It’s now almost 8 years since AMD revealed Ryzen, and Intel still can’t beat it.
That feels a slight bit unfair.
For non-gaming workloads, they’re basically sitting on par or better because of the giant pile of e-cores, and for single-threaded performance (on p-cores) they’re also on par to slightly ahead.
Sure, the x3d chips are the gaming kings and no argument here, but that’s not moving volume - even AMD is all-in on the datacenter side because their gaming/consumer side sales have evaporated into nothingness.
Intel’s problem isn’t an inability to design CPUs that are competitive, it’s an inability to create production-ready processes that are competitive with TSMC.
At some point they’re going to have to decide if spending endless billions on processes that aren’t competitive is the best use of their resources. Owning the ability to make your product is super important, but for certain market segments (client desktop and laptop) maybe going ‘fuck it’ and fabbing on the best process you can find so that your CPUs come out competitive is probably the way to go - and, honestly, is pretty much already what they’ve done with ARL.
I’d also maybe agree that the pricing is an issue: they’re not industry-leading anymore but they’ve kept that pricing which almost immediately makes them less appealing than AMD if you don’t need something Intel is offering you (like the accelerators in scalable Xeons or whatever). ARL immediately made me go ‘How much? What the bleep?’ when they announced pricing, because uh, they’re way off on what they really should be asking.
You make some good points. But I do want to make some comments.
non-gaming workloads, they’re basically sitting on par or better
The testing I’ve seen, the CPU does deliver decent performance numbers for productivity. but it does that at too high TDP, using above the MAX rate of 250 Watt, and despite the better node, it still has worse PPW.
The reviews I’ve seen rate the CPU from flop to meh, with some saying it’s released to early, because the platform is simply buggy.
So the Ultra 9 285K delivers slightly better in productivity than the Ryzen 9950X, but it does that at higher power consumption, despite the better TSMC 3nm production node. Where AMD is made on 4nm.
So it does have some wins, but I’ll maintain Intel doesn’t beat Ryzen overall, and Intel is only achieving this on the back of outside higher end production than AMD is using.
Intel’s problem isn’t an inability to design CPUs that are competitive, it’s an inability to create production-ready processes that are competitive with TSMC.
I’ll still say Intel is a bit behind on the design side, but yes manufacturing is where Intels future will probably be decided.
In the past, Intel always had the advantage of superior production, and could always power through squeezing out a bit extra from both design and process technology.
But with Arrow Lake, they fail to surpass AMD despite e newer gen process.
I agree 100% regarding the pricing, but from Intels viewpoint, they are selling the Ultra 9 285K CPU similar in productivity performance to Ryzen 9950X at about the same price. At least here in Denmark in retail they are very close with Ryzen being $30 cheaper. Other markets may have different prices. But for gaming the price is absolutely horrific 50% more expensive than the 7800X3D!!
Intel got caught off guard by the rise of advanced packaging, where AMD’s chiplet design could actually compete with a single die (while having the advantage of being more resilient against defects, and thus higher yield).
Intel fell behind on manufacturing when finFETs became the standard. TSMC leapfrogged Intel (and Samsung fell behind) based on TSMC’s undisputed advantage at manufacturing finFETs.
Those are the two main areas where Intel gave up its lead, both on the design side and the manufacturing side. At least that’s my read of the situation.
Intel got caught off guard by the rise of advanced packaging,
For servers and workstation yes, but the first 2 Ryzen generations were still 1 chip packages for desktop. And Intel lost marketshare quicker to AMD on desktop than on servers.
So while packaging was an important factor in higher end, Intel actually lost marketshare a lot slower on servers, where the packaging was the biggest benefit for AMD.
On desktop I think Intel lost because they’d stayed on 4 cores for too long. I remember personally complaining that my freaking phone had more cores than my desktop.
So I jumped to AMD as soon as I could afford the upgrade. 😀
On workstations AMD absolutely trounced Intel with Threadripper. Threadripper was amazing, a giant improvement of workstations, it’s sad that they don’t prioritize that segment anymore.
too high TDP, using above the MAX rate of 250 Watt
Agreed. Intel’s design philosophy seems to be ‘space heater that does math’ for some reason. That’s been true since at least 10th gen, if not before then. I don’t know if it’s just chasing benchmark wins at any cost, or if they’re firmly of the opinion that hot and loud is fine as long as it’s fast and no customers will care - which I don’t really think is true anymore - or what, but they’ve certainly invested heavily in CPUs that push the literal limits of physics while trying to cool them.
Intel always had the advantage of superior production
That really stopped being true in the Skylake era when TSMC leapfrogged them and Intel was doing their 14nm++++++++ dance. I mean they did a shockingly good job of keeping that node relevant and competitive, but the were really only relevant and competitive on it until AMD caught up and exceeded their IPC with Ryzen 3000.
about the same price
Yeah, if gaming is your use case there’s exactly zero Intel products you should even be considering. There’s nothing that’s remotely competitive with a 7800x3d, and hell, for most people and games, even a 5800x3d is overkill.
And of course, those are both last-gen parts, so that’s about to get even worse with the 9800x3d.
For productivity, I guess if you’re mandated to use Intel or Intel cpus are the only validated ones it’s a choice. But ‘at the same price’ is the problem: there’s no case where I’d want to buy Intel over AMD if they cost the same and perform similarly, if for no other reason than I won’t need something stupid like a 360mm AIO to cool the damn thing.
Absolutely, the 14nm process was leading when it was new, but the delays and ultimate failure of 10nm caused Intel to fall way behind. But before that from the very beginning of integrated circuits, Intel was the leader in manufacturing. From the late 70’s Intel when Intel made the i8086 they achieved an economic advantage, that enabled them to stay ahead pretty much consistently in manufacturing.
In 2016 TSMC achieved parity with their 10nm equivalent to Intel 14nm with maybe a slight advantage over Intel, and after that it’s well known that TSMC continued quickly improving past the points where Intel had failed, and TSMC became the leader.
I should have written always prior to 2016. Because it’s 8 years ago now, but before that, Intel had stayed on top for half a century. Despite for instance M68000 and Alpha were way better processor designs than anything Intel had.
there’s no case where I’d want to buy Intel over AMD if they cost the same and perform similarly,
I agree, the only reason I quote this, is because of the insane change in how Intel vs AMD is viewed compared to before Ryzen! Compared to AMD FX series, the Intel Core and Core2 were so superior, it was hard to see how AMD could come back from that. But when Ryzen was presented late 2016 it was clear to me they had something new and exciting. And they really elevated desktop performance after years of minor iterations from Intel.
Compared to AMD FX series, the Intel Core and Core2 were so superior, it was hard to see how AMD could come back from that.
Yup, an advantage in this industry doesn’t last forever, and a lead in a particular generation doesn’t necessarily translate to the next paradigm.
Canon wants to challenge ASML and get back in the lithography game, with a tooling shift they’ve been working on for 10 years. The Japanese “startup” Rapidus wants to get into the foundry game by starting with 2nm, and they’ve got the backing of pretty much the entirety of the Japanese electronics industry.
TSMC is holding onto finFET a little bit longer than Samsung and Intel, as those two switch to gate all around FETs (GAAFETS). Which makes sense, because those two never got to the point where they could compete with TSMC on finFETs, so they’re eager to move onto the next thing a bit earlier while TSMC squeezes out the last bit of profit from their established advantage.
Nothing lasts forever, and the future is always uncertain. The past history of the semiconductor industry is a constant reminder of that.
True, but with AMD the problem was that they had serious deficits and were near bankruptcy.
And these technologies are getting more an more expensive. The latest tapeout Apple did for M3, is estimated to have cost $1 billion. That’s for tapeout alone!!!
We are at a point where this technology is so prohibitively expensive, that only the biggest global players can play.
The thing is, if Intel doesn’t actually get 18A and beyond competitive, it might be on a death spiral towards bankruptcy as well. Yes, they’ve got a ton of cash on hand and several very profitable business lines, but that won’t last forever, and they need plans to turn profits in the future, too.
That feels a slight bit unfair.
For non-gaming workloads, they’re basically sitting on par or better because of the giant pile of e-cores, and for single-threaded performance (on p-cores) they’re also on par to slightly ahead.
Sure, the x3d chips are the gaming kings and no argument here, but that’s not moving volume - even AMD is all-in on the datacenter side because their gaming/consumer side sales have evaporated into nothingness.
Intel’s problem isn’t an inability to design CPUs that are competitive, it’s an inability to create production-ready processes that are competitive with TSMC.
At some point they’re going to have to decide if spending endless billions on processes that aren’t competitive is the best use of their resources. Owning the ability to make your product is super important, but for certain market segments (client desktop and laptop) maybe going ‘fuck it’ and fabbing on the best process you can find so that your CPUs come out competitive is probably the way to go - and, honestly, is pretty much already what they’ve done with ARL.
I’d also maybe agree that the pricing is an issue: they’re not industry-leading anymore but they’ve kept that pricing which almost immediately makes them less appealing than AMD if you don’t need something Intel is offering you (like the accelerators in scalable Xeons or whatever). ARL immediately made me go ‘How much? What the bleep?’ when they announced pricing, because uh, they’re way off on what they really should be asking.
You make some good points. But I do want to make some comments.
The testing I’ve seen, the CPU does deliver decent performance numbers for productivity. but it does that at too high TDP, using above the MAX rate of 250 Watt, and despite the better node, it still has worse PPW.
The reviews I’ve seen rate the CPU from flop to meh, with some saying it’s released to early, because the platform is simply buggy.
So the Ultra 9 285K delivers slightly better in productivity than the Ryzen 9950X, but it does that at higher power consumption, despite the better TSMC 3nm production node. Where AMD is made on 4nm.
So it does have some wins, but I’ll maintain Intel doesn’t beat Ryzen overall, and Intel is only achieving this on the back of outside higher end production than AMD is using.
I’ll still say Intel is a bit behind on the design side, but yes manufacturing is where Intels future will probably be decided.
In the past, Intel always had the advantage of superior production, and could always power through squeezing out a bit extra from both design and process technology.
But with Arrow Lake, they fail to surpass AMD despite e newer gen process.
I agree 100% regarding the pricing, but from Intels viewpoint, they are selling the Ultra 9 285K CPU similar in productivity performance to Ryzen 9950X at about the same price. At least here in Denmark in retail they are very close with Ryzen being $30 cheaper. Other markets may have different prices. But for gaming the price is absolutely horrific 50% more expensive than the 7800X3D!!
Intel got caught off guard by the rise of advanced packaging, where AMD’s chiplet design could actually compete with a single die (while having the advantage of being more resilient against defects, and thus higher yield).
Intel fell behind on manufacturing when finFETs became the standard. TSMC leapfrogged Intel (and Samsung fell behind) based on TSMC’s undisputed advantage at manufacturing finFETs.
Those are the two main areas where Intel gave up its lead, both on the design side and the manufacturing side. At least that’s my read of the situation.
For servers and workstation yes, but the first 2 Ryzen generations were still 1 chip packages for desktop. And Intel lost marketshare quicker to AMD on desktop than on servers.
So while packaging was an important factor in higher end, Intel actually lost marketshare a lot slower on servers, where the packaging was the biggest benefit for AMD.
On desktop I think Intel lost because they’d stayed on 4 cores for too long. I remember personally complaining that my freaking phone had more cores than my desktop.
So I jumped to AMD as soon as I could afford the upgrade. 😀
On workstations AMD absolutely trounced Intel with Threadripper. Threadripper was amazing, a giant improvement of workstations, it’s sad that they don’t prioritize that segment anymore.
Agreed. Intel’s design philosophy seems to be ‘space heater that does math’ for some reason. That’s been true since at least 10th gen, if not before then. I don’t know if it’s just chasing benchmark wins at any cost, or if they’re firmly of the opinion that hot and loud is fine as long as it’s fast and no customers will care - which I don’t really think is true anymore - or what, but they’ve certainly invested heavily in CPUs that push the literal limits of physics while trying to cool them.
That really stopped being true in the Skylake era when TSMC leapfrogged them and Intel was doing their 14nm++++++++ dance. I mean they did a shockingly good job of keeping that node relevant and competitive, but the were really only relevant and competitive on it until AMD caught up and exceeded their IPC with Ryzen 3000.
Yeah, if gaming is your use case there’s exactly zero Intel products you should even be considering. There’s nothing that’s remotely competitive with a 7800x3d, and hell, for most people and games, even a 5800x3d is overkill.
And of course, those are both last-gen parts, so that’s about to get even worse with the 9800x3d.
For productivity, I guess if you’re mandated to use Intel or Intel cpus are the only validated ones it’s a choice. But ‘at the same price’ is the problem: there’s no case where I’d want to buy Intel over AMD if they cost the same and perform similarly, if for no other reason than I won’t need something stupid like a 360mm AIO to cool the damn thing.
Absolutely, the 14nm process was leading when it was new, but the delays and ultimate failure of 10nm caused Intel to fall way behind. But before that from the very beginning of integrated circuits, Intel was the leader in manufacturing. From the late 70’s Intel when Intel made the i8086 they achieved an economic advantage, that enabled them to stay ahead pretty much consistently in manufacturing.
In 2016 TSMC achieved parity with their 10nm equivalent to Intel 14nm with maybe a slight advantage over Intel, and after that it’s well known that TSMC continued quickly improving past the points where Intel had failed, and TSMC became the leader.
I should have written always prior to 2016. Because it’s 8 years ago now, but before that, Intel had stayed on top for half a century. Despite for instance M68000 and Alpha were way better processor designs than anything Intel had.
I agree, the only reason I quote this, is because of the insane change in how Intel vs AMD is viewed compared to before Ryzen! Compared to AMD FX series, the Intel Core and Core2 were so superior, it was hard to see how AMD could come back from that. But when Ryzen was presented late 2016 it was clear to me they had something new and exciting. And they really elevated desktop performance after years of minor iterations from Intel.
Yup, an advantage in this industry doesn’t last forever, and a lead in a particular generation doesn’t necessarily translate to the next paradigm.
Canon wants to challenge ASML and get back in the lithography game, with a tooling shift they’ve been working on for 10 years. The Japanese “startup” Rapidus wants to get into the foundry game by starting with 2nm, and they’ve got the backing of pretty much the entirety of the Japanese electronics industry.
TSMC is holding onto finFET a little bit longer than Samsung and Intel, as those two switch to gate all around FETs (GAAFETS). Which makes sense, because those two never got to the point where they could compete with TSMC on finFETs, so they’re eager to move onto the next thing a bit earlier while TSMC squeezes out the last bit of profit from their established advantage.
Nothing lasts forever, and the future is always uncertain. The past history of the semiconductor industry is a constant reminder of that.
True, but with AMD the problem was that they had serious deficits and were near bankruptcy.
And these technologies are getting more an more expensive. The latest tapeout Apple did for M3, is estimated to have cost $1 billion. That’s for tapeout alone!!!
We are at a point where this technology is so prohibitively expensive, that only the biggest global players can play.
The thing is, if Intel doesn’t actually get 18A and beyond competitive, it might be on a death spiral towards bankruptcy as well. Yes, they’ve got a ton of cash on hand and several very profitable business lines, but that won’t last forever, and they need plans to turn profits in the future, too.