overtake Intel in manufacturing technology, and this has paid off.

While recognizing GlobalFoundries' role in producing 14nm and 12nm GPUs and CPUs, in particular for Zen and Zen+, which proved great successes in restoring AMD's reputation, it seems that the relationship between AMD and its oldest foundry partner will not go beyondthose nodes. With TSMC, AMD now has the weapons to fend off its biggest competitor, Intel.

 

Good news for AMD and consumers

Share this post


Link to post
Share on other sites

My only problem with the Zen architecture is the memory problem.  It is one of the reasons why I have always chosen Intel for all my CPU choices, up till now.  The fact that AMD has released the 2990WX is of great interest to me, due to its core count and price.  However, as always with AMD, there is a caveat.  In this case, memory latency, which is the biggest one, and I am not mentioning the other memory elephant in the room either.

I look at the design of the 2990WX and I marvel that it can even exist at all.  Then you look at how its designed and one wonders how designers thought it was a good idea to design it the way it is.  As I look over the horizon, there is no word whether there is an incoming fix for this in the future, which is a part that most workstation designers would kill to work with.  I guess, its just business as usual with AMD.

Share this post


Link to post
Share on other sites
On 9/9/2018 at 5:24 PM, strifus said:

My only problem with the Zen architecture is the memory problem.  It is one of the reasons why I have always chosen Intel for all my CPU choices, up till now.  The fact that AMD has released the 2990WX is of great interest to me, due to its core count and price.  However, as always with AMD, there is a caveat.  In this case, memory latency, which is the biggest one, and I am not mentioning the other memory elephant in the room either.

I look at the design of the 2990WX and I marvel that it can even exist at all.  Then you look at how its designed and one wonders how designers thought it was a good idea to design it the way it is.  As I look over the horizon, there is no word whether there is an incoming fix for this in the future, which is a part that most workstation designers would kill to work with.  I guess, its just business as usual with AMD.

Zen 2 is going to have a drastic change in the memory department.

Stock.PNG.6db6ce3265f0ce5c30edf46fdae0886b.PNG

Share this post


Link to post
Share on other sites

Oops phone edit results in repost 

Edited by Dasa

Share this post


Link to post
Share on other sites

What software are you looking at using with 2990wx that is latency sensitive?

 

Latency on Intel quad channel systems ìsnt great ether 

8706_23_amd-ryzen-threadripper-2990wx-29

Share this post


Link to post
Share on other sites
6 hours ago, Dasa said:

picture

You also have to take into account the architecture differences they've chosen. Micro arch's are a collection of tradeoffs. A sacrifice here for a plus there.

Edited by SceptreCore

Share this post


Link to post
Share on other sites

Right!  Crazy how small the chip was... Some found this in between the layers of the pcb.... We were all discussing it at work. 

Share this post


Link to post
Share on other sites
On 10/6/2018 at 12:33 AM, gamble said:

Right!  Crazy how small the chip was... Some found this in between the layers of the pcb.... We were all discussing it at work. 

Definitely not an "accident" ...

Share this post


Link to post
Share on other sites

pNLra0t.jpg

 

Hopefully that's not just under specific workloads due to something like the addition of avx512 that only some programs can make use of.

If its General IPC improvement and if they can squeeze some higher clocks out of 7nm things are looking good.

Edited by Dasa
  • Like 1

Share this post


Link to post
Share on other sites

If would be amazing if gaming was +5 on ipc and with like you said some more clocks.   Seems like clocks is the biggest thing from gain more ground in gaming.  AMD is pretty good about price for the amount of cores you get.  Hoping there is some OC fun to be had as xfr really does good job but would like more headroom. 

Share this post


Link to post
Share on other sites

AMD Unveils ‘Chiplet’ Design Approach: 7nm Zen 2 Cores Meet 14 nm I/O Die

amd-chiplet-678_678x452.png

AMD on Tuesday disclosed some additional details about its upcoming codenamed Rome processor based on its Zen 2 microarchitecture. As it turns out, the company will use a new design approach with its next-generation EPYC CPUs that will involve CPU ‘chiplets’ made using TSMC’s 7 nm manufacturing technology as well as an I/O die made using a 14 nm fabrication process.

AMD’s chiplet design approach is an evolution of the company’s modular design it introduced with the original EPYC processors featuring its Zen microarchitecture. While the currently available processors use up to four Zen CPU modules, the upcoming EPYC chips will include multiple Zen 2 CPU modules (which AMD now calls ‘chiplets’) as well as an I/O die made using a mature 14 nm process technology. The I/O die will feature Infinity Fabrics to connect chiplets as well as eight DDR DRAM interfaces. Since the memory controller will now be located inside the I/O die, all CPU chiplets will have a more equal memory access latency than today’s CPU modules. Meanwhile, AMD does not list PCIe inside the I/O die, so each CPU chiplet will have its own PCIe lanes.

Separating CPU chiplets from the I/O die has its advantages because it enables AMD to make the CPU chiplets smaller as physical interfaces (such as DRAM and Infinity Fabric) do not scale that well with shrinks of process technology. Therefore, instead of making CPU chiplets bigger and more expensive to manufacture, AMD decided to incorporate DRAM and some other I/O into a separate chip. Besides lower costs, the added benefit that AMD is going to enjoy with its 7 nm chiplets is ability to easier bin new chips for needed clocks and power, which is something that is hard to estimate in case of servers.

IMG_20181105_093504_575px.jpg

AMD Previews EPYC ‘Rome’ Processor: Up to 64 Zen 2 Cores

amd_zen4_678_678x452.png

At its Next Horizon event in San Francisco, AMD announced the fourth iteration of its Zen microarchitecture. The Zen 4 is currently in development, so the company does not share many details about it right now.

Right now, the company is sampling its codenamed Rome CPUs based on Zen 2 microarchitecture and made using TSMC’s first-generation 7 nm manufacturing technology (N7). After that, AMD plans to release a processor based on its Zen 3 architecture and these chips will be made using TSMC’s N7+ fabrication process that will take advantage of EUV lithography. Since Zen 4 microarchitecture is still in design phase, chances are that processors on its base will be made using a more advanced node, so think 5 nm, but keep in mind that any guesses today are speculations at best.

At its event AMD implied that the first CPUs based on its Zen 3 microarchitecture will ship in 2020, so it is natural to expect Zen 4 to reach actual products in 2021 or later. As for what to expect from the new microarchitecture, the company naturally promised higher performance and performance per watt when compared to prior generations.

Previously AMD has only discussed Zen 2 and Zen 3 microarchitectures, yet it is not surprising that the company will keep evolving its successful design in the coming years.

Edited by SceptreCore

Share this post


Link to post
Share on other sites

AMD 'Zen 2' IPC has 29 percent higher IPC

The news coming from Expreview mentions that AMD performed DKERN + RSA tests for integer and floating point units and that resulted into a score of 4.53, that was 3.5 for the first-generation Zen and that indeed is a 29.4 percent IPC. Of course, these are number from the manufacturer themselves and have to be observed with caution as other benchmarks have not been mentioned. So I have to add though, 29% "in certain workloads". Still, impressive.

Probably their best foot forward here... s what this might mean in games will hopefully be anywhere from 10-15%

  • Thanks 1

Share this post


Link to post
Share on other sites

Which means the improvement is evident in particular cases which might not apply to normal usage and not overall evident ...

Edited by Jeruselem

Share this post


Link to post
Share on other sites
12 hours ago, Jeruselem said:

Which means the improvement is evident in particular cases which might not apply to normal usage and not overall evident ...

Well in that particular workload it would have done soe FPU work. Zen 2 has doubled the FPU and fattened the pipelines and the front end to feed it... so you need to really take a guess on how much of that you can take out. Then how much of it even relates to a typical singlethreaded workload. It's not even an educated guess.

Share this post


Link to post
Share on other sites

 

 

Check out around 13:45 mark...  Man I cannot wait, Ill probably target the 7 series but that 9 series is crazy at those speeds/cores... Even the price is hard to believe if true... what will Intel do?  The TDP is very interesting too!

Edited by gamble
  • Like 1

Share this post


Link to post
Share on other sites

Awww gamble buddy... you beat me, I'm just watching this now and was about to post it. 

This guy has gotten pretty damn close to accurate before and I hope he is true now. I wasn't too sure about the rumours of AMD going 16 core for mainstream, as I think it might be pushing too hard to play against Intel... but hell... if I can have a mid-range 8 core CPU that can turbo up to nearly 5.0GHz and cost the same as an R5 2600X... then I don't see why the heck not! 😄 

We will see if this is accurate at CES in January. not long now...

Share this post


Link to post
Share on other sites

Is that not crazy saying 8 core mid range, low range as 6 core.... heck I might go that route too.  The clocks, price and tdp are amazing all around.  Can 7nm really be that good?  With that tdp it should stay relatively cool.  Boost clocks Im assuming will be only a few cores as it would be hard to believe that would be boost for all core. 

Curious as to what the new chipset will bring possibly be named x570.... I hope they have some real phases now too on these MB and not the fake half phases MB vendors been called out on recently.  

I know his sources have been pretty much spot on, but it is still hard to believe...  

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this  

×