AMD CPU Market Share Rises In All Segments Except Servers

[GOTSpectrum]

Whether you are an AMD fanboy or an Intel fanboy, one thing we can all agree on, competition is good for the consumer. Its great we are seeing AMD making inroads into the market as a whole, competitions means lower prices and better products. 

 

https://wccftech.com/amd-cpu-market-share-desktop-notebook-gains-q1-2019/?fbclid=IwAR0-SE67ovW_5YFj_xUho4AlfHpE6eGUzzqij7FdMw55C9ajgug4O4jfhAQ

 

Quote

t looks like AMD Ryzen processors have been a fruitful investment for team red, not only leading to strong financial quarters but also continually gaining CPU market share. According to the latest figures for Q1 2019, the AMD desktop CPU market share grew by 1.3% which now sits at 17.1% & the notebook CPU market share increased by 1.0% which is now at 13.1%. Compared to last year, the gain is 4.9% for the desktop side and 5.1% for the notebook side.

So we are seeing some pretty nice growth here, its about time AMD got what they deserved with all the work they put into Zen. 

 

Quote

The notebook segment is a major win for AMD as they delivered a really strong Ryzen 3000 series lineup at CES 2019 which is now available in notebooks from various OEMs.

This is somewhere where AMD has usually struggled in the past, so to see them expanding market share in the mobile segment is really something. 

 

Quote

n the server side, while AMD’s market share grew compared to the previous year by 1.9%, this specific quarter saw a decline of 0.3% which is largely due to the industry waiting for the 7nm EPYC Rome series processors which will be shipping in Q3 2019.

The Osborne effect, where the hype and knowledge of a near-future superior product kills interest in the current line-up. Lets just hope that we see some decent group with AMD newest offerings, higher density, greater IO options and potentially a better performance per watt than the competitor must be something that will interest many in the field.

 

All in all, as said before it is good to see the fruits of AMDs labour finally starting to bare, only time will if they can keep up this momentum long term. 

[sazrocks]

It would be interesting to see these figures in volume rather than just relative percentages. 

[exetras]

The Server or Enterprise market is slow to move. It's going to take time for AMD to penetrate it.

 

The last time AMD was part of the server market they left it with a generation of Opterons that could not compete with Intel and nothing to upgrade to. this made for a painful and expensive upgrade.

 

If Intel cant keep up with the Chip demand AMD could take a bite out of their market. Also people are still talking about Specter and all those vulnerabilities and the performance loss due to the fixes.

[GOTSpectrum]

6 minutes ago, exetras said:

The Server or Enterprise market is slow to move. It's going to take time for AMD to penetrate it.

 

The last time AMD was part of the server market they left it with a generation of Opterons that could not compete with Intel and nothing to upgrade to.

 

If Intel cant keep up with the Chip demand AMD could take a bit out of their market. Also people are still talking about Specter and all those vulnerabilities and the performance loss due to the fixes.

This is quite true, but with the recent large deployments we have been seeing for AMD over the next few years I think they are finally starting to see some headway. 

 

The fact that after Rome you can have developed a system, then when it's ready to deploy drop in a superior product with up to twice as many cores for a similar power(and cooling) specification must be tempting for some of the smaller data centre uses. 

[Trixanity]

AMD really needs to bring up their APU release schedule. It's a bit too far behind for my taste. Right now we appear to be looking at a early 2020 for 7nm while Intel is ready with 10nm Ice Lake for holidays this year. So they kinda need it out the door in Q3 to really nail Intel to the cross.

[GOTSpectrum]

1 minute ago, Trixanity said:

AMD really needs to bring up their APU release schedule. It's a bit too far behind for my taste. Right now we appear to be looking at a early 2020 for 7nm while Intel is ready with 10nm Ice Lake for holidays this year. So they kinda need it out the door in Q3 to really nail Intel to the cross.

But the thing is, APUs are a relatively small part of AMDs line up, not only that but the only reason they exist is sticking the mobile chips in a desktop form factor. 3000 series APUs will most likely still be quad core parts IMO. 

 

What I'd really like to see is an APU with say a Vega/Navi 20, 4gb hbm and 4c with SMT. would be nice for sff with enough power to smack down basically anything has to offer GPU wise. Hell if they wanted to go the NUC and gaming laptop route they could cut out AM4 100% and just major the die bigger do you can have a 6 core with say Vega 24. 

 

Or the other option just take the next gen console APUs, tweet as little as possible and sell them for high end laptops.

 

I don't expect any of this to happen but I can dream!

[Trixanity]

3 minutes ago, Ben Quigley said:

But the thing is, APUs are a relatively small part of AMDs line up, not only that but the only reason they exist is sticking the mobile chips in a desktop form factor. 3000 series APUs will most likely still be quad core parts IMO. 

 

What I'd really like to see is an APU with say a Vega/Navi 20, 4gb hbm and 4c with SMT. would be nice for sff with enough power to smack down basically anything has to offer GPU wise. Hell if they wanted to go the NUC and gaming laptop route they could cut out AM4 100% and just major the die bigger do you can have a 6 core with say Vega 24. 

 

Or the other option just take the next gen console APUs, tweet as little as possible and sell them for high end laptops.

 

I don't expect any of this to happen but I can dream!

They need the APUs precisely for the mobile market. I'm not talking their ridiculously small desktop segment.

Also, the 3000 series is already out. Again, precisely because it's late so the name is meant to line up with the rest of their 2019 stack.

 

It remains to be seen if 7nm allows two CCXs in the APUs. They kinda need it for competing against Intel. It will almost certainly be monolithic due to costs. Same reason HBM isn't happening outside some experiment with packaging because it's risky to pour money into an uncertain product (still doubt it'll happen anytime soon though).

 

I also doubt they can sell console products. They're developed on behalf of and in collaboration with partners. Other than the need to tweak the design for general use, I imagine Sony/Microsoft have exclusive rights to the product because otherwise AMD could just shop it around to everyone.

 

Fact remains that AMD needs to significantly expand their APU business.

[GOTSpectrum]

6 minutes ago, Trixanity said:

 doubt they can sell console products. They're developed on behalf of and in collaboration with partners. Other than the need to tweak the design for general use, I imagine Sony/Microsoft have exclusive rights to the product

I doubt this to some degree, the fact both consoles have such similar sillicon proves that they don't have total exclusivity. So if they can sell don't and Microsoft such similar sillicon there's no reason they couldn't sell a slight variation to other vendors. Like the Apu that is used in the Chinese 'game console' Subor Z is probably quite similar to what we will see in next gen consoles. 

[GOTSpectrum]

9 minutes ago, Trixanity said:

Fact remains that AMD needs to significantly expand their APU business.

also, the rain we aren't seeing large APUs is memory bandwidth. I think with ddr5 things could change. But feeding 6-8 cores along with a GPU over dual channel DDR4 just isn't all that feasible.

[Morgan MLGman]

I actually work in very large data centers and I've spoken to various people and administrators about using AMD for servers. It's true that some people did not even know that AMD has a great new enterprise-grade CPU lineup and that those CPUs bring many benefits such as lower TCO (total cost of ownership) which is caused by the lower power consumption of those CPUs as well as allowing for smaller form factors etc.

One guy actually told me that they chose Intel because of "support", but when I asked what support Intel gave him for the past 4-5 years he couldn't find a single example. I think with the second generation EPYC might start gaining track in enterprise environments because it had its first tests and more companies start using it.

It just takes time in this branch of IT, so I wouldn't be too upset with those figures just yet.

[Trixanity]

21 minutes ago, Ben Quigley said:

I doubt this to some degree, the fact both consoles have such similar sillicon proves that they don't have total exclusivity. So if they can sell don't and Microsoft such similar sillicon there's no reason they couldn't sell a slight variation to other vendors. Like the Apu that is used in the Chinese 'game console' Subor Z is probably quite similar to what we will see in next gen consoles. 

The console APUs aren't just a CPU, GPU and some memory slapped together. Lots of special sauce is used. There are differences between them despite the first glance making them appear largely the same.

 

The next gen consoles will be significantly different to Fenghuang. We already know that.

 

Just to be clear. There is certainly room to use the R&D spent on console chips for other things but they'd have to change some things regardless of what they do. And if I'm not mistaken it's Sony and Microsoft ordering the chips from TSMC, not AMD, so it's AMD selling a design to them and then getting kickbacks from retail sales.

[Guest]

Great work AMD. Now pull your shit together with your graphics cards.

[cj09beira]

58 minutes ago, Trixanity said:

AMD really needs to bring up their APU release schedule. It's a bit too far behind for my taste. Right now we appear to be looking at a early 2020 for 7nm while Intel is ready with 10nm Ice Lake for holidays this year. So they kinda need it out the door in Q3 to really nail Intel to the cross.

for now they cant they need time to make things right, apus on mobile are harder to do than desktop cpus, as they need to get a cpu and a gpu working right and together, and they need oems to be on board, which for now have mostly not been great,

so they first release the product they control then latter on work on the apus, it sucks to consumers to have to wait but thats all amd can do, amd also has to balance it with radeon launches as that also needs dev time 

[Trixanity]

35 minutes ago, cj09beira said:

for now they cant they need time to make things right, apus on mobile are harder to do than desktop cpus, as they need to get a cpu and a gpu working right and together, and they need oems to be on board, which for now have mostly not been great,

so they first release the product they control then latter on work on the apus, it sucks to consumers to have to wait but thats all amd can do, amd also has to balance it with radeon launches as that also needs dev time 

It's ultimately AMD that's losing. It's a high volume market that they're not competing properly in.

[GOTSpectrum]

3 minutes ago, Trixanity said:

It's ultimately AMD that's losing. It's a high volume market that they're not competing properly in.

But realistically it's not a high volume market. People are mostly happy with either a standard iGPU or a CPU+GPU situation. 

 

The APU would have to offer something vastly superior, significantly lower cost or much better battery life, which at the moment, due to limited OEM interest there's not going to be the vastly lower cost.

 

Also the fact that it chews up your system memory is something that puts many people off APUs.

[cj09beira]

4 minutes ago, Ben Quigley said:

But realistically it's not a high volume market. People are mostly happy with either a standard iGPU or a CPU+GPU situation. 

 

The APU would have to offer something vastly superior, significantly lower cost or much better battery life, which at the moment, due to limited OEM interest there's not going to be the vastly lower cost.

 

Also the fact that it chews up your system memory is something that puts many people off APUs.

the is talking about the laptop market, which can become a very large market, though penetration will be dificult, 7nm would help.

 

10 minutes ago, Trixanity said:

It's ultimately AMD that's losing. It's a high volume market that they're not competing properly in.

sure its not ideal, but amd only has a limited amount of money and time, and they choose to focus in the diy and server markets first 

[GOTSpectrum]

6 minutes ago, cj09beira said:

the is talking about the laptop market, which can become a very large market, though penetration will be dificult, 7nm would help.

I'm aware,

 

What I was trying to say is the current AMD offerings sorry cover the vast majority of the laptop market.

 

The percentage of laptops that have higher than  4 cores and GPU performance grater than that what is on offer from AMD is quite small.

 

In the end the laptop market is dominated by cheap 300-500USD devices.

[BigDamn]

Zen2 will really determine how far AMD is going to take this. If Ryzen 3000 is as good as the rumors make it out to be (ha), Intel will have a tough time making up the ground it's lost.

 

Heh, watch AMD takeover the CPU market and Intel takeover the GPU market. Wouldn't that be some fun fuckery

[GOTSpectrum]

9 minutes ago, BigDamn said:

Zen2 will really determine how far AMD is going to take this. If Ryzen 3000 is as good as the rumors make it out to be (ha), Intel will have a tough time making up the ground it's lost.

 

Heh, watch AMD takeover the CPU market and Intel takeover the GPU market. Wouldn't that be some fun fuckery

Even if it isn't as good as the rumours. Let's say they get just a clockspeed bump up to 4.5 with an efficiency bump meaning you can have 70% more cores, that are higher clocked got the same power. That will still massively eat into intel if the price is right.

 

On top of this AMD really only need one improvement for data center and HPC, and that's FPU/AVX performance. If they can match intel on that, clock for clock they're golden. They already have more IO and more density and lower TCO(total cost of ownership) due to lower power and initial cost. You have to remember that if you can save 25w, you are not only savings 25w on the chip, but also needing 25w less cooling potential, and we already know cooling is one of, if not the most expensive part of running a data center. Also most server grade hardware is usually clocked 2-3ghz for max efficiency so zen can already do that with ease.

 

But in the end they only manage an Intel style a 2-5% IPC gain, along with the FPU improvements I mentioned above and the clockspeed/efficiency improvements of the 7nm node they are on to a winner. Throw on top of that the potential 12-15% IPC gain any well it's going to become very interesting. 

[Fnige]

3 hours ago, Ben Quigley said:

Or the other option just take the next gen console APUs, tweet as little as possible and sell them for high end laptops.

THE PORTABLE XBOX _{(or playstation)}

[strajk-]

As someone that comes from the Automation Industry (Process Control System Developer), I can confirm that the process of Enterprise and Server is very slow in comparison to the average Client and Workstation.

 

For example, the rule of thumb whenever a project has been accepted in that field, has always been "we will only know the outcome in two years", so whatever you do, you will only know if it is a success or not after a minimum of two years.

 

And quite honestly AMD has been doing fine in all segments, just recently did they start to get results/contractors in both Server and Enterprise.

[mr moose]

1 minute ago, strajk- said:

As someone that comes from the Automation Industry (Process Control System Developer), I can confirm that the process of Enterprise and Server is very slow in comparison to the average Client and Workstation.

 

For example, the rule of thumb whenever a project has been accepted in that field, has always been "we will only know the outcome in two years", so whatever you do, you will only know if it is a success or not after a minimum of two years.

And if it ain't broke don't fix it, ergo servers and process controllers don't get upgraded until they are literally holding production back.

[strajk-]

2 minutes ago, mr moose said:

And if it ain't broke don't fix it, ergo servers and process controllers don't get upgraded until they are literally holding production back.

Oh yes definitely, once one has been deployed they won't change them, unless there are plans to further improve upgrade the facility (which actually happens quite often).

The contractors sell everything in a package, hardware, software, control systems, controllers and everything around it, and those contractors do prefer picking up whatever is both cheaper and reliable.

 

Personally I have been recommending AMD to our clients (except for portable panels due to lack of good APUs), and they seem to have been doing the same to their peers too, so it wouldn't surprise me if in one or two more year AMD will dominate in sales, specially considering the shortage + price gauge on Intel side.

[Doobeedoo]

It will take time in that space though. Some companies wanted to see how first gen was. Also waiting for new gen with twice the cores especially. 

[BuckGup]

23 hours ago, exetras said:

The Server or Enterprise market is slow to move. It's going to take time for AMD to penetrate it.

 

The last time AMD was part of the server market they left it with a generation of Opterons that could not compete with Intel and nothing to upgrade to. this made for a painful and expensive upgrade.

 

If Intel cant keep up with the Chip demand AMD could take a bite out of their market. Also people are still talking about Specter and all those vulnerabilities and the performance loss due to the fixes.

A lot of the new HPC being built are using EPYC CPUs now