News AMD Patent Hints At Hybrid CPU To Rival Intel's Raptor Lake CPUs

[Admin]

AMD's patent outlines a hybrid computing architecture with clusters of both big and little cores.

AMD Patent Hints At Hybrid CPU To Rival Intel's Raptor Lake CPUs : Read more

 

[JayNor]

"The patent explains..."

patent application...

 

[hotaru.hino]

Seems like hybrid processors in PC space are the future after all.

Granted you could say that Zen 2/3 is already power efficient, given that if I use Power Saver on Windows, the CPU cores sit at around <5W according to HWiNFO.

 

[ezst036]

At first I was not happy with the idea of hybrid processors. My thought has been why would I want so many(16 or 32 say) cores when half or more may be small cores when I could have a full 32 big core goodness? The small cores, I don't really want them.

However over the last year with manufacturing capacity issues being so front and center I suspect the writing is on the wall. And if there's more than 10 or 12 big cores anyways then its probably alright to have other small ones along for the ride. I've come to accept this.

The reality of fab capacity woes kind of make big little inevitable.

 

[everettfsargent]

"The chips allegedly feature high-performance Zen 5 cores and low-powered Zen '4D' cores. Unless AMD has been diligently working behind the scenes, it's unlikely that Strix Point will make it to the market in time to compete with Intel's Alder Lake chips that may launch in late 2021 or early 2022. "

Ba ha ha ha ha ...

First, Alder Lake just has to substantially beat the 5950X across the board. Period. Full stop. Anything short of that, and we already know that the type of workload will determine Alder Lake's success at heterogeneity and success at efficiency. In other words, it is bound to be a dog in certain workloads wrt 5950X (the reverse of this, for AMD, not so much so imho). Oh and Intel may announce a PR laden Q4 launch day, but we all know that that means Q1 2022 as readily available in the retail channel to end users as tray or boxed parts.

 

[littlechipsbigchips]

This is good for notebooks and tablets , but I dont see it go anywhere in desktops and workstations.

 

[escksu]

This is good for notebooks and tablets , but I dont see it go anywhere in desktops and workstations.

It will work for desktops and workstations and even servers. You have to look beyond consumer market and look at corporate sector. There are many cases where desktops are used instead of notebooks and many are on 24/7. So, the power savings can be quite substantial. Even servers, they consume alot of power even at idle, esp. when you consider a data center.

So, if by moving certain tasks to power efficient cores, it will help to save power.

 

[jeremyj_83]

It will work for desktops and workstations and even servers. You have to look beyond consumer market and look at corporate sector. There are many cases where desktops are used instead of notebooks and many are on 24/7. So, the power savings can be quite substantial. Even servers, they consume alot of power even at idle, esp. when you consider a data center.

So, if by moving certain tasks to power efficient cores, it will help to save power.

This wouldn't really work on servers as >99% of hosts run hypervisors. The whole idea there being to run multiple smaller virtual servers to maximize the CPU load. In this case you do not want your virtual hosts running at idle most of the time. In fact you want them running 50-75% CPU usage all the time. That will keep your CPU ready percentage low and maximize your investment.

For things that don't need the maximum performance, then it could work. Firewalls would probably be a good choice for something like this in the server market.

 

[Howardohyea]

"The chips allegedly feature high-performance Zen 5 cores and low-powered Zen '4D' cores. Unless AMD has been diligently working behind the scenes, it's unlikely that Strix Point will make it to the market in time to compete with Intel's Alder Lake chips that may launch in late 2021 or early 2022. "

Ba ha ha ha ha ...

First, Alder Lake just has to substantially beat the 5950X across the board. Period. Full stop. Anything short of that, and we already know that the type of workload will determine Alder Lake's success at heterogeneity and success at efficiency. In other words, it is bound to be a dog in certain workloads wrt 5950X (the reverse of this, for AMD, not so much so imho). Oh and Intel may announce a PR laden Q4 launch day, but we all know that that means Q1 2022 as readily available in the retail channel to end users as tray or boxed parts.

Unless the new i9 12900K have at least 12 big cores and numerous small cores, it won't stand a chance even with the 5950X. Unless Intel's bringing their HEDT cheaper than ever.

 

[NeoMorpheus]

Go AMD!

Dont let that dirty and anti-consumer company (Intel) to ever get back on top.

Man, I never desired a company to die as much as Intel, well, add Nvidia for the same reasons.

 

[hotaru.hino]

Go AMD!

Dont let that dirty and anti-consumer company (Intel) to ever get back on top.

Man, I never desired a company to die as much as Intel, well, add Nvidia for the same reasons.

So you want an anti-consumer company to die by way of establishing the anti-consumer practice of having an effective monopoly?

 

[NeoMorpheus]

So you want an anti-consumer company to die by way of establishing the anti-consumer practice of having an effective monopoly?

Yes and others will rise to fill the void.

Why would we want the same bad actors around when others could potentially give us the consumer better choices and treatment?

I already know what Intel and Nvidia will do if they get back on top (in intel case) or continue expanding their monopoly (Nvidia case), so I prefer someone new.

 

[hotaru.hino]

Yes and others will rise to fill the void.

Why would we want the same bad actors around when others could potentially give us the consumer better choices and treatment?

I already know what Intel and Nvidia will do if they get back on top (in intel case) or continue expanding their monopoly (Nvidia case), so I prefer someone new.

If it were that easy, how come someone else hasn't risen up to take on any of those companies?

 

[TerryLaze]

This wouldn't really work on servers as >99% of hosts run hypervisors. The whole idea there being to run multiple smaller virtual servers to maximize the CPU load. In this case you do not want your virtual hosts running at idle most of the time. In fact you want them running 50-75% CPU usage all the time. That will keep your CPU ready percentage low and maximize your investment.

For things that don't need the maximum performance, then it could work. Firewalls would probably be a good choice for something like this in the server market.

Well you would want that but what if it's a slow day and there is nobody doing any work?
But even the way you think of it, if that 25-50% CPU usage that isn't being used can be made to use less power because the big cores can be switched off while the small cores are still fast enough to turn on the big cores when they are needed, then it's still a big win if you multiply it by however many CPUs these servers have.

 

[TerryLaze]

Unless the new i9 12900K have at least 12 big cores and numerous small cores, it won't stand a chance even with the 5950X. Unless Intel's bringing their HEDT cheaper than ever.

I don't know if you noticed the last almost 4 years, but intel doesn't even try to compete against ryzen in multithreaded, why sell more for less money if you could do the exact opposite?! they went as far as to reduce cores in the last gen at the same price, intel doesn't give a flying duck about what ryzen does because intel is selling better than ever and is making twice the money that they did before zen came along.
They don't need to beat the 5950x ,AMD has to be able to match the amount of small dual and quad core CPUs that intel is able to sell.

 

[jeremyj_83]

Well you would want that but what if it's a slow day and there is nobody doing any work?
But even the way you think of it, if that 25-50% CPU usage that isn't being used can be made to use less power because the big cores can be switched off while the small cores are still fast enough to turn on the big cores when they are needed, then it's still a big win if you multiply it by however many CPUs these servers have.

That isn't how this works. Sure there can be some virtual servers that are able to get by on only small cores, ie DNS/DHCP or file servers, however, the vast majority of servers are not those. Take for example your production ERP server. That is running 24/7 and having delays in switching between high and low performance cores matters. Plus software like ERP has specific requirements. On top of that you are already over provisioning your CPU resources in a virtual environment, for example your server has 32 virtual CPU (all CPU threads are vCPUs), you will probably be allocating out 48 vCPUs to the VMs.

 

[Soaptrail]

At first I was not happy with the idea of hybrid processors. My thought has been why would I want so many(16 or 32 say) cores when half or more may be small cores when I could have a full 32 big core goodness? The small cores, I don't really want them.

However over the last year with manufacturing capacity issues being so front and center I suspect the writing is on the wall. And if there's more than 10 or 12 big cores anyways then its probably alright to have other small ones along for the ride. I've come to accept this.

The reality of fab capacity woes kind of make big little inevitable.

Initially I thought it would be limited to mobile form factors like laptops but i do see advantages to little cores on desktops. Chips can be engineered knowing the heat from little cores will not get so hot so we could get faster big cores or just better designs to dissipate the heat better than today's 12 and 16 core CPU's.

However this will be trickier to implement that going from mono and dual cores to quad cores as the system has to be selective about what processes go where.

 

[jeremyj_83]

I don't know if you noticed the last almost 4 years, but intel doesn't even try to compete against ryzen in multithreaded, why sell more for less money if you could do the exact opposite?! they went as far as to reduce cores in the last gen at the same price, intel doesn't give a flying duck about what ryzen does because intel is selling better than ever and is making twice the money that they did before zen came along.
They don't need to beat the 5950x ,AMD has to be able to match the amount of small dual and quad core CPUs that intel is able to sell.

Intel did reduce the cores in the current gen, but that was due 100% to heat and power. They could not have make Rocket Lake as a 10c/20t CPU and kept power and heat low enough.

 

[hotaru.hino]

However this will be trickier to implement that going from mono and dual cores to quad cores as the system has to be selective about what processes go where.

I can't imagine this being hard to implement on Windows, considering Linux and Darwin (iOS) would have to be big.LITTLE aware.

If anything else, it's just an evolution of the turbo boosting/power saving algorithms already in place and Windows' scheduler is aware of the concept of "preferred cores." Well, supposedly is aware of it.

 

[spongiemaster]

That isn't how this works. Sure there can be some virtual servers that are able to get by on only small cores, ie DNS/DHCP or file servers, however, the vast majority of servers are not those. Take for example your production ERP server. That is running 24/7 and having delays in switching between high and low performance cores matters. Plus software like ERP has specific requirements. On top of that you are already over provisioning your CPU resources in a virtual environment, for example your server has 32 virtual CPU (all CPU threads are vCPUs), you will probably be allocating out 48 vCPUs to the VMs.

Not sure why server scenarios matter. Has Intel announced hybrid designs for any future server platforms? As of now, this is only getting rolled out to desktop systems. When you think of how many 100's of thousands of office desktops there are and what workloads they are typically using, the power savings will add up. Why use a 30W+ core to read emails and browse the internet and another one for background tasks when it can be done with a couple 5-10W cores just fine?

 

[jeremyj_83]

Not sure why server scenarios matter. Has Intel announced hybrid designs for any future server platforms? As of now, this is only getting rolled out to desktop systems. When you think of how many 100's of thousands of office desktops there are and what workloads they are typically using, the power savings will add up. Why use a 30W+ core to read emails and browse the internet and another one for background tasks when it can be done with a couple 5-10W cores just fine?

What I said was in response to an earlier message saying.

It will work for desktops and workstations and even servers.

In regards to office desktops this could save a lot of power, or the difference could be minimal. Reason for the minimal savings deals with the ability of modern CPUs to down clock to almost 0 power usage anyways. Then they burst up a bit when needing more performance. It will be interesting to see how much savings it could have in a desktop that is at idle. Most power savings will probably be when the CPU is running around 5-10% continuously. IE areas where the big cores aren't fully needed to run things so instead of a big core running at 10W the little core is running at 5W.

 

[spongiemaster]

I don't see a hybrid configuration being useful for servers, which is probably why we haven't heard anything from Intel in that space. As you said, if your servers aren't running with a decent load most of the time, and are sitting idle, you probably bought the wrong server.

I would not expect a huge energy savings for an individual desktop with this. However, when you multiply a little by thousands of systems in a company, it adds up really quickly.

 

[hotaru.hino]

I don't see a hybrid configuration being useful for servers, which is probably why we haven't heard anything from Intel in that space.

The only thing I could imagine is you could have the smaller cores running the "bare metal" software so that it frees up the bigger cores for other purposes.

 

[deesider]

This tech isn't intended for desktops, and certainly not intended for servers. It's all about portable devices - laptops and tablets. Every year the market share of laptops vs desktops increases, and AMD know that they need to compete directly with both Intel and ARM in that space.

I've worked with a company that has over 100,000 employees and another with over 15,000, and in both every office worker was issued a Lenovo laptop. The high performance Ryzens are great and everything, but that isn't what the market is after. That's why Intel still outsells AMD 7 to 1.

The big.little configuration will also enable the perceived desire for always-on laptops that constantly update your emails and notifications just like a mobile device does. Only ARM devices have been used for that so far.

 

[TerryLaze]

Intel did reduce the cores in the current gen, but that was due 100% to heat and power. They could not have make Rocket Lake as a 10c/20t CPU and kept power and heat low enough.

Intel used to have an AVX offset and they could have used that again to reduce heat and power as much as they wanted, they could even hardcode it into the CPU so that reviewers wouldn't be able to run AVX at full throttle no matter what.