AMD chipping away at Intel's server CPU market share as it sees record growth

nanoguy

Posts: 1,355   +27
Staff member
The big picture: The overall CPU market saw the largest year-over-year increase in shipments in the last 25 years, with Intel gaining some x86 market share in the desktop space while losing some of its grip on the server segment to AMD and its Epyc CPU series.

Intel is stuck between a rock and a hard place, with the first being the dominating force of Arm in the mobile space and the second being an ambitious AMD that has gradually shed its underdog skin to become a serious contender in desktop and laptop CPUs, and an increasingly competitive rival in the data center as well. The Lisa Su-powered AMD saw a record first quarter with almost double the revenue compared to the period last year.

According to analysis by Mercury Research, AMD saw the fastest growth in server sales in over fifteen years, taking a strong bite into Intel's dominance in hyperscale computing hardware. Specifically, it grew to just under nine percent market share on the back of strong Epyc CPU sales. President of Mercury Research Dean McCarron notes this was a major contributor to the increase in average selling price for AMD CPUs.

This would also explain, at least in part, why Intel's Data Center Group delivered less than stellar results for the first three months of the year, down 20 percent compared to the same quarter of 2020.

Likewise, the first quarter of 2021 was not as successful for AMD in areas like consumer CPUs for laptops and desktops. The company has been struggling to meet demand as it depends on TSMC at a time when the foundry has a growing backlog of orders and hopes to overcome component shortages sometime in 2022. Intel capitalized on this by pricing its entry-level CPUs more aggressively and effectively grabbing some market share back from AMD for the first time in three years.

The Mercury Research report notes that the CPU market has seen significant momentum over the last year as every quarter registers a new record in the total amount of sales.

The first quarter of the year usually sees lower sales volume, but the first quarter of 2021 saw a 41 percent year-over-year increase, the highest recorded for the CPU market in 25 years.

Chipmakers are the biggest winners in this story, with global semiconductor revenue projected to grow to $522 billion in 2021 despite the industry's chip shortage.

Permalink to story.

 
41bc4e6bdded27e9cb501320a98f3ec4e5ee1be04445dde56dd091943de2c562.gif
 
Long term this is where AMD should be focusing their efforts imo: data centers. Let's not forget that while right now they mostly only need to think of intel as the more immediate competitor, ARM is trying to get into the market too so not letting up means they'll be better positioned if ARM gets more competitive.
 
The only reason that EPYC's slaughtering of Xeon is taking so long is that the turnover rate in the data centre space isn't exactly lightning-quick. There's a LOT of cost involved in changing hardware at that level so nobody wants to do it more quickly than is really necessary.
 
I'm all for competition in the market pushing new architectures and one upping each other. Faster computing will benefit everyone from the top of the stack to bottom.
 
The only reason that EPYC's slaughtering of Xeon is taking so long is that the turnover rate in the data centre space isn't exactly lightning-quick. There's a LOT of cost involved in changing hardware at that level so nobody wants to do it more quickly than is really necessary.
Also, you will be surprised as to how little the people that make the decisions know about hardware.

They know Intel but they dont know AMD, so they will buy Intel regardless of being a worse purchase in all aspects.

There are legitimate cases, like Avid certification of a server chip, for example or the programs already locked into Nvidia CUDA.
 
Last edited by a moderator:
Also, you will be surprised as to how little the people that make the decisions know about hardware.

They know Intel but they dont know AMD, so they will buy Intel regardless of being a worse purchase in all aspects.
I believe that this may have been true 20 years ago but I don't believe it now. They KNOW the AMD name, everyone in the tech industry does. Remember that for YEARS, the most powerful supercomputer in the world was the Cray TITAN which used Opteron CPUs before the Chinese ARM-based supercomputers surpassed it. For years after that, the TITAN was still the most powerful supercomputer in the USA. That makes the AMD name VERY well known in the server space and these managers go to companies like Cray (Now HP) or Dell, tell them what they want and/or need and leave it to the experts to design a solution.

These corporations rely on getting the best performance-per-watt and the engineers that design these machines are VERY aware of what's out there and what each company brings to the table. The old wisdom of name recognition hasn't applied in a VERY long time, far longer than the existence of Ryzen. In fact, it was the AMD Opteron SERVER CPU that caused Intel's monopolistic hold on the industry to break.

The AMD Opteron was so superior as a server CPU compared the Intel Xeon at the time that Michael Dell himself told Intel to stick it and started building servers with Opterons because his competitors were getting HUGE design wins over him simply because he was only using Xeons. Of course, by that time, the damage had been done and AMD couldn't continue, despite having the superior product. Michael Dell had waited too long to stop enabling Intel's criminal activities and the Opteron died as a result.

Make no mistake, the AMD Opteron was a VERY successful CPU during its time but AMD almost went bankrupt and couldn't compete with Intel at that level anymore. Hell, they could barely compete at the consumer level, but we all now know very well why that happened (and it's wasn't because of AMD).

A lack of name recognition hasn't been applicable for decades.
 
The only reason that EPYC's slaughtering of Xeon is taking so long is that the turnover rate in the data centre space isn't exactly lightning-quick. There's a LOT of cost involved in changing hardware at that level so nobody wants to do it more quickly than is really necessary.

Yep, my office W10 VM still runs off Haswell Xeons and there's been no indication that hardware will be replaced anytime soon. Loooong turnover times.

Edit: just checked and they're Broadwell. But they're still not going anywhere.
 
Last edited:
I believe that this may have been true 20 years ago but I don't believe it now.
Oh trust me and believe it. Wont name my company, but multi-billion media corp.

We were on a meeting discussing several aspects and you have to see everyone's eyes when I suggested AMD systems.

My division has a certain level of independence, so I started moving to AMD, but in the end, the higher ups are the ones that can come back and say "company standard is Intel" and I'm done there.

Edit Found this, which applies perfectly to our convo:

309518855b4497d1613f97832686cbaf443f2c26e5c683b5377457bfc33904b8.jpg
309518855b4497d1613f97832686cbaf443f2c26e5c683b5377457bfc33904b8.jpg
 
Last edited by a moderator:
The only reason that EPYC's slaughtering of Xeon is taking so long is that the turnover rate in the data centre space isn't exactly lightning-quick. There's a LOT of cost involved in changing hardware at that level so nobody wants to do it more quickly than is really necessary.
This is true. In addition, I feel some companies are also trying to dip their toes into AMD hardware from a testing perspective just to see how well and reliably it performs, relative to their existing setup. So I feel if AMD is able to keep up with the pressure on Intel, its possible to see an accelerated pickup rate over time, assuming that AMD can maintain the supply. But one thing is for sure looking at recent earning results, and that is AMD is starting to gain acceptance in every PC segment.
 
Yep, my office W10 VM still runs off Haswell Xeons and there's been no indication that hardware will be replaced anytime soon. Loooong turnover times.

Edit: just checked and they're Broadwell. But they're still not going anywhere.
Of course not. This is because when they DO upgrade, it will be literally hundreds, if not thousands of CPUs across the entire corporation all at once. That's a lot of money to drop at one time so they make sure it's worth it.

Truth be told, they often go overboard with how long they wait to upgrade and cost the corporation extra money as a result of the reduced efficiency caused by computers that are too slow. It's a terrible practice but an inefficiency cost is much easier to hide from ignorant corporate presidents and/or shareholders.
 
Oh trust me and believe it. Wont name my company, but multi-billion media corp.

We were on a meeting discussing several aspects and you have to see everyone's eyes when I suggested AMD systems.

My division has a certain level of independence, so I started moving to AMD, but in the end, the higher ups are the ones that can come back and say "company standard is Intel" and I'm done there.

Edit Found this, which applies perfectly to our convo:

309518855b4497d1613f97832686cbaf443f2c26e5c683b5377457bfc33904b8.jpg
309518855b4497d1613f97832686cbaf443f2c26e5c683b5377457bfc33904b8.jpg
That's a sad state of affairs but I can assure you that not all companies are like that (as shown by AMD's extremely rapid market penetration). I suppose that as long as ignorant boomers hold the reins of these corporations, technology will never truly be taken seriously. Not all boomers are tech-morons (I should know because I was originally taught by one who wasn't) but the vast majority of them are.

As that generation disappears and is replaced by more Gen-X'ers, computer technology will be taken FAR more seriously and seen as the critical part of doing ANY business that it really is. Remember that back in the day, people who understood computers were considered abnormal and referred to as "nerds" but now, people who don't understand computers are considered abnormal and called "dinosaurs".

Your corporation's "Intel Standard" won't stand if your competition is able to leverage the advantage that AMD offers to outperform yours. No corporation that has a CEO who is so arrogant and rigid that they can't see past their own noses does well, no matter how big they are. Once they stop innovating, it's only a matter of time before they lose their relevance. This is because in the modern business world, complacency is the first symptom of death.

Corporations that once dominated huge industries have learnt this the hard way like IBM, GM, Ford, Chrysler and Blockbuster. As much as I hate Apple, I've always had to hand it to them that they never stopped innovating. Intel stopped innovating in the CPU market for a period of less than ten years and look what happened. If Intel hadn't become involved in things like SSDs, networking, etc. it's very possible that Intel would be in the same predicament that AMD was in 2015. Even though Intel is still remarkably profitable, they've become a laughing stock for most of the tech industry and their CEO paid the price for his lack of dedication to innovation.
 
This is true. In addition, I feel some companies are also trying to dip their toes into AMD hardware from a testing perspective just to see how well and reliably it performs, relative to their existing setup. So I feel if AMD is able to keep up with the pressure on Intel, its possible to see an accelerated pickup rate over time, assuming that AMD can maintain the supply. But one thing is for sure looking at recent earning results, and that is AMD is starting to gain acceptance in every PC segment.
Absolutely. One thing that I think we don't realise enough is just how many people don't even know what CPU they have in their PCs and craptops. If they were sold an AMD-based or Intel-based machine, they wouldn't know. They just know that it "plays YouTube". :laughing:
 
Of course not. This is because when they DO upgrade, it will be literally hundreds, if not thousands of CPUs across the entire corporation all at once. That's a lot of money to drop at one time so they make sure it's worth it.

Truth be told, they often go overboard with how long they wait to upgrade and cost the corporation extra money as a result of the reduced efficiency caused by computers that are too slow. It's a terrible practice but an inefficiency cost is much easier to hide from ignorant corporate presidents and/or shareholders.
Won't mention the name of my company but I see the loss of productivity and increase in support costs every day. I shake my head at some of the insanely old computers we still have running. Most of the employees could get far more work done on a new lower end machine with dual monitors and an ssd, and save the company more money in a month than they would spend in the new equipment. Try explaining that to a manager who can barely operate a computer though and their eyes glaze over. The incompetence in management in these large corporations is maddening.
 
I personally think that AMD is having some 'issues' at the 7nm node. and wonder how that fares for the 5nm node. Their Epyc sales are back ordered, which means that some will not be sold as companies refuse to wait. There is no gen 4 Threadripper, which is usually a modified server chip, the mods being mostly higher frequencies. Can they not get the frequencies to scale? Who knows. They DO keep a tight lid on leakage. (Latest rumor - Threadripper, code named Chagall (!) in August)

IBM has developed the tools for 2nm nodes... Who will be the first to byte? (Pun intended)

And, there are still the issues that TSMC themselves are facing. These will not affect Intel.

I am NOT raining on AMD's parade, just pointing out that which should be obvious to those that can see.
 
I personally think that AMD is having some 'issues' at the 7nm node. and wonder how that fares for the 5nm node. Their Epyc sales are back ordered, which means that some will not be sold as companies refuse to wait. There is no gen 4 Threadripper, which is usually a modified server chip, the mods being mostly higher frequencies. Can they not get the frequencies to scale? Who knows. They DO keep a tight lid on leakage. (Latest rumor - Threadripper, code named Chagall (!) in August)

IBM has developed the tools for 2nm nodes... Who will be the first to byte? (Pun intended)

And, there are still the issues that TSMC themselves are facing. These will not affect Intel.
There are no 7nm issues, other than capacity is constrained for several reasons. Gen4 Threadripper is not around since there is not really need for it (3rd gen is more than good enough) and since 7nm is on shortage, AMD rather use chiplets for Epyc. Simply put: capacity is problem, nothing else is.
 
Back