Question Alder Lake - Official Thread

[Hulk]

With the release of Alder Lake less than a week away and the "Lakes" thread having turned into a nightmare to navigate I thought it might be a good time to start a discussion thread solely for Alder Lake.

 

[Asterox]

The gaming power draw isn't so bad. On the plus side, the included HSF must be pretty stout if it allows the CPU to sustain operation @ 122W.

That is less interesting , in relation to the fact that 4/8 i3 12100F under CPU load heats up just as 6/12 i5 12400F.

 

[Zucker2k]

Combined with unlocked memory tuning this will make DYI very fun indeed. FUN times, love the competition.

Which competition? Surely not in overclocking?

 

[Heartbreaker]

Combined with unlocked memory tuning this will make DYI very fun indeed. FUN times, love the competition.

Though it only works on a handful of expensive boards with external clock generators, and likely only for a short window in time.

 

[coercitiv]

Though it only works on a handful of expensive boards with external clock generators, and likely only for a short window in time.

I was just about to post that, ADL-S has the option to use either an internal or an external clock generator, and value boards will likely use the less expensive option.

 

[JoeRambo]

Though it only works on a handful of expensive boards with external clock generators, and likely only for a short window in time.

Some B660 motherboards supposedly have them + there is obviuos market opportunity for Asrock's of the world. It's not like Celeron 300A OC worked on every motherboard either, obviously DYI users usually have no trouble navigating information and making decisions for themselves.

 

[Mopetar]

Actually, I agree with you. Those are not bad. Its the E-core thing, and the thread director, etc...and I don't need a gaming proc, I need CORES !!! I have even recommended a 12700k for a gamer or two, on this forum ! But yes, the mobo prices and DDR5 is another issue.

I think that gaming benchmarks generally tended to favor DDR4, so that isn't as much of an issue for people who are just looking for a gaming rig or aren't running any of the applications that tend to benefit from DDR5. The new chipsets from Intel along with the less expensive boards are also starting to hit the market now so I think it's a lot less of an issue in terms of pricing.

I'd really like to seen Intel mix up their offerings in future products. I'd like to see a line of CPUs that only has 4P cores, but up to 32E cores. Obviously that isn't something that every customer will be able to take advantage of, but it would actually give Intel something that could compete in the same space that Threadripper now completely owns.

 

[Markfw]

I think that gaming benchmarks generally tended to favor DDR4, so that isn't as much of an issue for people who are just looking for a gaming rig or aren't running any of the applications that tend to benefit from DDR5. The new chipsets from Intel along with the less expensive boards are also starting to hit the market now so I think it's a lot less of an issue in terms of pricing.

I'd really like to seen Intel mix up their offerings in future products. I'd like to see a line of CPUs that only has 4P cores, but up to 32E cores. Obviously that isn't something that every customer will be able to take advantage of, but it would actually give Intel something that could compete in the same space that Threadripper now completely owns.

A threadripper with E-cores ? I don't think it could touch existing threadripper, let alone future ones. ALL the cores are strong, and 64 strong cores will demolish 32 weak ones in anything.

 

[Mopetar]

A threadripper with E-cores ? I don't think it could touch existing threadripper, let alone future ones. ALL the cores are strong, and 64 strong cores will demolish 32 weak ones in anything.

Really Intel can't compete with the full lineup because it would require a massive die akin to one of their server chips and they're only just starting to apply something akin to AMD's chiplet based approach that might make such a product feasible. We haven't seen how Sapphire Rapids will ultimately perform, but even if it does do well it still won't quite match AMD in terms of core count.

The real problem for Intel though is that their P cores are too big in terms of area. Even if they can theoretically outperform AMD, they won't be able to compete on raw core count due to space limitations. The E cores actually solve this problem and with Alder Lake we saw that 4 E cores only take up about the same amount of space as 1 P core.

The tile based approach from Intel also doesn't help them compete easily against the full range of Threadripper CPUs. Because each tile contains some of the IO, it's going to require using 4 tiles to get a full set of PCI lanes and memory channels. On the other hand even the Threadripper parts with fewer chiplets can still offer the same memory channels and PCIe lanes because the IO die handles that. A 3945WX only has 2 CCDs, but it still has 8 memory channels and 120 PCIe lanes, just the same as the 8 CCD 3995WX.

 

[IntelUser2000]

The real problem for Intel though is that their P cores are too big in terms of area. Even if they can theoretically outperform AMD, they won't be able to compete on raw core count due to space limitations. The E cores actually solve this problem and with Alder Lake we saw that 4 E cores only take up about the same amount of space as 1 P core.

In HEDT there will be Sapphire Rapids-X with full fledged Golden Cove cores and 56 of them so they won't have a performance problem.

Sure it'll be large but they aren't going to be charging just $500 per chip anyways.

 

[Mopetar]

In HEDT there will be Sapphire Rapids-X with full fledged Golden Cove cores and 56 of them so they won't have a performance problem.

Sure it'll be large but they aren't going to be charging just $500 per chip anyways.

Intel might be able to get close, but AMD won't stand still. Even if Threadripper isn't a major priority for AMD and Intel can close the gap, AMD won't sit still and there will be Threadripper parts with more than 64-cores eventually and then Intel is back to where they started all over again.

Golden Cove isn't a good solution for a variety of reasons. The full 56-core version has a 350W TDP according to some leaks and you have to drop down to 44-cores to get a 270W TDP that matches up more close with the 280W TDP in AMD's 64-core variants. Each of the 14-core tiles is also quite large, around 400 mm^2. That means a lot of silicon to be able to sell parts with lower core counts if they want to be able to offer a full number of memory channels and PCI lanes on those parts.

AMD can offer full PCI lanes and memory channels regardless of chiplet count because that functionality is part of the IO die. AMD's cores also take up a lot less area which means they can get far more chiplets per wafer compared to Intel, even though Intel isn't making monolithic dies any longer. Intel would be a lot better off trying to compete against AMD by having a part that uses a large number of E cores which take up far less die space and are more energy efficient. Golden Cove is not a silver bullet here.

 

[CHADBOGA]

With AMD seemingly ultra focused on servers, it looks like Intel is going to take a lot of desktop share back from AMD in 2022.

 

[Markfw]

With AMD seemingly ultra focused on servers, it looks like Intel is going to take a lot of desktop share back from AMD in 2022.

If I were a betting man, I would put up $10,000 to you and say NO. Look at the stats as to how Intel is selling at mindfactory (I think thats it) ! 30% to 70% AMD in Nov 2021 ? Or was it Dec.

Anyway, think what you want....

 

[Mopetar]

With AMD seemingly ultra focused on servers, it looks like Intel is going to take a lot of desktop share back from AMD in 2022.

It probably depends on when Zen 4 comes out and whether or not AMD cuts Zen 3 prices in the lead up to the launch of Zen 4. And as odd as it may sound, even though they're moving to a new node AMD might have better supply with Zen 4 since they don't need to split the wafers between CPU, GPUs, and console APUs like they've had to with Zen 3. Intel probably gets a momentary bump, but I don't think it lasts for the entire year.

 

[CHADBOGA]

If I were a betting man, I would put up $10,000 to you and say NO. Look at the stats as to how Intel is selling at mindfactory (I think thats it) ! 30% to 70% AMD in Nov 2021 ? Or was it Dec.

Anyway, think what you want....

Intel's cheap motherboards and lower end CPU's either haven't come out yet or have only just been released, so pointing to Nov 21 is beyond ridiculous.

We can revisit this prediction by the end of the year or when we have figures for the end of the year and I suspect you will be happy that you aren't a gambling man.

 

[CHADBOGA]

It probably depends on when Zen 4 comes out and whether or not AMD cuts Zen 3 prices in the lead up to the launch of Zen 4. And as odd as it may sound, even though they're moving to a new node AMD might have better supply with Zen 4 since they don't need to split the wafers between CPU, GPUs, and console APUs like they've had to with Zen 3. Intel probably gets a momentary bump, but I don't think it lasts for the entire year.

If Zen 4 is DDR 5 only, then I doubt you will see great adoption till some time in 2023.

 

[TheELF]

If I were a betting man, I would put up $10,000 to you and say NO. Look at the stats as to how Intel is selling at mindfactory (I think thats it) ! 30% to 70% AMD in Nov 2021 ? Or was it Dec.

Anyway, think what you want....

Mindfactory sells around 20 thousand CPUs on average per month while the OEMs sell around 70 million PCs per quarter average...
The words drop and hot stone come in mind here.
And while AMD gains traction in OEM systems as well, the huge majority of them are still intel based.

Ryzen 5000 makes Intel's Rocket Lake look like a paper launch in latest Mindfactory data but overall AMD CPU sales also hit new low

AMD has punished Intel in the latest set of Mindfactory sales data, with Vermeer Ryzen 5000 chips like the Ryzen 9 5900X and Ryzen 5 5600X helping make Rocket Lake look like a paper launch. However, overall CPU sales figures for Team Red have plummeted since the Zen 3 desktop chips were released.

www.notebookcheck.net

PC vendor shipments worldwide by quarter 2023 | Statista

Lenovo shipped 16.21 million personal computers during the fourth quarter of 2023, surpassing HP's 13.95 million shipments in that same quarter and retaining its position as market leader.

www.statista.com

 

[IntelUser2000]

Intel's cheap motherboards and lower end CPU's either haven't come out yet or have only just been released, so pointing to Nov 21 is beyond ridiculous.

This is very true. If I was buying Alderlake, just 2 weeks ago I wouldn't have considered it at all. I'd have loved to go DDR5, but the prices are ridiculous. Then the motherboard prices are ridiculous too. I want maximum $200 CDN, not $300 and $400!

Could have been a perfect time to get consumer Optane DIMMs considering the cost of DDR5.

People say GPU prices are expensive but its like "go DDR5 or a dGPU?"

It's better now, but $160 for a cheapest motherboard is still too expensive. What if I want a bare one for $80 to go with 8GB DDR4?

 

[coercitiv]

Intel will definitely flood the market with value parts and they will inevitably get market share back, just not the share they crave for.

• mobile will see lower loss, AMD will be much more resilient here and we'll keep seeing strong adoption from major OEMs
• on the DYI front AMD will stay competitive at the top, probably winning some of the flagship battles

In other words AMD will probably yield the best results with the limited supply they have: steady mobile market share and performance mind share. So, to paraphrase Pat Gelsinger, AMD won't be in the rear view mirror, and they won't be in the windshield either: they'll be riding shotgun and giving Intel instruction on how to drive or else...

Pat had a very bombastic way of telling people he has managed to stop the massive internal bleed and Intel is now in intensive care, aiming for a full recovery.

 

[IntelUser2000]

Intel should be at a significant disadvantage in mobile especially at the U space since AMD has a substantial MT performance lead. Yet Tigerlake-U did quite well.

Alderlake significantly reduces the MT gap, but makes you wonder if it matters if AMD couldn't dent Intel's share with the massive MT lead. Rembrandt's battery life advantage might be a bigger deal here.

Pat had a very bombastic way of telling people he has managed to stop the massive internal bleed and Intel is now in intensive care, aiming for a full recovery.

Sounds like a well thought and at the same time hilarious overview of the market.

 

[Mopetar]

If Zen 4 is DDR 5 only, then I doubt you will see great adoption till some time in 2023.

They've got the advantage of launching when they want to. At best we see Zen 4 in early July which is still half a year away. They could even hold off for an autumn launch if necessary, but DDR5 will have had that much time to work itself out.

If AMD already had committed to DDR5 as the only option I would assume that they're working with manufacturers to ensure that there's sufficient supply to cover their launch.

 

[Khato]

Intel should be at a significant disadvantage in mobile especially at the U space since AMD has a substantial MT performance lead. Yet Tigerlake-U did quite well.

Alderlake significantly reduces the MT gap, but makes you wonder if it matters if AMD couldn't dent Intel's share with the massive MT lead. Rembrandt's battery life advantage might be a bigger deal here.

I do frequently debate what the typical workload is for <28W laptops? In particular how much the MT performance aspect matters in that segment? I'm sure it's consequential to a subset of users, but what portion of the total market care about anything other than ST performance and battery life? Metrics where Intel is either ahead or effectively tied with AMD. Arguably it's the implementation of the laptop by the OEM which matters more in the majority of the metrics, and there Intel tends to maintain a major advantage both through engineering engagement with the OEMs and programs like Athena.

 

[CHADBOGA]

They've got the advantage of launching when they want to. At best we see Zen 4 in early July which is still half a year away. They could even hold off for an autumn launch if necessary, but DDR5 will have had that much time to work itself out.

If AMD already had committed to DDR5 as the only option I would assume that they're working with manufacturers to ensure that there's sufficient supply to cover their launch.

Haven't AMD indicated that Zen 4 is a 2nd half of the year product?

Whilst that means it could technically ship in July, products that get labelled as 2nd half, tend to be later than that.

 

[CHADBOGA]

So good energy efficiency, but not as good as Zen 3.

So, they contradict themselves... Well, lets bring this up in the appropriate forum.

It doesn't matter what they say as much as what we can see with our own eyes from their testing results graphs.

The energy efficiency differences in my opinion for desktop users isn't significant enough when comparing a 12700K and below from Intel, to their AMD counterparts within roughly the same price range.

 

[Mopetar]

Haven't AMD indicated that Zen 4 is a 2nd half of the year product?

Whilst that means it could technically ship in July, products that get labelled as 2nd half, tend to be later than that.

Yeah, they have a window from July to December. I would imagine that they're basing some of that on DDR5 availability as well as their board partners. They might be ready to launch as soon as July, but could choose to delay if necessary. You are right that H2 usually does mean later rather than sooner, but it could also be that AMD doesn't want to actually commit to anything at this point even if they could be ready with enough Zen 4 CPUs by July. Honestly AMD is in a good enough place right now where I hope they do wait until there's good board and DDR5 availability so that it's a smooth launch.

 

[epsilon84]

If I were a betting man, I would put up $10,000 to you and say NO. Look at the stats as to how Intel is selling at mindfactory (I think thats it) ! 30% to 70% AMD in Nov 2021 ? Or was it Dec.

Anyway, think what you want....

I can't see how Intel won't be gaining marketshare when they actually have a competiive product and have priced it aggressively against the competition.

The bulk of desktop CPU sales are in the lower end segment, a market AMD has effectively abandoned since they never renewed their 3100/3300X quads and their $200 offering is an aging 3600 that is being dominated by the i5 12400.

Of course, AMD can always respond with their own price cuts, but as things stand, its basically guaranteed that Intel will be gaining marketshare in the desktop space IMO.