novaBBS - soc.culture.china - Does an AMD Chiplet Have a Core Count Limit?

Hello,

I am a white arab from Morocco, and i think i am smart since i have also
invented many scalable algorithms and algorithms..

Does an AMD Chiplet Have a Core Count Limit?

And it is related to my following thoughts and writing:

More of my philosophy about the logical analogy with queries to the DNS
root nameserver and to the domain’s nameserver and more..

I have just written the following, read it carefully:

----

More of my philosophy about Bus Hierarchies with Distributed Memory...

I have to explain more since i am specialized in parallel programming
and concurrency, you have to know that AMD Threadripper Pro and such
CPUs also uses Bus Hierarchies of many buses with Distributed Memory
with many memory controllers so that to scale, and it uses snooping on
each bus of the many buses, so even if it doesn't use the scalable
Directory-Based Cache Coherence, it scales well by good placement that
reduces global bus traffic and latency.

----

So i think that Bus Hierarchies with Distributed Memory scales well by
good placement that reduces global bus traffic and latency, so it is
by logical analogy like using caches in a DNS service so that to reduce
latencies when doing queries to the DNS root nameserver and to the
domain’s nameserver, so this is why i have also posted about how to
speed much your internet connection by using a much faster
DNS server like the very fast Cloudflare DNS service that i think
also uses caches so that to reduce efficiently latencies, read about it
here:

https://www.internetsociety.org/blog/2018/04/cloudflare-launches-enhanced-dns-service/?gclid=CjwKCAjwmK6IBhBqEiwAocMc8k3_XxAWyDhVv73RPGhCqJ2_2fftZm8XznO2qkDhHFXYCLMwtZI7VhoCHJMQAvD_BwE

This is why you are noticing that your internet connection is getting
much faster by using Cloudflare DNS service, read about it
in my following writing:

---
More philosophy about the much faster DNS server..

I invite you to look at the following video on how
to configure internet protocol to use new DNS servers
that are much faster and this will speed much your internet
connection, here is how to do it:

https://www.youtube.com/watch?v=kqnvrjgyEMc
-----

More of my philosophy about AMD Ryzen Threadripper PRO and Nvidia V100
PCIe (Volta) and more..

I think that in the very near future CPUs will be able to replace GPUs
in specialized works of GPUs, since we can approximate it by noticing
below that the 32 cores or 64 cores Next-Gen Sapphire Rapids Xeon with
High Bandwidth Memory will be near the memory bandwidth and near the
GFLOPS of Nvidia V100 PCIe (Volta) powerful GPU and it will come with a
competitive price, read all my thoughts below and do your comparison so
that to notice it.

And I invite you to look at the following spec of AMD Ryzen Threadripper
PRO 3975WX 32-Core CPU:

https://www.techpowerup.com/cpu-specs/ryzen-threadripper-pro-3975wx.c2315

And look carefully at the following benchmark:

https://www.xcelerit.com/computing-benchmarks/insights/benchmarks-intel-xeon-scalable-processor-vs-nvidia-v100-gpu/

So as you are noticing that the spec of Nvidia V100 PCIe (Volta) 16 GB
is 7,014 GFLOPs (double), and AMD Ryzen Threadripper PRO 3975WX 32-Core
CPU is around 6,451.2 GFLOPS, but look carefully at the price of Nvidia
V100 PCIe (Volta) 16 GB that is 7124 US dollars:

https://www.amazon.ca/PNY-TCSV100MPCIE-PB-Nvidia-Tesla-v100/dp/B076P84525/ref=pd_di_sccai_3?pd_rd_w=AmXIj&pf_rd_p=e92f388e-b766-4f7f-aac1-ee1d0056e8fb&pf_rd_r=77B7DWXEVBM5VSXT4NZG&pd_rd_r=27e26b6a-c0a6-4558-8a68-97e286ba6213&pd_rd_wg=HxaUi&pd_rd_i=B076P84525&psc=1

And look at the price of AMD Ryzen Threadripper PRO 3975WX 32-Core CPU
that is 2790 US dollars:

https://www.newegg.ca/amd-ryzen-threadripper-pro-3975wx/p/N82E16819113677

So i think that AMD Ryzen Threadripper PRO 3975WX 32-Core CPU is
competitive in performance and price for the GFLOPS with Nvidia V100
PCIe (Volta) 16 GB.

I have just read the following interesting article about AVX512

On the dangers of Intel's frequency scaling

https://blog.cloudflare.com/on-the-dangers-of-intels-frequency-scaling/

So as you have just noticed by reading the above article that you have
not to use AVX512, because it heats a lot the CPU cores, so what Intel
is doing is to reduce a lot the speed of the CPU cores, but this is not
good for performance. So what i advice is to avoid AVX2 or AVX512
and choose to use AVX that has not this problem.

More of my philosophy about Bus Hierarchies with Distributed Memory...

More of my philosophy about AMD Threadripper Pro multicores CPU..

You can read more about my education and my way of doing here:

Here is more proof of the fact that i have invented many scalable
algorithms and algorithms:

https://groups.google.com/g/comp.programming.threads/c/V9Go8fbF10k

I am specialized in parallel programming and concurrency and think i
know more about multicores CPUs, and the AMD Threadripper Pro multicores
CPU has many controllers that send on the many buses of NUMA nodes by
doing snooping, so it is not Directory-Based Cache Coherence, but it is
an efficient way that efficiently reduces snooping since it is
parallelized, and here is more of my thoughts about this subject:

More about data centers and about NUMA multicore servers..

I have just written the the following:

--------------------------

I have just written the following:

---
About Snooping vs. Directory-based coherency..

Performance Scalability of a Multi-core Web Server

As you notice above that the Address bus saturation causes
poor scaling! And the Address Bus carries requests and
responses for data, called snoops, and more caches mean more sources
and more destinations for snoops that is causing the poor scaling.

So to solve the problem of poor scalability above, you have to use
Directory-based coherence that is a mechanism to handle Cache coherence
problem in Distributed shared memory (DSM) a.k.a. Non-Uniform Memory
Access (NUMA).

---

And you have to know that Directory-Based Cache Coherence is scalable.

I invite you to read the following interesting article about AMD
Threadripper Pro CPU and you will notice that it is a very good CPU
that supports ECC memory and is effectively a faster version of AMD’s
EPYC, limited for single CPU workstation use:

AMD Threadripper Pro Review: An Upgrade Over Regular Threadripper?

https://www.anandtech.com/show/16805/amd-threadripper-pro-review-an-upgrade-over-regular-threadripper

And read the following:

AMD's Zen 4 EPYC could get HBM upgrade to fight Intel Sapphire Rapids

And read my following writing about Intel Next-Gen Sapphire Rapids Xeon
with High Bandwidth Memory:

Intel to Launch Next-Gen Sapphire Rapids Xeon with High Bandwidth Memory

"Current eight-channel DDR4-3200 memory designs, for example, have a
theoretical maximum of 204.8 gigabytes per second, which pales in
comparison to GPUs which quote 1000 gigabytes per second or more. GPUs
are able to achieve higher bandwidths because they use GDDR.

At the very least, we expect the equivalent of up to 8-Hi stacks of
HBM2e, up to 16GB each, with 1-4 stacks onboard leading to 64 GB of HBM.
At a theoretical top speed of 460 GB/s per stack, this would mean 1840
GB/s of bandwidth, although we can imagine something more akin to 1 TB/s
for yield and power which would still give a sizeable uplift."

Russia To Build RISC-V Processors for Laptops: 8-core, 2 GHz, 12nm, 2025

Read more here:

Click here to read the complete article

Subject	Author
Does an AMD Chiplet Have a Core Count Limit?	World-News2100

You will contract a rare disease.

interests / soc.culture.china / Does an AMD Chiplet Have a Core Count Limit?