VEGA AS4161 : 64-bit High performance Quad core Microprocessor, RISC-V based

[snappercayt]

Summary

VEGA AS4161 : 64-bit High performance Quad core Microprocessor

Quotes

Quote

 a quad core out-of-order processing engine with a 16 stage pipeline for high performance compute requirements. The processor also supports single and double precision floating point instructions, and MMU for Linux based applications. 

My thoughts

 Quad core, based on RISC-V, has Arduino Shield compatible board like ARIES v2.0 (more info here) what is not to like here? not much talk about it anywhere, I hope LTT picks it up and can promote this young venture so it can get some test samples out in the wild for more tinkerers.

Sources

1. video link 

2. the website link https://vegaprocessors.in/dhanush64.html

 

[Bombastinator]

Risc-V is a totally different architecture from x86, Like ARM. It is used as an embedded processor in stuff like smart devices quite frequently. There is a decent chance this thing is really small and low power.  A pi4 is also 4 core. Them calling it a Vega processor is confusing as that is the same name given to an older line of AMD GPUs for atx machines which iirc is not risc-v based.  So there may be two Vegas.  One is a risc-v cpu, and the other is an AMD gpu.

[snappercayt]

16 minutes ago, Bombastinator said:

Risc-V is a totally different architecture from x86, Like ARM. It is used as an embedded processor in stuff like smart devices quite frequently. There is a decent chance this thing is really small and low power.  A pi4 is also 4 core. Them calling it a Vega processor is confusing as that is the same name given to an older line of AMD GPUs for atx machines which iirc is not risc-v based.  So there may be two Vegas.  One is a risc-v cpu, and the other is an AMD gpu.

Yes, Radeon RX Vega series is a graphics processor line from AMD based on GCN5 and Vega is family of Soft IP which has different variants from IoT capable to more network focused microprocessors developed by C-DAC which is based on RISC-V. I did not say they are same. It is a new quad core microprocessor and has not been talked about much. 

[Bombastinator]

16 minutes ago, snappercayt said:

Yes, Radeon RX Vega series is a graphics processor line from AMD based on GCN5 and Vega is family of Soft IP which has different variants from IoT capable to more network focused microprocessors developed by C-DAC which is based on RISC-V. I did not say they are same. It is a new quad core microprocessor and has not been talked about much. 

Fair enough.  I just wanted to make sure there was no confusion considering the venue.

[porina]

56 minutes ago, Bombastinator said:

Them calling it a Vega processor is confusing as that is the same name given to an older line of AMD GPUs for atx machines which iirc is not risc-v based.  So there may be two Vegas.  One is a risc-v cpu, and the other is an AMD gpu.

AMD were hardly original in picking Vega themselves, as I think to the masses it is better known as a star in the sky, like Polaris before it. If you use long existing references for naming chances you wont be alone. Another recent example is we have Ampere, nvidia's current GPU architecture, and the company Ampere making ARM server CPUs. Both took inspiration from some dead guy. Sorry, my historical electrical people knowledge isn't so great, but he did something important to have a fundamental unit named after him.

 

 

On the chip itself, will see how it goes. I would love to have a desktop level RISC-V system to play. Given even initial Pi could run desktop Linux, something similar to that would suffice.

[Bombastinator]

42 minutes ago, porina said:

AMD were hardly original in picking Vega themselves, as I think to the masses it is better known as a star in the sky, like Polaris before it. If you use long existing references for naming chances you wont be alone. Another recent example is we have Ampere, nvidia's current GPU architecture, and the company Ampere making ARM server CPUs. Both took inspiration from some dead guy. Sorry, my historical electrical people knowledge isn't so great, but he did something important to have a fundamental unit named after him.

 

 

On the chip itself, will see how it goes. I would love to have a desktop level RISC-V system to play. Given even initial Pi could run desktop Linux, something similar to that would suffice.

The problem is they’re both computer chips and the vega gpu system is better known.  Your point about unoriginal naming schemes is taken though.

[Brooksie359]

10 hours ago, Bombastinator said:

Risc-V is a totally different architecture from x86, Like ARM. It is used as an embedded processor in stuff like smart devices quite frequently. There is a decent chance this thing is really small and low power.  A pi4 is also 4 core. Them calling it a Vega processor is confusing as that is the same name given to an older line of AMD GPUs for atx machines which iirc is not risc-v based.  So there may be two Vegas.  One is a risc-v cpu, and the other is an AMD gpu.

I wouldn't be surprised if AMD made them change the name as they likely already trademark Vega. 

[Brooksie359]

9 hours ago, porina said:

AMD were hardly original in picking Vega themselves, as I think to the masses it is better known as a star in the sky, like Polaris before it. If you use long existing references for naming chances you wont be alone. Another recent example is we have Ampere, nvidia's current GPU architecture, and the company Ampere making ARM server CPUs. Both took inspiration from some dead guy. Sorry, my historical electrical people knowledge isn't so great, but he did something important to have a fundamental unit named after him.

 

 

On the chip itself, will see how it goes. I would love to have a desktop level RISC-V system to play. Given even initial Pi could run desktop Linux, something similar to that would suffice.

Well I mean apple isn't exactly a unique name but that doesn't change the fact that if you try and name something apple in the tech industry you would get sued immediately. 

[porina]

1 hour ago, Brooksie359 said:

Well I mean apple isn't exactly a unique name but that doesn't change the fact that if you try and name something apple in the tech industry you would get sued immediately. 

There are things Apple Inc do because they have to under trademark law, and there are things they do because they have huge wedges of cash to throw at lawyers. It doesn't always go their way.

 

https://en.wikipedia.org/wiki/Apple_Corps_v_Apple_Computer

 

1 hour ago, Brooksie359 said:

I wouldn't be surprised if AMD made them change the name as they likely already trademark Vega. 

You can have similar sounding/looking trademarks as long as they're in different product areas. I'm not trademark lawyer so I don't know how close or far apart that has to be to become a problem.

[williamcll]

This thing can't even hit 500Mhz, I don't think it would run most applications at all. 

[hishnash]

40 minutes ago, williamcll said:

This thing can't even hit 500Mhz, I don't think it would run most applications at all. 

while I agree it is not going to be a powerhouse the clock speed is not a good metric of that, you could build a RISC core that out perfumers a 6ghz x86 chip while running at 500Mhz since but focusing on a massive IPC, the simpler instruction set of RISC solutions makes building a ultra ultra wide core more possible but I do not expect they have done that here.  

This chip will be a fairly simple in order only narrow core without any speculative out of order support so will be very poor for regular desktop compute but completely fine for controle systems, sensors etc. 

[Bombastinator]

5 hours ago, porina said:

There are things Apple Inc do because they have to under trademark law, and there are things they do because they have huge wedges of cash to throw at lawyers. It doesn't always go their way.

 

https://en.wikipedia.org/wiki/Apple_Corps_v_Apple_Computer

 

You can have similar sounding/looking trademarks as long as they're in different product areas. I'm not trademark lawyer so I don't know how close or far apart that has to be to become a problem.

The question becomes are these the same or different areas though.  One could argue either way.

[05032-Mendicant-Bias]

9 hours ago, williamcll said:

This thing can't even hit 500Mhz, I don't think it would run most applications at all. 

Features like branch prediction and out of order are worth a lot more than frequency. X86 had decades to work out something good. ARM, especially the Apple version of the core too is ahead.

Still, by specsheet, this is a real processor that can theoretically run a full linux, given proper hardware level drivers for the kernel if it works as advertised. I think it's a lot of work to make the linux scheduler take proper advantage of this architecture, and I suspect most companies that adopt the processor IP won't release the kernel open source.

 

 

[snappercayt]

22 minutes ago, 05032-Mendicant-Bias said:

Features like branch prediction and out of order are worth a lot more than frequency. X86 had decades to work out something good. ARM, especially the Apple version of the core too is ahead.

Still, by specsheet, this is a real processor that can theoretically run a full linux, given proper hardware level drivers for the kernel if it works as advertised. I think it's a lot of work to make the linux scheduler take proper advantage of this architecture, and I suspect most companies that adopt the processor IP won't release the kernel open source.

 

 

It is a very real processor! and I think I found some real competitor for it too, some Single Board ARM devices and bad boys like this V-Raptor SQ nano, which was in the news few years here back but the render looks so good. In an application like network and storage is best starting point for a Linux capable system.

 

[Franck]

On 8/28/2022 at 4:44 AM, Bombastinator said:

Risc-V is a totally different architecture from x86

Despite being different it is programmable like a typical FPGA so you can write all x86 instruction on it and you have a x86 ~ish chip.

[Bombastinator]

1 hour ago, Franck said:

Despite being different it is programmable like a typical FPGA so you can write all x86 instruction on it and you have a x86 ~ish chip.

If you can do that.  It’s probably illegal though.  IP and stuff.  AMD64 is owned by AMD.  Intel licenses it.  If the company that makes the chip can acquire such a license why didn’t they do it already?  Also Aren’t fpgas multiple times more expensive than regular CPUs?  Might be a new tech or something where such things got suddenly much cheaper I suppose.  Also with an artificial market like x86 chips they may be so overvalued that a company might be able to sell an fpga for the same price as an x86 cpu and still make money.  Just not nearly as much.   I have no idea what x86 CPUs actually cost to make.  For all I know it’s a couple of bucks, they just sell them for $500 because they can. 

[Franck]

53 minutes ago, Bombastinator said:

 Might be a new tech or something where such things got suddenly much cheaper I suppose

RISC V are not that expensive, way less than typical FPGA. They are more targeted to embed device and they can be really cheap like 10$ but you can get beefier in the upward of 400$

 

[Brooksie359]

17 hours ago, porina said:

There are things Apple Inc do because they have to under trademark law, and there are things they do because they have huge wedges of cash to throw at lawyers. It doesn't always go their way.

 

https://en.wikipedia.org/wiki/Apple_Corps_v_Apple_Computer

 

You can have similar sounding/looking trademarks as long as they're in different product areas. I'm not trademark lawyer so I don't know how close or far apart that has to be to become a problem.

But that's my point. The Vega cpu and the vega gpu are at the end of the day processors so they are super close compared to what is even normally required. Usually they don't even have to directly compete but have to be in the same market segment like computer hardware. 

[Bombastinator]

Just now, Brooksie359 said:

But that's my point. The Vega cpu and the vega gpu are at the end of the day processors so they are super close compared to what is even normally required. Usually they don't even have to directly compete but have to be in the same market segment like computer hardware. 

But one is a gpu for x86 and one is a cpu that is not.  I suspect this will come down to whether Nvidia feels they are too close.

[Brooksie359]

17 hours ago, porina said:

There are things Apple Inc do because they have to under trademark law, and there are things they do because they have huge wedges of cash to throw at lawyers. It doesn't always go their way.

 

https://en.wikipedia.org/wiki/Apple_Corps_v_Apple_Computer

 

You can have similar sounding/looking trademarks as long as they're in different product areas. I'm not trademark lawyer so I don't know how close or far apart that has to be to become a problem.

Also why do you say it doesn't always go their way and then link a legal dispute that went their way lol 

[Brooksie359]

2 minutes ago, Bombastinator said:

But one is a gpu for x86 and one is a cpu that is not.  I suspect this will come down to whether Nvidia feels they are too close.

That is not how this works. It would be like if Intel named its new cpu rtx 7. Yeah they would immediately get sized by nvidia even though one is a gpu and the other is a cpu. 

[Bombastinator]

5 minutes ago, Brooksie359 said:

That is not how this works. It would be like if Intel named its new cpu rtx 7. Yeah they would immediately get sized by nvidia even though one is a gpu and the other is a cpu. 

Might depend on what the cpu is for.  If it’s an embedded chip for something like tire pressure sensing it becomes problematic.

[porina]

20 minutes ago, Brooksie359 said:

Also why do you say it doesn't always go their way and then link a legal dispute that went their way lol 

Apple the computer company did not get their way in most of the battles listed on that page.

[Brooksie359]

3 minutes ago, porina said:

Apple the computer company did not get their way in most of the battles listed on that page.

Ok but it supports my justification even more. You have two companies in different market segments suing each other over a trademark that is super generic. In what way does this prove that AMD can't sue for the Vega trademark? I mean they are even in the same Industry so it's way more reasonable for AMD to sue them over the Vega name. 

[porina]

1 minute ago, Brooksie359 said:

In what way does this prove that AMD can't sue for the Vega trademark?

The main point I was trying to make was that there are rules in place for trademarks, and at the end of the day it is up to the legal system to decide where the lines are and if they have been crossed. I don't claim to understand it well enough to say if AMD should or need to take any action because of it, or if the new CPU should or shouldn't have used Vega also.