Intel Letting AMD Lead?

[phoon]

I'm first and foremost a fan of competition. We all know who's leading in all around performance and I'm happy to see AMD make a comeback. I've been daily driving an i7-6700k for a long time now (since launch day) and its time to upgrade. If I had the budget right now, I would 100% go AMD (5900X) I really don't see any reason to go Intel or to wait for tiger lake.

 

This is were I got curious, hearing about how tiger lake mobile CPUs "leaks" show that Intel is still behind (Not that I follow it much) got the Intel fan boy inside me thinking: "Is Intel letting AMD lead? Do they want this to happen for 'healthy competition'?"

 

Interesting discussion topic:

 

Is Intel purposefully not releasing consumer CPUs that are on par with AMD's (for some marketing reason) or are they ACTUALLY behind and there's nothing they can do about it?  

 

Also, if you think that Intel is falling behind because AMD is advancing faster than them, why do you think this is happening? Is it a managerial issue, or is the engineering team lacking?

 

Do you disagree with my statement entirely?

 

I'm interested to hear. 

[Eigenvektor]

3 minutes ago, phoon said:

Is Intel purposefully not releasing consumer CPUs that are on par with AMD's (for some marketing reason) or are they ACTUALLY behind and there's nothing they can do about it?

I don't think it would be a wise marketing decision to purposely let your competitor "win" or take away your market share. It will reduce your income and the ability to compete in the future. Of course this is Intel, they have a huge amount of money already, so it's not going to hurt them immediately but it's still not something I can see them doing on purpose.

 

Intel botched the transition to 10 nm, which is why they're still stuck at 14 nm for the most part. I'd say its impressive they can still keep up as well as they are. Once they transition nodes properly I'm sure they'll be back in the game.

[The_russian]

4 minutes ago, phoon said:

Is Intel purposefully not releasing consumer CPUs that are on par with AMD's (for some marketing reason) or are they ACTUALLY behind and there's nothing they can do about it?  

I don't think it makes sense for Intel to be hiding some better CPUs right now and instead intentionally releasing worse products. It doesn't make much sense for a business to intentionally not make as much money as they can. You specified for a marketing reason, I'm sure you've seen what Intel's reputation has become the last few years, it wouldn't make sense for them to do that intentionally.

 

7 minutes ago, phoon said:

Also, if you think that Intel is falling behind because AMD is advancing faster than them, why do you think this is happening?

In my opinion as someone who is not very informed on the technical aspects of how CPUs are designed and improved, it seems like complacency. AMD wasn't much of a competition for a while there and Intel probably got complacent with releasing minor improvements instead of pushing as hard as they could for better performance and efficiency.

[aDoomGuy]

They are behind and there are several reasons for that. Two of them are that Intel have been too arrogant and complacent also another reason is that they doesn't care too much about consumer CPU. Intel's real market is server CPU and they are still doing ok there I think.

 

100% managerial I think. Resting too much on their throne without competition and suddenly boom. The revolution.

[HelpfulTechWizard]

Itnel has fallenfallen behind. They sat on top for like 15 years, and forgot what competition was.

[Eigenvektor]

2 minutes ago, aDoomGuy said:

…also another reason is that they doesn't care too much about consumer CPU. Intel's real market is server CPU and they are still doing ok there I think.

The server market moves a lot slower than the consumer market. Intel has a huge amount of trust, which is why companies use them despite having worse performance. It'll take some time before AMD has gained enough reputation and more companies start to use them in servers (it's also possible Intel has better support and/or other features that make them a better choice for enterprise still).

[mahyar]

well intel should gain some market share back IF alder lake (12th gen) is as promised

[mahyar]

Just now, Eigenvektor said:

The server market moves a lot slower than the consumer market. Intel has a huge amount of trust, which is why companies use them despite having worse performance. It'll take some time before AMD has gained enough reputation and more companies start to use them in servers (it's also possible Intel has better support and/or other features that make them a better choice for enterprise still).

like that server that linus talked about

[porina]

15 minutes ago, phoon said:

Is Intel purposefully not releasing consumer CPUs that are on par with AMD's (for some marketing reason) or are they ACTUALLY behind and there's nothing they can do about it?  

The very simple matter of it is that Intel's manufacturing plans fell apart somewhere around 2015 and they've not managed to fully recover since. Remember tick tock? The idea was that you'd alternate between generations of improving manufacturing process, and the other more focused on architecture. Broadwell, Intel's first 14nm product, was the first to show signs of problems. It was very late that was a limited release just so they could say they did it at all with only two mainstream desktop CPUs released on that architecture (more HEDT ones were available). Skylake came along and everyone forgot about it. Then... today's problems really hit, which is why we're still on 14nm on desktop today. 10nm was totally broken. So they've had no choice but to rehash 14nm products since then. That's not to say there are no improvements made to 14nm, but you can't get as big a jump as with moving to smaller nodes.

 

It is flat out wrong to say Intel were arrogant, complacent or similar. They simply could not release a much better desktop product if they wanted to for most of that time. Only recently with Tiger Lake have they had a 10nm product that could be somewhat competitive on the desktop, if it wasn't for their limited production capability. 10nm is so late now, it makes no sense for them to pour billions into increasing capacity, when 7nm is going to replace it by the time they'd have 10nm running. They have the designs, just not the ability to make them. I also think many people on this forum fail to appreciate how much work is needed to put a CPU design on silicon. Different fabs have very different design rule sets. It isn't like sending a job to a printer. It needs to be laid out for the particular fab which is no trivial work.

 

Rocket Lake will be interesting, as it will be the first new architecture on desktop since Skylake. It is a catch up generation. I don't think anyone really expects them to take leadership, and Intel themselves don't either. It was earlier in the year, where the CEO gave out a longer view. These things take time to implement, and it can still go wrong.

 

AMD on the other hand, had everything they tried go right in the same period. They bet on TSMC which worked out really well. They're aggressively scaling their designs. There is no guarantee of that, but it happened. No one can predict the future.

[Happiness_is_Key]

39 minutes ago, Eigenvektor said:

Intel botched the transition to 10 nm, which is why they're still stuck at 14 nm for the most part. I'd say its impressive they can still keep up as well as they are. Once they transition nodes properly I'm sure they'll be back in the game.

Good point. Could it be that they're heavily looking into the 7nm lithography process to come back in? I'm not sure if you can go straight from 14nm to 7nm, but if you could, you might also be able to go to 5nm as well. That's a BIG jump, I know.

 

I don't follow CPU rumors that much at all, though, I might get into it. Also, what's up with AMD's documentation on their website? Is it just because they're 'new' or a redesign to the site? Long time Intel user here....

[Eigenvektor]

3 minutes ago, Happiness_is_Key said:

Good point. Could it be that they're heavily looking into the 7nm lithography process to come back in? I'm not sure if you can go straight from 14nm to 7nm, but if you could, you might also be able to go to 5nm as well. That's a BIG jump, I know.

Afaik Intel's 10 nm process is very comparable to TSMC's 7 nm process. The nanometer label has become a marketing term over the years, so you can't really use the numbers to compare between different fabs. Not sure they'll try to switch nodes again before they ironed out the issues they had.

 

I don't remember the exact details, but I think Intel tried to also switch to EUV and introduce different material (InGaAs?) on top of going from 14nm to 10 nm which didn't turn out as intended. So basically they bet big on their new process and it didn't pan out, which is what stumped them for a long time. So I don't think its fair to say they're not trying to compete at all. They bet on the wrong horse and now they're paying the price.

 

15 minutes ago, Happiness_is_Key said:

I don't follow CPU rumors that much at all, though, I might get into it. Also, what's up with AMD's documentation on their website? Is it just because they're 'new' or a redesign to the site? Long time Intel user here....

I'd say Intel is simply more "professional" about it. Their online documentation for CPUs is great. Probably because they are much bigger on the enterprise side (and have more resources).

[Happiness_is_Key]

18 minutes ago, Eigenvektor said:

Afaik Intel's 10 nm process is very comparable to TSMC's 7 nm process. The nanometer label has become a marketing term over the years, so you can't really use the numbers to compare between different fabs. Not sure they'll try to switch nodes again before they ironed out the issues they had.

 

Oh, cool! I'm still learning here.... I know Apple uses TSMC which probably takes up quite a bit of their manufacturing capacity, but I heard they have their own dedicated facility for that. I know AMD is trying to secure one in regards to the shortages (besides the scalpers) of everything like that. So, Intel does have their own branch for lack of a better word for manufacturing (meaning they don't invest in a semiconductor manufacturing company)?

Quote

I don't remember the exact details, but I think Intel tried to also switch to EUV and introduce different material (InGaAs?) on top of going from 14nm to 10 nm which didn't turn out as intended. So basically they bet big on their new process and it didn't pan out, which is what stumped them for a long time. So I don't think its fair to say they're not trying to compete at all. They bet on the wrong horse and now they're paying the price.

Good on Intel though on trying to innovate (to be clear, not saying AMD doesn't). It's just interesting to see everyone kind of trying their own thing.

 

Quote

I'd say Intel is simply more "professional" about it. Their online documentation for CPUs is great. Probably because they are much bigger on the enterprise side (and have more resources).

Got it. Another great point. When checking out AMD, I usually have to refer to a different site like wikichip.org for example for specs on specific things. I just like to dig into the details,, haha.  Thank you!

[porina]

34 minutes ago, Happiness_is_Key said:

Good point. Could it be that they're heavily looking into the 7nm lithography process to come back in? I'm not sure if you can go straight from 14nm to 7nm, but if you could, you might also be able to go to 5nm as well. That's a BIG jump, I know.

The current understanding is that 10nm is so delayed, basically if they were to ramp it up it'll conflict with the 7 nm introduction which is a separate development project, and not tied to 10nm working. So they just need 10nm working enough to help them out until 7nm comes online. News earlier this year was that 7nm is also delayed, but it is hard to tell if it is just something relatively minor, or if it is 10nm all over again.

 

Intel 5nm hasn't been openly discussed as far as I'm aware, but they'll surely be working on it in some way already so that it will be ready when the time comes in future years. Keep in mind it will be the process going against what will be TSMC 3nm due to their naming differences.

[Happiness_is_Key]

3 minutes ago, porina said:

The current understanding is that 10nm is so delayed, basically if they were to ramp it up it'll conflict with the 7 nm introduction which is a separate development project, and not tied to 10nm working. So they just need 10nm working enough to help them out until 7nm comes online. News earlier this year was that 7nm is also delayed, but it is hard to tell if it is just something relatively minor, or if it is 10nm all over again.

 

Intel 5nm hasn't been openly discussed as far as I'm aware, but they'll surely be working on it in some way already so that it will be ready when the time comes in future years. Keep in mind it will be the process going against what will be TSMC 3nm due to their naming differences.

Ah, I recall hearing something about in an LTT video or maybe WAN show. I wonder if Intel could try to advance AFAP (As Fast As Possible), and then back track somehow. As they have control of all the variables, that might work, but many others have more knowledge than I do on that. Super interesting nonetheless! Thank you as well for your reply.

[porina]

39 minutes ago, Happiness_is_Key said:

I wonder if Intel could try to advance AFAP (As Fast As Possible)

You can be sure they're doing what they can in a way that makes sense. There's an old joke about that kinda applies here: if a woman takes 9 months to produce a baby, how long does it take 9 women? Not that I'm saying making silicon is like making babies, but there are things that simply can't be done faster by throwing more money at it. 

[phoon]

On 12/5/2020 at 1:12 PM, Eigenvektor said:

Afaik Intel's 10 nm process is very comparable to TSMC's 7 nm process. The nanometer label has become a marketing term over the years, so you can't really use the numbers to compare between different fabs. Not sure they'll try to switch nodes again before they ironed out the issues they had.

 

I don't remember the exact details, but I think Intel tried to also switch to EUV and introduce different material (InGaAs?) on top of going from 14nm to 10 nm which didn't turn out as intended. So basically they bet big on their new process and it didn't pan out, which is what stumped them for a long time. So I don't think its fair to say they're not trying to compete at all. They bet on the wrong horse and now they're paying the price.

Oh my, you sure know your stuff bro. As a computer engineering major this subject is my #1 interest. It ties in to my gaming and pc hardware passions.

 

Mind sharing where you keep yourself up to date? 

[phoon]

On 12/5/2020 at 12:22 PM, porina said:

Rocket Lake will be interesting, as it will be the first new architecture on desktop since Skylake. It is a catch up generation. I don't think anyone really expects them to take leadership, and Intel themselves don't either. It was earlier in the year, where the CEO gave out a longer view. These things take time to implement, and it can still go wrong.

Do you think that Intel's 2022 CPU lineup will be well ahead of AMD or not?

[ShrimpBrime]

Alder Lake is to feature up to 16 cores. 10nm DDR5

It will be a combination of Golden Cove and Gracemont cores.

Supposedly, the 8 larger Golden Cove cores will support Hyper Threading while the Gracemont cores will not. So in total 24 threads. This will be Socket LGA 1700. 

 

By the time Intel releases 7nm superFin, it will be 2022-2023 and AMD will have moved to 5nm by then.

 

We won't know who has a faster processor next year until numbers are released. (For now, AMD is better)

 

 

[Eigenvektor]

3 hours ago, phoon said:

Mind sharing where you keep yourself up to date? 

That's a good question I read ArsTechnica and TechPowerUp with some regularity. The occasional article on Tom's Hardware, GamersNexus and other tech news sites. When stuff interests me, I'll read their linked sources or simply use Google to find additional information.

 

Depending on the site, it can be interesting to read the comments. Less so for TPU where it is mostly flamewars, but e.g. ArsTechnica has people who are actually knowledgeable (some of them being scientists and engineers actually working on the stuff the article is about).

[porina]

7 hours ago, phoon said:

Do you think that Intel's 2022 CPU lineup will be well ahead of AMD or not?

I don't claim to predict the future accurately. We can be sure that AMD will try to keep their development momentum going, and that Intel are doing what is realistic within their power to both catch up and pass AMD in due course. Either side may do better or worse than expected. My best guess for now is that Alder Lake might release before Zen 4, and as such it could overtake Zen 3 architecture for some time, but Intel will not dethrone overall single socket performance simply by AMD having more big cores.

 

7 hours ago, ShrimpBrime said:

Supposedly, the 8 larger Golden Cove cores will support Hyper Threading while the Gracemont cores will not. So in total 24 threads.

My gut feeling is that for any performance workload it's best to ignore the small cores altogether. It's just an 8 core CPU. What will be more interesting is when they get around to new HEDT offerings to flesh out higher core counts. Given Ice Lake Server is probably going to be a year late by the time it comes out (from memory Intel first spoke about it being an early 2020 product), it might be rather late to use that as basis. They don't have anything else (known) in that area though.