Discussion Intel current and future Lakes & Rapids thread

[Geddagod]

That s only for 4C/8T SKUs, so in cheap laptops whose not that much is expected.

Ye, AMD is only screwing the poor folk. That's fine.

What Inel is doing is as if AMD stuck Zen 2 without SMT and Zen 4 cores within a same SKU, think about it...

Why think about hypothetical hybrid setups? AMD is already releasing hybrid products without differentiating the cores.

Guess that i did hit a nerve...

Nah I'm playing

If MTL was that wonderfull we would know about it currently,

It's almost as if the product hasn't launched yet.

Intel s marketing dpt would be occupied spreading infos about the thing rather than smearing AMD

You know, the timing also coincides when AMD is going to do a major refresh with the new "8000" series since we are coming into 2024.

seems that they are keeping mum on the subject for a good reason,

Because it hasn't launched yet?

if MLID is half right then it will barely match a 7840U/HS, the graph he displayed show 25-30% better perf/Watt at isoperf for MTL in respect of a 13900H...

Even if you take the 13700h graph and shift it by 20% from the node itself you aren't getting anything much better than matching Zen 4 in perf/watt for reasonable TDPs.

Intel obfuscates the architecture used in their products. The 13400KF uses both RPL and ADL and it depends on the stepping as to which you get,

Holy moly the 5% difference in performance between steppings is making me so mad grrrr

hat is far worse than AMD informing the consumer through the product name which Zen is in the chip.

Can't wait to buy 7000 series Zen 2 products right along side... also 7000 series Zen 4 products.

but calling it worse than what Intel do is just total nonsense.

It is far, far worse. If your best argument is that Intel 7 dies are being snuck into Intel 7+ skus... then you don't have much to stand on.

 

[H433x0n]

Guess that i did hit a nerve...
If MTL was that wonderfull we would know about it currently, Intel s marketing dpt would be occupied spreading infos about the thing rather than smearing AMD, seems that they are keeping mum on the subject for a good reason, if MLID is half right then it will barely match a 7840U/HS, the graph he displayed show 25-30% better perf/Watt at isoperf for MTL in respect of a 13900H...

I don’t get this argument, why would Intel spoil their launch date by giving away all of the details before hand?

They didn’t have finalized clocks until literally last week, how are they supposed to give performance data before then?

 

[Saylick]

I don’t get this argument, why would Intel spoil their launch date by giving away all of the details before hand?

They didn’t have finalized clocks until literally last week, how are they supposed to give performance data before then?

Idk, maybe they just pull an AMD a la the RDNA 3 launch, i.e. claiming N31 was 1.5x - 1.7x the performance of N21. Look how that turned out!

 

[Abwx]

I don’t get this argument, why would Intel spoil their launch date by giving away all of the details before hand?

They didn’t have finalized clocks until literally last week, how are they supposed to give performance data before then?

Because previously we had infos well before launch, and certainly not one week before, if not one hour since we still have nothing at this point...

 

[Geddagod]

Because previously we had infos well before launch, and certainly not one week before, if not one hour since we still have nothing at this point...

cap

 

[Hulk]

Have you ever bought a computer, built a computer, bought a laptop, or any other device with an x86 cpu and didn't know what core you bought, how many were in there, and the stock clocks? Confounding, disjointed naming scheme or not, I always knew exactly what I was buying.

I know, I know. But the average consumer isn't like "us" here in the forum. Yes, that's true. They also can't tell the difference between an i3 and and i9 in their day-to-day web browsing and Word work. That is what Intel and AMD are counting on. And they are counting correctly

 

[gdansk]

Have you ever bought a computer, built a computer, bought a laptop, or any other device with an x86 cpu and didn't know what core you bought, how many were in there, and the stock clocks? Confounding, disjointed naming scheme or not, I always knew exactly what I was buying.

A few years ago AMD sold 8 core Ryzen 5 1600. So even if they did their research the buyers were likely a bit confused.

 

[lightisgood]

1) Gaudi2 is way behind H100 for training.

However...

NVIDIA Shows Intel Gaudi2 is 4x Better Performance Per Dollar than its H100​

NVIDIA Shows Intel Gaudi2 is 4x Better Performance Per Dollar than its H100

In a stunning twist, NVIDIA shows the Intel Gaudi2 is roughly 4x better performance per dollar than the H100 in its MLPerf Training results

www.servethehome.com

In actual fact, Gaudi2's cost/unit is roughly 0.5x on H100.
So, Gaudi2 is 8x better perf/cost than H100.
The competitiveness of MI300 is be really poor.

2) MI300X can scale much better than both.

MI300X is 1H24 product.
H200 and Gaudi3 are 1H24 products, too.

 

[Henry swagger]

https://twitter.com/x/status/1732643274085798344

sapphire rapids is doing very good.. granite rapids should continue the good work

 

[Timorous]

Can't wait to buy 7000 series Zen 2 products right along side... also 7000 series Zen 4 products

That is the thing though, they are not really 7000 series. You have the 7020 series, 7030 series, 7035 series 7040 series and 7045 series.

Holy moly the 5% difference in performance between steppings is making me so mad grrrr

Stupid argument. If the issue is multiple architectures in the same product year family with a clear indicator as to which arch is in the product then it is objectively worse for the 13400KF to have different architectures based on the stepping which is not usually an important differentiator.

Like I said you can make an argument that the product year should not be the 1st digit but in no world is a model naming convention where the architecture is encoded in the name worse than one where it is not.

 

[Hitman928]

However...

NVIDIA Shows Intel Gaudi2 is 4x Better Performance Per Dollar than its H100​

NVIDIA Shows Intel Gaudi2 is 4x Better Performance Per Dollar than its H100

In a stunning twist, NVIDIA shows the Intel Gaudi2 is roughly 4x better performance per dollar than the H100 in its MLPerf Training results

www.servethehome.com

In actual fact, Gaudi2's cost/unit is roughly 0.5x on H100.
So, Gaudi2 is 8x better perf/cost than H100.
The competitiveness of MI300 is be really poor.



MI300X is 1H24 product.
H200 and Gaudi3 are 1H24 products, too.

To address the headline, which is apparently all you read as they already took the half price of Gaudi into account, it's a bit of click bait from servethehome (unfortunately). When you scale to many more GPUs, the perf/GPU and thus perf/$ goes way down because you don't get anywhere near linear scaling with many more nodes. The same would be true for Gaudi2 if they tried to scale it to the same number of GPUs. Nvidia marketing is just doing Nvidia marketing things and putting out an unfair comparison to claim H100 is many multiple times faster than Gaudi2 and kind of hoping no one notices the detail that they are using many more H100s in the comparison. So good on Patrick for calling out Nvidia's deception, just wish he did it in a not so click baity way.

Now, if we just look at a comparison where Gaudi2 and H100 have the same number of GPUs, what do you get?



So H100 is 2x faster for only 2x the money? That's pretty good as usually you have to pay a steep premium for the best performance. Why would anyone buy Gaudi if it's half the performance and doesn't even give a perf/$ advantage? So then MI300X comes in with the same to better performance as H100 in training, which is way faster than Gaudi2, faster performance at inference, much better scalability, and actually offers a perf/$ advantage. H200 launches in 6 months with zero compute improvement, it's only upgrade is to switch to HBM3e for more capacity and higher bandwidth, both of which will still be behind MI300X. Additionally, HBM3e is only going into volume production next year so supply will most likely be very limited. H200 is not an MI300X killer.

As far as Gaudi3, Intel is claiming 1.5x the memory and 2x the compute. That would, in napkin math, put it roughly on par with H200 and roughly equal with MI300X in compute but the MI300X will still have a memory advantage. Any idea when Gaudi3 actually launches? How close is it going to be to B100 (which should be another big jump over Gaudi3)? I think there's a reason Intel is killing the Gaudi line after 3, it just hasn't been able to compete like Intel needs it to.

I'll leave it at that, if you feel the need to continue trying to spin the narrative against MI300, I suggest you go do it in the appropriate thread in the graphics section where much of this has already been discussed.

 

[Geddagod]

That is the thing though, they are not really 7000 series. You have the 7020 series, 7030 series, 7035 series 7040 series and 7045 series.

Literally no general consumer will be aware or tbh, even care, about what "architecture" it uses.

Stupid argument.

Why, cuz it's so true?

. If the issue is multiple architectures in the same product year family with a clear indicator as to which arch is in the product then it is objectively worse for the 13400KF to have different architectures based on the stepping which is not usually an important differentiator.

That's not an issue though.

Like I said you can make an argument that the product year should not be the 1st digit

Make an argument? This is just you downplaying this a ton. The product year should never be the 1rst digit, and doing so is just being massively misleading to a consumer base who has recognized the 1rst digit as the "generation" for years and years now.

but in no world is a model naming convention where the architecture is encoded in the name worse than one where it is not.

It is, when the model naming convention where the architecture isn't encoded, also doesn't have the year in the 1rst digit.

 

[Hitman928]

Make an argument? This is just you downplaying this a ton. The product year should never be the 1rst digit, and doing so is just being massively misleading to a consumer base who has recognized the 1rst digit as the "generation" for years and years now.

Ok, so which generation was Coffee Lake again? If I buy a 10th gen mobile core processor, which architecture am I getting?

I get that making the first digit correspond to the year of release isn't all that consumer friendly, but that is basically what OEMs have asked for so AMD has obliged and come up with a system (although not clear without knowing the decoding process) to do so while still giving info on architecture and performance. Intel's system allows them to name things whatever they want each year and mix architectures within the same naming system with no indication as to what you are getting. Neither are helpful to the average end user. The average user just cares about what kind of performance they can expect from each model with some kind of scale which they both kind of accomplish but neither is great at that either. Where AMD is now using older architectures to fill in cheaper models, Intel uses severely castrated CPUs which might as well be an older architecture for the performance drop. Outside of that it just really comes down to personal opinion if you'd rather have some sort of algorithm you can know to decoding the model name or you think it doesn't matter and you should just do your research before hand to know what you are getting.

 

[Geddagod]

Ok, so which generation was Coffee Lake again? If I buy a 10th gen mobile core processor, which architecture am I getting?

The naming schemes to the naked eye looked quite different. Different number of digits, placement of letters earlier, it's not as good as I wanted it to be but it also appears to be a one-off. They went back to normal with TGL/ADL/RPL.
But that naming wasn't that great either- which is why moving to the Ultra moniker actually is pretty good tbh. Makes the distinction even more clear.

but that is basically what OEMs have asked for so AMD has obliged an come up with a system (although not clear without knowing the decoding process) to do so while still giving info on architecture and performance.

And we all know OEMs love looking out for the consumer as well

Intel's system allows them to name things whatever they want each year and mix architectures within the same naming system with no indication as to what you are getting.

...information which general consumers don't know/care about.
And the people knowledgeable enough to care about that, would almost certainly also put in the effort to google it (what architecture a product is) if they had too.
Also, when else other than ICL has Intel mixed archs in the past for same segment client?

The average user just cares about what kind of performance they can expect from each model with some kind of scale which they both kind of accomplish but neither is great at that either.

One is clearly more misleading than the other

Where AMD is now using older architectures to fill in cheaper models, Intel uses severely castrated CPUs which might as well be an older architecture for the performance drop.

Intel segments E-core only models as N series, with the letter N literally infront of the digits. And then there's the usual i3, i5, etc etc tiers as well.
Also Intel uses older models to fill in cheaper models too... even when ADL and RPL were selling they still had a ton of TGL chips in lower end laptops

Whatever small benefit AMD's naming scheme has over Intel's for including the specific architecture in the digits, is overshadowed by the huge negative in that it has the year as the first and most prominent digit- the digits consumers are accustomed too and will look at primarily- when buying hardware.

 

[Hitman928]

The naming schemes to the naked idea looked quite different. Different number of digits, placement of letters earlier, it's not as good as I wanted it to be but it also appears to be a one-off. They went back to normal with TGL/ADL/RPL.
But that naming wasn't that great either- which is why moving to the Ultra moniker actually is pretty good tbh. Makes the distinction even more clear.

And we all know OEMs love looking out for the consumer as well

...information which general consumers don't know/care about.
And the people knowledgeable enough to care about that, would almost certainly also put in the effort to google it (what architecture a product is) if they had too.
Also, when else other than ICL has Intel mixed archs in the past for same segment client?

One is clearly more misleading than the other

Intel segments E-core only models as N series, with the letter N literally infront of the digits. And then there's the usual i3, i5, etc etc tiers as well.

Whatever small benefit AMD's naming scheme has over Intel's for including the specific architecture in the digits, is overshadowed by the huge negative in that it has the year as the first and most prominent digit- the digits consumers are accustomed too and will look at primarily- when buying hardware.

All of your arguments against AMD's scheme and for Intel's are basically negated by your own statement reply to my statement

Intel's system allows them to name things whatever they want each year and mix architectures within the same naming system with no indication as to what you are getting.

...information which general consumers don't know/care about.
And the people knowledgeable enough to care about that, would almost certainly also put in the effort to google it (what architecture a product is) if they had too.

If the general consumer doesn't care and those who do care can just google it, why is this a discussion at all? You're saying it doesn't matter that Intel has no scheme because people either don't care or can just google it, but AMD having a scheme is worse? Why can't people just continue to not care or google it with AMD's model numbers?

When I was talking about castrated CPUs, I wasn't even talking about E-core only SKUs but simply severely cut down CPUs filling the budget/entry segments. Rather than do that, AMD slots in older generations so instead of getting 1P4E or 2P4E (6T and 8T) at the bottom of the ADL mobile stack, you get 4C8T at the bottom of the Ryzen stack, but it's Zen2 architecture to hit the cheaper price point.

I would much prefer if Intel and AMD kept the first number being the architecture, but that's not going to happen with OEMs wanting new numbers every year and AMD and Intel not having new architectures every year to give them. AMD basically said we'll come up with a numbering system to accommodate and Intel basically said we'll call things what we want, when we want. I can easily see calling out both, I can't see how you can defend Intel's approach over AMD's though, that's where things get silly to me.

 

[Geddagod]

All of your arguments against AMD's scheme and for Intel's are basically negated by your own statement reply to my statement

What?

If the general consumer doesn't care and those who do care can just google it, why is this a discussion at all?

The general consumer doesn't care about the specific architecture used, but it's still misleading to market older architectures as a brand new generation, because the perf overall with those older architectures of those "new" chips still sucks overall, and are easily confused with much better performing chips from different generations because the first digit is the same. It's extremely misleading.
When Intel mixed in ADL/RPL silicon, you got the same perf from either variants. When CML/ICL was mixed into the same generation, the naming differences were much larger than a single digit in AMD's naming scheme.

When I was talking about castrated CPUs, I wasn't even talking about E-core only SKUs but simply severely cut down CPUs filling the budget/entry segments. Rather than do that, AMD slots in older generations so instead of getting 1P4E or 2P4E (6T and 8T) at the bottom of the ADL mobile stack, you get 4C8T at the bottom of the Ryzen stack, but it's Zen2 architecture to hit the cheaper price point.

Intel does the same thing with TGL chips when ADL/RPL were being produced. Difference is that they called those 11th gen, and ADL/RPL 12/13th gen. They didn't rebrand TGL chips into the 12th and 13th gen series.
And for the really low end, Intel makes sure they don't muddy the name of their generations with super low performing chips, even if it's the same "generation" as higher performing variants, by labeling them celeron and pentium, both of which have vastly different naming schemes than the regular chips.

I can easily see calling out both, I can't see how you can defend Intel's approach over AMD's though, that's where things get silly to me.

Intel's approach is drastically less misleading. AMD's approach is better in the aspect is differentiates architectures, but Intel has a much bigger, and much more important advantage, in not using the year as the first digit. That's way more relevant than the slight advantage AMD has in differentiating architectures.

 

[eek2121]

If the general consumer doesn't care and those who do care can just google it, why is this a discussion at all? You're saying it doesn't matter that Intel has no scheme because people either don't care or can just google it, but AMD having a scheme is worse? Why can't people just continue to not care or google it with AMD's model numbers?

End users don’t care. It is just a subject for industry nerds to get passionate about. I have said it before and I will say it again: As long as 8020 chips are “better” in some way over 7020 chips and feature no large regressions, most users won’t care except those wanting to argue about it.

AMD’s scheme is fine. If anything, Intel needs to work on their own. See the 13950k 14900k.

 

[Timorous]

but it's still misleading to market older architectures as a brand new generation, because the perf overall with those older architectures of those "new" chips still sucks overall, and are easily confused with much better performing chips from different generations because the first digit is the same. It's extremely misleading.

Does it? The steam deck holds up great vs the Rog Ally when running at 15W despite the huge gulf in on paper specs. Battery also lasts a lot longer too.

Architecture, TDP, Cooling are all critical to overall performance. You can put a top of the line Intel CPU into a product but if you do a bad job on cooling it will perform worse than cheaper variants in a better chassis, same happens with GPUs.

Buying a laptop is a mine field. At least with the AMD naming scheme you know what architecture you are buying and it also indicates when a Laptop was manufactured because a laptop with an 8000 product will be newer than one with a 7000 product. There may not be much difference in performance but there might be other feature differences like Wifi 6e or BT or HDMI 2.1 or DP 2.0 and so on. Or maybe there won't because the OEM used the same chassis/mobo as the old model and just dropped in the new CPU variant.

 

[eek2121]

Does it? The steam deck holds up great vs the Rog Ally when running at 15W despite the huge gulf in on paper specs. Battery also lasts a lot longer too.

Architecture, TDP, Cooling are all critical to overall performance. You can put a top of the line Intel CPU into a product but if you do a bad job on cooling it will perform worse than cheaper variants in a better chassis, same happens with GPUs.

Buying a laptop is a mine field. At least with the AMD naming scheme you know what architecture you are buying and it also indicates when a Laptop was manufactured because a laptop with an 8000 product will be newer than one with a 7000 product. There may not be much difference in performance but there might be other feature differences like Wifi 6e or BT or HDMI 2.1 or DP 2.0 and so on. Or maybe there won't because the OEM used the same chassis/mobo as the old model and just dropped in the new CPU variant.

The Steam Deck has the “Apple Advantage” where Valve controls the software and hardware. That is one of the reasons why it performs as well as it does. When you can optimize the software while tweaking the hardware, you get great results.

I do hope Valve updates the SoC when new consoles drop, however. 4-8 Zen 2 -> Zen 3 cores could help system responsiveness, particularly in desktop mode. A Steam Deck with 8 Zen 3 cores at 3.5-4ghz, 32gb of RAM, and 12 RDNA3 CUs @ 2-2.5ghz would be a decent upgrade.

 

[Hitman928]

What?

The general consumer doesn't care about the specific architecture used, but it's still misleading to market older architectures as a brand new generation, because the perf overall with those older architectures of those "new" chips still sucks overall, and are easily confused with much better performing chips from different generations because the first digit is the same. It's extremely misleading.

You seem to be making the same argument Intel was in their slides. Older architectures suck because they are older, buy newer because they are newer! So when someone bought 10th gen comet lake mobile, they got fooled by Intel because Ice Lake was the proper new architecture and the 10th gen CML was older (despite having the first name be the same and a higher number overall) and sucked. Oh wait, that doesn't count because if you knew how to decode it, there were more differences in the model number, unlike AMD where if you look past the first number, there aren't differences, oh wait. . . Maybe it will be better with MTL and RPL being the same generation because the first number will definitely not be the same there, except it is. But that doesn't count because Intel is adding Ultra to MTL. What does Ultra mean? No idea but you know Ultra is better than a number telling you exactly which gen it is I guess?

RPL generation having ADL also applies. You say that doesn't count because the performance difference was minimal (not always true, especially for gaming) but who decides what that threshold is? Again, at best this comes down to personal preference but any logical complaint you make against AMD's scheme can be made against Intel. Any defense you give Intel can be used to defend AMD's scheme. They both are not consumer friendly or all that transparent. Neither can really be defended because both are based on the same desire by OEMs to sell new and shiny model numbers than any technical reason. That's it.

Lastly, I'm speculating here, but I wouldn't be shocked at all if Intel starts doing even more of this themselves with the way new node development is progressing. I think you're going to see a lot more of older gens mixed in as refreshes to cover the less expensive segments as the cost to produce on each node continues to rise.

 

[igor_kavinski]

Has to be bait lol

Lightisgood meet Henry. Henry meet Lightisgood. Oh wait. Maybe it's the same person!

 

[Geddagod]

Does it? The steam deck holds up great vs the Rog Ally when running at 15W despite the huge gulf in on paper specs. Battery also lasts a lot longer too

Luckily for us, perf at 15 watts isn't the only TDP all CPUs are locked at.

Buying a laptop is a mine field.

AMD just made it much more confusing for most people.

At least with the AMD naming scheme you know what architecture you are buying

No general consumer will care about that. The reality of the situation is that AMD's newest generation suddenly became a lot more confusing for customers- with a lot more depth and complexity than really needed. If a guy recommends to his friend- wow my new AMD 5 tier laptop is running super fast, and then the other guy goes out and buys a 7520, well tough luck lol.

You seem to be making the same argument Intel was in their slides. Older architectures suck because they are older, buy newer because they are newer!

AMD's new naming scheme introduces a ton more variability into their newest generation, that's simply a fact.

So when someone bought 10th gen comet lake mobile, they got fooled by Intel because Ice Lake was the proper new architecture and the 10th gen CML was older (despite having the first name be the same and a higher number overall) and sucked. Oh wait, that doesn't count because if you knew how to decode it, there were more differences in the model number, unlike AMD where if you look past the first number, there aren't differences, oh wait. . .

A single digit change in the middle of the decoder is way less obvious than the change in letters, number of digits, and even then I said I didn't like ICL+CML's naming scheme either.
I also mentioned how it appeared to be a one off.

Maybe it will be better with MTL and RPL being the same generation because the first number will definitely not be the same there, except it is. But that doesn't count because Intel is adding Ultra to MTL. What does Ultra mean? No idea but you know Ultra is better than a number telling you exactly which gen it is I guess?

Because it's a clear and obvious label. Even normies can say, wow look at my new Core Ultra i9 I got. Normies won't be saying, wow look at this 7000 series ryzen 5 I got, with Zen 3/Zen 2/Zen 4. Ultra is literally infront of the numbers too lmao. Idk how much more obvious you can make it for normies.

RPL generation having ADL also applies. You say that doesn't count because the performance difference was minimal (not always true, especially for gaming)

For the tiers they were mixing the silicon for? RPC is esentially just GLC with more cache and some structure placement, and the differences between the two are extremely minimal. Pretending this is on the same level of mixing Zen 2/Zen 3/Zen 4 chips is just disingenuous.

Any defense you give Intel can be used to defend AMD's scheme.

Except that's the thing, you literally can't. TGL isn't 13th gen. Zen 2 chips are part of the 7000 series though. The only defense against this is that RPL used ADL silicon for the low end, but the difference between that and Zen 2 and Zen 4 being under the same generation is drastic.

Intel's naming system allows them to do the same thing AMD is doing btw, they just haven't done it except with RPL, and even then the difference between the two products they mixed is insanely small. And even then, they got large and deserved backlash for it. AMD's naming scheme means they are forced to show the differences in generations, but the placement of where that number is esentially means for normies, they won't pay attention to it at all, and end up being worse off because of it.

AMD's naming conventions are simply worse for the end consumer and more misleading than Intel's. The difference is day and night. At best, AMD has a theoretical advantage in showing the arch while Intel doesn't, but in practice, AMD is the one misleading customers more with more drastic mixing of archs and products, because the general consumer just isn't that observant.

 

[Hulk]

Just a couple days and all of our MTL questions will be answered friends.

My questions are as follows:

1. Will it be possible to shut down the GPU and CPU tiles and web browse, work in Office, and do other light compute work with only the SoC tile active for the majority of the time?
2. Regarding the above scenario, what is the actual power saving for this mode of operation compared to previous gen?
3. In the same power/thermal envelope how does performance compare to previous gen and the competition? I mean specifically testing CPU power package at 10 Watts, 12 Watts, 15 Watts, 20 Watts, 28 Watts, 35 Watts, 45 Watts... in a variety of applications so we can finally have an apples-to-apples comparison of mobile parts.

Without performance at various CPU package power our endless efficiency debates that go in circles will continue.

 

[igor_kavinski]

1. Will it be possible to shut down the GPU and CPU tiles and web browse, work in Office, and do other light compute work with only the SoC tile active for the majority of the time?

If you mean if the user will be able to shut down the tiles, I don't think so. The Thread Director and/or Windows will decide when to power off the unneeded tiles.

Browsing/office work, the tiles may shut down if you are idle (like thinking what to do or what to type or reading).

 

[Hitman928]

Luckily for us, perf at 15 watts isn't the only TDP all CPUs are locked at.

AMD just made it much more confusing for most people.

No general consumer will care about that. The reality of the situation is that AMD's newest generation suddenly became a lot more confusing for customers- with a lot more depth and complexity than really needed. If a guy recommends to his friend- wow my new AMD 5 tier laptop is running super fast, and then the other guy goes out and buys a 7520, well tough luck lol.

AMD's new naming scheme introduces a ton more variability into their newest generation, that's simply a fact.

A single digit change in the middle of the decoder is way less obvious than the change in letters, number of digits, and even then I said I didn't like ICL+CML's naming scheme either.
I also mentioned how it appeared to be a one off.

Because it's a clear and obvious label. Even normies can say, wow look at my new Core Ultra i9 I got. Normies won't be saying, wow look at this 7000 series ryzen 5 I got, with Zen 3/Zen 2/Zen 4. Ultra is literally infront of the numbers too lmao. Idk how much more obvious you can make it for normies.

For the tiers they were mixing the silicon for? RPC is esentially just GLC with more cache and some structure placement, and the differences between the two are extremely minimal. Pretending this is on the same level of mixing Zen 2/Zen 3/Zen 4 chips is just disingenuous.

Except that's the thing, you literally can't. TGL isn't 13th gen. Zen 2 chips are part of the 7000 series though. The only defense against this is that RPL used ADL silicon for the low end, but the difference between that and Zen 2 and Zen 4 being under the same generation is drastic.

Intel's naming system allows them to do the same thing AMD is doing btw, they just haven't done it except with RPL, and even then the difference between the two products they mixed is insanely small. And even then, they got large and deserved backlash for it. AMD's naming scheme means they are forced to show the differences in generations, but the placement of where that number is esentially means for normies, they won't pay attention to it at all, and end up being worse off because of it.

AMD's naming conventions are simply worse for the end consumer and more misleading than Intel's. The difference is day and night. At best, AMD has a theoretical advantage in showing the arch while Intel doesn't, but in practice, AMD is the one misleading customers more with more drastic mixing of archs and products, because the general consumer just isn't that observant.

You keep referring to consumer confusion as if consumers have any clue what any of the model numbers mean. They don't. AMD's scheme is no more confusing to them than Intel's because they have no clue what Intel's is either. There is 0% chance a non-informed user would know which model was newer or better between an xg7-1065z and an xg7-10700. Can you tell me which of those models is the newer generation and what is the performance and efficiency difference between them? How much better is an x6-1165 versus an x5-1145? Is an h5-700 better or an l5-500 better? It's meaningless to someone who isn't informed on the underlying technology.

Both schemes suck in different ways, that's it. You can have a preference for one or the other, but neither can be justified because the motivation is pure marketing.

BTW, the difference between RPL and ADL, as I mentioned, isn't always that small. Just look at the gaming difference between the desktop SKUs. Mid-range RPL gave you above top of the line ADL performance. In laptops, RPL could be significantly more efficient as well. But hiding which generation you are getting you say doesn't matter here because it doesn't meet your own personal threshold for performance to justify a clear distinction.



If you look at the bottom of the range (the only place Zen2 is used), do you really think it's an advantage that a 1P4E 13th gen ADL CPU with 1.6/1.2 GHz base frequencies is truly ADL versus a 4C/8T Zen2 with a 2.4 GHz base frequency? I would take the Zen2 every time, despite it being, "old". Maybe you would prefer the 1P4E ADL, that would be your preference but that's all it is.

This is such a silly argument that I'd rather not spend any more time on. There's a reason Intel's slide show was almost universally criticized and it's not because people just want to hate on them, it's because they are pointing a finger at AMD (with reason) but hoping no one notices they are pointing right back at themselves too.