News Mockup of AMD's Zen 4 Raphael CPU Appears With Potential Specs

[Kamen Rider Blade]

Because GPUs are still built on monolothic processes. If/when AMD makes a chiplet version of their GPUs, I'm willing to wager they'll only make one configuration per die because it's the most cost-efficient way to do things.

I'm betting that AMD will counter Intel and attack every segment of their market at every price point and every SKU configuration.

AMD will become flexible in it's configurations that anything Intel can do, AMD will have a equal-core count solution and higher core-count solution at better price than Intel.

 

[InvalidError]

And they're not doing that. That's the point of their chiplet strategy. Rather than make dies for each market segment, build the smallest feasible die and tack on more as needed. If the "smallest feasible" is an 8-core die, then so be it.

Not the smallest feasible, smallest practical.

I'm sure it would be feasible to make CCDs smaller, but then overheads like IOD ports, power balls, cutting losses to separate the dies, etc. would end up wasting a lot of wafer space per core on top of the latency-related issues of having a bunch of chiplets. There are practical limits to making things smaller.

 

[everettfsargent]

Sorry, but four cores are so DOA on the desktop. There are no current generation four core desktop CPU's from either Intel or AMD at the moment. Think 6/8/12/16/20/24 for the desktop going forwards.

 

[hotaru.hino]

I'm betting that AMD will counter Intel and attack every segment of their market at every price point and every SKU configuration.

AMD will become flexible in it's configurations that anything Intel can do, AMD will have a equal-core count solution and higher core-count solution at better price than Intel.

I would be willing to believe you.

If AMD owned 16 foundries that solely make their own product.

 

[Kamen Rider Blade]

I would be willing to believe you.

If AMD owned 16 foundries that solely make their own product.

They don't need to, the Ryzen Chiplet strategy is flexible enough that they can customize it as needed to counter Intel.

It's already doing great and they just have to continue with the Full Court Press across all market segments like they were doing Pre-COVID.

Leave no stone un-turned, leave no product segment un-challenged.

Remember how Ryzen 2000 & 3000 were performing, it was giving Intel quite the run for it's money.

 

[hotaru.hino]

It's already doing great and they just have to continue with the Full Court Press across all market segments like they were doing Pre-COVID.

Why? What's the point when they're making a killing on what they're doing now?

While I can't find a figure on how much it costs to make a chip, making one simply to tackle the lower to lowest end (AMD seems to have dropped sights on where the Atom, Celeron, and Pentium live) I can't see as a lucrative enough venture to consider unless currently AMD is desperate. While sure, it won't be the same cost as making an entirely new architecture, there's still fixed costs associated with development, production, and testing of a product, and I'm pretty sure those still cost well into the 8 figure range. If you're going to spend 8 figures to make products you sell for <$70, you better be able to flood the market with it.

 

[InvalidError]

While I can't find a figure on how much it costs to make a chip, making one simply to tackle the lower to lowest end (AMD seems to have dropped sights on where the Atom, Celeron, and Pentium live) I can't see as a lucrative enough venture to consider unless currently AMD is desperate.

Another question to ponder: what would the Atom, Celeron and Pentium be like today if Intel had a working high-volume 7nm process? At roughly triple the process density vs 14nm, chances are everything would get core count bumps.

 

[Kamen Rider Blade]

Why? What's the point when they're making a killing on what they're doing now?

While I can't find a figure on how much it costs to make a chip, making one simply to tackle the lower to lowest end (AMD seems to have dropped sights on where the Atom, Celeron, and Pentium live) I can't see as a lucrative enough venture to consider unless currently AMD is desperate. While sure, it won't be the same cost as making an entirely new architecture, there's still fixed costs associated with development, production, and testing of a product, and I'm pretty sure those still cost well into the 8 figure range. If you're going to spend 8 figures to make products you sell for <$70, you better be able to flood the market with it.

The point is to deny Intel access to revenue at all product segments.

From Top to Bottom, in all markets.

You can't be expected to grow as a company in the x86 market and CPU market if you aren't tackling all aspects at all price points.

Every x86 CPU sale that goes to AMD and NOT to Intel denies Intel of revenue.

attacking across the product portfolio from top to bottom was the right call when AMD did it with the 2000 and 3000 series.

Intel was suffering from lower to usual sales during those time periods Pre-COVID.

It'll be the right move moving forward in taking marketshare and increasing revenue for AMD while denying Intel as much as possible.

That's why you build the product stack.

 

[InvalidError]

The point is to deny Intel access to revenue at all product segments.

You can't be expected to grow as a company in the x86 market and CPU market if you aren't tackling all aspects at all price points.

Companies exist to make a profit. Publicly traded companies are duty-bound to maximize profit for investors. Maximizing profits means focusing product manufacturing on the highest-margin products possible. It makes absolutely zero sense for AMD to waste any effort trying to appeal to the low-end while it cannot even cope with high-end demand and AMD's manufacturing side issues may not be relenting much for another year or two.

You can grow perfectly fine by giving up the low-end when you are already unable to keep up with high-end demand. Can't grow any faster than you can get stuff made.

 

[hotaru.hino]

From Top to Bottom, in all markets.

You can't be expected to grow as a company in the x86 market and CPU market if you aren't tackling all aspects at all price points.

AMD doesn't have to have a product to answer to everything Intel has to offer. They only have to make superior products in areas that matter.

But if you truly believe AMD has to have a product to answer everything Intel has to offer, then AMD is pretty much dead. Not only does AMD not have the capacity to make a product for everything Intel has to offer, but they're not even close to having half the products and services Intel offers by comparison.

 

[Kamen Rider Blade]

Companies exist to make a profit. Publicly traded companies are duty-bound to maximize profit for investors. Maximizing profits means focusing product manufacturing on the highest-margin products possible. It makes absolutely zero sense for AMD to waste any effort trying to appeal to the low-end while it cannot even cope with high-end demand and AMD's manufacturing side issues may not be relenting much for another year or two.

I've already stated that during the COVID-19 times, when there is a chip shortage, that they focus on the high end of their product stack.
You keep re-iteresting the same points, and I agreed with you that during this trying time, that they should do that.

But in the LONG TERM, they're going to target the low end. Just like they did in the past with the Ryzen 1000, 2000, 3000 series.

Right now is a special case and circumstance, I'm talking LONG TERM, in the future when things get back to normal.

You can grow perfectly fine by giving up the low-end when you are already unable to keep up with high-end demand. Can't grow any faster than you can get stuff made.

AMD was growing just fine in the 1000, 2000, 3000 series era; Pre-COVID 19 & Chip shortage.
AMD was doing alot of market-share damage to Intel along with Mind-share damage.

AMD doesn't have to have a product to answer to everything Intel has to offer. They only have to make superior products in areas that matter.

And they're already doing that.

But if you truly believe AMD has to have a product to answer everything Intel has to offer, then AMD is pretty much dead. Not only does AMD not have the capacity to make a product for everything Intel has to offer, but they're not even close to having half the products and services Intel offers by comparison.

AGAIN, I'm talking about Post COVID-19 & Chip Shortage, how hard is this for you to understand?
Why do you keep focusing on the immediate and not on the future when things get back to normal?

And yes, AMD will have to grow in terms of products & services to compete agains what Intel offers, that's understandable given AMD's position currently.

 

[InvalidError]

But in the LONG TERM, they're going to target the low end. Just like they did in the past with the Ryzen 1000, 2000, 3000 series.

AMD never 'targeted' the low-end, the low-end has always been an outlet for high-end defects. In the 3000-series, the defect rate dropped and the 3100, 3300X and 3500 vanished from the market never to be seen again and AMD hasn't released any new low-end parts since, not even refreshed APUs after the 3400G from three years ago.

AMD has been practically MIA from the low-end for the better part of two years, which would be BEFORE covid started.

AMD only catered to the low-end in the 1000-3000(G) series because it had no reputation to stand on to get sales moving and Zen 1/1+ were pretty crappy. It had no choice because most enthusiasts wouldn't give it the time of day.

 

[cryoburner]

And don't forget that on 7nm, Ryzen 3000 series had 4 Core CCX's as well in the form of the Ryzen 3 3300X .

The "Smallest Feasible" die would be a 4-core die like last gen with the Ryzen 3000 series.

The Ryzen 3300X did not use a special 4-core die. Both it and the 3100 utilized the exact same 8-core chiplets put into all of their other higher core-count processors, utilizing chips that didn't meet the requirements for those processors on enough of their cores. All Ryzen non-APU processors since the 1000-series have thus far contained at least eight cores, just with some disabled for certain models. For the 3000-series and prior, those cores were organized into two mirrored groups of four within each chip. The 3100 had half of its cores disabled in each group, while the 3300X had one entire group disabled.

Another thing to remember about those processors is that they were only available in very limited quantities for a short time, and disappeared almost entirely shortly after launch. That's probably because they were only released as a way to get rid of a stockpile of partially-defective chiplets that weren't usable for the higher-margin parts, rather than being a specific market AMD wanted to target. Ever since the 2000-series, AMD has been primarily using monolithic APUs for their lower core-count desktop processors targeting MSRPs less than $200. The muti-chip packaging approach may not even be ideal for processors targeting low price points.

 

[cryoburner]

AMD only catered to the low-end in the 1000-3000(G) series because it had no reputation to stand on to get sales moving and Zen 1/1+ were pretty crappy. It had no choice because most enthusiasts wouldn't give it the time of day.

I wouldn't say the original Zen architecture was bad. IPC and efficiency-wise, it was roughly on-par with Intel's processors in most applications, with mainly just memory access speeds and maximum clock rates holding it back a bit. And as for the processors themselves, models like the 1600 with its 6-cores and 12-threads arguably made the competing 4-thread i5s look bad.

Of course, that was more down to AMD offering significantly more cores and threads for the money rather than the architecture itself, and if they had been able to get a bit more performance-per-core out of those processors, I imagine they would have charged more for them. Ultimately though, the big improvements to their architecture over what they had previously brought them close enough to Intel on a per-core basis to make them a viable option for most use-cases, so that first-generation architecture arguably served its purpose of being decent enough to make compelling products out of.

 

[InvalidError]

I wouldn't say the original Zen architecture was bad. IPC and efficiency-wise, it was roughly on-par with Intel's processors in most applications

IMO, AMD didn't come close to achieving parity until Zen 2 and 1/1+'s only saving grace was offering two extra cores for a given price and product tier.

Motherboard manufacturers also treated Ryzen somewhat like a joke product through most of the first two generations with most boards leaving much to be desired until manufacturers got numbers for Zen 2 and decided to get serious about it.

 

[hotaru.hino]

AGAIN, I'm talking about Post COVID-19 & Chip Shortage, how hard is this for you to understand?
Why do you keep focusing on the immediate and not on the future when things get back to normal?

Because TSMC is at freaking capacity already. COVID-19 didn't create a chip shortage. It created a demand surge. I mean sure, TSMC can create more plants, but it'll be at least a few years before they become operational.

I don't know why you have such a hard on for AMD needing to stick their fingers in everything other than you want to see Intel crash and burn. But if this is all what you see, I don't see a point in carrying this discussion anymore.

 

[InvalidError]

Because TSMC is at freaking capacity already. COVID-19 didn't create a chip shortage. It created a demand surge.

There was already going to be a demand surge regardless of covid due to AMD and others launching all of their new stuff at about the same time though. Covid simply made something that was already going to be bad worse. And then you have to add all of the supply chain disruption from lock-downs in various places, power outages, droughts, etc. TSMC expects to get ordered to reduce production or possibly outright shut down some of its Taiwan fabs this summer due to low water reserves during Summer. Building more fabs in Taiwan won't help with that!

 

[Kamen Rider Blade]

Because TSMC is at freaking capacity already. COVID-19 didn't create a chip shortage. It created a demand surge. I mean sure, TSMC can create more plants, but it'll be at least a few years before they become operational.

I don't know why you have such a hard on for AMD needing to stick their fingers in everything other than you want to see Intel crash and burn. But if this is all what you see, I don't see a point in carrying this discussion anymore.

Do you really want another 20-30 years of Intel dominance?

Intel needs to be knocked down a few pegs, especially after their Anti-Trust shenanigans.

The best way to do that right now is for AMD to succeed alongside TSMC.

That means they have to grow on ALL FRONTs, that means denying Intel of revenue through competition.

TSMC has to gain more fab capacity, AMD needs more market share.

We need to reverse the uncalculable damage that Intel did during the first time when AMD gained the upper hand and Intel used their Monopoly powers to screw over AMD.

I'm sorry, but I don't want to support Intel's convicted Monopolist arse and they need to pay for their crimes.

The various Anti-Trusts from governments around the world haven't penalized Intel enough IMO.