AMD Gains Ground in Data Center, Laptop CPU Markets (tomshardware.com) 49
AMD increased its market share in data center and laptop CPU segments during Q2 2024, according to a new report from Mercury Research. The company captured 24.1% of the data center CPU market, up 0.5% from the previous quarter and 5.6% year-over-year. In laptops, AMD's share rose to 20.3%, a 1% increase quarter-over-quarter and 3.8% year-over-year. The company's revenue share in laptops reached 17.7%, indicating lower average selling prices compared to Intel.
Intel maintained its overall lead, controlling 78.9% of the client PC market. In desktops, Intel gained 1% share, now holding 77% of the market. AMD's data center revenue share hit 33.7%, suggesting higher average selling prices for its EPYC processors compared to Intel's Xeon chips. AMD earned $2.8 billion from 24.1% unit share, while Intel made $3.0 billion from 75.9% unit share.
Intel maintained its overall lead, controlling 78.9% of the client PC market. In desktops, Intel gained 1% share, now holding 77% of the market. AMD's data center revenue share hit 33.7%, suggesting higher average selling prices for its EPYC processors compared to Intel's Xeon chips. AMD earned $2.8 billion from 24.1% unit share, while Intel made $3.0 billion from 75.9% unit share.
Not surprising (Score:2)
Intel hasn't had a truly competitive datacentre product in a very long time. They live off niche applications where their on-package accelerators can make a difference and on discounts to select customers (which these days is pretty much anyone that asks for a discount). It's getting to the point that Intel TCO is so bad that nobody wants to touch their hardware. While Emerald Rapids was a step in the right direction, it was far too late to market to matter (and was little more than an update to Sapphire
Re: okay (Score:2)
Re: (Score:3)
Right now, all of Intel's 13th and 14th generation desktop chips that use more than 65 watts of power have a bad habit of self destructing:
https://www.theverge.com/24216... [theverge.com]
AMD's security issue can probably be fixed with a microcode update, but Intel's latest issue is probably going to require a recall of certain chip models. Not good for the bottom line.
Re: 'Sinkclose' Flaw in Hundreds of Millions of AM (Score:1)
Re: (Score:2)
AMD claims they can patch that, and it's it's a management engine flaw (rather than a sidechannel attack), it likely won't come with a performance penalty.
Anyone who examined the situation realized that you need bare metal access and the ability to tamper with the UEFI to carry out the exploit, making it extremely difficult to carry off.
Re: (Score:3)
Intel is practically giving away hardware at this point to maintain marketshare. It isn't working. Once they start having to sell dcg products that require EUV equipment, Intel is cooked. They have a major volume problem for Intel 3 and 20a, and it isn't clear if things will improve for 18a.
Re: okay (Score:2)
Last I looked, AMD CPUs were still cheaper than the Intel parts they compete with. But I haven't been looking at DC parts... Are their server CPUs really marked down that much right now?
Re: (Score:2)
Officially, no. But if you are buying from a preferred vendor or are buying in bulk as a hyperscalar (think ODM channels), you will get a major markdown. All you really have to do is ask for a discount. Intel's dcg margins are awful, and have been for awhile now.
About the only way to pay full price for a Xeon is to place is all rregular orders through retail channels.
Re: (Score:2)
That was meant to say "small irregular orders".
Re: (Score:2)
Last I looked, AMD CPUs were still cheaper than the Intel parts they compete with. But I haven't been looking at DC parts... Are their server CPUs really marked down that much right now?
Virtualisation has really been the big driver towards Intel in the DC, if you have a hypervisor farm, they need to be the same type of proc, you cant mix Intel and AMD... You shouldn't even be mixing too many different generations of CPUs, you can but you'll be limited to the features of the oldest CPU. As for who's the best, AMD has always had a strong offering but Intel has always had the lions share of the sales.
As for desktop, AMD have had the best bang for buck for ages. For my latest gaming boxen (
Re: (Score:1)
Intel still outsells 3:1
You would imagine that market share like that would be highly lucrative. Intel profits should be soaring.
Have you seen Intel losses? The stock price? $19.34.
They've cut prices so hard they're going broke. And AMD is still making gains.
Pretty bad place to be.
Re: (Score:2)
You can't compare it to anyone else's stock price, except by comparative performance.
Intel's revenue is over double AMDs, and they don't have a vibrant GPU business on top of that.
There is no indication that they're "handing out CPUs for free and going broke"
Re: (Score:2)
Look at Intel's quarterly statements for the last year or two and get back to us.
Re: (Score:2)
The argument was that they're not making any money- they're making shit tons. Hand over fist. Twice as much as AMD, and change, without a vibrant GPU market.
In terms of profit, Intel has never done terribly well, because it is always dumping Billions into potential projects that generally end up being dead ends, or too early for their time.
Re: (Score:2)
Re: (Score:2)
Chipzilla is going to be fine.
Intel's had some bad years, but they're so large they actually can weather it. They're in no danger of disappearing. AMD, though, was. The only reason we didn't see AMD disappear was because Intel pawned Sony and Microsoft to AMD for the PS4/5 and Xbox.
AMD's marketshare is still relatively small. They've made gains, yes, but it's not like Intel hasn't been in a position where AMD had the superior product (the K-5 through to Athlon series, for example). And Ryzen is wildly succ
Re: (Score:2)
Chipzilla is going to be fine.
I agree. Intel is a strategic asset for the US. If worse comes to worse the US government will prop up Intel. Intel won't be allowed to fall.
But they're in a bad place, as I've said. They really must execute on the "Intel 3" node, on-time and without any horrendous missteps. That could get them out of the dive their currently in.
I have hope. I believe in Gelsinger and that what's happening now is the consequence of years of R&D neglect and other mismanagement he's actively correcting. The onl
Re: (Score:2)
Intel 3 is here and it has one product: Sierra Forest. Intel is skipping over it for most client CPUs and all mainline Xeons, using it only in the low-volume Sierra Forest and the uninspiring upcoming client product Arrow Lake-U (and even the it's only used in the compute tile).
For reference, Sierra Forest launched in April. Unless you had a prior contract with Intel requiring delivery on a deadline, you couldn't get one. You still can't buy one anywhere.
Re:Not surprising (Score:5, Interesting)
The solution to Intels problems is to spin off their fabs, To stop being a vertically integrated company. For their fabs to become rent-a-fabs. For the cpu designers the freedom to use the best fabs no matter who is running them.
But with all that, the brand Intel becomes the next Motorola, just an asset thats traded around.
Re: (Score:1)
But with all that, the brand Intel becomes the next Motorola
You mean AMD, I'm sure.. Right?
Because Intel has a major fabrication problem- that can't be denied... but its CPU architecture is still *fucking excellent*, comparatively speaking.
Motorola, they are not.
Re: Not surprising (Score:2)
"but its CPU architecture is still *fucking excellent*, comparatively speaking."
Compared to what? Certainly not AMD.
Re: (Score:1)
You always say this stupid shit, and then refuse to back it up with anything but an "uhhh, because I know these things"
Currently, they're neck-and-neck in a performance per clock metric.
In non-top-of-the-line chips, efficiency is about comparable, with some Intel parts taking the win, and some AMD parts taking the win- ultimately, coming down to both of them have highly efficient parts that they punish to hit performance benchmarks they've selected for sale.
Re: Not surprising (Score:3)
"In non-top-of-the-line chips, efficiency is about comparable,"
In their so-called top of the line chips Intel is only barely matching AMD by using vastly more power and running at a much higher temperature, reducing CPU lifespan. We do not measure performance with budget CPUs like you are doing here as an Intel fanboy. Now run along and play in traffic.
Re: (Score:1)
Now run along and play in traffic.
Lame response. Try harder. He's still right about you.
Re: (Score:1)
Lame response. Try harder. He's still right about you.
Not trying hard to try hard here, friend. I think we'll have to disagree.
Re: (Score:2)
In their so-called top of the line chips Intel is only barely matching AMD by using vastly more power and running at a much higher temperature, reducing CPU lifespan.
They're both doing this.
Both the 14900K and the 7950X only lose about 20% of their performance for 50% of their total power usage. That means both chips are being cranked way past what is sane for the sake of hitting a marketing mark, as I said in the first reply.
We do not measure performance with budget CPUs like you are doing here as an Intel fanboy. Now
There is no we on your side of the fence.
I make purchasing decisions, and unsurprisingly, the market seems to agree with me.
While AMD continues to make gains, they're still a minority player in the datacenter, and not by a small amount.
As I im
Re: (Score:2)
That means both chips are being cranked way past what is sane for the sake of hitting a marketing mark, as I said in the first reply.
Eh?
Or you know people want fast chips and are prepared to pay the power to get it. I got a 7950x3d and I don't want an arbitrary mark in efficiency, I want it fast.
Re: (Score:2)
Or you know people want fast chips and are prepared to pay the power to get it.
That's not mutually exclusive.
I got a 7950x3d and I don't want an arbitrary mark in efficiency, I want it fast.
The argument can be made for some_intel_here as well.
Re: (Score:2)
That's not mutually exclusive.
You dais they're being cranked way past sane: I want it as fast. That's not insane.
The argument can be made for some_intel_here as well.
This isn't about AMD vs Intel so much as being cranked way past what's sane. I've got a 4090 in this machine so the power use is often through the roof. Cranking the CPU way up and it running quite hot is definitely worth it.
Re: (Score:2)
You dais they're being cranked way past sane: I want it as fast. That's not insane.
Fair.
I'll caveat the claim with, "way past sane for the purposes of evaluating the efficiency of the microarchitecture"
This isn't about AMD vs Intel so much as being cranked way past what's sane. I've got a 4090 in this machine so the power use is often through the roof. Cranking the CPU way up and it running quite hot is definitely worth it.
I don't disagree with you in the slightest. Trading efficiency for horsepower is a perfectly valid desire.
Keep in mind that the argument you jumped in on had context. Particularly:
In their so-called top of the line chips Intel is only barely matching AMD by using vastly more power and running at a much higher temperature, reducing CPU lifespan.
Re: (Score:2)
OK fair enough.
Re: (Score:2)
The power curves being similar on these things is no coincidence, ya dumbshit. Of course they try to make the fastest chips and of course things land about where the market will bear the added cost of the lower yields inherent in the binning of the best parts.
And the 65W desktop market is very price sensitive, while the laptop market has bigger drivers like gpu and display quality
Re: (Score:1)
By "marketing mark" do you by chance actually just mean "market" ??
lol.
There's a reason both Intel and AMD have marketed new chips as a rebrand of the old, just clocked a bit higher and taking significantly more power.
They're setting the market, not you. It's cute that you think you are, though.
The power curves being similar on these things is no coincidence, ya dumbshit.
That was literally my point, ya dumbshit.
Of course they try to make the fastest chips and of course things land about where the market will bear the added cost of the lower yields inherent in the binning of the best parts.
Again, my point. If you ha actually read the conversation, that was my reasoning for discluding the top-of-the-line chips from comparison.
And the 65W desktop market is very price sensitive, while the laptop market has bigger drivers like gpu and display quality
Objection: speculation.
Further, how in the fuck is that relevant at all? We weren't even talking abou
Re: (Score:1)
Re: (Score:2)
Re: (Score:2)
We're talking datacentre here, and Intel is in the shutter. Emerald Rapids is slower than Genoa and uses more power. Turin and Turin-dense are nearing release. Intel is screwed.
Re: (Score:2)
See: SQL benchmarks.
There's a reason we're still using Intel.
Emerald Rapids actually surpassed Genoa (Score:2)
I wouldn't say Emerald Rapids was just a step in the right direction, it's the first time in quite a while that Intel leaps ahead of AMD in data centre performance. Sapphire Rapids was basically as fast as Milan and came close to the time AMD released the much faster Genoa. But a year later, Emerald Rapids is actually faster than AMD's last years release. That is as long as you have a modern image with the latest drivers - I tried Google's high clock rate c4 Emerald Rapids with an old CentOS webserver image
Re: (Score:2)
Where do you get that crap?!? In no way did Emerald Rapids outperform Genoa except in some biased Intel slide.
Re: (Score:2)
We have Genoa and Emerald Rapids servers in our cloud (GCP), they run side by side so I can compare performance.
Re: (Score:2)
https://www.phoronix.com/revie... [phoronix.com]
https://www.tomshardware.com/p... [tomshardware.com]
Re: (Score:2)
Yeah, those are rather meaningless. Of course a 96 core EPYC is faster than a 64 core Emerald Rapids, by a lot. Duh.. But in the cloud we pay per vCPU not per processor! So the only 3 things I care about are single-core performance (because there are tasks, like web server requests, which cannot be further parallelised), multi-thread *per vCPU* and price *per vCPU*. I have been using Google's c4 since they were in private preview early this year. At first they were underperforming in multi-thread and it tur
Re: (Score:2)
Of course a 96 core EPYC is faster than a 64 core Emerald Rapids, by a lot. Duh..
Funny enough, not in coherent workloads.
Install Postgres, and watch your 96 core EPYC struggle to keep up with your 64 core Xeon.
This is because the cross-CCD coherency cost is ~130ns, or double and change that of a Xeon.
Work within a CCD, or work that doesn't care much about coherency at all does well on the EPYC, though.
Re: (Score:2)
This is because the cross-CCD coherency cost is ~130ns, or double and change that of a Xeon. Work within a CCD, or work that doesn't care much about coherency at all does well on the EPYC, though
Well, yes, because EPYC's L3 is then much faster if your workload fits in that CCD - it's a classic design choice that offers a benefit along with a drawback. Genoa doubled the window each core gets, so that improves things, but Postgres is indeed the one major benchmark often cited as an example of that drawback the EPYC L3 design brings. I am not sure if it's other databases too, we use mysql and it is not affected the same way AFAICT (well... it is one of the less "sophisticated" RDBMSs), but in general
Re: (Score:2)
We see about 2:1 performance there on a *rapids vs $your_epic_here for any amount of threads greater than a CCD.
Re: (Score:2)
And that if you need 96 cores of firepower that isn't doing a pretty limited set of work (RDBMS, large scale software traffic forwarding and analysis [DPDK]) then the EPYCs are a superior offering.
If you are doing one of those workloads, EPYC really kind of sucks. In a world where the Xeon and EPYC had a similar price/core, then there would simply be no contest. The Xeon is the better part. Unfortunately, Intel isn't stupid
Re: (Score:2)
Intel hasn't had a truly competitive datacentre product in a very long time.
This is a very interesting claim. Many would believe that Intel hasn't had a product that could compete in a technological sense, and I wouldn't necessarily disagree. However, the financial history is overwhelming that Intel has dominated the last decade in terms of dollar sales. AMD has never been close. And given the choice between dominating technology or sales, I assume that any company in their right mind would always choose sales.
Simply thinking that Intel's overwhelming dominance even during a mi