Intel Doesn’t Think the Gaming Market Is Big Enough to Justify a Core CPU with 3D V-Cache: “Comes with Trade-Offs, Certain Disadvantages, or Co...

[Tsing]

Intel appears to have an answer to AMD's 3D V-Cache technology, but it won't be coming to client products, such as Intel Core processors for laptops and desktops, anytime soon, according to new statements shared by Florian Maislinger, Tech Communications Manager at Intel.

See full article...

 

[Slag-King]

Translation: we have no idea how the hell AMD did this. I have a feeling that the lunar/arrow lake modular design is constrictive to the point that they have no way of adding the 3d cache,

I don't know...the 3d cache certainly benefits photo editing software, video encoding, science-specific workloads, neural engines, etc. These are not gaming scenarios.

 

[magoo]

Market speak for we cant compete. There are millions of "gamers". Ask AMD.

 

[Grimlakin]

Yea but intel doesn't want their consumer CPU's to step on their HEDT or server CPU's.

AMD's answer is advanced chiplet design and better interconnects allowing 96+ core server side CPU's.

Got FOMO buy a 16 core desktop that can do all our servers can.

Intel's answer... If you need that drop 40k on a HEDT solution chumps...

One of them is making money.

 

[Peter_Brosdahl]

I freely admit, despite how much I really love my 5800X3D and look foward to another 3D Cache processor that it's not the end-all be-all. We can get by w/o but the when a $460 part (7800X3D) can compete and beat in some areas a $589 part (14900K) then someone at Intel needs to be asking themselves what is the best way forward because brand loyalty only goes so far. It only gets worse when throw in temps, TDP, and cooling differences between the two.

Meanwhile the 285K, not the worst and should get better with BIOS/driver updates, at $700 makes even less sense for gamers and it's clear that Intel doesn't have anything for them right now.

I don't have anything against Intel but I sure wish they'd get over their identity crisis because I don't believe they know who they are anymore. Meanwhile I love the 13900HK in my laptop. It's a beast and puts a smile on my face while gaming at 1440p 100-200 Hz.

 

[Riccochet]

Gamers want the best for gaming. Imagine that. And Intel isn't it. Not saying Intel isn't "good enough", because they are. But if I'm building something that's going to be 95% used for gaming I'm going with the best option for gaming. Not something that's OK at gaming and obliterates the competition at .ZIP'ing files and transcoding video. I don't care if it takes an extra 5 seconds to zip a file.

 

[theblackangus]

<snark>
Q. So the V-Cache trade-offs are worse than the trade-off from ARL?
..........
<sound of crickets>
..........
A. Well you see...

</snark>
(Yes I know its not that simple an answer)

 

[LazyGamer]

Got FOMO buy a 16 core desktop that can do all our servers can.

Intel's answer... If you need that drop 40k on a HEDT solution chumps...

Intel has no problem in the productivity space; in the server space, they're more looking to solve different problems.

It's games with cache dependencies (most of them) that they're refusing to cater to.

 

[Niner51]

Intel has no problem in the productivity space; in the server space, they're more looking to solve different problems.

It's games with cache dependencies (most of them) that they're refusing to cater to.

I can already see Intel not worrying about gaming performance with Arrow lake. I agree they don't seem to know how to compete in that segment like they used to and when they tried to keep up they failed and basically had to do a recall of sorts with the 13th/14th Gen cpu's. I just can't understand with their resources that they couldn't figure things out.

 

[LazyGamer]

I just can't understand with their resources that they couldn't figure things out.

I think we can agree that, generally, Intel 'knows' what they'd need to do from a product standpoint, IMO. The challenge is always getting such a product actually built and shipping it and it's a business problem.

AMD seemed to address it unintentionally by building something that's excellent for certain enterprise applications - i.e., databases IIRC - and then having internal engineering leadership push for a consumer release. And that this was even possible due to the modularity of their architecture, such that they didn't need to fabricate different silicon, and instead they had to make a change in packaging.

I'm betting that Intel, now with their 'tile'-based architecture, might just now be in a position to respond. I'm much less certain that they'll actually be able to ship such a product, though

 

[Grimlakin]

I think we can agree that, generally, Intel 'knows' what they'd need to do from a product standpoint, IMO. The challenge is always getting such a product actually built and shipping it and it's a business problem.

AMD seemed to address it unintentionally by building something that's excellent for certain enterprise applications - i.e., databases IIRC - and then having internal engineering leadership push for a consumer release. And that this was even possible due to the modularity of their architecture, such that they didn't need to fabricate different silicon, and instead they had to make a change in packaging.

I'm betting that Intel, now with their 'tile'-based architecture, might just now be in a position to respond. I'm much less certain that they'll actually be able to ship such a product, though

If you live in the world of databases... part of the issue here is if they release a relatively dense core count consumer CPU that has the cache... that will eat into their Server space users.

When you consider that high core count CPU's drive up the cost of database servers and many of them don't need the massive core counts on offer... you start to understand.

a dual proc 32 core per socket server all cores enabled.. with SQL enterprise.. is a massive cost from licensing. Close to 1/2 a million dollars. PER server. (some work with offline/online servers can be done to keep licensing costs in check.)

If businesses could get the same kind of ram capacities they need out of a dual consumer grade system.. they would swap to that. Easily.

(and lets be real. only very large databases will need over 512GB of ram and consumer CPU's are supporting 256gig a socket today. It's the ECC stuff that keeps people away from consumer in the AMD space and that has work arounds.)

 

[Riccochet]

In all fairness this is the first "tiled" silicon CPU that Intel put out. The 1000 series Ryzen weren't all that good. 2000 series saw a good jump, and 3000 was a large jump. AMD just has years of product improvement under their belt. Intel will catch up, possibly.

 

[Grimlakin]

In all fairness this is the first "tiled" silicon CPU that Intel put out. The 1000 series Ryzen weren't all that good. 2000 series saw a good jump, and 3000 was a large jump. AMD just has years of product improvement under their belt. Intel will catch up, possibly.

Non the core Ultra series launched around November of last year. This SHOULD be their second generation of tiled CPU's.

 

[LazyGamer]

In all fairness this is the first "tiled" silicon CPU that Intel put out. The 1000 series Ryzen weren't all that good. 2000 series saw a good jump, and 3000 was a large jump. AMD just has years of product improvement under their belt. Intel will catch up, possibly.

Well, AMD is doing chiplets, with relatively slow, power draining connections through PCB. This has its disadvantages, namely in terms of latency; prior to the X3D series, AMDs Ryzen was the more-affordable second tier for gaming as Intel's monolithic designs were just more effective (if also more power hungry under load).

It's only with the additional cache that AMD has been able to mask the inefficiencies inherent in their chiplet layout. And aside from the wacky latencies Intel is seeing with Arrow Lake, they'd still need to do something similar; DRAM, as fast as these new CUDIMMs can be run (8000+), is still an order of magnitude slower than even slow L3 cache.

 

[Riccochet]

Non the core Ultra series launched around November of last year. This SHOULD be their second generation of tiled CPU's.

These 200 series are the first desktop parts.

 

[Grimlakin]

These 200 series are the first desktop parts.

That is true but lessons learned from the previous gen are what's important. We can't forget about each gen. If you don't think AMD is learning from the laptop chips that's silly. Efficiency is paramount! It's sad we don't expect Intel to learn from their generational CPU's even if it's Laptop, Desktop, or Server.

 

[Zarathustra]

I don't think "big enough" is the problem.

The PC gaming market has never been bigger than it is today. We are in an absolute PC gaming renaissance. (or at least were during the pandemic, the market has retracted a little since then as many upgraded at the same time, and now are waiting for their next upgrade)

The problem for them is that the gaming market doesn't have as deep pockets as the Enterprise market does, and it is competing for the same silicon production capacity,

Intel wants to enter niches where they can maintain a performance lead and charge a premium. They don't want to compete head to head where AMD has already cemented themselves with a strong product.

And this is part of the problem.

In order for the free market to work properly, you need 3-5 viable competitors making the same thing competing against each other, but instead big companies like to act like cartels, carving themselves out niches and not touching each-others turf.

And I don't know what you do about that. Break everyone up?

 

[Grimlakin]

I don't think "big enough" is the problem.

The PC gaming market has never been bigger than it is today. We are in an absolute PC gaming renaissance. (or at least were during the pandemic, the market has retracted a little since then as many upgraded at the same time, and now are waiting for their next upgrade)

The problem for them is that the gaming market doesn't have as deep pockets as the Enterprise market does, and it is competing for the same silicon production capacity,

Intel wants to enter niches where they can maintain a performance lead and charge a premium. They don't want to compete head to head where AMD has already cemented themselves with a strong product.

And this is part of the problem.

In order for the free market to work properly, you need 3-5 viable competitors making the same thing competing against each other, but instead big companies like to act like cartels, carving themselves out niches and not touching each-others turf.

And I don't know what you do about that. Break everyone up?

If you think the FTC isn't looking at Nvidia, Microsoft, Apple, and other trillion dollar companies as something that might be operating illegally you're silly. Not to say they aren't getting fat ****ing checks from the same organizations to look the other way.