Speculation: Ryzen 4000 series/Zen 3

Page 84 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Shivansps

Diamond Member
Sep 11, 2013
3,851
1,518
136
Someone has any info regardless if Vega on APUs had too many rops? i really wast expecting a ROP cutdown on top of the CU cutdown.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,542
14,496
136
It’s real, it is just not what some people think it is and no one wants it.
Intel does not have a 64 core CPU, Where did you see that they do ? Not even a rumor that I have seen.

And look at the comment "trollllolololol"
 
  • Like
Reactions: Ajay

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,542
14,496
136
Why does this link say 64 core, 64 threads ? Its an Intel web page about that CPU ?

Not to mention, it came out in 2016 and is discontinued now ? and was 1.3 ghz, max turbo 1.5 ghz ? I can see why it failed. And 215 watt for that pathetic performance ?
 
Last edited:

itsmydamnation

Platinum Member
Feb 6, 2011
2,764
3,131
136
Why does this link say 64 core, 64 threads ? Its an Intel web page about that CPU ?

Not to mention, it came out in 2016 and is discontinued now ? and was 1.3 ghz, max turbo 1.5 ghz ? I can see why it failed. And 215 watt for that pathetic performance ?
its the "ultimate" successors to larrabee , its got 2x 512bit vector units per core but the cores where in order( if i remember correctly) thus the SMT4 to sustain throughput.
 

Arkaign

Lifer
Oct 27, 2006
20,736
1,377
126
Yeah IIRC it's basically a bunch of dressed up Atom cores, with about a 1/8th level of performance relative to 'real' Core series performance. So, this big "64 core" pile of stanky Phi ass is more like an 8 Core Xeon with super low clock speeds outside of a need for tons of low performance threads lol (which has got to be a pretty rare situation).
 
  • Love
Reactions: spursindonesia

lobz

Platinum Member
Feb 10, 2017
2,057
2,856
136
Why does this link say 64 core, 64 threads ? Its an Intel web page about that CPU ?

Not to mention, it came out in 2016 and is discontinued now ? and was 1.3 ghz, max turbo 1.5 ghz ? I can see why it failed. And 215 watt for that pathetic performance ?
you may not remember this, because it's not a "real CPU" in the traditional sense. It was the HPC messiah. Guess how it's turned out! :D
 

eek2121

Platinum Member
Aug 2, 2005
2,930
4,026
136
Intel does not have a 64 core CPU, Where did you see that they do ? Not even a rumor that I have seen.

And look at the comment "trollllolololol"

Written by one of AnandTech's writers. Not fake, but definitely trolling.


The old Phi (hot garbage) is what came to mind with 64c/256t

Larrabee was quite a novel design. It was designed originally as a GPU. The Pentium M cores used in the prototypes at the time were inexpensive to manufacture.

The real issue was execution. Unfortunately, Intel never figured out how to capitalize on this design.

The idea of massive amounts of x86 cores is not dead, and there are use cases.

One realization that came to mind when the 3970X and 3990X came out is just how close we are to real time software ray tracing, for example.

I expect AMD to explore their own version of Phi or Larrabee at some point. Imagine having 512-1024 general purpose Zen cores at your fingertips. Even at low clockspeeds, this would be a potent solution for some use cases.
 

eek2121

Platinum Member
Aug 2, 2005
2,930
4,026
136
Yeah IIRC it's basically a bunch of dressed up Atom cores, with about a 1/8th level of performance relative to 'real' Core series performance. So, this big "64 core" pile of stanky Phi ass is more like an 8 Core Xeon with super low clock speeds outside of a need for tons of low performance threads lol (which has got to be a pretty rare situation).

I will point out that you just described what GPU workloads are. ( Many low performance cores performing highly threaded workloads).

There is a theory that eventually AMD and Intel will move away from the concept of GPU/APU/CPU to a tightly integrated solution where you have a single chip that performs all of those duties and more. The reason is that CPUs, GPUs, and other components duplicate, either directly or indirectly, many functions. When that happens, we will begin to see a true computing renaissance.

Some are speculating that Intel's oneAPI is the beginning of an early push in that direction.
 

Arkaign

Lifer
Oct 27, 2006
20,736
1,377
126
Written by one of AnandTech's writers. Not fake, but definitely trolling.



Larrabee was quite a novel design. It was designed originally as a GPU. The Pentium M cores used in the prototypes at the time were inexpensive to manufacture.

The real issue was execution. Unfortunately, Intel never figured out how to capitalize on this design.

The idea of massive amounts of x86 cores is not dead, and there are use cases.

One realization that came to mind when the 3970X and 3990X came out is just how close we are to real time software ray tracing, for example.

I expect AMD to explore their own version of Phi or Larrabee at some point. Imagine having 512-1024 general purpose Zen cores at your fingertips. Even at low clockspeeds, this would be a potent solution for some use cases.

I see your idea, but didn't this largely get superseded by GPGPU? I know Nvidia keeps expanding the feature sets of their SKUs for emerging DC and AI stuff. Given their interviews, it seems like Xe is focused on this as far as Intel is concerned. Even for work like Crypto, GPU is astonishingly more efficient compared to even fat cores, to say nothing of little stuff like Zacate/Atom scale.

AMD is poised to get back into it with Navi 20 it seems.
 

eek2121

Platinum Member
Aug 2, 2005
2,930
4,026
136
I see your idea, but didn't this largely get superseded by GPGPU? I know Nvidia keeps expanding the feature sets of their SKUs for emerging DC and AI stuff. Given their interviews, it seems like Xe is focused on this as far as Intel is concerned. Even for work like Crypto, GPU is astonishingly more efficient compared to even fat cores, to say nothing of little stuff like Zacate/Atom scale.

AMD is poised to get back into it with Navi 20 it seems.

GPUs to this day still have limited instruction sets. There are many types of tasks that cannot easily/efficiently be performed on a GPU.

I expect both AMD and Intel to work towards convergence within the next couple of decades.
 
Mar 11, 2004
23,073
5,554
146
GPUs to this day still have limited instruction sets. There are many types of tasks that cannot easily/efficiently be performed on a GPU.

I expect both AMD and Intel to work towards convergence within the next couple of decades.

Yeah, its likely one of the reasons why Intel is having another go at GPUs. AMD talked about that being one of the driving forces for the design of Bulldozer, with their plan to eventually start adapting GPU and CPU side by side using the same pipeline. Which that obviously failed but its because of the flaws in Bulldozer and being so different was never going to work for AMD (heck it often couldn't even work for Intel). I believe ARM is even talking about this type of thing these days.

It'll be really interesting to see how things develop.
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,686
1,221
136
Bulldozer => Multithreading architecture that achieves consistent performance of chip multiprocessing and the compute density/efficiency of coarse-grain multithreading.
Fusion => Greater CPU/GPU/Multimedia integration, in which the farthest they got in R&D is shared L3 cache between the three. While, in pathfinding they got as far as the "Fusion" core. Which fused all three in a four-threaded SMT processor with three distinct pipelines; One x86 superscalar pipeline, one VCN vliw pipeline, and one GCN/RDNA long-simd pipeline.

Don't confuse the two. Don't worry about the SMT4, it is a very common theme with AMD to pathfind w/ SMT4.

Also, the x86 pipeline used was Family 16h-derived. As Bulldozer was never used in pathfinding projects in regards to fusion.
 
Last edited:

PhonakV30

Senior member
Oct 26, 2009
987
378
136
Anyone knows what sort of latency you could expect from LPDDR4-3200?

AMD-Ryzen-5-4500U-CPUZ_2.png


3200Mhz - Cl 32 (!!!) - 29 - 34 - 68

Horrible Latency.My Memory timing is 16-18-18-38-56.
 

RetroZombie

Senior member
Nov 5, 2019
464
386
96
One realization that came to mind when the 3970X and 3990X came out is just how close we are to real time software ray tracing, for example.
Just remembering that at the early of the 3d graphics accelerators the game engines also done software based 3d rendering and all done by the cpu.
So the cpu surprisingly had enough power to do it all, but at the cost of image quality more than a frame rate problem.

For example with quake 1 i could use the integrated id software rendering and also change the engine to run on the windows opengl (microsoft one) even without a graphics card accelerator, the game would run with extreme image quality levels (even looking better than the dedicated gpu rendering) but at about one frame per second instead of around 40fps with the id software based solution.

We might be heading into the time where with so many cpu cores to spare that software based solutions could start to appear (in the cloud for example), where it's cheapper and more efficient to use cpu only solutions.
 

Gideon

Golden Member
Nov 27, 2007
1,625
3,650
136
Some very interesting new AMD Patents (most for Zen 3 but some for GPUs)

It looks like the memory controller was heavily modified in zen 3, also some Micro-op cache improvements. Clock-stretching and an voltage droop detecion (improved somehow? as this has been in since steamroller cores).

I hope the memory latency improves as even with the chiplet-approach current AMDs memory controller has at least 10-15ns left on the table. Clocks should also improve (Just temper the expectations on the ST Boost front, MT will probably improve more)

Underfox said:
 

DisEnchantment

Golden Member
Mar 3, 2017
1,601
5,780
136
Some very interesting new AMD Patents (most for Zen 3 but some for GPUs)

It looks like the memory controller was heavily modified in zen 3, also some Micro-op cache improvements. Clock-stretching and an voltage droop detecion (improved somehow? as this has been in since steamroller cores).

I hope the memory latency improves as even with the chiplet-approach current AMDs memory controller has at least 10-15ns left on the table. Clocks should also improve (Just temper the expectations on the ST Boost front, MT will probably improve more)
Mmmmm....
I also have been following the USPTO website closely. But I am not sure how he discovers those are going to make it to Zen 3
 
  • Like
Reactions: HurleyBird

JPB

Diamond Member
Jul 4, 2005
4,064
89
91
Courtesy of Hexus

AMD Ryzen 7 4800HS appears in Time Spy CPU benchmarks

An encouraging benchmark run for the AMD Ryzen 7 4800HS APU has been unearthed in the 3DMark Time Spy benchmark online results browser. First of all you might be asking what is the Ryzen 7 4800HS? At the launch of the Ryzen 4000 Series mobile APUs AMD revealed details about the Ryzen 7 4800H, and this 4800HS model appears to be an optimised top-binned 35W version of that mobile APU, destined for premium gaming laptops.


d522a424-8eb6-4bfe-ae7b-e224671846e7.jpg


Twitter's Tum Apisak found the AMD Ryzen 7 4800HS in Time Spy benchmarks online and has listed its CPU score alongside friends and rivals as below:
Time Spy CPU Score
  • R7 4800HS - 8,730
  • R7 4800H - 8,350
  • R7 3700X - 10,180
  • R7 2700X - 8,600
  • R5 3600 - 7,300
  • R5 3600 - 7,150
  • i7-9700K - 8,200
24d1419c-5be0-4e90-afea-355bca4dfe1b.jpg


The above table is all the more remarkable when one considers the TDPs of the chips that are listed. The AMD Ryzen 7 4800HS is a 35W part, the standard 4800H a 45W part. Compared to a previous gen desktop part, the powerful 105W AMD Ryzen 7 2700X, the new 4800HS still comes ahead in this CPU score benchmark run. All the aforementioned CPUs have a 8C/16T configuration. The sole Intel entry in the comparison list above is the 8C/8T Core i7-9700K, which has a 95W TDP and a Time Spy CPU Score of 8,200.

d29acf70-a401-4aaf-8f7e-2887d4ee0d80.jpg


Time Spy CPU Scores that leak out like this might have a lot of variance due to other system components and so on, however, it is still a tantalising indicator that there are some exceedingly beefy AMD powered laptops on the way.
Lenovo touts AMD Ryzen 9 4900U powered laptops
On the topic of AMD 4000 series mobile APUs, Lenovo was recently spotted showing off its latest Yoga Slim 7 family of 2-in-1 laptops with choice of AMD or Intel processors. Notebook Italia recorded a video of the exhibition which claimed one of the laptops came with an AMD Ryzen 9 4900U inside.

The source has since blurred out the AMD processor info in the video(embedded above), saying it was a typo on the Lenovo specs list. It is hard to be sure if this is the case, or if AMD exerted some pressure on its partners to keep the Ryzen 9 4900XX APUs under wraps for now. Interestingly the official Yoga Slim 7 (14-inch, AMD) product page still lists the processor as an AMD Ryzen 4000 Series, with choice of "up to AMD Ryzen 9," available for users to configure.