News Nvidia Claims Arm Grace CPU Superchip 2X Faster, 2.3X More Efficient than Intel Ice Lake

ezst036

Honorable
Oct 5, 2018
519
397
11,920
As an ARM chip, I'm more interested in how Grace stacks up against another ARM chip.

Apple's M1. (Max/Ultra/etc) Looking forward to benchmarks some day.
 
Sep 5, 2021
3
0
10
ARM chips from Nvidia are going to surpass x86 CPUs from Intel and AMD in data centers like how Apple Silicon has surpassed them in consumer desktops and laptops.
 

Liquidrider

Honorable
Nov 25, 2016
19
4
10,515
ARM chips from Nvidia are going to surpass x86 CPUs from Intel and AMD in data centers like how Apple Silicon has surpassed them in consumer desktops and laptops.


If you believe that then you are ignoring the industry standards. AMD's Epyc line-up is on a whole new playing field, which is probably why Nvidia isn't comparing ARM to them. While Industry does change Nvidia has a long road ahead. Switching from x86 to ARM isn't a like turning on a switch.
And I would love to hear your explanation of how Apple surpassed Intel & AMD in the consumer market? Surpass in what exactly? Windows still dominates the consumer market by a wide margin.
 

spongiemaster

Admirable
Dec 12, 2019
2,273
1,277
7,560
If you believe that then you are ignoring the industry standards. AMD's Epyc line-up is on a whole new playing field, which is probably why Nvidia isn't comparing ARM to them.
Except Nvidia did compare their chip to Epyc during the original announcement. The ones they currently use in their DGX systems. Seems unlikely Nvidia would replace the Epyc CPU's in their current system with slower ARM CPU's.
 
Jan 3, 2022
67
13
35
The various permutations of WRF are real-world workloads commonly used for benchmarking, and many of the modules have been ported over for GPU acceleration with CUDA.

I'm very interested to see if Saphire Rapids will use same GPU acceleration with new Intel OneAPI and how will this compare with CUDA.
 

jp7189

Distinguished
Feb 21, 2012
323
187
18,860
Except Nvidia did compare their chip to Epyc during the original announcement. The ones they currently use in their DGX systems. Seems unlikely Nvidia would replace the Epyc CPU's in their current system with slower ARM CPU's.
Of course they would. More NVDIA components equals more profit to them.
 

jp7189

Distinguished
Feb 21, 2012
323
187
18,860
ARM chips from Nvidia are going to surpass x86 CPUs from Intel and AMD in data centers like how Apple Silicon has surpassed them in consumer desktops and laptops.
Chuckle even their best simulations are showing a loss in reality. 144 Grace cores vs 72 <real> intel cores. 2x the cores = 2x the performance. BUT intel has a much slower socket to socket interface and much slower RAM. That tells us that the <future> fundamental Grace core is slower than existing Intel cores.
 

JamesJones44

Reputable
Jan 22, 2021
620
560
5,760
If you believe that then you are ignoring the industry standards. AMD's Epyc line-up is on a whole new playing field, which is probably why Nvidia isn't comparing ARM to them. While Industry does change Nvidia has a long road ahead. Switching from x86 to ARM isn't a like turning on a switch.
And I would love to hear your explanation of how Apple surpassed Intel & AMD in the consumer market? Surpass in what exactly? Windows still dominates the consumer market by a wide margin.

What "Industrial standard" are you speaking of? In the server/backend/cloud/DataCenter world ARM has been supported for a long time, almost all software developed today for the server land runs on both x86 and AArch. If you question it, go ahead and take a look at Docker Hub and see how many containers of top software run on both.

x86 in server land is in real trouble due to the performance per watt issue. Every watt AWS, G-Cloud, Azure saves is money in the bank. Desktop land is a different story, but in server land there is no "standard". Just look at how much market share it has lost to ARM already there since 2020. Heck we even have customers with private DataCenter running ARM now, something that 3 years ago would have been crazy to think of. I'm sorry, but the days of required architecture are over, it's either best performance or best performance per watt and that is all that matters now.
 

Kamen Rider Blade

Distinguished
Dec 2, 2013
1,280
810
20,060
x86 in server land is in real trouble due to the performance per watt issue. Every watt AWS, G-Cloud, Azure saves is money in the bank. Desktop land is a different story, but in server land there is no "standard". Just look at how much market share it has lost to ARM already there since 2020. Heck we even have customers with private DataCenter running ARM now, something that 3 years ago would have been crazy to think of. I'm sorry, but the days of required architecture are over, it's either best performance or best performance per watt and that is all that matters now.

Server Revenue says otherwise. Team x86 is doing fine and the gap is growing between Team x86 & Non-x86
fqNiAgN.jpg


Non-x86 Server Revenue is on a Downward Trend:
A49oNrM.jpg

IBM eats up a size-able portion of the Non-x86 Server Revenue.
 
What "Industrial standard" are you speaking of? In the server/backend/cloud/DataCenter world ARM has been supported for a long time, almost all software developed today for the server land runs on both x86 and AArch. If you question it, go ahead and take a look at Docker Hub and see how many containers of top software run on both.

x86 in server land is in real trouble due to the performance per watt issue. Every watt AWS, G-Cloud, Azure saves is money in the bank. Desktop land is a different story, but in server land there is no "standard". Just look at how much market share it has lost to ARM already there since 2020. Heck we even have customers with private DataCenter running ARM now, something that 3 years ago would have been crazy to think of. I'm sorry, but the days of required architecture are over, it's either best performance or best performance per watt and that is all that matters now.
You say "almost all software developed today for the server land runs on both x86 and AArch." What software are you talking about? While you can get Windows Server or SLES to run on ARM, a lot of applications won't. For example, SQL Server & SAP HANA don't support ARM so you still need x86 servers to run them and those are two widely used Databases. Maybe the application server that attaches to the DB will work, but it doesn't make sense to have the split in uArch in the data center. VMware only supports ARM as a fling. That means it is only for evaluation and not production use and using it means you won't get support from VMware if there is an issue. That is just an example of 3 HUGE data center applications that will not run on ARM. Until the huge software companies support ARM across the board ARM will be nothing more than a niche in the server world.

In terms of power use between ARM and x86 we actually don't know if ARM is more efficient. Since things like Graviton2 are custom chips, the TDP isn't released to the public.
 

JamesJones44

Reputable
Jan 22, 2021
620
560
5,760
You say "almost all software developed today for the server land runs on both x86 and AArch." What software are you talking about? While you can get Windows Server or SLES to run on ARM, a lot of applications won't. For example, SQL Server & SAP HANA don't support ARM so you still need x86 servers to run them and those are two widely used Databases. Maybe the application server that attaches to the DB will work, but it doesn't make sense to have the split in uArch in the data center. VMware only supports ARM as a fling. That means it is only for evaluation and not production use and using it means you won't get support from VMware if there is an issue. That is just an example of 3 HUGE data center applications that will not run on ARM. Until the huge software companies support ARM across the board ARM will be nothing more than a niche in the server world.

In terms of power use between ARM and x86 we actually don't know if ARM is more efficient. Since things like Graviton2 are custom chips, the TDP isn't released to the public.

Nobody runs Windows Server in the cloud, no one. Even local data centers, hardly anyone runs Window Sever anymore, old Electric and Water Utilities is about it, same with SQL sever. Almost everyone is moving to Kubernetes which is Linux. As for software that runs on ARM, here is a small list. Elastic Search, Oracle DB, My SQL, Radius, Kubernetes, Docker, Nginx, VMWares EXSi has supported ARM for a very long time, if you are talking about VM Ware Desktop, no one uses that on the sever side. As for programming languages, Node, Java, .Net, Go, Rust, Kotlin all support running under ARM. This means your of your popular development languages for server side development can be built with ARM, in fact, it so simple as long as you have a Docker Build File for any of these languages it's a simple argument passed to the Docker Build command to build for both x86/aarch. Again, go to Docker Hub and look for yourself.

So basically any software that is used in a modern stack (aka cloud native https://docs.microsoft.com/en-us/dotnet/architecture/cloud-native/definition). If your stack is 20 years old then maybe they still use Windows Server, SQL Server and god forbid anything made by SAP. Here is an awesome line direct from Microsoft in the Cloud Native definition "Containers support both Linux and Windows workloads. The Azure cloud openly embraces both. Interestingly, it's Linux, not Windows Server, that has become the more popular operating system in Azure.".
 
Last edited:
Nobody runs Windows Server in the cloud, no one. Even local data centers, hardly anyone runs Window Sever anymore, old Electric and Water Utilities is about it, same with SQL sever.
Where did you get that information? I actually run a data center that does cloud hosting as well. About half of our servers are Linux and the other half are Windows. As of May 2021 SQL Server still has a 16.36% market share which is good for 3rd place behind Oracle & MySQL.

VMWares EXSi has supported ARM for a very long time
ESXi only supports ARM as a fling. Flings are apps and tools built by VMware engineers and community that are intended to be explored. That is 100% NOT production support.
 
Seems unlikely Nvidia would replace the Epyc CPU's in their current system with slower ARM CPU's.

going forward they probably going to replace all their DGX with their own ARM CPU even if Epyc end up being faster. main thing with DGX is majority of the processing are done by the GPU anyway. with Hopper nvidia add things like DPX support which in the past are either being done by CPU or FPGA. so they probably will just going to trim their CPU design for what they need instead of using more general purpose x86.
 

JamesJones44

Reputable
Jan 22, 2021
620
560
5,760
Where did you get that information? I actually run a data center that does cloud hosting as well. About half of our servers are Linux and the other half are Windows. As of May 2021 SQL Server still has a 16.36% market share which is good for 3rd place behind Oracle & MySQL.


ESXi only supports ARM as a fling. Flings are apps and tools built by VMware engineers and community that are intended to be explored. That is 100% NOT production support.

16% and shrinking. I do a lot of server side development, I can't remember the last time a client demanded Windows and SQL Sever. Old Banks, Utilities, etc used to, but these days they have even warmed to Linux which doesn't have the same limitations you get from Windows Server as far as architecture goes.

Just look at the number of AWS instances that are available in both Arm and x86 and while GCloud and Azure don't have them yet, they both are planning to bring them online later this year. My point is, there is not "standard" in the cloud as far as instruction architecture goes. It's moving to architecture agnostic and anyone who believes otherwise is simply blind.
 

Kamen Rider Blade

Distinguished
Dec 2, 2013
1,280
810
20,060

JamesJones44

Reputable
Jan 22, 2021
620
560
5,760

Thank you.

I think the note in that article is what people need to pay attention to.

The reason is that Amazon Web Services is using its homegrown Graviton and Graviton2 Arm server processors in a significant percentage of its server installations, and we think Microsoft, Oracle, and Tencent are also deploying a fair number of Altra Arm server chips from Ampere Computing as well. It’s enough to fill in the gaps, and so is the significant – and growing – revenues that Inspur is getting from its Power server line, which is distinct from IBM’s own Power Systems.

This is a trend that is going to continue, ARM servers are showing up in the cloud a lot and they don't track revenue from AWS using their own Graviton processors. G Cloud, Azure, etc. are all bring ARM servers online in the cloud and while their performance isn't as high as say an Ice Lake instance, they are good enough for most applications and cost less both for the app developer and for the provider. Many popular services (Postgresql, Redis, Spring Boot, etc.) already support ARM with minimal effort. That wasn't the case 5 years ago, but it is now. That was my original point about there not being a "standard" on the cloud side.
 

Kamen Rider Blade

Distinguished
Dec 2, 2013
1,280
810
20,060
This is a trend that is going to continue, ARM servers are showing up in the cloud a lot and they don't track revenue from AWS using their own Graviton processors. G Cloud, Azure, etc. are all bring ARM servers online in the cloud and while their performance isn't as high as say an Ice Lake instance, they are good enough for most applications and cost less both for the app developer and for the provider. Many popular services (Postgresql, Redis, Spring Boot, etc.) already support ARM with minimal effort. That wasn't the case 5 years ago, but it is now. That was my original point about there not being a "standard" on the cloud side.
Hopefully, somebody figures out a way to factor in the # of Chips Sold/Made for their specific variant of ARM server CPU's used in their specific slice of the Cloud.