Intel's CPU market increased

[maflynn]

Intel grabs CPU market share back from AMD for first time in 3 years

Mercury Research’s results for the fourth quarter of 2020 show that AMD lost some of its x86 market share despite having a good year overall. In desktop, the company fell from 20.1 percent in the previous quarter to 19.3 percent

I do think the situation is temporary I think Gamer Nexus hit it on the head of the nail, AMD, is focusing on growing its server segment, it has contracts to provide sufficient quantities for the PS5 and Xbox, and TMSc can only produce so many units. To put it another way, AMD is spread thin right now, and that permitted Intel to grab some marketshare.

 

[jrichards1408]

Is AMD venturing into the M1/arm CPU segment?

 

[maflynn]

Is AMD venturing into the M1/arm CPU segment?

What do you mean by that?

 

[Erehy Dobon]

An arguably better analysis of the current AMD-Intel CPU situation was presented a few days ago by AnandTech's Cutress:

Best CPUs for Gaming: February 2024

www.anandtech.com

who often writes from a non-gaming perspective.

Unlike Intel and Nvidia, AMD is shipping massive amounts of APUs alongside CPUs and GPU, notably as SoCs in the 2 (2.5?) newly launched videogame consoles: Xbox Series X (and Series S) and PlayStation 5.

AMD is constrained by putting all of their eggs in one basket -- TSMC's 7nm process node -- which is why they are getting squeezed really tight right now versus better availability of CPUs in summer 2020 (the Amazon.com top CPU list showed this).

The Ryzen 3700X is currently going for $45 more than MSRP/SEP from August 2020. The newer 5600X can be found on the street around $330; I happened to score one at the $299 MSRP last month.

However, this is all spilling into other component pricing. Motherboard pricing is unappealing and the graphics card situation is appalling.

The ghetto $65 Radeon RX 550 2GB card I bought in August as a cheapo GPU (since my CPUs don't have integrated graphics) are currently going for $180-200.

Cutress's analysis is interesting in that many of his recommendations from 2020 (at various price points) are no longer valid due to recent price increases.

 

[I7guy]

Don't know if it's temporary or not. With the hype surrounding the 11th gen and Alder Lake, the tide may be swinging back toward Intel.

Although, for internet bragging rights, Intel has to beat AMD in single core performance in benchmarks.

 

[lcseds]

Maybe I have not seen enough reports but you have to go to the Ryzen 9 series to come close to, or beat the i7-10700K. I don't think the measurements on various websites are consistent while taking "core count" into consideration. Depends on the website as to whether the Intel is slower than Ryzen in any test. I think if you want one processor to beat another, you can find a website that tests one way or another.

 

[maflynn]

Don't know if it's temporary

I think Intel still needs to up their game, and while their newer offerings are a step in the right direction, I think AMD has the momentum at the moment.

They were caught with their pants down and sat on the 14nm technology for far too long. I think they probably really didn't feel the need to invest in actually fixing or getting to 10nm, and that allowed AMD to sneak in and over take them

 

[pshufd]

Don't know if it's temporary or not. With the hype surrounding the 11th gen and Alder Lake, the tide may be swinging back toward Intel.

Although, for internet bragging rights, Intel has to beat AMD in single core performance in benchmarks.

I've looked at Rocket Lake and am quite unimpressed. But I'm not a gamer. I've looked at Alder Lake and am not impressed so far. Let's see benchmarks and let's see if they can actually deliver for 2021. Their track record on schedules have been poor for the past decade.

 

[I7guy]

I agree Intel has been napping, either through arrogance or ignorance and AMD snuck up and put Intel to shame at least in the court of public benchmarks. We know what Intel needs to do, and whether Rocket Lake or Alder Lake is enough grab back some of what they let AMD take, is anybody's guess.

Having said that, for the pc builder, when choosing a cpu today there are many options, many system boards, many combos and it's enough to make one's head explode.

 

[Erehy Dobon]

Rocket Lake will not solve Intel's problems. It's still on the 14nm node. Intel has stayed competitive with AMD in recent years by tossing efficiency out the window.

Because they are on such a large old process node, the transistors generate huge amounts of heat. The large die sizes generally limit Intel from introducing consumer desktop CPUs with more than 8 cores.

Alder Lake -- on the 10nm process -- is the first Intel desktop CPU generation to give them a chance at recouping some of the ground they have squandered to AMD.

Intel's real chance at true progress is Meteor Lake (scheduled for 2023) which is expected to be produced on Intel's 7nm process as well as TSMC's 5nm process.

Only the naive believe Intel is going to win everything back in one generation. I expect AMD to hold onto their lead for the next 3-4 years. Intel's best chance of catching up to AMD is if the latter stumbles (just like Intel did). Not so sure that will happen under Dr. Lisa Su's leadership.

Intel lost its crown because of years of poor leadership. Any fan of team sports can empathize.

 

[Erehy Dobon]

Having said that, for the pc builder, when choosing a cpu today there are many options, many system boards, many combos and it's enough to make one's head explode.

As someone who had a twenty-year hiatus between custom PC builds I don't find it particularly difficult.

You always start with your usage case and budget. From there you prioritize features. Memory support, I/O speed, networking, thermals, power, etc.

There are a lot more choices today than twenty years ago but by being clear with your usage case and budget, you can easily eliminate over half of the motherboards and CPUs out there. If you throw in thermals, you will probably eliminate 85% of the CPUs.

It is not that difficult if you have a clear idea about what you are trying to accomplish. "What software do I want to run? How well does it need to run? What do I want to pay for this?" These are the same guiding questions that I asked myself in the late Nineties. I'm sure they're the same questions NASA scientists asked in the Sixties.

I built nothing but Intel PCs in the late Nineties and switching to AMD last year wasn't particularly confusing or weird.

While the number of motherboards and CPUs have definitely increased some decisions have basically vanished. I don't have to worry about NIC (Network Interface Cards) or soundcard decisions. Same with external I/O since USB has standardized pretty much everything. Storage decisions are "you get what you pay for".

With Microsoft's WHQL system, it's far easier to get reliably stable videocard drivers.

LOL, perhaps the biggest decision in 2021 is "To bling ? or not to bling ??? That is the question." (Addressable RGB or not.)

?

 

[pshufd]

As someone who had a twenty-year hiatus between custom PC builds I don't find it particularly difficult.

You always start with your usage case and budget. From there you prioritize features. Memory support, I/O speed, networking, thermals, power, etc.

There are a lot more choices today than twenty years ago but by being clear with your usage case and budget, you can easily eliminate over half of the motherboards and CPUs out there. If you throw in thermals, you will probably eliminate 85% of the CPUs.

It is not that difficult if you have a clear idea about what you are trying to accomplish. "What software do I want to run? How well does it need to run? What do I want to pay for this?" These are the same guiding questions that I asked myself in the late Nineties. I'm sure they're the same questions NASA scientists asked in the Sixties.

I built nothing but Intel PCs in the late Nineties and switching to AMD last year wasn't particularly confusing or weird.

What's difficult is getting parts. Especially if you want a 5900X or 5950X.

 

[pshufd]

Rocket Lake will not solve Intel's problems. It's still on the 14nm node. Intel has stayed competitive with AMD in recent years by tossing efficiency out the window.

Because they are on such a large old process node, the transistors generate huge amounts of heat. The large die sizes generally limit Intel from introducing consumer desktop CPUs with more than 8 cores.

Alder Lake -- on the 10nm process -- is the first Intel desktop CPU generation to give them a chance at recouping some of the ground they have squandered to AMD.

Intel's real chance at true progress is Meteor Lake (scheduled for 2023) which is expected to be produced on Intel's 7nm process as well as TSMC's 5nm process.

Only the naive believe Intel is going to win everything back in one generation. I expect AMD to hold onto their lead for the next 3-4 years. Intel's best chance of catching up to AMD is if the latter stumbles (just like Intel did). Not so sure that will happen under Dr. Lisa Su's leadership.

Intel lost its crown because of years of poor leadership. Any fan of team sports can empathize.

Intel will have to execute on Alder Lake. In the meantime, AMD isn't standing still. They're talking about 28% to 40% IPC gains in Zen 4.

 

[I7guy]

As someone who had a twenty-year hiatus between custom PC builds I don't find it particularly difficult.

You always start with your usage case and budget. From there you prioritize features. Memory support, I/O speed, networking, thermals, power, etc.

There are a lot more choices today than twenty years ago but by being clear with your usage case and budget, you can easily eliminate over half of the motherboards and CPUs out there. If you throw in thermals, you will probably eliminate 85% of the CPUs.

It is not that difficult if you have a clear idea about what you are trying to accomplish. "What software do I want to run? How well does it need to run? What do I want to pay for this?" These are the same guiding questions that I asked myself in the late Nineties. I'm sure they're the same questions NASA scientists asked in the Sixties.

I built nothing but Intel PCs in the late Nineties and switching to AMD last year wasn't particularly confusing or weird.

While the number of motherboards and CPUs have definitely increased some decisions have basically vanished. I don't have to worry about NIC (Network Interface Cards) or soundcard decisions. Same with external I/O since USB has standardized pretty much everything. Storage decisions are "you get what you pay for".

With Microsoft's WHQL system, it's far easier to get reliably stable videocard drivers.

LOL, perhaps the biggest decision in 2021 is "To bling ? or not to bling ??? That is the question." (Addressable RGB or not.)

?

Prior to the Nov/Dec build I last built a pc in 2009ish. The choice then was easy. Was very confusing not having followed the market in many years. (I probably should have asked a question, just like maflynn did, at least to get started, but did not think of it.)

I was "confused" about why I would want an AMD vs Intel given my use case. In truth, I could have flipped a coin. I ended up getting a 10900K, but it was for my own "bragging" rights, definitely overkill for my use case, which is a couple of games(doom eternal and MSFS) and MS office productivity tools. I could have save a couple of bucks an gotten an 10700K, which would have been equally as fine for my needs.

But you are correct about other things like sound cards, wifi, ethernet and USB has standardized almost everything in terms of connectors.

But I still believe it can be confusing to the newbie given the dizzying array of processors and boards.

 

[Erehy Dobon]

What's difficult is getting parts. Especially if you want a 5900X or 5950X.

High-end CPUs, some high-end motherboards and all GPUs are currently difficult to acquire. Choosing them isn't difficult, it's just the current availability. This was not the case even as late as September-October 2020.

The difficult decision for a system builder today is whether not to cave into the scalpers and throw the budget out the window. Or to exercise patience and possibly wait anywhere from 6-12 months before you can get your hands on your desired components. Of course, there will be newer, better parts so it'll be an agonizing chase for components for the next 2-3 years.

Semiconductor demand outstrips supply by an estimated 30% and the industry can't just flip a switch and double 7nm part production.

I put together my first recent build (the NZXT H1 case) in June of last year, then moved to a second build which was mostly finished by late August. There were articles on some of the PC sites saying "Now is a terrible time to buy a graphics card." They were SO wrong. I paid $499 (full MSRP) for the reference Nvidia GeForce RTX 2070 SUPER. Today used cards are easily going double that price.

I've resigned to the reality that I probably won't get my paws on an RTX 3080 Founder's Edition card at MSRP in 2021. By 2022, maybe something else will be the hot item (like a 3080 SUPER). I am damned lucky to have my 2070 SUPER which is enough for ultra 1440p gaming right now.

 

[pshufd]

High-end CPUs, some high-end motherboards and all GPUs are currently difficult to acquire. Choosing them isn't difficult, it's just the current availability. This was not the case even as late as September-October 2020.

The difficult decision for a system builder today is whether not to cave into the scalpers and throw the budget out the window. Or to exercise patience and possibly wait anywhere from 6-12 months before you can get your hands on your desired components. Of course, there will be newer, better parts so it'll be an agonizing chase for components for the next 2-3 years.

Semiconductor demand outstrips supply by an estimated 30% and the industry can't just flip a switch and double 7nm part production.

I put together my first recent build (the NZXT H1 case) in June of last year, then moved to a second build which was mostly finished by late August. There were articles on some of the PC sites saying "Now is a terrible time to buy a graphics card." They were SO wrong. I paid $499 (full MSRP) for the reference Nvidia GeForce RTX 2070 SUPER. Today used cards are easily going double that price.

I've resigned to the reality that I probably won't get my paws on an RTX 3080 Founder's Edition card at MSRP in 2021. By 2022, maybe something else will be the hot item (like a 3080 SUPER). I am damned lucky to have my 2070 SUPER which is enough for ultra 1440p gaming right now.

I chose the i7-10700 as that was a good fit for me from Intel but I found out later that the 5900x would have been an even better fit. I just didn't know how far ahead AMD was.

I'm not buying from scalpers. There's always even the risk that you'll get a brick instead of a chip. If the chip is defective, there may be no real vendor warranty.

Right now it's a race between AMD and Apple for me. If Apple introduces an M1X, I'll go with that as I really want macOS. If AMD can deliver parts (CPU, GPU, MB), then I could go with them and do a macOS VM or hack. I'd expect either solution to last between five and ten years.

President Biden wants a commission to study supply chain issues. Intel and AMD want R&D dollars. I don't know if we can add Fabs in the US but it would be years before anything came online. So we have shortages. Similar to vaccines. I'll just keep my eyes open for parts. Maybe the thing to do is buy them when they are for sale from a legitimate seller. Then build a stockpile of parts to do a build when enough parts come in. The GPU situation is ridiculous. At least Intel parts are available on the CPU side.

 

[Erehy Dobon]

But I still believe it can be confusing to the newbie given the dizzying array of processors and boards.

One big difference for system builders between the late Nineties and today is the great coverage and tutorials that are now available. Twenty years ago there were a handful of websites (Tom's Hardware, Anandtech, a few others that shut down) and some paper magazines (like Maximum PC). Online tutorials on building a custom PC were very few and far between.

Back then you REALLY needed to do your homework to avoid pounding your head against a brick wall for hours trying to get a $30 NIC to work on Red Hat Linux.

 

[pshufd]

Back then you REALLY needed to do your homework to avoid pounding your head against a brick wall for hours trying to get a $30 NIC to work on Red Hat Linux.

I did that a long time ago trying to get the network up on an Oracle Enterprise Linux installation. The solution was to get a the small list of supported NICs and then buy one.

 

[lcseds]

I just ordered a i7-10700KF and an Asus Z590 system board from Amazon at normal pricing. Plus Samsung SSD, memory, etc. The only thing I cannot get right now is a video card. I will have to use the GTX 1070 from the old build. It will do for the time being.

 

[Erehy Dobon]

Then build a stockpile of parts to do a build when enough parts come in.

This is what I am doing. I have two functional builds. I have two incomplete builds that are expensive doorstops.

Periodically I buy a new component when it goes on sale but I've already noticed that RAM prices have increased about +25-30% in the past couple of months. My guess is that NAND flash chips will also increase in price as the new generation of videogame consoles will eat up more of these.

I do realize that whatever I buy/stockpile today will be last year's tech shortly.

 

[pshufd]

This is what I am doing. I have two functional builds. I have two incomplete builds that are expensive doorstops.

Periodically I buy a new component when it goes on sale but I've already noticed that RAM prices have increased about +25-30% in the past couple of months. My guess is that NAND flash chips will also increase in price as the new generation of videogame consoles will eat up more of these.

I do realize that whatever I buy/stockpile today will be last year's tech shortly.

It might also fetch a nice profit it you sell the parts.

If I buy Zen 3 and it takes six months to get the parts, then it's only another nine months to Zen 4. This is why it may just be easier to buy a prebuilt M1X system. No fuss, no muss. But no pride of construction.

I saw a monitor review - 27 inch Dell 4k vs LG Ultrafine. The reviewer said that the Dell was on the expensive side. It was $650 - I was shocked. I bought an earlier generation for $350 two years ago. The model I bought is also $650. Computer components are crazy expensive today. It very well might be easier to just buy a 27 inch iMac given that the LG Ultrafine 27 inch 5K monitor alone us $1,300.

I saw a used Dell Inspiron that I was considering bidding on a few weeks ago. He wanted $450 but I was trying to work on a different deal which fell through. The system had a GTX 1060 and those go for well over $200 these days. I could have tossed in an old video card to sell the system and keep the 1060 for a new build. Crazy what goes through my mind when it's hard to get parts.

 

[Erehy Dobon]

While I do need another CPU, I'm less concerned about that. Normal gaming doesn't require a lot of CPU horsepower, everything I do is completely GPU bound.

That's an important reason why I stuck with 65W TDP processors (an 8-core 3700X and a 6-core 5600X).

I have two great PC builds and my Mac mini 2018 (and a pedestrian 10th gen Intel Windoze notebook that I don't even count). I don't NEED another PC at this time but it would be nice to have that RTX 3080 because in the not-too-distant future I'd like to upgrade my living room TV to a 4K LG model.

The GPU situation is appalling.

 

[pshufd]

While I do need another CPU, I'm less concerned about that. Normal gaming doesn't require a lot of CPU horsepower, everything I do is completely GPU bound.

That's an important reason why I stuck with 65W TDP processors (an 8-core 3700X and a 6-core 5600X).

I have two great PC builds and my Mac mini 2018 (and a pedestrian 10th gen Intel Windoze notebook that I don't even count). I don't NEED another PC at this time but it would be nice to have that RTX 3080 because in the not-too-distant future I'd like to upgrade my living room TV to a 4K LG model.

The GPU situation is appalling.

I don't do gaming outside of Chess. I want something with string single-core performance for responsiveness and something that can run highly inefficient trading software quickly. I prefer 65 Watt CPUs but will go up to 105 - I would seldom use a lot of power but want it available when I need it. I have a GTX 1050 Ti and that's enough for my needs.

 

[Erehy Dobon]

LOL, I run Fidelity Active Trader Pro on a PC I haven't even mentioned here: a ghetto Apollo Lake mini PC that I picked up for $160-170 (I forget) with Intel HD 500 graphics. It's definitely unimpressive but since I'm not day trading (I mostly use it to monitor up to four securities) it does its job at a good price. I think the integrated GPU uses 1.5GB of the total 6GB RAM.

 

[pshufd]

LOL, I run Fidelity Active Trader Pro on a PC I haven't even mentioned here: a ghetto Apollo Lake mini PC that I picked up for $160-170 (I forget) with Intel HD 500 graphics. It's definitely unimpressive but since I'm not day trading (I mostly use it to monitor up to four securities) it does its job at a good price. I think the integrated GPU uses 1.5GB of the total 6GB RAM.

I run 15 RT charts with multiple studies per chart - it hunts for signals for me. It normally uses 13% of the i7-10700. It takes about five minutes to startup. I also use Think or Swim displaying 88 real-time charts and it uses about 2% of the i7-10700. Graphics isn't an issue with these programs - CPU is. Think or swim will use up to 24 GB of RAM if you configure it for that much and I know people that use a Mac Pro with a couple of high-resolution monitors to do complicated studies.